Thread got out of sync:>>2957310>I'm new to electronics. Where to get started?It is an art/science of applying principles to requirements.Find problem, learn principles, design and verify solution, build, test, post results, repeat.Read the datasheet.>OP source:https://github.com/74HC14/ohmOPbake at page 10, post in old thread>Comprehensive list of electronics resources:https://github.com/kitspace/awesome-electronics>Project ideas:https://hackaday.iohttps://instructables.com/tag/type-id/category-technology/https://adafruit.comhttps://makezine.com/category/electronics/>Books:https://libgen.is/>Principles (by increasing skill level):Mims III, Getting Started in ElectronicsGeier, How to Diagnose & Fix Everything ElectronicKybett & Boysen, All New Electronics Self-Teaching GuideScherz & Monk, Practical Electronics for InventorsHorowitz and Hill, The Art of Electronics>Recommended software tools:KiCAD 6+CircuitmakerLogisim Evolution>Recommended Components/equipment:OctopartLCSCeBay/AliExpress sellers, for component assortments/sample kits (caveat emptor)Local independent electronics distributorsladyada.net/library/procure/hobbyist.html>Most relevant YouTube channels:EEVblogW2AEWMoritz Klein>microcontroller specific problems?>>>/diy/mcg>I have junk, what do?Shitcan it>consumer product support or PC building?>>>/g/>household/premises wiring?More rules-driven than engineering, try /qtddtot/ or sparky general first>antigravity and/or overunity?Go away
idea: DIP plushie
>>2964853What, like this?
I have never changed soldering iron tips.I just always rub it on the golden metal wool stuff until it wets again.I also always solder at the highest temperature setting (480C)
people give conflicting advice on discharging capacitors. some badasses say it's fine to just short them with a screwdriver. but for sensitive circuits it's not just about personal safety and cosmetic damage to the screwdriver from sparks, but you don't want to risk damaging the capacitor itself. a 180 ohm 5 watt resistor would discharge it within a fraction of a second so it's not even that inconvenient to use if you have it on hand?
>>2964975> will not be in continuous use during workdays> IPC standards is mostly for durability and inspectability, not really important for hobby> if something has a bad connections your time is already worthless since nobody is paying you for it, might as well grab that multimeter and find the issue> you are probably not using a suitable sized tip anyway so just cranking up the temperature might act as a crutch in many cases> with high enough temp you often almost doesn't need to wet out the tip because you will still put enough heat into the joint so allow the solder to wet jointFor hobbyist this is honestly fine, it's a hobby... Do whatever you you enjoy and works for you. Until the day you run into some issue and then it's super simple to look up some proper procedures to fix it.
I got a fnirsi cartridge-style soldering iron and I'm amazed at how quickly it heats up. From cold to melting solder in 2-3 seconds. Way faster than my ceramic-heater hakko.Huge fan so far, but I can't recommend or comment on its reliability because I just got i.
>>2964982> people give conflicting advice on discharging capacitorsIt's like EMI but for midwits. Something mysterious that you can rant about the potential dangers about. Only it's incredibly easy to avoid or design away in the case of capacitors.> but you don't want to risk damaging the capacitor itself.If it's critical and sensitive - design proper bleed circuitry so you don't have to short it.> not even that inconvenient to use if you have it on handIf it solves your (potential and/or imagined) problem and is is easy and/or convenient, why are you even asking about it?
>buy weller soldering iron>it was actually QCdis paying 10x the price worth it? yes
>>2964982Then just discharge it with a resistor if it's easy. For what it's worth most capacitors will survive because they either have enough internal inductance/resistance to limit the discharge rate or they're designed for fast discharges anyway. I regularly discharge 100 kV caps that are >$2000 each using a shitty relay or even just a plastic stick with copper braid at the end and have never damaged one doing it.
>>2964999What tips does it use? T245 like the HS02 or DWS200? They’re fast, yeah. And you can use name-brand JBC tips if you have issues with tip longevity. The HS01’s SH72-style tips are also still decently fast, but because it’s a chinky tip design you’re stuck with what comes out of china.If you do have an HS01/02, what’s the ergonomics like? They look a fair bit chunkier compared to the TS100-style irons, but the cap that fits over them is nice looking. On the other hand, that Miniware TS1M looks really nice.>>2965009>buy noname chinky iron>find free online schematic for it reverse-engineered by russians>put custom open-source firmware on it>can buy tips from multiple different manufacturers and suppliers>mosfet dies after 5 years, look at schematic and fix itFeels good to easily be able to do whatever the hell I want on what I bought and paid for.
I want one of those soldering stations with arms and a magnifying glass. Are there any that aren't shit? Most of the reviews are terrible`
>>2965044>soldering stations with arms and a magnifying glassreputable soldering iron manufacturers dont create such carnival monstrosities
>>2965068I mean, let's not pretend people aren't bolting articulating magnifying glasses to their workbenches
>>2965069>workbenchesmakes more sense to bolt them to your headthe lenses follow your eyes (and tilt) everywhere automatically
>>2965025I got the HS-02 which takes C245 tips, far as I can tell. Ergonomically it's nice, sort of tri-lobed, easy to grip. The USB-C cable that it comes with is noodly, also nice.I just did another soldering test from cold and it melted the solder in under 2 seconds. I'm still blown away, I honestly don't think I'll ever use my hakko again, and I don't care at all that cartridge tips are more expensive.
>>2965087100W is better than most stations, I’m using a T12 station and it’s 72W or so, which is enough in most cases without having to boost the tip temp. Takes more like 8s to melt, which I’m guessing means it has more thermal mass. Because I’m already using T12s I bought an Si012 as my portable iron, it was the only T12 iron with both a USB C socket and a DC socket. At 20V USB PD though it’s only 50W, which can feel lacking at times. They could improve the iron by giving it a third set of contacts to take SH72 tips (which are lower resistance), in addition to TS/ST and T12 tips. IronOS would be nice too.I still think someone should make a portable iron that can shunt 200W to a T245 cartridge, or even 400W to the extra fat JBC cartridges, with an XT60 connector moulded into the back. USB C could be a dongle if there’s no room, better than doing it the other way around.Or for network admins, a PoE-powered soldering iron!
>>2965088> 200WThat’s a lot if power. You’d need AI assist with the premium monthly plan for your digital AI soldering iron.
>>2965107Dimmer iron anon again? Stay your mocking keyboard fingers til you can provide a real argument. If I can power through D2PAKs and PowerPak-SO8s, then hot-swap tips to work on an 0402, I see no downside. The Miniware TS1M seems to up to 200W with a very convenient form-factor, though the nice T245 pencil stands are larger than the station itself at this point. You supply your own power supply, but that’s pretty easy for hobbyists with all sorts of power bricks lying about, or RC technicians with their big lipos.
Never soldered anything in my life that worked afterwards.I need to solder 2 small joints for TSOP flashing.Will conductive paint marker/pen do the job? It just needs to last long enough for the flashing.
>>2965117You’d be better off with a programming spring-clip or a custom pogo-pin jig. The latter is doable if you can 3D print such a jig.I suspect conductive paint/ink/varnish/glue won’t have low enough resistance, but you could still give it a try. Just make sure you don’t short anything with it. And that you can clean it off afterwards.
>>2965117Learn how to solder.
>>2965117https://www.youtube.com/@jkgamm041/videos
fucking inductors are getting really tiny
>>2965117You're likely not using an appropriate amount of flux.That's the number one mistake people make and you cannot get away without using flux.
I need a "3rd hand" for soldering. I'm mostly interested in working with small PCBs so I went and looked at what's available for cheap. Which of these would you pick? Which of these is the most useful and the best?
>>2964845Recommendations for benchtop variable power supply? $100-200 range
>>2965198Blue Tack putty.
>>2965199what specs?
>>2965199Best value for money in that range is Riden's RD60xx range.I got a RD 6018, it does 60V and 18A so 1080W max output.The RD 6018 unit is a DC to DC converter, fed from a DC power supply.You can buy just the RD 6018 front panel unit solo, an empty case with prepared wires you just have to screw in, the loose power supply, or an entire kit that takes 5-10 minutes to assemble with nothing but a screwdriver.I got the kit with the front panel and the case:https://aliexpress.com/item/1005001488070904.htmlBut bought the PSU from a different vendor which was cheaper:https://aliexpress.com/item/1005003293483757.html (60V 20A variant)Most of these brick power supplies are adjustable, the one I bought was advertised as 60V, but I adjusted it up to 68V I think, which is the max input the RD6018 can handle. The max output I get on the RD6018 is now 62V or so. They got a little potentiometer screw near the terminals for adjustment.It's pretty accurate and has pretty low ripple and the power is clean and stable, even the snobs at EEVBlog and the like don't dare to shit on it.That should be well within a 200 bucks budget even with shipping and it's pretty much the beefiest option you can get before dipping into highend professional stuff.I have to say though I rarely use the max power of this thing, and having 2 or 3 lab power supplies of lower capacity would be kinda nice. sometimes you need 3.3V 5V 12V all at once but only 1-2 Amps total and want current limiting on each voltage.
>>2965198Any of those will work for soldering wires together. All but 3 won’t be useful for holding a PCB as you work on it. When soldering on a PCB you will need to put some force on the iron, and those flexible arms just bend. You need something that locks into place, like 3, or like a conventional PCB vice. I use an articulated PCB vice (TH1980 from Jaycar) which clamps well enough and allows me to spin it around to easily work on both sides of a board. You can also 3D print these. But it’s also handy to have some alligator clip bendy arms to hold wires onto the board while you solder them, some magnetic-base gooseneck arms should sit on the base of most PCB vices.>>2965211Actually for digital stuff, instead of current limiting, I think you’re often better off with overcurrent shutoff. With a manual reset button to enable the output(s) again. It wouldn’t need to be continually adjustable, having a few settings from 50mA up to 5A would probably be enough. Ideally if any rail goes over its threshold, all rails would shut off. Seems like a simple project to make as a front-end to a spare ATX power supply, though maybe that’s overkill. You get +/-12V that you could use for analogue circuitry, but that negative rail is very low current and it might be kinda noisy.
>>2965199What for? If it’s load testing, you want a high power CC/CV supply, and noise doesn’t matter so SMPS is fine. If it’s digital electronics, then you’d probably be fine with fixed rails and with relatively low power draw. If it’s for analogue electronics, you’d want something split-rail and low-noise, so a multi-tap transformer with a conventional linear regulator array. If you need low-noise and high-power with continuously adjustable voltage and/or current, that’s when you look towards what I call autoranging-transformer linear supplies, they’re big and heavy compared to switching supplies for the same power output. Your picture there is probably such a supply. They might also be harder to kill.Fried a switching supply the other day, I was putting it in series with another supply to get more voltage and wanted to test the short circuit current, but there was no internal protection diode. Smoke.
>>2965211>>2965216>>2965219Just go with Korad KA600
>>2965177When would you ever need such a small ferrite inductor? Ferrites are only good up to a few MHz, at which 120nH gives you just a few ohms. The open core design compared to thin film inductors is also probably an issue.
I cant tin my soldering iron
>>2965239You have to do a few things to prolong their life. First, use RMA flux and 60/40 or 63/37 solder. Second, don't use the tip to pry or gouge. Its only purpose is to get things hot. Third, temp control. Fourth, when you're finished soldering, clean the tip and load it with a blob of solder to protect it from oxidizing.
>>2965242Another thing is keeping the tip clean while you work. A very slightly dampened sponge and/or brass wool works well. Some people will bark at you for suggesting a damp sponge because it thermally stresses the tip, but you don't need to use it every time you make a joint.
>>2964845On this blown variac, can I do the following solderjob without causing too much of an issue besides a dead spot on the knob? I'd also un-short/trim back the clearly shorted windings in the middle.
>>2965239Wipe the tip off with a facial/toilet tissue while it's hot. Does it look white and shiny? Does it look like the plating has flaked off? Photos.>>2965249If that's the winding direction, yes. But you may want to skip a winding and do the bodge elsewhere if that's where the wiper runs across.
>>2965249>without causing too much of an issue a better idea is to short it all as shown in yellow- you wont get a dead spot, causing harmful voltage spikes, just an invariable spot- you're out of the way of the brush or slider, thus preventing harm to it
>>2965219Good advice.Also, I’m finding a variac more and more useful.For old wall-warts and other linear power supplies with transformers in them, you can adjust the output voltage to anything you want.The things are robust as hell, and so sometimes you just want AC. They’re also great for fine-tuning mains powered things like fans and soldering iron wattage.I think, in retrospect, a variac should be one of the first things you should get when entering into this hobby.
>>2965238It's hard to appreciate weird niche components until you really really need one
>>2965269They’re big and heavy, and quite expensive if you get them new. Either expensive from a local shop or expensive to ship from an e-commerce vendor due to the weight.For 99% of use cases you can use a TRIAC dimmer, a multi-tap transformer or two, or an adjustable output DC supply. All together those will be cheaper and lighter. Only for sketchy high power AC loads might you need one, like tesla coils or vacuum tube circuits.
>>2965263You'd also need to sever the two remaining wires, else you'd get some shorted windings. And I don't know how well the wiper will slide across those cut windings. You could cut across the yellow lines (and solder them at the cut ends), but then solder a piece of copper flat-bar to one side to be a nice smooth thing for the wiper to slide across.On this topic, don't the wipers in these things intrinsically contact multiple windings at once, hence making a few shorted windings at any one time?Actually I just looked this up:https://www.youtube.com/watch?v=KkXW0ouTkNwApparently they do and it's not too significant. In the comments Mike's Electric Stuff is mentioning how a higher power variac had deliberately significant resistance across the carbon brush to minimise the shorted winding current.
>>2965318>also need to sever the two remaining wiresau contraire, mon frèredoing so would cause an open circuit and the *whole* thing would stop working correctly
>>2965331Well you have to keep one connection around the transformer, but I was thinking it would be better to be at one end. If you use one of the middle windings as the connection then break the other to stop from creating a shorted winding.It might actually be worth peeling the broken wires out of the gungy lacquer on the top and side and soldering on a solid-core bodge wire for each winding. Doesn't matter if it's bulky and ugly over there, so long as it's nice and smooth where the wiper goes.
>>2965177Combine it with the world's smallest microcontroller, world's smallest diode and world's smallest transistor to create the world's smallest switch mode power supply.
How do I keep potentiometers from sliding around if they don't have the little tabs on them? Glue? I tried lock washers once and it didn't even work
>>2965341by filing out an oblong hole if the threaded shaft has flats on it, or gluing a bracket flush to the back of the panel if the potentiometer body isn't roundhard to imagine any pot lacks a locking feature, unless it's strictly designed to be pcb mounted
>>2965198>>2965198I got 4 and its ass. The light is very dim. When I put my iron in the holder, the tip hits the plastic and melts it. The attachment to the main support is always wobbly. The arms are fixed length apart so its hard to get the correct position, and there's only 2 of them.Pony up some extra cash to get either 1 or 2. Idk how 3 works.
i want to get my dad a soldering iron for christmas, his hobby is repairing old radios, a couple years ago I got him an ersa 0920BD i think, he said it worked well but he somehow bricked the tip.you guys got any recommendations
Retard coming through, new to building shit. Been working on a gps based gokart racing datalogger for a while and now stole this circuit from an Amazon tachometer for RPM loggingIt works, works great, just no fucking idea howSo you wrap a wire around the ignition coil 3-5 times, and the other end is placed near the antenna (little 5x5mm piece of steel)is a half loop/horseshoe configuration How the fuck does this work
>>2965374>How the fuck does this workInduction.
>>2965198I use blue painter's tape to hold things to the desk
>>2965374fucking magnets
>>2965374common-emitter amplifier. wrapping the wire around the coil "inductively couples" the two, essentially creating a transformer.
>>2965392get out of here with your learns
This is CNC controller anon again, I'm trying to wire up an emergency stop switch with this circuit I came up with and I'm having problems. The controller board inputs are all logic low, they get a signal when you short them to ground so I thought I could use a transistor to short the input pin when the switch is on. BUT when this is hooked up the control board is always reading an input in both the on and off position... and I'm pretty sure my transistor is fried, I'm reading 80kOhm between base and emitter and 14MOhm between collector and emitter. So I have some questions:1. Would this circuit work if the transistor was ok or do I need a 'pull down' resistor?2. Did this circuit fry the transistor or was it likely damaged from before?3. Should I be using an entirely different circuit?
>>2965428What if I added a 10k or higher resistor between +5V and the collector?
>>2965356I am going to ignore what worked well and recommend you temperature regulated irons:If brands don’t matter, anything between:>Yihua 928D III>KSGER T12>Aixun T3AIf brands do matter:>Hakko FX600 series, Weller W61, Goot PX-335/PX-201>Hakko FX888D or equivalent, we’re getting into pricy town compared to your first giftIf gimmicky irons are fine:>Pine64 Pinecil + power supply>Miniware TS1M T245 + power supply>Ifixit FixHubGood luck>>2965428When you say the control board is always receiving an input, do you mean it’s always being pulled down, or it’s never being pulled down?If never, it may be that the saturation voltage of your BJT is higher than the threshold voltage for a logical low. A logic MOSFET like a 2N7000 should avoid this issue. A dodgy transistor could also cause this, since the control board’s pin works by sourcing current. It’s also feasible that a 10k base resistor is too high, but probably not.If it’s always getting pulled down, then definitely a dodgy transistor. A pull-down resistor shouldn’t be necessary for a BJT’s base, unless the relay board logic input is sourcing current somehow.Resistance measurements are useless for non-ohmic devices like transistors and diodes, you’re better off measuring voltages while it’s turned on. Especially the base-emitter and collector-emitter voltages in both switch states.You could simplify the circuit by having a switch that pulls either control board’s pin pin or relay board to 0V, with a pull-up resistor on the relay board, and LEDs up to 5V from each. Maybe a diode on the control board’s input too if it’s not able to handle 5V.
>>2965356Also can you not just buy him some replacement tips for his existing iron? Then the main gift could be a desoldering iron like a Yihua 929D 3, they’re handy for repairs.
>>2965456you can't easily buy replacements for that iron where we live>>2965452ok thank I'll look into these
>>2965356get at least two tips for him instead of a new iron...and every time he fucks one just order anotherthis fag is right >>2965456 a desoldering iron its really nice to have
I'd like to DIY an IR blaster from a kit. I have a bunch of electronics around the place that use remotes and, honestly, most of the remotes suck. Some take coin batteries and have shit range. Others are just annoying to track down. I'd like one to rule them all. Something like the Flipper Zero is overkill. I don't want to spend $200 on it as I just need IR.My phone does not have a built in IR blaster and I don't want to use my phone to control it. I can solder but I am not a programmer. Is there anything that fits my needs?
>>2965476I looked into this a while back and the IRremote library for arduino is compatible with every protocol under the sun, and it can both receive and transmit these. In the example codes that come with the library, there's sketches to record IR signals and send them to your computer over the USB serial connection. But if you want to make it as good as you posibly can, personally I'd use an ESP32 to host a local web interface, allowing you to connect to its IP with your (phone) web browser and have a convenient UI without something else to hold. It could also be much easier to customise, for adding new remote signals without having to add new buttons or whatever. This also frees you up to have a powerful blaster plugged into wall power.See projects like these:https://iot-kmutnb.github.io/blogs/mini-projects/esp32_ir_remote/https://github.com/e-tinkers/esp32_ir_remote
>>2965452Those are quite outdated recommendations.Sequre S99 - around 30 bucks on aliexpress/banggood/ ...original JBC C245 tip with 2.5 Ohms for 150W - ebay, around 20 bucks https://www.ebay.de/itm/312729462360cheap tips from china have 5Ohms, it will "only" do 75W with those150W USB-C power supply of your choice>>2965460>you can't easily buy replacements for that iron where we liveYou realise you can order stuff online? Are you Amish or something?
>>2965229>30V>5A>90 bucksyeah nah thx.
>>2965452I replaced the transistor and added a 10k 'pull up' resistor from 5V to the collector and it works now.
>>2965476>>2965480The M5Stack Cardputer looks like a good candidate. For $30 it has a built in IR blaster as well as a ton of other stuff.
>>2965345They know everyone breaks the tab off anyway so they stopped making it like that.I'm currently trying to get the thinnest double sided tape I can find and then cut circles out of it. 7mm hole surrounded by a mm or two of adhesive would work perfectly.
>>2965520>>2965345forgot picture. I guess I could switch to those mini pcb pots but they cost more and don't have solder lugs
>>2965483Yeah, I guess T245 USB C irons are probably a better recommendation than the Pinecil, but no IronOS, no schematic, >cheap tips from china have 5Ohms, it will "only" do 75W with thoseI have limited experience to the contrary buying standalone chinky T245 tips, but I agree that buying name-brand tips is probably the way to go unless you can confirm what you're buying from a sufficiently transparent vendor.But you might actually be better off with 5 ohm tips. Those USB C power supplies max out at 5A from what I've seen, so a 28V 140W supply is better matched to a 5 ohm tip than a 2.5 ohm tip. In both cases you'd have to PWM the element to lower the average current to 5A, but a 5 ohm tip would be 125W, while a 2.5 ohm tip would be only 62.5W. It would be better if instead of PWM it told the PD supply to adjust down the voltage, I don't know to what extent USB C irons use PPS.The 2.5 ohm tips are better for getting the full 200W out of a 24V supply.>>2965488It's a linear supply, they'll always be more expensive than a switched-mode one. Even the chinky linear supplies only go down to $65 or so. Many consider the low noise and robustness of them to be worth the price, though I think you don't often need your CC/CV load-testing supply to be the same as your low-noise supply, particularly if you want split-rails.>>2965499For $3 you get an ESP8266 module with an IR LED on it>>2965522The flat sides of the metal piece atop the FR2 board is something you could make a locking feature out of. Same for the protruding flat sides of the FR2 board itself.
Why on earth does adafruit resell ws2812 LEDs but call them "neopixels"Seems like pic related
>>2965577She can't justify the jewish price gouging without adding rainbow faggotry to the description and advertising it to the weirdos on Hack-a-dong.
>>2965531>For $3 you get an ESP8266 module with an IR LED on itThat would require they program stuff for it, right?>>2965577This: >>2965581 They overcharge for everything. I made the mistake of buying an analog joystick from them for $20 just to find out its the same model that everyone else sells for $5-$7.
>>2965581>>2965583It steams my hams how much they profit from keeping hobbyists ignorant. So much of what they sell has insane markup and it's marketed in the most cringe, irritating way.But I guess it works.
>>2965392That part I understand, I don't entirely understand how its tripping the transistor to dump the line
>>2965593>I don't entirely understandit's not magic, it's magnetism, which is higher than magicthe ignition coil gets pulses of current, each one creating a magnetic field the added winding captures said fieldplacing a metal antenna next to the winding captures some of that field, producing a voltagenegative portion of these pulses are killed off by D1positive portions turn on the Q1
>>2964959Is he still available somewhere? I wanted to buy him but I can't find it anywhere probably because adafruit discontinued making them
Anyone got any recs for a cheapo microscope for SMD soldering?
>>2965621You want a binocular one for depth perception.I recently bought a FYSCOPE 3.5x-90x head of ali for 109€ on sale.https://de.aliexpress.com/item/32949932295.htmlThe head with focus length adjustment was 31€ on sale.https://de.aliexpress.com/item/1005005117568194.htmlI've build the base myself, it's just a 32mm pipe in a base plate and I didn't see any I liked for a reasonable price.You still need a ring light to clamp to it, I bought a polarizing one but it doesn't polarize, so cannot recommend the one I got.Polarizing is nice for getting far superior contrast, non-polarized sucks.Idk if that's "cheap" enough for you but that's as cheap as I've found for something that has no latency and depth perception which both seems quite important.
>>2965638>binocular one for depth perceptiontrash. you're bent over inhaling solder fumes trying to see shit through some tiny tubes.much better to put it on a screen.
>>2965583>>2965592I'm absolutely certain that a portion of that money goes to political terrorist orgs like BLM and antifa.
>>2965667No doubt, just look at their nail paint alone in their promo videos
>>2965638Not him, but that’s four times what I spent on my soldering station. I think binocular microscopes are the way to go for soldering with them because of the lack of latency (along with a fume extractor), while digital systems are at least more cost effective for inspection and allow you to take nice macro photos. What zoom levels are you finding useful? 90 seems kinda overkill.I’ve been wanting to make a digital microscope moving arm, with a lamp and fume extracting duct built in, and some diy rotatable polarised discs for the light and microscope. But I suspect I’d need the same kind of rigid mount for the microscope camera as your optical system, as opposed to a more mobile swinging arm like I’d want for a lamp. A wider depth of field isn't really an option at such extreme zoom levels. But I guess it is an option to try to make a highly rigid locking articulated arm, so long as it has a few mm of fine focal adjustment for the microscope itself. Maybe counterweights too.
>>2965692>Maybe counterweights too.It would be cooler if it was stabilized with a gyroscope.
>>2965693Gyro would work for rotational stability, but that's easy to do with rigid mounting on linear guide rails. It's more the linear stabilisation that's important. And if you're fighting against gravity with a gyro on a cantilever, you need one of those active gyro mechanisms that measures the precessive force and uses that connected to a motor.
>>2965646much better to not be poor.
Is the Digilent 410-380 microSD Card Slot compatible with the ESP32 S3? The supply voltage is indicated as a range from 2.7 V to 3.6 V. I'm pretty new to electronics, so I have the need to ask before buying a unit (link for details: https://mou.sr/4qhBqdy)
>>2965638>>2965646>>2965744>>2965621Frens, I finally acquired it! Got back home like an hour ago. The microscope is fucking mint and it's exactly the kind I wanted: AmScope Simul-Focal Stereo.It's used but like new. No dents, no nicks no flaws of any kind. The stand is the upgraded extra heavy-duty kind and not the one that it comes with.Bought it off a guy who's a legit motherboard repair guy! Self-employed, been doing it for over 30 years. He only deals with computer stores and companies and not regular customers. Super cool dude. We talked for like an hour about electronics and business. He specializes in Macs because that's where the money is but he has worked on many laptops.Dude ended up gifting me a really decent soldering iron with a shitton of tips and about a dozen Mac motherboards to practice on. Some even work fully. Oh, and he gave me an LED light that I just found out that it's expensive af.Anyway, an amazing purchase and a super new fren contact acquired. It's better than Christmas.Anyway, what else do I need? Any suggestions/tips? It's my first microscope. Never used one before (I worked with a shitty USB cam).
>>2965745it appears to be a passive adapter so it should cause any issues.>>2965748very nice anon. i use a trinocular amscope at work and its amazing.>what else do I need?what else do you have? what else are you trying to do?a fume extractor would be nice, if you dont have one.
>>2965749thank you for your input, anon
>>2965749>>2965750shouldnt*, in case it wasnt obvious
>>2965751I was aware, no worries
>>2965749>very nice anon. i use a trinocular amscope at work and its amazing.I've never used a microscope before. Never even looked through it in person. It's about 100x better than a digital one. You can actually have a depth perception.>what else do you have? what else are you trying to do?Repair laptops and gaming consoles.>a fume extractor would be nice, if you dont have one.Yeah, I don't have one of those. I saw one guy make one from stuff at the hardware store. Might go with that for the time being.Anyway, pic relates in the soldering iron he gifted me. It's brand new and came with two dozen tips and solder wire.
>>2965755>Repair laptops and gaming consoles.fume extractor, hot air station, flux, wick, gold wool, uh... a multimeter probably. oh, and copper/aluminum tape for blocking hot air. and stencils + solder paste if you plan on doing reballing.>Anyway, pic relates in the soldering iron he gifted me. It's brand new and came with two dozen tips and solder wire.find the tip that looks like pic rel. its called a "bevel" or "mini-wave" tip. its the best tip. use only it.
>>2965762>fume extractor, hot air station, flux, wick, gold wool, uh... a multimeter probably. oh, and copper/aluminum tape for blocking hot air. and stencils + solder paste if you plan on doing reballing.Thanks. Took notes. Will grab all of these things. Any recommendation on which flux to get?>find the tip that looks like pic rel. its called a "bevel" or "mini-wave" tip. its the best tip. use only it.Found it! He gave me like 5-6 of that type in various sizes. It has to be his fav tip or something. I basically have a whole line of Mechanic tips now.I've always used the type that looks like a flathead screwdriver... it came with an old soldering iron that I have and never changed it for something better.
>>2965646>you're bent overUh no you sit normally, there's a mirror to bend the light in the microscope.>inhaling solder fumesYou can't afford an extractor?>much better to put it on a screen.So you have never actually tried to solder anything, I see.The screens are only useful for showing things to apprentices or an audience, you can't work with them due to lack of depth perception.
>>2965692>Not him, but that’s four times what I spent on my soldering station.It's my most expensive piece of soldering kit, but I started doing SMD shit of a size where working without a magnification is really a huge pain in the arse and my eyes aren't getting better with age.I simply don't see any cheaper solution that's adequate.>while digital systems are at least more cost effective for inspection and allow you to take nice macro photos.If you spend 50-60 bucks more you get a camera port on them, "trinocular" but I have no need for that.>What zoom levels are you finding useful? 90 seems kinda overkill.The working distance is 165mm at 3.5x zoom and that is already OK to work on most things. But i find you can zoom in at like 10x or 15x and still have easily over 100mm working distance and enough field of view to work with.I think the 3.5x-45x configuration is plenty good for soldering, but the cheapest kit of head + attachment lenses I found included a 2nd lens so you can go up to 90x.The real range(s) of this set are 3.5x-45x and 7x-90x and switching between those requires swapping the bottom lens.
>>2965748I saw people on youtube do cheap polarizing film mods on these, you should check that out. Polarized light bumps the contrast like crazy, super useful.
>>2965777Dark-field illumination exposes a lot of defects too. Good for finding broken traces and corrosion damage, among other defects.
>>2965748I would have thought louis would have given you a lifetime’s supply of flux
>>2965577I think it's a holdover from almost 15 years ago when individually addressable LED strips were somewhat hard to find. They had documentation in english and shipped from US. I almost bought some but the price was outrageous even then.That reminds me, anybody remember when the big thing was like 10 x 10 x 10 cubes made of 5mm LEDs with the legs soldered together?
>>2965084I ordered the most popular set on Amazon for ~$20 and the only delivery option was Christmas Eve. Sorry random delivery driver
>>2965748Cool stuff. I've seen those ring lights on alibay for under $30, but I don't know if you got a branded one. I also recommend the polarised technique, but from what I understand you want to be able to independantly rotate the light's film and the microscope's film. That might require a fair bit of funky work, especially if you don't have a 3D printer.>>2965775>but I started doing SMD shit of a size where working without a magnification is really a huge pain in the arse and my eyes aren't getting better with ageYeah fair enough, I don't exactly do any computer or phone work. My eyes are steadily getting worse, but for now at least it's nearsightedness in one eye that actually makes electronics easier. What isn't easy is having my workpiece 150mm from my eye without straining my back. I hold my breath, a fume extractor is a good idea.>If you spend 50-60 bucks more you get a camera port on them, "trinocular" but I have no need for thatI gathered that from the above posts, if anything I'd just want a lens assembly that only has a camera port. Mainly for inspecting boards after I've soldered them, or trying to reverse engineer them. I guess it's feasible to use a really long exposure and light the board from behind to see through it.>But i find you can zoom in at like 10x or 15x and still have easily over 100mm working distance and enough field of view to work withGood to know.>>2965778Hm, I wonder how polarisation of the light or camera effects this? I know glancing reflections are preferentially polarised so it might not be that useful. I also wonder how useful it might be to light up the subject from different directions/polarities with different coloured lights, though I guess it's less useful with a binocular microscope.>>2965794LED cubes are still kinda cool because everyone wants volumetric displays to be practical, but it looks like other methods are better. See Ancient's videos on youtube. Also that stacked OLED by Sean Hodgins.
>>2965777>I saw people on youtube do cheap polarizing film mods on these, you should check that out. Polarized light bumps the contrast like crazy, super useful.I just watched Dave Jones' video on it and the conclusion is that you can just use a small light and come from the side when you want to read the chip info.I found this webm interesting tho.
>>2965765>Any recommendation on which flux to get?i just bought some chip quik 491, fwiw. i think ive been using 291, never used 491 before (its the halogen-free version of 291). i wanted to try something a little less lung-caustic. but anything from chip quik or amtech should be good. i like gel flex (syringe) for most stuff, but you can try others (liquid pens, paste) and see what you like.>>2965779lol i also thought of louis but theres no way he doesnt know who louis is.>>2965777we have an olympus with dark-field at work. ill post some examples when i get to the office.
>>2965846>i just bought some chip quik 491, fwiw. i think ive been using 291, never used 491 before (its the halogen-free version of 291). i wanted to try something a little less lung-caustic. but anything from chip quik or amtech should be good. i like gel flex (syringe) for most stuff, but you can try others (liquid pens, paste) and see what you like.I used ChipQuik flux a while back (not sure which number it was) and it was quite good but also very sticky. I'm not sure if all of them are sticky or not. They do make amazing low-melt solder too.Amtech is dead. I think the new brand name is Stirri.
>>2965845>I just watched Dave Jones' video on itMy sincerest condolences.>and the conclusion is that you can just use a small light and come from the side when you want to read the chip info.Well for people who have the patience to watch a 2 hour video on something that takes 30 seconds to explain, sure, go dick around with an light at just the right angle, I spend 3 bucks on polarizing filters.
I build a thing, but it's too nervous for my taste. These are self blinking leds on 1.5hz. I can't find any 5mm leds that are 0.5hz, only per 5000 on ali, or few listings on ebay in Australia and the US. Anyone know where to find them and ship to the EU with reasonable shipping cost?
I have an old Toyota Tacoma. The original HVAC blower motor resistor lasted 10 years. The first cheap replacement lasted 4 years. The next lasted 3 years. The newest replacement was a dud from the factory. Can't I just DIY one of these? Its just a couple of different values of resistors and a plate for cooling, right?The cheap ones are $15 and the OEM ones are fucking $70. Can't I just strip whatever is coating that plate with a wire wheel, drill a few holes, solder new resistors into the holes, and be done with it?What value of resistors would I need? How would I keep them from cooking themselves?
>>2965869Replace the cabin air filter and maybe the blower motor.
>>2965874I tested the blower motor with a meter and it is within spec. Works fine on full blast and hasn't popped the fuse. The filter is only a few months old and is clear.
>>2965869The whole squiggly line is the resistive element, and the little lump in the middle may be a thermal fuse. It might be as simple as replacing that thermal fuse, assuming it’s possible to solder to the resistive trace. Otherwise, you could try buying another cheap unit and sticking a heat-sink to either side with thermal epoxy, assuming there’s room.Is it some sort of ballast resistor or is it for heating the air?
>>2965900Fan speed control.
why don't they just print the value of the resistor instead of using that stupid color code. we have the tech to do it easily now. i never use the color code, i just measure loose resistors, but it takes time
>>2965911>i never use the color code, i just measure loose resistorsIt's good practice anyway. You never know how far off tolerance they'll be until you test them.
>>2965911more expensive I assume. They do print it for smd resistors & components
I was considering the idea doing a 3d printed diy hard drive enclosure thing for 2 dozen or more drives as an alternative to the extremely loud, power hungry and expensive even when used enterprise gear. Will a quality power supply from meanwell or whatever be clean and reliable enough to feed hard drives or am I better off just going with standard atx psus?
>>2965924Should be fine, both the PSU and hard drives should have power noise specs you can reference if you’re feeling a touch of the ‘tism. Or compare reliability specs to server power supplies. Maybe there’s some way of running parallel supplies for redundancy.But do you also need a 5V rail? Mean Well do make some dual rail supplies if you need that.
>>2965905Oh that’s strange, but I guess it’s reliable. Heat sinks would work if you want to keep them protruding from that plug thing. Otherwise the power resistors you use will be large and heavy enough that they need extra support for mounting. You’d probably need those big aluminium finned power resistors, rated at 20W or more.
Sooo. I need to design BLDC motor controller. Motor has sine BEMF (according to tongue test no oscope to test properly)What do I do? 3 phase bridge with 2 shunt? 3 shunt 1 shunt? I mean i am fine with BLDC commutation but in future i wanna FOC
>>29659322 shunt minimum for FOC, consider 3 shunt if space isn’t an issue because that’s what some controllers want. Know that low-speed commutation will require actual encoders, back-EMF won’t be enough if you’re only spinning a hundred RPMs.You have three real choices for design:>copy simpleFOC or VESC schematics and firmware to make your job as easy as possible, if not using their existing boards>use a dedicated BLDC control IC like the TMC4671 to do all the math for you, and configuring it with a custom programmed MCU real-time>doing it all from scratch with an MCU reading the currents and position, performing the Park and Clarke transforms, and outputting the PWM signals to the gate driversI went for option two and am mostly regretting it. Haven’t got around to programming it because I’m a codelet. Also I want to go with GaN FETs now. There’s nothing stopping you from keeping the logic side of option 1 unchanged, but altering the gate drivers and transistors to have a different power-rating or form-factor. If you do pick option 3, ensure that your MCU of choice has the speed and power to perform the matrix calculations, that it has the 3-phase PWM output capability (preferably from 6 pins with adjustable dead-times), and that it’s ADCs are fast enough if you don’t plan on using external ones. A floating point unit would be a good idea, though I bet you could manage with fixed-point math.I highly recommend doing a lot of research on ideal MOSFET selection and driving, especially if you’re trying to push for high powers. Most drivers will use bootstrap caps for high-side Nch driving, but some will have an integrated charge-pump that allows the high-side FETs to be on indefinitely. Parasitic turn-on and miller clamping exist, and you might need to care.What’s the project?>inb4 another zoomerslop smart dial
>>2965911Don't be mad if you keep doing electronics you'll just learn it by heart.>>2965915I thought it was because the resistors that use color code are cylindrical and the code can be read from any direction.Blog: Philips Oneblade shavers are nice for shaving but are built like disposable vapes, dropped it and it stopped working because metal connector from + battery terminal to PCB snapped, replaced with piece of stripped solid core 1.5mm2 wire. Worth it? Maybe but a new similar entry level one would be less than two replacement blades.
>>2965925Yeah not going to bother now with the power supply, should have looked before asking but even cheap power supplies cost as much as a midrange atx psu for equivalent current capacity and yes I would also need a 5v supply too so it's probably not worth it. Only advantage the power supply would have is less unneeded cabling but at a higher cost and potentially higher risk of damage to your expensive hdds full of data.
Has anyone worked with one of these? Repair shop that I visited yesterday doesn't even use hot air anymore and they only use this. Guy who I talked to was just a customer service guy so he couldn't tell me more. Is this the future?
Just spent an hour trying to get crimp connectors to work. Fuck these gay little niggers so fucking much. How difficult can it be to crimp something onto a wire? Need to do a load of automotive wiring but I can’t even get this shit to work on my kitchen table.
>>2965744>much better to not be poor.>posts cheap chink shit knockoff scope
>>2965970
>>2965974Tried a pair of those first but it was completely hit and miss whether I would get a decent crimp. Seems to be dependant on wire outer thickness or something.
>>2965970Get a crimp tool of this type. Honestly I don't get those generic "Abiko pliers" and how you'd supposedly crimp with any consistency using them. Or maybe it's a skill issue.
>>2965976Make sure you're stripping the wire to the right length, and lightly pinch the connector onto the insulation with your multi-tool to hold it in place before you crimp.
>>2965974>>2965978>>2965982Ok after another hour of seething I tightened the adjuster on the tool and the crimps seem a lot better so hopefully that’s all that was needed.
>>2965984>hopefully that’s all that was neededno, niggaa proper crimping tool tightens the joint just the right amountwith a manual tool like yours it all depends on your grip, so joints can be too loose or too tighta loose connection in a car is a huge headache to diagnose, and cars vibrate a lotmore than your biggest anal dildo
>>2965989Yeah I’ve got the other type of pliers as well that I picked up in my quest to actually get some decent crimps done but I can’t post a pic as I’ve been stopped from posting images on an imageboard apparently. Also why is it so difficult to find one size of uninsulated spade connector to suit all the different wire sizes. Makes it a pain to switch between wire sizes.
>>2964845how do I figure out a circuit for an electro magnet powered by some rechargable batteris. I want the electro magnet part at the end of a broom handle with the batter maybe a foot or two above it with a trigger stile switch higher up. Only needs to lift about 8 oz at a time. Resistance, capacitors, diodes, wire gauge stuff like that?????
>>2966011>Resistance, capacitors, diodesno need for any of that; just battery + switch + electromagneteasiest way to make the electromagnet is to take apart an old iron transformerhttps://www.youtube.com/watch?v=WhX3BW_Hez4then stick all the E-shaped metal plates together, and use the secondary coil to create magnetic fieldchoosing batteries and coils requires some engineering, or trial and errormy guess is you'd use a battery voltage equal, or slightly higher, than the transformer's secondary voltage
Frens, any input on what makes mosfets a drop-in replacement? I want to replace the 50W rated already-swapped-once ME60N03 that drives the hotend in my 3d printer for something with more power. It seems like either the 2N06L07B (210W) and SL60N06 (100W) are almost identical, except for slightly different RDS (og ME60N03 0.0085 Ohm vs 0.007 Ohm vs 0.0115 Ohm) and Vgs(th) (og 3 vs 2 vs 4v). Any clues on whether either would work?
>>2966022The lower RDS practically is the the additional power rating.What failed on the old one? If it was the gate, maybe spurious noise or static electric discharge got it and not the power dissipation so it might need a snubber/diode/tvm protection.Heatsink it it better.I never buy just one replacement on anything because of the shipping cost. I always buy at least 1 extra one, and usually, nowadays, at least 10 if they're generally useful. Mosfets are generally useful, you’ve got the simplest possible application: insensitive positive temp co, slow switched resistive load.
>>2966024>What failed on the old one?Me being a retard and getting a 70W heater, the mosfet blew open and would've burnt my house down if I weren't near it - one of these days i'm going to add a relay to kill the psu if the hotend's getting power when it shouldn't be>Heatsink it it better.Will do, thanks fren> buy in bulkThat's the plan, since I'd be paying more for the shipping than the mosfet itself. I did the same with the replacement ones, I've still got a few spare, but I want to switch to a bambu clone heatend, and these come with 60W heaters, thus the potential mosfet change
>>2965964What is it?>>2966014Putting DC voltage through a transformer will result in a lot more current flowing than the equivalent AC voltage. You instead measure the DC resistance and estimate the (V^2)/R power rating of your electromagnet based on a thermal resistance that you estimate from maximum temperature and surface area compared to known components. For that size of transformer I’d guess 10-20W, depending on how well heat flows from the coil to the core.>>2966022Voltage rating significantly above the maximum voltage it will see, 30V is fine for a 24V system like a 3D printer. Current rating equal or greater and resistance equal or lower, it’s harder to reason around these two as they depend on the thermal arrangement. Threshold voltage roughly equal or lower, but if you know it’s being driven with 10V then a threshold voltage of 5V instead of 4V isn’t going to matter. Gate charge should be equal or lower, this dictates possible switching speed. If the switching speed is low (e.g. <10kHz) then this doesn’t matter as much. You can calculate the switching speed based off the gate current (limited by a resistor probably) and the gate charge, this should be significantly less than the reciprocal of the switching frequency. Also pinout should be the same, but most TO-200 are GDS.>>2966028Huh, I put a 70W heater in my Ender 3 V2 and it’s handled it just fine.
>>2966030>Huh, I put a 70W heater in my Ender 3 V2 and it’s handled it just fine.It was one of the first mods I did (not counting the fixes, sincerely, fuck the anycubic vyper) and there's a 90% chance I flattened the wires running to the heater with the heatsink and shorted them out. > Gate charge should be equal or lowerSeems to be the only not matching factor in my case, since both the mosfets I listed are 45 and 50nC instead of the stock 22nC. Thanks for the tip on the switching speed calc, I might get around to doing that if the new mosfets don't arrive beforehand - I'll probably just slap them on and see if they work
>>2966030>What is it?After looking it up, tt's an IR laser that heats up computer motherboard parts. It's extremely precise and allows you to desolder or solder parts without affecting neighboring parts on the mobo. It also comes with a library of preset temps for components so you can't damage/melt even the hard to solder shit like connectors that have plastic in them. Oh, and it also has a thermal and a regular camera so you can use it or diagnosing faults too.I've never seen it working and never tried it out. I'm wondering if anyone here has worked with one. The phone repair shop I visited had one and I wrote down the brand & model number but didn't see it working.
>>2965973>cheapIt's not cheap. That;'s just the price of a mounting bracket, retard.
>>2966035and how much is that?
>>2966037NTA but it got me curious, and from their site that machine is currently $1500/€1300, which is way less than I thought. Cool machine ngl, but I'm a bit concerned about the safety of it, considering there's a laser and no protections.
>>29660371500 canadian on ali express
well so this just happened. i had my synth, played a lot during the afternoon. live in appartment, and i permanently have static issues, like i walk a bit and then everytime i pick the door theres literally sparks and sort of hurts, its intense. so, i walk to grab some shit, then went back to the synth and there it happened a fucking static pop when i grabbed a metallic knob. have in mind those knobs are only metal on outside all the rest is plastic, its like a slice of some shitty ass tin, then all the rest is plastic except the screw. and then all the arrangement (synth, pedals) started to light leds intermitently, measured supply, from 9v it dropped to 2.5v. so, something shorting. it was the synth. opened it up, a fried 074 very hot and a 072 that even being cold and all was simply dead. changed the ICs and everything was normal again.BUT HOW IN THE FUCK HOW WAS THIS CAUSED. my fucking hand produced a static pop in an ISOLATED piece of metal outisde of the instrument and this produced a mini EMI nuke that fried these two stupid chips? makes zero sense.can anyone explain?
>>2966053Do you have a carpet? Do you live in a very dry environment? Do you wear polyester clothes? If you said yes to all three then congratulation you live in the land of the triboelectric effect.As for the synth, it didn't need a direct connection, and the spark might have coupled capacitively with the circuit inside and nuked the two opamps with a massive voltage spike. If you're constantly getting zapped then you may want to do something about the three aforementioned conditions.
>>2966046>1500 canadian on ali express>>2966043Thats actually a good deal.Any reviews?
>>2966053>can anyone explain?next time you touch the fridge, see how far that spark can fly thru the air. i used to get 0.5 to 0.75 inches.
A chip like this died in my audio amp. I know it's this chip because it failed for hundreds of other people and my symptoms are the same. I ordered the chip before even opening the amp. I opened it today and the shit's small af. The chip is ~11mm on the side. How tf do I replace it? I thought a soldering iron was gonna do but now I'm not so sure given how many tiny legs it has.Help pls.
>>2966063drag soldering or hot air
>>2966063If you don't have a hot air station then you'll need some Chipquik solder alloy. After you pull the chip, clean the pads with desoldering wick. Place the new chip, solder two pins at opposite corners to hold it in place, and drag solder the rest of the pins. Clean with isopropyl, inspect for bridges, etc.
>>2966063>>2966065and you definitely need flux. Don't even try without it. Use leaded solder (Kester 44) in either 60/40 or 63/37, and liquid flux (GC Electronics or MG Chemicals). A bevel tip works best for drag soldering.
>>2966063There is no point in replacing the chip.The chip didn't die randomly, there is a cause which killed it, like a faulty power supply circuit.You have to find and fix that first, or the new chip you put in will blow up again right after you install it.
>>2966064>drag soldering or hot airThanks. I can learn how to drag solder. I don't have hot air.>>2966065>If you don't have a hot air station then you'll need some Chipquik solder alloy. After you pull the chip, clean the pads with desoldering wick. Place the new chip, solder two pins at opposite corners to hold it in place, and drag solder the rest of the pins. Clean with isopropyl, inspect for bridges, etc.Thanks anon!! After some searching, this is a low-melt solder alloy, right? I just apply it and them move the iron around the chip until I can flick it off?>>2966066>and you definitely need flux. Don't even try without it. Use leaded solder (Kester 44) in either 60/40 or 63/37, and liquid flux (GC Electronics or MG Chemicals). A bevel tip works best for drag soldering.Thank you. I will order flux tomorrow.>>2966076>There is no point in replacing the chip.>The chip didn't die randomly, there is a cause which killed it, like a faulty power supply circuit.>You have to find and fix that first, or the new chip you put in will blow up again right after you install it.It's a well know fault with this chip (Cmedia in the pic). Mine lasted for over 2 years. It actually still works but you have to heat it up with a hair dryer first and get it hot and then while it's hot it will continue running! Only if you shut it off and it cools down, it will stop working. I never unplugged mine for years and the one time it shut down due to a blackout, it would not come on.I have no idea why that happens.
>>2966078>this is a low-melt solder alloy, right? I just apply it and them move the iron around the chip until I can flick it off?Exactly. You should use Kapton tape or foil tape to protect the components surrounding the chip. If you don't have a suction tool then slide it toward the top edge of the board until it's off of the pads.
>>2966078Could be a broken bond wire inside the chip from thermal stress. Try sticking a heat-sink to it or any nearby hot components once you’ve swapped it out, might save you having to replace it in the future.
>>2966080>flux>low melt solder>regular solder>Kapton tape.anything else I'm missing?Also, came across this on chinese sites... it's under $35 USD. Cheaper than the accessories above. But would it work??
>>2966082>Could be a broken bond wire inside the chip from thermal stress. Try sticking a heat-sink to it or any nearby hot components once you’ve swapped it out, might save you having to replace it in the future.TY! Thats' a really good idea.I actually found the post that clued me in about why mine failed. This guy is one of the people who works for the amp/dac company who made the product. Here's what he says. Not sure what's going on desu. I'm just a rank amateur when it comes to electronics. >That's an interesting question... and I don't have a specific answer.>>I can tell you that failure of the C-Media chip is one of the more common faults we see in the DC-1.>It doesn't go by batches or anything like that - that particular chip just seems to be more prone to failure than many other components.>It could be because the USB port is more subject to stress - because it is used to connect two digital circuits - and because it accepts power from two different devices which it sits between.>So, if there's any sort of ground issue, or voltage differential between the two, that chip is right in the middle of it.>Or it could simply be that the CM6631 is internally very complex and so is more sensitive to failure.>>The Streaming USB input on the XMC-1 uses the same chip - and we've seen relative few problems with those.>(However, in all fairness, I don't think proportionally that many people use the USB Stream input on the XMC-1.)>(And, in contrast, MOST people who have a DC-1 are using the USB input.)
>>2966083>But would it work??Probably, but I've never used that specific one myself. Does the PCB you're working on have a heavy copper pour and/or large ground plane around the chip?
>>2966085>Does the PCB you're working on have a heavy copper pour and/or large ground plane around the chip?No idea. It's a black PCB. Mine looks exactly like the pic.
>>2966086It's rated at 700W so I think it will work, but I've never used that particular hot air gun so take it with a grain of salt.
>>2966030>You instead measure the DC resistance and estimate the (V^2)/R power rating of your electromagnet based on a thermal resistance that you estimate from maximum temperature and surface area compared to known components. For that size of transformer I’d guess 10-20W, depending on how well heat flows from the coil to the core.Is there a recipe for something simple? I recall as a kid just raping a large nail with wire and connecting to battery for a quick and easy elctromagnet. Is there some projects book or online article that lays out a relatively simple example. Seems like 20 watt for a smaller battery is a lot.
>>2966098Thats 20W maximum. Nothing stopping you from putting less power into the electromagnet by using a lower voltage. Just know that V = I*R and P = V*I and you can calculate the rest.>raping a large nail with wireNon-consensual solenoid winding, is it? A nail+wire method will pick up a few paper clips, maybe a small screw, but probably not whatever 8 ounces is. To maximise the force of a solenoid, you need:>to maximise the amps multiplied by number of windings>to maximise the difference in magnetic reluctance of the magnetic circuit when the object is picked upMaximising the amp-turns is easy, just add more windings or put more current through them. More current without overheating will mean thicker wire. So both of these paths mean adding more volume of wire, the best path to this is using enamel-coated “magnet wire”. Use thicker wire and you can handle more amps but with less turns, use thinner wire and you get the opposite. Ultimately they’re equivalent, the same volume of solenoid windings will be able to produce the same amp turns for the same power dissipation, but at different voltages. So if you’re winding your own, you’ll have to choose a total length of wire that results in a sensible power dissipation at your chosen voltage. 3.7V (4.2V max) from a single lithium cell might be what you want to aim for, but a holder of several AA or C or D batteries is also worth considering. Power dissipation is the limiting factor, larger coils can dissipate more heat without overheating.Cont.
>>2966098>>2966109Cont:Maximising the change in reluctance is easy to think about when you consider the horseshoe magnet. Magnetic fields travel easily inside iron, and not well through air. So a horseshoe magnet with nothing completing its loop will have a high reluctance, but when it picks up a nail the field flows through that nail to complete the circuit, so then it has a low reluctance. The difference between these two states is where the force comes from. No calculations needed here, you just have to consider the geometry. Instead of a straight nail, you’d be better off with a nail bent into a U-shape, with both the ends exposed. That way the magnetic force doesn’t need to go all the way around the top through air. As well as a “n” shaped core, a “m” shaped core with the winding around the middle stalk also works, like the aforementioned “E core”. Better than a 2mm thick nail would be a 4mm thick nail, as thicker iron allows more magnetic field to flow. Better than a stainless nail would be an iron nail, as iron has more magnetic permeability. Even better is electrical steel or ferrite, which is what transformers are made of. Apparently cutting a toroidal ferrite core in half is difficult, I guess it is a ceramic.So whether you pick an existing magnetic core from a transformer or motor, or cobble one together from stuff you have lying about, the ideal geometry is something like a horseshoe or “m”, thats as thick as possible while still having room to wind wire on, and made out of the most magnetically permeable substance possible. Magnets themselves aren’t that good. If you do cobble something together, wrap it in electrical tape so the wire doesn’t scratch off its insulation.You could feasibly be winding hundreds of turns of thin enamel wire to get something strong enough, hence why an existing transformer core + winding is so appealing. But still give it a try with twenty windings to see how heavy a something it can pick up.
>>2965856Use regular LEDs and a suitable Nucleo-64 or -144 board with enough GPIO for controlling them
>>2966078low-melt solder is nice because it minimizes the risk of damage to the chip. but most alloys are a bitch to work with compared to good old 60/40 leaded solder. their Sn42Bi58 bismuth solder is decent, but you still need plenty of flux. the indium stuff may seem tempting but it's a bitch to get to actually wet the surfaces
>>2965937>What’s the project?E-scooter motor controller. Because I don't like idea of paying $50 for shit i want, if I can burn $10 worth of FETs. Technically 1 shunt would work too, but ADC timing and some dirty hacks to move PWM phases.... Bleh. Other thing is microcontroller choice. I don't want STM32 because... you dont ask login for fucking IDE and brick $2 stlinks. So China WCH it is. But which one?V006 has 1 opamp that can switch its input and has PGA to adjust gains, but bitch runs at 48 MHz and has very castraded core (no division, no fpu). V203 has slower ADC, 1 MSa instead of 3 MSa. Built in op-amps need external resistors. But bitch runs at 144 MHz and has least castrated core RV32-IMAC
Where the fuck do I get tin-bismuth paste for fucks sakes, fucking aliexpress banned it
>>2966162digikeymouserfarnell
Hello, you guys. I would like to ask a simple question about OP AMP design. I asked an AI about where is the best place to put the coupling cap in an op amp being used for audio. I want to make a non inverting audio amplifier and I also have a pot acting as my volume control. I was connecting the coupling capacitor after the pot. From what I have read this is not the best practice. The best practice would be to connect the coupling cap before the volume pot. The reason being because this ensures a 0 DC voltage on the middle pin of the pot? In my case, I was not doing that. I was connecting the cap after the pot so there's a good chance that some DC level was passing through my volume pot. There are other things which the place of the pac will affect. FOr example, it will form some kind of filter with the impedance of the non inverting pin. If I connect the capacitor after the vol. pot. the frequency cut will vary with my volume, something like that? But if I place the cap before the pot. then now the cap also takes in consideration the resistor impedance? Can someone more knowledgeable confirm this tome?
>>2966167>I asked an AIdon't trust the output of bullshit generatorspost a drawing of your circuit
>>2966166expensive expensive expensive
>>2966141Get some experience with those micros before you commit to them. Ideally get ones that work on the CH32V003Fun stack by CNLohr. WCH do make some ICs specialised for motor driving like the CH32M007 and M030, but they have integrated gate drivers so might not be ideal. SimpleFOC does support quite the variety of chips that you may want to consider:https://docs.simplefoc.com/microcontrollersGoing for an ESP32 seems sensible.>>2966167A properly designed split-rail op-amp circuit has no DC bias to get rid of anyhow.A cap before a volume pot that's wired up as a voltage divider shouldn't influence frequency response. The cap and pot act like a high-pass filter, and the full resistance of the pot is the resistance to ground, that never changes. I'd be simulating it before asking AI.
>>2966190>Get some experience with those micros before you commit to them.I mean... it is just STM32 with ARM core replaced with RV core. Sorta. Kinda. DependsIf Chinese docs aren't clear, ST docs would prob explain it. >CH32V003Fun stack by CNLohrHm... I think I should use just native HAL stuff, because driving motor is time critical and abstractions would just make it harder to understand why on earth MOSFETs blew up 100th time today>WCH do make some ICs specialised for motor driving like the CH32M007 and M030I checked them, but issue is that they aren't high voltage enough. 48V variant can work with 50V max, and I think any BEMF spike into DC bus will damage it on 36V (30-42V li-ion) system ESP32, idk. Would work probably. But I have never dealt with anything like that.
>>2966199Experience with the toolchain I mean.
After all the talk about the scopes, I've been looking at getting one and I'm wondering if one needs to buy the expensive camera that mounts in that tube at the top or is it possible to somehow mount a phone so you can film what you're watching with a phone camera?Or maybe mount it on one of the ocular lenses? Is that possible?
I want a setup where I can turn on a power strip and have a bunch of devices turn on at once, including some that are powered over USB and need to communicate with my desktop. I was thinking I could >buy a powered USB hub>run it through a USB m-f cable that cuts the power line and leaves data>plug the USB hub's 120V adapter into the power stripso when I turn on the power strip, the hub gets power and sends it to the downstream devices and starts talking to my desktop. But what actually happens is the hub powers downstream devices but won't communicate with the desktop. Presumably the IC is hardwired to the upstream USB 5V and the barrel jack can't send power to that net. Is there any reason I shouldn't open it up, short the barrel jack to the input USB 5V net, and cut the 5V line of the input connector? Or is there an easier way to do what I'm after?
I'm trying to design a SMPS using CM6800 using Fairchild AN33 https://www.ic72.com/pdf_file/m/164313.pdfI'm aware its obsolete.I got stuck on (12).Where does 6.8V for the error amplifier output come from? The block diagram suggests it's 2.5V when the PFC output is 0 and less when the PFC output is positive.A 200uA output of a multiplier is mentioned alongside R1, this can only concern the gain modulator.The only current path is through I_Sense via 3.5 kR. I_Sense can be expected to be negative.Can anyone spot what I am missing? It seems to me the author pulled a whole lot out of their ass all of a sudden.I can only see R1 to be a function of I_AC V_RMS and VEAO and I_SENSE for the condition where I_SENSE and V_FB are lowest (high current and load) I_AC and V_RMS are highest.But (12) does not reflect that.>>2966267>presumablyTest that.>Short the barrel jack to the 5VIs the thing really on a 5V mains Adapter?>easier wayMake a USB T, data from PC to hub, 5V but no data from hub downstream port.I believe the reason behind this is the hub potentially being at a completely different reference potential than the PC. Does the hub power on if it's only on the PC but no power? If not why not connect it to the PC normally?
>>2965970I keep having to say it:Crimp tools are one of the few tools where quality really matters, they might work fuck all, or be a pain in the ass with the connectr sticking or assuming a rather odd shape.Now your tool looks like Knipex, which should be fine quality and you're trying to crimp open barrel type connectors.It sort of looks like your pliers are made for them. But make sure to check. Crimping tools or inserts are highly specific and only crimp one size of one specific type of connector. The wire size is specific to it again.Understand how open barrel crimp connectors work. Usually you'll find two sets of 'wings'. The ones towards the connector end are supposed to be crimped to the wire. Those at the rear hold on to the insulation and do not make electrical contact. So strip the appropiate length and insert the wire accordingly. You can always try to help the shape along just a tiny bit before making the connection. I'll try to find a tool like the ones you show and see if I can make a connection...
>>2966286I even found extra shitty connectors and extra shitty chinese wire strippers with such jaws.Observe how far it is stripped back and how it is placed. Do not twist the strands.
>>2966288The jaws are probably soft asf but especially rough machined. This causes sticking of the connectors or even tearing. Results will suck.
Fascinating. It is even worse than I expected. Observe how sloppy tolerances lead to the connector getting sucked in on either sides of the anvil, making the crimp on the insulation insufficient.The crimp on the conductor is not fully formed either because the McDonalds arches the pliers should have aren't really there in the appropiate gate.>>2965978 #This is the wrong tool btw, it's for insulated connectors.
Similarly cheap ratcheting style crimping tool. Notice the anvils being wider on one side, the dies are larger on that side too. This is where the insulation goes naturally.Not perfect but much much nicer already.
>>2966284>Test thatI'm trying to avoid opening it up until I'm sure it's not going to be easier to return it>Is the thing really on a 5V mains Adapter?12V. So not shorted to the barrel jack, really, but to the 5V on the output USB side>Does the hub power on if it's only on the PC but no power?Yeah, frustratingly.You might be right, it may well be easier to put together some Frankenstein cable. But I'm trying to stay as COTS possible with the hope of saving time and having something pretty in the end
>>2966284Update:Much hasn't changed since then. They still do not explain how VEAO ever goes above the non inverting input when V_Boost doesnt go negative.>>2966300>Opening itThat is not what I mean. With the hub powered and nothing connected, can you either measure 5V out at the port facing upstream or can youn plug anything in that gives an indication of 5V?>To the 5V on the output sideOkay so we're on the same page just that I suggested you can do it externally.>Frankenstein cableI'm kind of having a hard time telling you nothing can go wrong. USB is differential, so technically it does not matter where the reference voltage is.But ICs will only widthstand so and so much. Also the lines might be terminated other than series.It begs the question:Is there no voltage on the upstream chip because the designer expects the PC on a different reference potential and both sides are galvanically isolated? It would make sense in that case to simply power the upstream IC from the PC and not add a DC DC to the BOM.Can you please see if you have continuity from the upstream shield to the downstream shield?I believe if this is the case it should be safe to do this.
>>2966301Forgot: https://www.manuallib.com/download/pdf14/FAIRCHILD-FAN4800-COMBO-CONTROLLER-APPLICATIONS-MANUAL.PDF
>>2966168Do I put the cap before or after the volume pot.?]The circuit is very simple and it's exactly like pic related shows with the adition of a coupling cap in the non invertin input and a capacitor across the Vcc pin and the Ground pin. Simple, right? But where is the optimal place to put the cap? I'm not an electronic engineer dude. I'm a dog, do you feel me, bro? Anyway, thanks.>>2966190Don't you think simulating circuits are for pussies? I mean no offense, but in theory or in the simulation everything always seem nice and stuff, but then when you try the thing in real life shit just happens, you know? Murphy's law or something. I'm tired of simulations. That's why I did buy an oscilloscope. So I can do the shit in real life.
>>2966301Update:Maybe they're just fucking dumb and never bothered putting the symbol the right way round until now. I'll just fly with this FAN4800 as its not obsolete and the documentation checks out so far. Maybe the parts are different but I doubt it. They're just dumb.>>2966304It won't matter and you could really just leave the cap out. If I were you I would put it 'before' the pot, that is to the left of it.When you design for HF and in this case audioF, it's common practice to put caps at either the output or input or both of modules you design. Because fuckwits. Sure this often means you have redundant caps when you just wire some amp to some mixer but meh.The AC impedance your circuit 'sees' will be the same. Those elements are in series. It will see 10k-j(wC)-1. But if for some reason your audio circuit sources DC the cap will prevent it from frying itself, frying the pot, or just wasting power. If your source depends on sinking of DC at the output, then do it in that module, don't rely on the next module.Cap at Vin.
>>2966260I was looking at something more like pic related, just a monocular microscope lens with a camera mount atop. But the swingy lamp arms are probably too unstable for me to use comfortably, so I'd probably get one of the focus adjusting doodads like in your picture, either mounted on a vertical and horizontal rod like in your pic, or on some sort of diy linear rod assembly. Ideally it would be some sort of cool kinematic arm where I hold a button and it unlocks the motion, and I can effortlessly move it with one hand to wherever I need it, and when I take my hand off it locks again. I'd attempt to do it with electrically actuated brakes and counterweights, but not with motorised axes.>>2966267Seems doable, but I'd use a dedicated USB isolator instead. That way if it's a homelab-use USB hub you're not going to fry your PC. You can even buy isolated USB hubs with a DC power input, that's what I'm using. Often they have isolated DC-to-DC converters in them, but you could just desolder that part, after ensuring that the hub will feed power back out its USB input to the other side of the isolator. Even if it doesn't, you can use a normal USB lead to feed 5V and GND to it without monkeying around inside an off the shelf hub.The only thing I'd be worried about with your method would be if the voltages on the USB data pins might be out of allowable range for a USB hub IC with no voltage on its rails. Chances are it has protection clamping diodes, but those will be putting the hub IC into momentary brownout. Just seems like improper design.>>2966284VEAO = 6.8V in (12), I assume it's a worst case scenario maximum voltage for when the output is in saturation. For open-loop operation. The datasheet for the CM6800 suggests its maximum operating voltage is 6V, but I don't know if that means the maximum it's allowed to see or the maximum it will produce.
>>2966304Oh and>simulating circuits is for pussiesReally nigga? You're dicking around with a OPamp audio amplifier and tell people simulating shit is superfluous?Simulations are way faster than whipping up a prototype and faster than calculus. Especially if breadboarding or similar wont cut it. You're also looking at way too many variables to get anywhere, using an experimental approach, when working on any circuit that has at least some complexity and at least some quality requirement. Most audio amplifiers have neither, as they're trivial and the human ear is the limiting factor.Not everyone wants to blow up a big FPGA or such for every attempt to at solving a multi variable problem.
>>2966310>VEAO = 6.8V in (12), I assume it's a worst case scenario maximum voltage for when the output is in saturation. For open-loop operation. The datasheet for the CM6800 suggests its maximum operating voltage is 6V, but I don't know if that means the maximum it's allowed to see or the maximum it will produce.Yes,surely the value is representative of an extreme case,but if picrel here >>2966301 was correct (VBoost is the output of a PFC stage, on a cap, a polarized one usually) then how would VEAO ever drop below 2.5V let alone negative.
>>2966053At work I accidentally zapped my computer a couple times and it would instantly reboot. I always ground myself first now by touching the screw at the lightswitch faceplate.
>>2966301>>2966313>They still do not explain how VEAO ever goes above the non inverting input when V_Boost doesnt go negative.If FB is lower than 2.5V, the VEAO will be high, maybe arbitrarily high depending on the feedback and however the fuck a transimpedance amplifier works. If it were an op-amp it would hit the rails. FB is the division of the output of the PFC stage, I can't remember if PFC wants to run a DC output or just one that fluctuates less. Either way I'm fairly sure that no current should flow when the AC voltage is zero-crossing, hence FB will likely dip low anyhow.>>2966304Oh it's not an op-amp, it's a power amp.Take a look at the internal schematic, pic related. Seems like if they say it can work with no capacitor (and presumably a signal on either side of the 0V rail), then it's self-biasing with that 10k pot there. Technically this topology could even work with negative input voltages. Here's a thought: try it without a cap, and also with the cap in two different positions. Choose whatever sounds better. I think before would be better.>Don't you think simulating circuits are for pussies?Better than AI. For something simple and linear like idiot-checking a frequency response, there's nothing wrong with sims at all. Programming AIs improve themselves by running sandboxed code for training feedback without human intervention. If EE AIs could do the same with spice models we could see them becoming actually useful with regards to novel circuit troubleshooting or design. What matters when it comes to real ICs is that the spice model is as accurate as possible, which is why you'd use a conventional algorithm with the help of a text comprehension LLM to convert datasheets into spice models. I expect it could work perfectly well for transistors and linear ICs, same for digital logic, but less so for stranger ICs like switching converters. The "startup" transient is usually enough to get your sim out of any unstable equilibria.
>>2966301>measure 5V out at the port facing upstream?>continuity from the upstream shield to the downstream shield?I'll take a look next time I'm poking at the thing>>2966310>isolated USB hubs with a DC power inputThat's a class of product I didn't know existed. Interesting. Does yours have the behavior I'm looking for, where the upstream doesn't power downstream and it needs its independent DC supply?
>>2966321>transimpedance amplifierthanks thats what I was missing, over here the circles are shown at the output so I didnt think much of it
>>2966324>Does yours have the behavior I'm looking for, where the upstream doesn't power downstream and it needs its independent DC supply?No it has the isolated DC-to-DC converter on the board, but you could easily desolder that. It's just 4 through-hole pins. It's the big black rectangles in pic related, I own both.
>>2966324>That's a class of product I didn't know existed.I'm just speculating the designer wants to be sure even in cases where the thing is somehow not referenced to PEN.I got an ADUM3160 on a board with a DC-DC and opto here on my desk. Very important thing. It could also be that the hub is supposed to do that job.>>2966324>Does yours have the behavior I'm looking forMine? I don't have one (?). But I know at least one colleague who specifically told me his powered hub was isolated up to a few kV. So like the ADUM just fancy.
>>2966324>>2966326Oh sorry im tired. Replied to others stuff.Either way the pic pretty much shows something like an ADUM.>>2966328
>>2966324>>2966326Looking into it more, they use the ADuM3160, which goes up to 12Mbps, which may not be enough for your uses. But the datasheet has this table, which seems to clearly say that states when one side is powered and another is not are fine, and will not result in any adverse effects.
>>2966330Doesn't that say if either side is not powered it either goes into high Z or a disable state?Aka not working.What I do know: The shield is not connected through. So measuring for continuity through the hub will provide clues.
>>2966333Yeah. Both sides need to be powered independently for it to work. The PC side needs 5V from the PC, while the isolated side needs 5V from somewhere, be it an isolated DC-to-DC converter, or a seperate 5V DC supply like from a wall adapter.
>>2964845I need small and cheap Sodium-Ion BMS, does anyone have recommendations or should I just go with a chinksite cheapo unit like this?https://de.aliexpress.com/item/1005007587146335.html
>>2966353That might be fine, just be sure all the specs line up properly and take the maximum current rating with a grain of salt. Especially the undervoltage cutoff threshold, those can be kinda low on some chinky BMSs, and make sure the quiescent current isn't too high for your use-case.Depending on the project you might want to be looking for a higher-end BMS, maybe there are generic BMSs on which you can program the voltage thresholds in order to work with whatever battery chemistry you desire.
>>2966354Thanks!I'm just building some DIY dashcam. Its supposed to run as long as possible without using my starter battery when the alternator is off, I want a small Sodium-Ion pack because they can run on hotter temps during summer, lower temps in winter and don't just explode when undercharged.
>>2966304a cap just after Vin will remove DC bias and should work fine so long as it's at least 1 µF or so, since 1/(2*pi*10e3*1e-6) ≃ 16 Hz. to make it truly idiot proof you probably want a non-polarized cap for that like a film cap or ceramic. think "idiot connected -12V DC on the input" levels of proofingI don't think the 250 µF cap on the output should be necessary if the input DC bias is removed with a cap, unless the LM386 has very high input offset voltage. but it doesn't really hurt either except for cutting off the very lowest frequencies. but you can just increase it to 1000 µF or so to get 20 Hz through to a 8 Ohm speaker
>>2966355Are you doing 12v? Why not just do deep cycle lead acid at that point? Like 1.3-4.5Ah. Don't need a BMS, easy to replace if it ever dies, easy to charge from the alternator, etc.
>>2966355going to agree with anon >>2966358your car already has all the 12V bits. just hook a small deep cycle SLA to some line that's powered when the ignition is on and a resistor to limit charging current. if you're feeling fancy you can put a low voltage cutoff between that SLA and the dashcam so you don't kill the former if you don't run the car for a long time. oh and don't forget a fuse
>>2966353Know that sodium ion has a very wide voltage range, that "12V" battery will range from 7.2V up to 15.8V using that BMS. So ensure your circuit can handle those voltages. Also that seems to be a protection BMS, but not a balancing BMS. So not the best idea for longevity. Balancing BMSs are usually a bit more expensiv than ones that just do protection. This one:>https://www.aliexpress.com/item/1005010263606134.htmlLooks to do balancing.>>2966358>>2966359Hey if you use an 18Ah or bigger you can probably use thick wires and a bypass button across the charging resistor to jump-start your car from it.
>>2966358>Are you doing 12v?Yes>deep cycle lead acidBecause I haven't thought about doing that yet. My thinking was that I could just build two battery packs and when I'm not driving around I just charge the 2nd unit at home and swap them when I get a push notification that the pack in the car is reaching a low charge state.>>2966359I'll go back to the drawing board, Thanks Anons, both of you!
>>2966361>push notificationdo zoomers really
>>2966360>Know that sodium ion has a very wide voltage rangeYes, that's why I was looking for a sodium-ion bms and not just using a li-ion one.>that "12V" battery will range from 7.2V up to 15.8V using that BMSI was thinking about adding a buck/boost converter + power / voltage smoother. The board I'm running will need one regardless because its 5v.>Also that seems to be a protection BMS, but not a balancing BMS. So not the best idea for longevityFug.>>https://www.aliexpress.com/item/1005010263606134.html>Looks to do balancing.Thanks!>Hey if you use an 18Ah or bigger you can probably use thick wires and a bypass button across the charging resistor to jump-start your car from it.So the battery pack becomes a secondary starter battery if the lead acid one dies?
>>2966362Yes, yes we fucking do because the boomers, xoomers and moomers let in a gorillion gibs-me-dats and we now have to GPS+Camera surveil our shit because of that. While at it, I can at least add some comfort features.
>>2966363>So the battery pack becomes a secondary starter battery if the lead acid one dies?Yes. Might as well be built into the vehicle and automatically charge as you drive it,but is otherwise disconnected. Actually you’d be better off if such a battery doesn’t have a parasitic load on it at all, so having your jump-starting battery also be your dashcam battery is probably not the best idea.Buck converters have non-trivial quiescent current, you’ll want to really optimise the power section. Ideally you wouldn’t need one and could run the cam at native voltage, or slightly above and use an LDO. Cant use 2S sodium ion? Not that I saw any BMSs for that, but again you might be able to get a programmable one, be that with software or jumpers/resistors.If it’s a conventional dash cam you’re just feeding power into the USB socket of, you might find that it draws a lot more current when powered this way than when it’s running on its internal lithium battery. So feeding an internal 3.2-4.2V might be more sensible, assuming it has such a battery in the first place.
>>2966369>having your jump-starting battery also be your dashcam battery is probably not the best ideaI don't really need a "jump-start" battery, my starter battery is just fine, the car is a small 2l gas engine from 90's so it can sit for half a year and start just fine.I just don't want a small computer + 2 cameras to drain it.>Buck converters have non-trivial quiescent current, you’ll want to really optimise the power sectionYeah, that's what my gripe about it was as well. >Ideally you wouldn’t need one and could run the cam at native voltage, or slightly above and use an LDO. Cant use 2S sodium ion?There are no 2S systems of which I know, which is why I'm fucking around with a buck/boost converter. This led me to go fuckhueg 4S4P or 4S8P 10Ah or 18Ah cells, but at that point two 210Ah prismatic cells would be lighter but then again there's no 2S BMS which is why I'm spinning around with this.>Not that I saw any BMSs for that, but again you might be able to get a programmable one, be that with software or jumpers/resistors.Probably, but I haven't looked yet.>If it’s a conventional dash cam you’re just feeding power into the USB socket ofUhhh, no, it is a completely custom DIY project that is not just a dashcam but a to create a on board computer system that also has dashcam integration.
billionaires are stocking up on silversmost electrons are silver how fucked are we?
>>2966387>most electrons are silver
>>2966307>>2966321>>2966357Thanks.>>2966312My point is that in simulation you just connect the circuit and everything works fine because it is digital. As long as you follow the schematics and the software is updated you will have no problem in simulating circuits. But try buying one of these LM386 chips. It is obsolete now, but try buying one and physically building the circuit on a small breadboard. Maybe you can make it work without too much of a problem because you are so fucking good at electronics, but thesse little chips are very sensitive to crappy grounding so these chips oscillate like crazy. I will use a software, like proteus, to build the board. Thanks.Overall? I'm indeed impressed with the engineering of this small chip. I know it is obsolete and stuff. But still, it is amazing that someone way back in the day had the means to create such an useful little chip. I mean, one little chip and you get 0,5W of power? You hook two of these things together and you have 1W audio amplifier easy. I mean, not so easy because the chips oscillate like crazy if you are not careful with the layout. I also like how it self bias the output to half your power supply.
From what I know, one should avoid mixing Digital and Analog signals on the same PCB, does that hold true for ground planes as well? The internet experts and literature are split on this. I'm tending towards creating a single plane just for analog ground.The design I'm working on right now (am this tard btw >>2966371) calls specifically for an analog ground (AGND). It is the MIPI/CSI-2 Connector from the camera sensor module to be specific, the output of which is to be serialized and sent over a hf cable to another board near an SBC that deserializes the input and feeds it into the MIPI/CSI-2 port of said SBC.
>>2966354This.When I buy a module like that from a chink supplier, I always look up the detector IC and the FETs exacting specification, because the vendors like to just spam buzzwords like 'Li-Ion, LiFePo4, LiPO...'. I also hook up a variable power supply and check at which voltage the UVP and OVP pins change state.You can always whip something up. Those things are easy. But also don't get fooled into believing you 'need' a BMS or balancer or both. Those requirements are application specific. Initial balancing or even simple resistor networks go a long way.Don't get fooled into believing a resistive BMS generally doesn't do. It solely depends on your energy budget.So to whip something up:Think of the thing as having 3 jobs:Lockout charging OR discharging when under- OR overvoltage.Lockout when overcurrent.Balance.This means you generally need one shunt, which maybe the FET itself, you need voltage sensing on every series element and you need back to back FETs that handle the current and voltage. You may need current limiting and maybe multiplexing devices (resistors, caps, inductors, xformers, switches) for balancing. Dumb resistors shunting series elements are fine.You can either get any of the ICs specific to your cell count or daisy chain things like the Hycon ICs you often find on those boards.You CAN even daisy chain single battery boards, but:They often need to see charger voltage once after assembly to get out of a latch.Every FET will have to individually widthstand the pack voltage and every FET on resistance contributes to on state losses.Better OR the outputs and use two beefy FETs.You will not beat the chinese in pricing but it's important to understand what you have and why you are using it.
>>2966426This is very often unavoidable. Integration is not a bad thing.Many mfgs of dual domain chips will provide layout examples. Those are generally proven and it's good advice to stick to it.Understand EMI and coupling, galvanic and capacitive.Ground planes are primarily a concern for galvanic coupling. Imagine you have an analog input, say to an amplifier, and the non-inverting input tied to a trace that is defined as ground. If you now have a source of digital noise on that same trace, which has an impedance, the digital current noise will cause a voltage drop across that trace and your amplifier will amplify this signal.So make it a GND plane, right?Always a good idea but it wont behave the same across the entire spectrum as it does for DC and it still is a non zero impedance that carries this current noise.This is why you will often want to split ground planes or use star point grounding, where GND pins do not share a path to ground or the shortest possible.Use capacitors. Read up on EMI and PCB design.In many practical applications you will find that you can just tie everything that has GND in it to GND on one big plane and you'll notice no ill effects. it really depends on the precision and requirements of and for the thing youre designing.
>>2966427>it's important to understand what you have and why you are using it.That's reasonable.So you're saying I should just build it myself? To understand what I'm even doing.
/ohm/ and /mcg/ should do a long-running collaborative mega project that leads to wherever it wants to go (probably weapons n sheeeit).
>>2966411The idea of simulation is not that you had a sort of software that you feed your entire schematic while omitting alot of real world factors and it just goes BRRRRT and tells you how it will go.With circuit simulators you have to 'ask' the simulator a specific question, understand how to model it and model it pecisely. You will then get a partial answer to one of the thousand questions you have about your design.>The chips oscillate like crazyWhich is a good example for one of those specific questions you may ask a simulator. You model not only your lumped components and tie them together through magical zero impedance lines, you instad model all distributed elements, your TX-lines, noise and what not and sweep for different parameters to see how it behaves. But I doubt you'll need to simulate to come up with a layout that stays clear of Barkhausens swing park.
>>2966431Depending on if I can get my hands on the material for it (which does not look like it), I'm either going to use FPD-Link IV or V3Link. The former uses mixed signal chips that take 6gbps signals, serialize them to 7.55GHz signals, yeet them across a coax-cable to a deserializer, deserialize it back to an 6gpbs signal and move that into a computer, all of which is very fucking hard; the latter is the same, but about half the GHz.I'm currently trying to design a board that takes the video-stream from the camera and translates it into the 7.55GHz signal. After that I want to design the deserializer board.https://www.ti.com/product/DS90UB971-Q1https://www.ti.com/product/DS90UB9722-Q1orhttps://www.ti.com/product/TSER953https://www.ti.com/product/TDES954
>>2966432>So you're saying I should just build it myself? To understand what I'm even doing.You may, depends on how you learn. But thats not my point.I mean, if you were to ask reddit, everyone someone mentions a battery, the entire userbase will be circlewanking about >muh housefire>BMS>active balancer>energy efficiency of a balancerwhen often it is not needed or redundant. Also adding complexity to a system must make arguably safer since any added complexity that does nothing else makes it less safe.Example of the top of my head:I want to run 1 LED as a flashlight of off 1 cell. I will manually charge the cell using my bench supply. The LED forward voltage is higher than the cell discharge voltage. The LEDs bond wire is, with respect to the current capacity of the cell, fusible.Do I NEED a BMS?I can not deep cycle it because the LED will stop conducting.If the LED will somehow fail short the battery will easily open the short in no time.I can not overcharge it as it happens externally.>But what if a short develops at a different point in the circuit?Well, a BMS can not either safe you from a short across battery anode and cathode.And yet some guy on the internet with no tech education will feel smart enough to repeat the mantra they heard somewhere: Use a 1S BMS it is unsafe.
>>2966439>fucking hardYoure implementing it. The heavy lifting has been done by the people making the IC, which is why you can do this for a hand full of tendies.Microwave engineering is microwave engineering and as old as anything else. Idk the margins of the technology your're suggesting but I will just assume it's got some reasonable headroom.>Getting stuff like this and thatAgain: Idk in this case specifically but usually: No. Not in single piece quantity. I used to always just ask my boss if I could take things from the lab that didn't look like we were going to sell them to anyone after development work. Another way I used to use to obtain crazy expensive and nice things was through university. Sure it helps to be enrolled or at least know the guys in the circuit design dept. They have not only a budget but contacts at TI, Analog, Onsemi, Infineon etc. and those will often send a free sample for educational purposes.After all they want Johnny, once hes got a job, remember how it's done using the parts he knows.So yeah, if worried: Brush up your EMI and PCB design knowledge.
>>2966440>I mean, if you were to ask reddit, everyone someone mentions a battery, the entire userbase will be circlewanking aboutThat's why I'm here and not on reddit.>when often it is not needed or redundant. Also adding complexity to a system must make arguably safer since any added complexity that does nothing else makes it less safe.Eh, I wanted to take some load off of my plate by making it easier (buying premade).>I want to run 1 LED as a flashlight of off 1 cell. I will manually charge the cell using my bench supply. The LED forward voltage is higher than the cell discharge voltage. The LEDs bond wire is, with respect to the current capacity of the cell, fusible.>Do I NEED a BMS?Not in that situation lmao. Just a resistor.>I can not deep cycle it because the LED will stop conducting.Yes, but Sodium-Ion batteries are deep-cycle-able anyways (which is why I wasn't too concerned about the safety of the batteries)>If the LED will somehow fail short the battery will easily open the short in no time.True>Well, a BMS can not either safe you from a short across battery anode and cathode.Shorted externally? When that happens, you fucked up by not building a proper enclosure. Internally? The cell producer fucked up by not aligning the roll OR you fucked up by not building a proper enclosure which led to the cell being punctured. But also not a concern for sodium-ion batteries I guess. I know, I know, you're making an argument about using ones brain.>>2966444TRIPS CHECKED>Youre implementing it.I know. But I'm not an electrical engineer, I'm a "computer scientist" (over glorified fag ass programmer) with a knack for engineering stuffs.>Microwave engineering is microwave engineering and as old as anything elseCurrently reading the TI guides for mmWave streaming: https://www.ti.com/tool/TIDA-020093CONT. in next reply
>>2966444 (still checking)>>2966447CONTI can get my hands on the FPD-Link IV chips, that's no issue at all, Mouser bless. Getting all the official TI Documentation is my issue lmao. Currently digging through their site to see what I can get, but I'll ask my professors if they have connections after Christmas-break.>Brush up your EMI and PCB design knowledge.On it, I've read the altium stuff for high speed pcb design, watching some YouTubers. If you have some book recommendations for me, I'd read those immediately.
>>2966448They don't have a reference design, application note or eval board out there? Most often they have any of that, if not for the specific product then for one that is very closely related.I mean if they have a product they have done that work anyways so why not publish it.>load off my plate by buying premadePleasendon't get the impression I have, at any time, recommended you DIY. I'm just saying don't cargo cult. Understand and get what you need. And check what you got delivered, do not trust the specs.
>>2966449>They don't have a reference design, application note or eval board out there?They do, but its hidden behind NDAs because it is "cutting edge" and they sure won't give me access to it:https://www.ti.com/tool/DS90UB971-Q1EVMhttps://www.ti.com/tool/DS90UB9702-Q1EVM>Most often they have any of that, if not for the specific product then for one that is very closely related.V3Link stuff is open though.>I mean if they have a product they have done that work anyways so why not publish it.Because more moneys for TI.>Pleasendon't get the impression I have, at any time, recommended you DIY.No worries mate, I like doing stuff on the side to learn just a little more. I'll look into it and if its too much I'll at least take the knowledge with me.>I'm just saying don't cargo cult.Yeah, I'm trying to avoid that like the plague>Understand and get what you need. And check what you got delivered, do not trust the specs.I've already ordered some camera sensors but I'm not 100% sure if they're not trying to scam me. I'll definitely test those.
>>2966450Are you sure this is the way to go then? Are you sure you're not setting yourself up for demot?
>>2966450>V3Link stuff is open though.Holy shit, I just realized something. Pic rel.>>2966451In what way?
>>2966456See there you go. So now is there a reference design of any sort?
>>2966457There sure is: https://www.ti.com/lit/an/spradh2a/spradh2a.pdf?ts=1766387920503andhttps://www.ti.com/product/DS90UB953-Q1#tech-docsGonna download it all real quick and then read it.
>>2966461This then will answer your questions regarding layout, always bottom pages.https://www.ti.com/lit/ug/snlu224d/snlu224d.pdfYou'll come to realize that almost all products industry makes are reference designs, cheapified, cobbled together, and sometimes with minor external circuits to 'do thing'.
>>2966463Yeah, V3Link is literally just FPD-Link III without the automotive hardening = cheaper
I still owe OP something so OP please if you readint this say hi.
>>2966472I'm still wondering, video feed at 6Gbps is nothing youl want to continously save. Whats the plan? Do you want to keep a rolling memory of the last N seconds to save conditionally in case of event?This whole thing reminds me of my VHDL courses in uni, it's pretty easy to just slap a Xylinx and implement a jogger detector. This would also fix the moral issue of surveilance of public spaces.
>>2966477>jogger detectorMY SIDES>Do you want to keep a rolling memory of the last N seconds to save conditionally in case of event?Yesn't, I'll run it on a Rock 5T with at least 16GB of RAM and openCV, footage will be inside a RAMside ringbuffer, only saving to permanent storage under certain conditions like>Sudden acceleration>High non-sudden lateral forces >Motion>Sounds>Capture button has been pressed>Someone says "Ok Garmin, Video Speichern"Video quality will vary depending on different conditions like alternator state (12v sense wire for ignition), time of day, time since engine shutdown, etc.I think the lowest I'd want is probably a 480p picture once a minute if the car has been sitting for more than a day.Only when the car is sitting and the jogger detector triggers, I want it to jump back to 4K60, save to disk, stream to my home server, notify me via push notification (writing an app for that as well btw), set a GIS geo-fence with minutely location updates and reset the engine shutdown timer.>This would also fix the moral issue of surveilance of public spaces.Its not just a "moral issue" here in germany, its literally banned by law because of our hipocrite moralfag boomer moron goverment. Its ok to record to a ringbuffer that keeps overwriting itself and to only save footage permanently if the car has been disturbed.
>>2966483I just had the idea to trigger a save-event when the hazard lights are turned on.>>2966477Also I forgot to check those dubs. Checked.
I'm a noob who's probably way overthinking thingsHow would you design a mobile device's power supply where, when it's plugged into USB, the USB both powers the device and charges its battery, then when USB is unplugged, power is drawn from the battery? Battery provides 3.7V, USB provides 5V when plugged in, both pass through a stepdown regulator to 3.3V before powering anything on the board.My idea is to use a LM66200 as a 2:1 OR. USB power is on VIN1, so when it's plugged in the LM66200's high status pin triggers a mosfet that allows power to flow to the positive terminal of the battery through a MCP73831 (only chosen because that's the IC that adajewt uses on their little battery charger for the very battery I'm interested in using on the board).When the USB is connected, I'm not sure how to prevent the power flowing to the battery from also circling back to the LM66200 on VIN2, which will likely lead to behavior I don't want. Maybe just another mosfet?
>>2966484>CheckedThanks, always good to know you appreciate digits.>>2966483>Video SpeichernI just knew it. Same problem here.>>2966486This solved problem is generally reffered to 'power path controller'. You can look it up and either implement proven circuits or get a one stop shop IC.To me this sounds like:Get a TP4056, in parallel with your regulator and to USB. That way if USB then charge and 3.3V.Use a diode to get 3.3V supply on the regulator from the battery positive if not USB. Maybe another Diode is needed to not power the TP4056 from battery creating infinite energy. This way batt will not see 5V but reg will always see 5V or Batt. A capacitor on 3.3V out might help with the state change.Check the datasheet of a TP4056 if anythin says no. I doubt it.
>>2966487Wanna add: Ideal diodes are popular.https://cdn.hackaday.io/images/958851644869893286.png
>>2966486PFET between battery and load, gate tied to USB VBUS.
>>2966487>>2966489Thanks so much. I figured it was solved, I just could not condense my thoughts into a googleable phrase that would return what you're referring to.>>2966490Figured some sort of FET would do the job as well
>>2966361>just charge the 2nd unit at homeWhy would you ever not just wire it up to charge off the car’s ignition switch?>>2966440Could you explain what you mean by using resistors instead of a balancing BMS? If he’s going to be charging it hundreds of times, I think active balancing is probably a good idea. It’s just a few dollars and it prevents your no-name chinese cells from drifting in capacity and dropping your effective battery size by 10% or more. Along with something that prevents you from overcharging or overdischarging any of them. Short circuit protection is a good idea too in case a wire gets damaged. i.e. a BMS.>>2966474HiThe photo is from where i work, I designed those 3d models and we printed them out.>>2966486Be careful, the maximum current of charging two things at once might be too much for some USB supplies. Ideally you can control the charge current based on what the USB supply can provide.
>>2966495Gotcha, very good point. The MCP will be set so that the battery (3.7V, 350mAh) receives 200mA while charging, and all the other shit on the board will collectively consume way less than 100mA (a low-power STM, a few SPI peripherals that collectively consume like <10mA at idle which will be the case while charging, other shit I have yet to add). Shouldn't ever come close to the 0.5A the USB 2.0 should provide but testing can always prove me wrong
>>2966495So I'm at that part of the cycle where I am here again.I told you you'll get a full report on the LC EDM.Finished my thesis, got a straight A which is nice. Got more time now to finish the actual product.Turns out there was a stupid mistake I dont really want to go on about but I fixed that just after handing in, on the day where I defended my work. Now it works real nice.
>>2966501I have plans to publish a follow up article with the fix for the dumb dumb and plans.So Yiang and Kunieda published this abomination:https://www.researchgate.net/publication/369473297/figure/fig1/AS:11431281154026686@1682646730934/Principle-of-conventional-LC-generator_Q640.jpgObviously this is a boost converter. It is resonant, has no diode. And it is coupled capacitively to the spark gap.Yiang and Kunieda like to pretend they were getting massive Coss from the transistor and distributed elements.Its easy to calculate the output amplitude and waveform here. Imagine Coss. The energy stored in L1 is a result of I and L which is a function of Vin and Ton. The Amplitude of the output is the same energy stuffed into Coss. In other words: The ratio of L and C describes the ratio between I_L and Vout. Bigger L bigger I smaller Coss more outch. Easy. The waveform is sinusoidal and the frequency is derived from the L1 Coss resonant frequency.
>>2966504So during that half period of f_LC where the voltage across the spark gap remains elevated (it's perfectly coupled as you are looking at a resistive capacitive divider with finite capaitance and infinite resistance) it might arc over, which is a function of various things but also a random time element.If and when it arcs over, the coupling capacity C1 is suddenly grounded. The voltage divider is now C1 and ~ 1R .This results in Coss being discharged into C1. So if Coss was equal to C1 you'd end up with half the charge, half the voltage and a quarter of the energy in every capacity, the other half in the gap.Make C1 larger and the energy transfer soon approaches 1.Currents and instantanous power are stupidly high here. Youre discharging 1.2 kV or more across 1R. Once current drops below holding current the gap goes high impedance again and luckily through some magic the charge on the floating end of C1 is restored... somehow... maybe a secondary arc or bipolar arc.Whatever energy is left in Coss and C1 (all of it if no spark) goes back through L1 to source at resonance.
>>2966505So what this really is is a new take on the good old RC oscillator, at least concerning the business end. We're charging a cap, to a much higher voltage, and we may not be holding it there but we hope it arcs over in one half period of LC.Now I never talked to them but my boys Yiang & Kunieda seem to be making mistakes.Also their scope is 'micromachining' fuck that. we go big or go home.Occult 'stray capacitances' arent going to cut it if we want to stuff commonly cited energy per discharge in there without breaking every transistor known to man and probably everything else (I tried ofc).Also the resonant frequencies implied by the traces published by our gook friends here dont add up, not with the inductance values for L1 they quote and any reasonable estimate for the stray capacitance or calculated stray capacitance (from Vout I_L and L as above). You almost think they consistently confuse uH with mH and their incredible stray capacitance is part of the L1. This also explains the next plunder:Lin Jiang et al. / Procedia CIRP 113 (2022) 226–231 page 227https://www.sciencedirect.com/science/article/pii/S2212827122013518Implies the majority of the discharge current was inductor current. "...which means the peakdischarging current may exceed the inductance current i L,although iL is dominant."That's what originally got me interested, as that sounds like it's working as a current source. BS!
>>2966508It stands to reason that the spark gap length is a function of peak V_out, as sparks only occur if the gap is shorter than the equivalent isolation distance for V_out. So IF you get a spark it will always be roughly pi/4 of the LC period after switching off the boost converter, when the output voltage is at it's peak, which means ALL energy has transferred into the cap and L1 can in fact not contribute shit. You can only come to this conclusion if your stray capacity is between your inductor windings and you dont think but simply measure.It gets worse: With the random time element governing discharge formation discharges will be normally distributed around a point shortly AFTER peak V_out. Where it drops.Consider the first derivative: At 0 pi charge is transferred from L into C, cos(0) is 1, charge over time or current is highest, at pi/4 cos is 0 and sin is 1, the output is high, no current into or out of the stray capacitance. Later again cos will be negative, current is negative and since it is the only conductive path the current through L is reversed, so if any I_L is detrimental at that point.So anyways I figured it cant hurt to fix a few things and see if we can go big.(cont.)(unless you want me to STFU)
>>2966448>If you have some book recommendations for me, I'd read those immediately."Electromagnetic Compatibility Engineering" by Ott"High Speed Digital Design" and "High Speed Signal Propagation" by Johnson"Grounding and Shielding" by Morrison
>>2966501Oh cool, congrats on the A. Great to hear that it's working well.>>2966505>through some magic the charge on the floating end of C1 is restoredI can only assume a bipolar arc, I doubt the leakage currents through the fluid are significant (though the photo looks like there are fluid droplets on that sparkyball). That or their caps are breaking down, which would be funny. Despite their simple diagram, I assume they had some method of feedback to keep the system resonant? The implication is that C1 is much larger than the load capacitance, hence their resonant frequency will depend a lot on the workpiece geometry, which changes throughout the cut. Makes me wonder if a dedicated resonant SMPS controller IC could be used here instead of MCU finagling.Well, despite this being a resonant topology, they're definitely not driving it as one since the oscillations die out after each pulse. The inductor current goes slightly negative (back into the positive rail) after the discharge, but this energy is being dissipated long after the discharge has ceased.6% tool wear ratio seems pretty good though. Still probably too high for nontrivial toolpaths like I'd like to use for general-purpose milling though. Also it says their tool is 300mm diameter copper, which I suspect is a typo. Peer review eh? I can't see anything describing "detritus evacuation fluid flow", for want of a better description.>>2966513I'm not 100% following, but please do continue.
>>2966513So I did the needfull research. Changes I ended up with in the final design:Instead of relying on occult 'stray capacitance' a 100nF WIMA FKP1 was connected across the inductor. Note that this might aswell go across the transistors CE or DS.L1 was chosen to be small asf. With good reason, we had the discussion here. 3.3uH, core saturates at almost 200A.Switch ended being up SiC FET, wee bit hot on the gate - 5/+20.The coupling capacitor C1 was done away with. Instead there id a second switch, low side of the spark gap, to maintain DC isolation unless a spark is expected. This switch, its periphery and alternatives to it are probably a cahpter of its ownThe FKP1 cap just barely survives. All other caps, like those chinese '50 khz resonant switching', generic PP caps etc. all overheat or explode in no time. Maybe I'll go through the documentation and pull some of the dV/dt on the capacitor up but it was mostly consistent with the RC discharge vor a cap across 1R starting at 1,2 kV.So why the small inductance: This thing is a switching power supply. All power must pass through L1. Since the inductance is frequency dependent reactance this means, that a greater L1, for a given frequency, admits less power into the work. Sure you can fight that by choosing a longer T_on for a given large inductance but:We want to define discharge energy per spark (usually <50mJ I designed for a bit more than 100mJ at max power) and the energy is proportional to L but proportional to the square if I. Which is again just another angle of looking at 'more power through at frequency'.Sooooo did it work?
>>2966518>Picrel is the experimental setup, everything I tested. This allowed two main configurations:One where it's coupled through C1 and another where it's coupled through Q2 (or DC blocked if you want).You'll probably wonder about the caps on Q2. Well, they are dimensioned so that, with the maximum discharge energy transferred to them, their voltage does not exceed Q2_D_DS. Because while it would be nice to be able to conduct experiments with Q2 actively quenching sparks it's not feasible. Q2 would have to widthstand kV, pass kA and low R_DS_ON. Even if you were to drop in such a part, power wise it would be laughably underwhelmed.Naturally you'll think: Well must Q2 be synced then? No :) . It could even be driven at a different frequency than Q1. Which begs the question: Can R replace Q2? Probably :), if R is dimensioned such that, in case of a short at the output, Q1_I_DS is limited to safe values. The caps across Q2 would provide the low impedance path for the spark, before discharging through some R. You'd just discharge C1 into the cap C3.>>2966518>I can only assume a bipolar arc, I doubt the leakage currents through the fluid are significantThe fluid can be and was tested as air, water (ionized) and oil. Anything really. Im not exactly sure about core demagnetization and the C1 discharging issue. Note that C1 is arbitrary and could be where Q2 is but I wanted to work with what Yiang & Kunieda have shown, inb4 'but thats not the same'.Again, you can work with RD networks, as in the pic, or just R across C1, so that during one period it manages to discharge but does not allow high short circuit currents.The switch thing makes the whole thing a non issue and the switch with driver is hardly any cost (inductor was 30 bucks) and when you want to implement actual controls it sure comes handy.So yeah it does work, I am just a bit stuck rn:EDM machines consist of several modules not only the generator. i Underestimated how important flushing is.(cont.)
>>296498299% of the time it'll be fine so you'll have morons insisting it'll always be OK.Three problems:All caps have a max current rating probably listed on the data sheet usually spec'd as constant.There are "big" caps with low ESRs for good Q and thin wires so if you short them they'll damage themselves. Been there done that with HV.The problem usually isn't the cap you're shorting the problem, in the RARE event it exists, is the other caps. So you got a linear reg for some analog thing (mic preamp whatever) and you short the cap on the output of the regulator... well that is shorting the output of the regulator which will try to draw the entire input cap thru the regulator... maybe protected against that, maybe not. Or there's three caps in parallel on the board and the traces between them are not quite short circuit rated but would have been OK with a "normal" ripple current. Whoops, I've blown things up this way.Also whatever the cap is connected to probably has some limit to dV/dT and you might find a way to exceed it, making life very exciting on the other side of the transformer. You can blow up switching power supplies this way if you're dumb enough.
>>2966520>100nF>The FKP1 cap just barely survivesNot using one of those big soviet surplus doorknob caps?>a second switch, low side of the spark gapI've always thought that might be worth doing, though I was thinking of doing so in conjunction with an additional high-side transistor in parallel with the inductor, which maybe could be used to quench sparks if the workpiece geometry doesn't allow for one.>>2966522Ah, so you could test both series and parallel capacitance, and came to the conclusion that parallel was better? Having AC coupling does seem to get in the way of the concept of an EDM supply. Is D3 expected to conduct in the reverse direction at all, like some sort of TVS?>flushingYeah, the fact that diy machines use a modified pressure washer suggests to me that it's quite important, probably for both EDM and ECM. You've at least got adaptive (and reverse) feed rates working though, right?
>>2966522And also my scope was the generator and I had to avoid scope creep. I couldnt afford to fuck around with pumps, filtering, motion control loop, EMI (it makes steppers just go) which was demotivational. I wanted a grandiose demo but no time for that.Flushing: A whole lot of sludge is produced and everyone says you got to move it out of the gap. Without the process starts fine and soon progress stalls, as the power is dissipated in the sludge, not the work.Motion: I tried the 3D printer. When I upped the power soon the ADUM3160 would randomly crap out and need resetting. I was worried about my machine, all my stuff is on there without backup and running it through the ADUM would make the machine imagine SD cards being plugged in and all that. Also you wont have video on screen when it's running.I have more time to work on that.I also wanted a good comparative study of several basic topologies. I am currently working on building:RC, LC (buck), switched R, switched L... and try to get figures of merit for each. Which means I need all that or a very smart design for an experimental setup that avoid all the problem and provides consistent setup and results.>>2966518>keep the system resonant?Nope. You just want that one half cycle.200 A in through L1, Q1 off, 200 A go to C2, dropping until the energy is in C2. If no spark L1 is now reverse biased ~ 1kV, huge current back to U_in, demagnetizes the core and recovers through D1. If spark then it's the same, only much less of everything. I dont think the core can be walked into saturation, as we both agree C1 must somehow regenerate, so we get some back through D1 and any DC component through the gap through C1... somehow. You could make a very funny very high impulse current converter like this. Just add a secondary to L1, with diodes that are reverse biased during the buildup of output voltage, maybe find a way to maintain this state even longer, tap the secondary when the current through D1 flows.
>>2966531>Ah, so you could test both series and parallel capacitanceNo no, I've always used C2. It is the primary energy storage. Without C2, trying to cram 100 mJ at the node where Q1_D connects, would mean sticking 100 mJ into pF maybe single digit nF. You can calculate that but somnething will give.Now C1, the coupling capacitor, was shorted in a second experiment and the short circuit behaviour was controlled by Q2.Q2 is a can of worms: Get the control scheme wrong and in case of a short on the spark gap current will escalate, as Q1 and Q2 hand it back and forth. If you want to avoid that Q2 must switch off the current through L1 and the spark gap, thus it must widthstand the output voltage. A requirement that isnt easily met, Q2 would have to do kV and kA, depending on normal or abnormal operation. The caps on Q2 save Q2 from high voltage, if Q2 must switch off the current through L1 and the shorted gap. It also means there's always some coupling. This is beneficial if you want, C3 is a lower impedance path than Q2s 2mR.>>2966535Picrel shows the output waveform on C1, when C1 is an inadequate part. The distortion is C1 physically distorting and capacity being reduced as voltage increases. Thats the resonant half wave. Gate of Q1 comes down, C2 shoots up, following the sine of the L1C2 tank.> Consider picrel.1 system idle, Vout=Vin2 Charging, Q1 active, Vout = Q1_D ~0V 3 boost action, half period of LC frequency4 recovery through D15 residual oscillation at LC frequency
>>2966538sparking, Vout seems to have been set at around 700V.
>>2966540same event(?) zoomed in.~ -10GV/s
>>2966542experimental setup, trying how ceramics, capacity derated for VDC, would fare.Nope :)Also my supervisors like the aesthetic of someone actually putting the work in and not having AI generate reports.
>>2966544I think I'll add this one and that's it. This is the discharge waveform when using C1 as coupling cap. Those oscillations are suspected to be a result of the caps not discharging efully but, as explained earlier, hitting a shelf where VC1=VC2 and also some parasitic inductance that makes for a bumpy ride. The method using Q2 and C3 produced less side effects and a subjectively steeper discharge but not entirely sure.
>>2966531>You've at least got adaptive (and reverse) feed rates working though, right?Nope, as mentioned I had a scope and needed to avoid scope creep. Everythibg outside the scope wastes time and wont be graded.I interviewed various people and came across the person who programmed the stack for G-EDM. He rewrote the GRBL controller, it's now got a line buffer for reversing. He counts missed sparks and advances when a preset number of missed sparks occurr.
>>2966535Buck might be nice and efficient and easy to drive, but it would require a high-ish voltage DC power supply in the first place.>>2966538>Q2 would have to do kV and kAAh, so kV when it's quenching and arc and kA when in normal operation. Seems like a triggerable spark gap might be a potential choice, though I guess it would erode the electrodes unless they were mercury or whatever. On this timing diagram, when is Q2 on and off?>>2966540>>2966542Impressive. What did you do to set the 700V? Combination of on-time and feed-rate?>>2966548Yeah I was talking to the GEDM guy in a youtube comments section and he said he'd add a 3D arbitrary toolpath mode eventually, if not for EDM for ECM.>counts missed sparksAh, not feedback from the spark voltage?
I have a multi tap doorbell transformer that has 24v@30va, 16@30 and 8@20 taps. Can I pull the specified power from each tap (80va total) or is it max 30va or whatever from all 3 taps combined? I'm sure this is a Google away but I'm not entirely sure how to format the question.
>>2966555>triggerable spark gapSolid state technology is so far now that many things, like IGBTs, SCRs, Thyristors etc. etc. are obsolete or relegated to absolute niche applications.I've evaluated all feasible fully controlled solid state candidates for Q1 and Q2. SiC MOSFET for Q1 and Si MOSFET for Q2 really. My intuition first said IGBT for Q2 and bobs your uncle. Turns out that the 3€ part that is Q2 at 1 kA still drops way less voltage than an IGBT.>>2966555>What did you do to set the 700V?I'll explain again: The relationship between L1 and C2 provide a 'transformer' that takes in a current and puts out a voltage, if you will. Peak V_out = I_L * sqrt(L1 / C2) . Derive this by equating stored energy, It also is intuitive: Higher current through a larger inductance into a smaller capacitance means higher voltage.So this L1/C2 also sets a fixed relationship for V_out to spark energy. Which isnt bad, when you think about roughing and finishing operations.Once the thing is built, T_ON for Q1 can be used to adjust the output voltage. It will be limited by Q1 current capacity towards the high end. Leaving that limit towards the low end you get to a point where you can start increasing the working frequency, as T_ON is reduced. At 100 kHz it would still produce meaningful output and machining actions. 20-40 kHz is sort of standard. Eventually the output will drop below a threshold of effecacy.>>2966555>On this timing diagram, when is Q2 on and off?The axis are marked and some diagrams show Q2_G, which is on when it's high. Again: Thanks to C3 this is not strictly a neccessity. But the impedance of Q2_R_DS_ON of 2mR parallel to C3 might just be a little lower than C3 alone. The gate waveform for Q2 shows that there definately is interference, when it arcs over, but it's probably less severe at the actual gate versus the probe and in any case it only ever drops to 10V which still is a good ON state. Worst case C3 catches it.
>>2966555>Ah, not feedback from the spark voltage?G-EDM man doesn't do that. I think I know why.I did get sidetracked from time to time, because it all seems liek a dependency loop, so I did build the classic thing where 2 comparators form a window comparator and give a direction input to one of those DIP16 bridge motor drivers. It worked suprisingly well, just a resistive divider and a few stages of RC filter and into the window comparator.There's several issues connected to the spark voltage concept and I sort of like the missed sparks counter. I just found the values sounded very conservative but it's what he determined through experiments and his stuff works well.>>2966555>ImpressiveI had her run at the 1.2kV the Q1 is rated for just fine but it's rare to get a good shot of events like discharges so I took those that look clear, not show the highest output.I also found that even at ~750V the surface finish on the hardened ball for example shows quite large craters.I'm hoping to run more and better tests soon, when I get competing designs finished.Currently working on a +/-90V SMPS with PFC because I'm a good boy so I can run topologies that require high input voltage without paying big laminated iron,
>>2966556A 30VA transformer will only provide 30VA total. This is because any transformer is a compromise between size, window area and core volume.For the same size more window area means less copper losses, more core volume means less core losses or at least better dissipation. The core designs are optimized for a good balance at the power rating. So at 30VA the core and the windings are loaded good.
>>2966556Likely 30VA total, maybe a bit more. Looks like the 30VA is likely to be limited by the thickness of the primary windings or the core itself, while the 20VA is limited by the thickness of the wire on the 8V secondary. There may be some aspect whereby the secondary winding thickness is a limitation on the 30VA windings, but I doubt you'd be able comfortably to pull more than 40VA from the whole thing. Absent any documentation saying otherwise, I'd stick to 30VA.>>2966557>The relationship between L1 and C2 provide a 'transformer'Ah, so instead of a purely inductive generator whereby the voltage increases until it makes a discharge (voltage subject to gap width), you store it in the capacitor before triggering the discharge with Q2. Q2 turns on ideally at the peak of the LC voltage curve. Short circuits need to be absorbed as heat in Q2, and ideally detected before sending more pulses, open circuit doesn't produce an excessive voltage spike and the oscillation gets damped by D1. So it should be pretty reliably not blowing up anything.>>2966559Of course your system wouldn't work with gap voltage feedback at all.>Currently working on a +/-90V SMPS with PFCAnd building it will be cheaper than buying it? Might as well go for GaN to get away with smaller filter caps and inductors.
>>2966559Found a 900V shot that made it into the paper.Funny: Observe how the diode drop (D1) during recovery manages to almost lift the gate even tho it's active clamped to -5V.
>>2966561>you store it in the capacitor beforeyes, all the energy can be expected to be stored in C2 when it arcs. The original paper by Yiang and Kunieda is wrong in this regard.>>2966561>before triggering the discharge with Q2.No, cant really switch that fast, if it wasn't for C3 I'd be worried about Q2 not having switched fully by the time it gets bonked. Q2 switches shortly before Q1 switches off and is kept on for at least one half Period of the L1 C2 frequency.At 100 kHz the thing still puts out 230V, still more than tripple of what common RC designs etc. are built to put out at 20 kHz or so. Here you start running into trouble since one half Period of the L1 C2 frequency is now a considerable portion of the entire cycle. If Q2 must remain on that long DC isolation of the output, against shorts, suffers. Then again at those lower energy levels Q2 probably wont need much time at all to dum p out whatever C3 caught and Q2 will hardly mind getting bonked with 200V in any state, as long as it's got C3 attached. Things I'll have to see about. I'm also nut sure at which frequency sparks dont properly quench anymore.
>>2966563I should add:Thanks to C3 the discharge can occurr at any time. Even when Q2 is off. Q2 will then have to clear C3 to prevent V_C3 being walked up to more than Q2_V_DS_max.
>>2966561>And building it will be cheaper than buying it?Forgot. Building is hardly ever cheaper than buying. Maybe my websrach-fu isnt good enough but I couldnt find ~500W 90V SMPS /PSUI am for a whipe now pondering wether buying a cheap computer PSU, calculating and replacing feedback stuff and power magnetics is the best way.
>>2966560>>2966561Ok thanks. I'm planning on using my doorbell transformer to power some security cameras (my house number light is also powered by the transformer, so it's a convenient power source close to the eves) and was curious on how many I could realistically power along with the doorbell camera. I'll stick with 2, should be good enough and well below 30va/w total.
>>2966565I was thinking of stacking some 48V power supplies, but yeah interfering with the feedback loop of an ATX supply would be great if you could do so without frying anything. The magnetics is definitely the part I’m least sure of, but if the datasheet for the transformer is descriptive enough it should be pretty doable. Worst case you just rewind the secondary of the existing core. I wonder if there’s a model of computer power supply with published schematics? By the OEM or by some Russians.That or making a pair of 500W DC-DC converters, I recommend the ISL81801.
>>2966487>Thanks, always good to know you appreciate digits.>Off by oneFML>I just knew it. Same problem here.Pressing F>>2966495>Why would you ever not just wire it up to charge off the car’s ignition switch?You mean, charging the pack when the car is running and the alternator is producing electricity? That's what I'm going to do as well, but my car sometime sits for weeks on end so I need to be able to charge the pack externally and replace it as needed as well.
>>2966495Sorry forgot about that.The cheapest or simples method to 'balance' supercaps or batteries is to just shunt every series element using a high value resistor.The higher a cells voltage the more current, or charge over time, it will dump through the resistor. It doesn't achieve perfect unity balance but much more than running no scheme and no scheme aswell as this is often enough.>>2966577>>2966593This will continously waste power.>uuhh me heckin efficienciesAll those questions always boil down to the specific requirements. Burning a watt is not always strictly wrong. If you're that close to a power source and charging happens frequently and you can just add capacity because the weight and volume hardly matters. Why not? >>2966593>weeks on endAlways first estimate the requirements. Then figure out the limiting factors. Battery too big, too expensive, well time to save power I guess, like sleep mode with ab interrupt and event - > camera. Or maybe 1 fps and movement - > full fps, event? Save last few seconds and everything after. >>2966577I hate the stacking PSUs, did sometimes and sometimes it works. But it is also that you have to first reverse engineer the output stage and check what you're about to do, it wont always work and can do funny things like walk the trabsformer into saturation and stuff.
How do I blink a LED?
>>2966660
>>2966662This is exactly what I was looking for. Thanks, Anon. Merry Christmas.
>>2966660how to blink leds sequentially without any mechanical part or miconcontrolerhttps://nitter.net/RueNahcMohr/status/1993046729844048034#m
>>2966664I glued a tiny magnet on top of a CD and placed a ring of reed switches above the disc.
>>2965602That makes sense nowAlso it works, but sending a bunch of noice back up to the mcuIf I started the engine before the datalogger, it wouldn't read the SD card, and what saved was partially corrupted lelTime for more learnin
>>2964845I have the circuit of >picrel.jpg and I want to calculate the cmrr, but there is a catch, the Vo=10Vi, Acm=100, so how can calculate the Adm in order to apply the 20log10(Adm/Acm) to know the cmrr?
I hacked up a power strip by removing all the guts and replacing it with a 10A relay and a pigtail to a 12V fan port in my computer so it can switch on a bunch of audio electronics, including a 2400W Crown pro audio amp and a 400W stereo amp. The only thing I worry about is the inrush current from all the filter caps in their power supplies. Should I replace the relay with a 30A?
>>2966664That kinda reminds me of Strowger switches used for some of the earliest automatic telephone exchange systems:https://en.wikipedia.org/wiki/Strowger_switchYou could probably implement a simplified version with something like a stepper motor and a rotary switch, however.
>>2966743idk but i'd replace that fixed gain amplifier with an op-amp differential amplifier circuit and calculate it that way>>2966770you added a freewheel diode, right?you could add a conventional two-relay power-resistor rc-delay soft-start circuit if there's room inside the power-strip, so long as no load would ever protest to that kind of series resistance like an smps might
>>2966793Yes of course. Well it turns out that the relay is stuck closed now. I can cycle the relay and hear the clicking but the amps stay on so I guess the contacts welded themselves or something.
>>2966804ssr time?
>>2966770ren't those things inrush current limited anyways? The wattage implies those things have a PFC stage anyways, which would actively limit inrush before the first caps.
>>2966660You'll need an oscillator, as >>2966662 implies. The most simple ways to do that is using relaxation ocillators, cross coupled pairs or ring oscillators. Many ways to skin a cat here, including integrated solutions like OPs, Comparators or NE555 (read the datasheet on that). Other than that: Appropiate power supply. That's it.
>>2964845are there any commonly available modules that are capable of>usb-c charging of 2S 18650 cells>onboard boost converter 9V minimum>integrated BMSi tried searching amazon and aliexpress but it's the usual chinglish gibberish where the majority of the listings have misleading descriptions and might have a boost converter but no BMS, or a BMS but not boost converter.
>>2966813I thibk alot of the injoinic chips are designed for similar stuff, I cant tell if thry will run as low as 2s but maybe have a look
>>2966813Some “power-bank modules” support multiple series cells and have a USB PD output that does 9V with the right trigger board, but I’ve only seen 3S, 4S, and 5S versions. I assume you have a space constraint otherwise you’d just use 3S with no boost converter anyhow.The more piecemeal method would be a 3S charging board, a BMS, and a boost board, which are probably pretty doable to obtain.
>>2966856> more piecemeal methodAnd that’s probably the best solution, especially if you want something that can output at the same time as charging.It drives me crazy when something can’t run while it’s being charged.
Is the input impedance (22k) of the inverting opamp mixer too low for the jfets to drive, and that's why people have experienced issues?also the dipshit that designed the board didn't include any bypass cap for the opamp. guy probably never even built or tested it
just letting you guys know that a nonstick frying pan on a cooktop works better than reflow ovens, purpose built hot plates, hot air stations, etc.
>>2964845I might have autisim
>>2966955No. This is autisim.
didn't realize my scope had the tek wizard built into itnice
>>2966955That's not autism, that's ADHD
>>2966970That's OCD.
>>2966970Picrel you posted is made by some artschool faggot with OCD. What he makes doesn't even work, it's just a bunch of chips and wires.If he was less of a fag and more of an autist, he'd make something that looks that way, and also actually works. Just cobble something together with logic IC's and counters and LEDs or some other basic fuck-fuck shit anyone can make
>>2966970>This is autisimNo, that's a computer generated image
>>2967000>>2967011Nah. I can teach you. You will learn.
>>2967024I always fail at that. Perhaps because I tend to cram things too much. I assume you use solid wire. What gauge?
>>2966577I gave it more thought. Maybe you're right and I should just use any high-ish voltage isolated DC supply I can find for cheap or free, factor the efficiency of a voost converter into the power requirement and just slap a boost stage.Why do you in particular recommend ISL81801 over say the usual SG3525 and some diskrete stuff? I dont particularly like chips that will need the rework station to solder and may need thermal vias.And I do kind of always have this 'problem' where I want to use components I have in stock and end up with a product that's more work and probably worse performance than just getting a highly integrated solution and order everything the final BOM exactly calls for. I just dob't like the feeling of paying more for buying single or low volume components and also this is exactly how you rack up boxes full of MOQ things you'll never use again.>>2967029All you really need is YV hookup wire and needle nose pliers.
Hello. Can someone with knowledge explai nto me why in pic related we use two non polarized capacitors before the transformer? Is the function of this block to bypass high frequencies to the ground and not allow them to go through the transformer?What about the stage after the bridge? Why get the Vcc from this point? Why not get at the end of the line, say between C39 and C40? Why the designer did this? Does it make a difference?
>>2967069look up X and Y capacitors
>>2967069It's just interference suppression.Note that regulations apply for capacitors between L and N or either of those and PE.
>>2967068>Why do you in particular recommend ISL81801It’s the most versatile-looking bidirectional H-bridge buck-boost controller IC I could find. Also sunk cost fallacy since I ordered five pre-soldered dev-boards that I haven’t managed to get working yet. It’s not the smartest choice for a single-purpose boost IC, but if you would get use out of 5 assembled boards for different purposes about the place it might be worth it. I intended mine for use with 12V lithium deep cycle batteries, for charging and discharging. But also to experiment with a split-H boost-buck topology.Otherwise I’d just go the shitty way and slap a gate drive circuit on the output of a TL494. SG3525? Never heard of it.Personally I’ve taken to only ordering THTs and SMD jellybeans and passives to stock up on. More specialist ICs, especially ones that are a pain to solder, I just get JLC to solder on for me. Not like I’m working with GHz.But don’t let my ramblings interrupt your holy duty of reconfiguring a 12V or 24V supply to output 96V instead.>>2967069I think those are meant to be Y caps not X caps. They got the right spirit at least.
>>2967075> tht and smd … together again
>>2967075SG3525 is very similar to the TL494, just a popular programmable oscillator with PWM, error amplifier, output stage and packaged with basics like UVLO and shutdown. I'd say the main difference is you get Q and Qbar outputs and more drive but one amplifier less, IIRC. So always when I want to do something quick and dirty, especially switching regulators or dc drives,but not go full discrete I grab that. It's a nice balance between integration and versatility / ease of use IMO.
>>2967095no more brother wars
>legendary stacked SMD resistor monster sandwich
>>2967168kek, source?
>>2967170gemini 3 "fast"
>>2967171and what was the question?
>>2967172basically i'm looking for low noise resistors, though the thermal noise for a given resistance value is the limiting factor and not the excess noise, there might be a slight benefit for higher bandwidth and better RFI suppression
Still working on the SMPS, whatever rpute I take when it comes to building it I'll have to wind the xformer.Does anyone have experience with power magnetics?I did all the math and while it all checks out there's a few physical principles I'm not sure I really got. But the equivalent circuit is a fine way to look at it do meh.But I was wonderung when it comes to my secondary, I naturally want to use the window area. So a ribbon or rectangular wire would be nice. Can I simply make a multi strand secondary where the strands lay in a flat layer, parallel to each other?
>>2967174A friend of mine once made a sealed box with a battery and amplifier circuit, it just confirmed the law. There was nothing to be gotten from one R over the other really.
>>2967224I have a bunch of smaller E-cores. To avoid buying before I know everything else works I guess I could just parallel several transformers to get what I want. But is there a reason why I should not just stack the halves, 3dp a bobbin for it and have one transformer? Idk but when I look at the prices an ETD or E core for 500 W in a half wave config, like 2 switch forward, with bobbin, wire and clamps and all probably costs 20 bucks or so.
>>2967224>I'll have to wind the xformerNo experience here (non-isolated is fine for me), I will just say that I’m deathly afraid of flyback converters because I don’t know how to characterise and validate their transformers, or how to figure out what specs I need in the first place. So any isolated converter I’ll ever design will use a push-pull or otherwise buck-like topology instead. I guess you can add an extra inductor or two to replace a flyback converter’s leakage inductance.Fuck it, just make a big dumb open-loop ZVS switcher.Smaller parallel circular windings won’t be any more spatially efficient than fewer larger circular windings. But they are easier to wind, especially around tight bends. No clue where to buy that flat enamel wire.
>>2964845Hey guys, I'm looking for a surface mount equivalent to the BC639. Is the BCP56 a good choice, or is there a better option?
>>2967095Anyone know where to get that style of lead/leg/whatever for a pcb like in that pic? I was thinking about it for a project that I might try in the future.
>>2967263No.
>>2967263>Anyone know dont be a lazy fuckif you dont know what something's called, circle it in the pic my guess is you're either looking for a ''pin header'' or ''DIP component carrier'' and you get both from Digikey, Mouser, Newark, or any of the retarded suggestions in the first post of this thread
>>2967266>>2967267No need to be such smart asses, I know what a pin header is, but I like THAT SPECIFIC STYLE in the pic that mimics an individual IC pin, not the header strips that you snap to length. I'm looking for it because it's for a component that would go into some equipment from the 70s and I am extremely autistic about the way that it would look even though I am the only one who would ever see it. If I could find some by just searching "header pins" I wouldn't have fucking asked. Apparently the ones used on the PCB in the picture went by the name of "Flip-Pin" and haven't been available for quite a while.
>>2967266i wonder what the edge of a multi-layer PCB would look like if polished? I should polish one of my boards up and see.
>>2967324>PCB would look like if polished?please stop, siryour brain has been adulterated by a techno virusuntreated, you will end up like this sad deranged loser
>>2967333>brown PCBsimagine the smell
Would this general idea work for driving an NPN switch from 12V while getting close to 12V at load terminals? I think the load would receive something like 10V at most if the NPN base was driven from the same 12V rail supplying the load. There are countless cheap, small low-power boost converter ICs available that could boost 12V up to 14-15V.
>>2967388are you sure you really need a high side driver? also, who not a MOSFET?
>>2967390>who not a MOSFETNPN needs just 1V to switch on. IIRC regular MOSFETs need at least 5-6V for low on resistance.
>>2967392use a P-channel device. and similarly with the BJT: use a PNP device instad
I have an electrical device which failed and I want to try and fix it.How do I tell if a diode has failed? I set my multimeter to the diode setting and realize that I have no clue what number should show up if its working. Also, do diodes tend to get really hot normally? And is it more likely that -assuming the diode is dead- the problem was the diode itself and had some kind of flaw or fault that made it's failure just a matter of time, or some other component or environmental factor that caused an otherwise healthy and normal, non-faulty diode to die? Ie, do I replace the diode or look for something else that's broken first?
>>2967409A diode is the electrical equivalent of a check valve- it allows current to flow in one direction but not the other. Make sure your probes are in the right sockets, set the meter to diode/continuity test mode and probe the diode. It should beep with the probes in one direction but not the other. Any number you see on the meter will be the forward voltage drop of the diode.
>>2967409>do I replace the diode or look for something else that's broken first?Look for burnt components, discoloration, and bad solder joints first, then check for shorts. What's the device and how did it fail?
>>2967412what happens when the number increases over several seconds until it goes out of range?>>2967415Nothing appears visibly burnt, several of the solder joints look less than perfect but nothing that should be a problem.The device is, uhm... Lets say its a DC motor and associated power supply that takes wall power in and is designed to operate in a high-vibration environment...
>>2967388Seems rather flawed. If you absolutely must do high-side switching (i.e. load must be ground referenced) then use a P-channel MOSFET as the other anon suggests, since it doesn’t require any boost conversion and has lower voltage drop. You could use a simple logic NPN or Nch transistor switched by the MCU with a pull-up resistor that controls the power Pch/PNP transistor. A BJT may need less voltage to turn on, but it needs a lot more current to turn on. The required boost converter would be much larger for a BJT as a result of this.Low-side current sensing is also probably going to be easier to measure from, the voltage drop usually isn’t an issue but it might mess with ground referencing.
>>2967424>what happens when the number increases over several seconds until it goes out of range?That means the multimeter’s probing current is charging up a capacitor, not going through a diode. In other words, the diode is not conducting in that direction. If you’re getting to a constant 0.3-0.7V or so in the other direction, the diode is fine.
>>2967431>That means the multimeter’s probing current is charging up a capacitor, not going through a diode. In other words, the diode is not conducting in that directionbut shouldn't it read 0 in one direction and 0.3-0.7V in the other? Or is it charging a capacitor just because its not isolated from the rest of the board and it's finding a way through some other way?>>2967424>>2967415>The device is, uhm... Lets say its a DC motor and associated power supply that takes wall power in and is designed to operate in a high-vibration environment...Also in case it wasn't clear its a "massager".
>>2967324IBM 50-layer z-series mainframe backplane.Sometimes, 48 layers just isn’t enough.
>>2967436>Or is it charging a capacitor just because its not isolatedYes. Have you tested the wall wart for proper voltage output? Another thing you can test is the motor coil resistance and the power switch.
>>2967441I kind of want to stand on that. >>2967426In most cases you're well advised to use a proper gate driver anyways and spend only a few cents more. Alternatively a simply boost with just N-ch, diodes and a cap always materializes fast and depending on what you do you can always whip up a gate transformer. >>2967240>isolated buckLike a coupled L buck buck? Why not include forward and 2 switch forward in your list?>characterize the transformer.Since I've spent a fucking week or so on the transformwr math already I can walk you through if you like. It is my understanding that the flyback transformer also is very forgiving: The air gap helps you avoid saturation and everything besides the air gap is really as usual. I have a large-ish flyback cire sadly it's no use for me rn. They scalr poorly with power requirement - which is why the one I have is big.>open-loop ZVSI so hate that thing just because the characteristics of the crowd it's associated with. And c'mon, its a cross coupled pair with feedback. Either way, if you want that thing to behave you still need to get the transformer right, then you need a PSU for the thing, which is gonna be a big one, etc.>no spatial efficiencyWhy? This creates-approximately, a flat, rectangular conductor. Assume you need 5 turns on the secondary and you'd size your wire so that 5 turns cover the bobbin top to bottom, your layer will be 6mm thick with lot's of air. Use 6p of 1mm wire instead, wrap it 5 times, you'r layer is only 1mm high. Sure it wont do the current a 6mm dia wire does but...>>2967258Understand the parameters that matter in your application, pull up the datasheets, make sure your application is within the window for all parameters and check the ones that really matter are a close match. >>2967324shitty its not homogenous.>>2967388This so stupid my man. Sorry. You got spare pins on Atiny? Just use another to clock a charge pump. >>2967392Since you already insist on sythesizing a voltage...
>>2967469>>2967392... you can have any voltage. Also enough to drive a high side N-ch MOSFET (32).-use a bootstrap, if it's either switching or you can allow it to - briefly - switch the output off to pump.-use a P-ch (I never do Peasants must be shunned). -use a free pin to charge pump-use a transformer-low side N-ch with 'bootstrap' or independent charge pump?How much current? I mean 2n7002 are 'on' at uh... 2V?
>>2967436An ideal diode should read open circuit - or infinite volts - in one direction and short circuit -or 0 - in the other.Real diodes usually read between 0.3V and 4V in forward conduction and a much higher breakdown value in reverse bias.A typical tester only tests up to 1 or 2V or so, it wont show the breakdown voltage but instead be satisfied that it reads open circuit and display that.It's not always possible to measure components inna circuit in isolation. The first step usually is understanding the environment, aka what else is directly connected, especially what may be shunting your device.A discharged capacitor will initially look like a short circuit to your meter and as the meter applies voltage to test the diode the capacitor will charge and the meter will creep up in voltage until the capacitor is charged to the test voltage.There's plenty of methods and experience to electronics repair and even professionals sometimes poke in the dark until it's too far to be saved.At any rate: If the vibrating device / massager is intended to be in physical contact with a person stay away from the transformer and anything bridging the primary and secondary side. Also make sure you understand what the fault was before putting it back to... work.
>>2967436>but shouldn't it read 0 in one direction and 0.3-0.7V in the other?No the meter reads voltage, not current. It should read 0.7 volts in one direction, because that’s as high as it goes before the diode starts conducting. But in the other direction it should read arbitrarily large volts, because it’s trying to put its test-current (e.g. 1mA) through the circuit but it maxes out at just a volt or so. You need to bring the reverse voltage over 100V to get most diodes to start conducting. In this case it reads something like “OL” or “-”. If the diode mode reads close to 0V (less than 0.1V) it indicates a short-circuit, and a dead diode (or something else is shorting between those two nodes).>>2967469>gate driverEh, if he’s considering switching a load with an NPN in the first place he probably won’t mind a fractional watt of switching loss. Pull-up resistor gate driving is fine enough for 3D printer mainboard heating elements. But it really does depend on the load.>isolated buckI said buck-like, IIRC forward converters operate until the same kind of mode. What matters as far as I understand is that energy isn’t stored in the magnetics, which you need a flyback transformer for. As I said, I stick to non-isolated designs.>The air gap helps you avoid saturation and everything besides the air gap is really as usualGood to know. I’ll put a few sheets of paper between the halves of an EI core.>you still need to get the transformer rightTried to make one run at 1kHz once and it didn’t. I hear everyone just runs them at normal 100kHz+ and they just work. With unnamed ferrite toroids.
What is it with those SPICE models, that just don't work with ngspice?https://www.vishay.com/en/product/83740/tab/designtools-ppg/They give you a binary .lib and a txt. The frontend for ngspice provided by KiCAD calls for the .lib.Once that is loaded you should be able to select a model from the file, in case there is several, and then assign the symbol pins to pins defined in the model.My issue is: It often does not even find a model in there. The txt file will not load / crash.
>>2967488FML this is how short my attention span has become
>>2967488solved. encoding problem or filename. Vim would just show what looks like hex. I got the Ascii and put it in a new BJT.lib. The model is gonna be shit anyways.
>>2967458>Have you tested the wall wart for proper voltage output?no wall adapter, it does any and all rectification and regulation inside the handle. It's got a transformer in there and a capacitor that makes it plausible to see it rectifying it, but no big obvious full bridge rectifier chips.I'm almost thinking it might be easier to just remove the board completely and wire the motor to some cables that could be connected to a bench supply. That way I don't have to worry about what broke and why, and if it still doesn't work that means the motor is bad in which case any other tinkering would have been pointless. The thing WAS heating up pretty substantially prior to failure.
>>2967497If I had to guess, it's probably the motor going bad. You could hard wire it just to test it. If it makes funny noises or gets hot then it's the motor.
>>2967497Are you saying you have not tested the output to the motor?In doubt you could remove motor and drive it off an external source. But testing what the motor is supplied with should tell you.>no rectumfryerMany appliances use universal motors. Those will run on AC just fine. A transformer may just be a convenient way to isolate the motor and perhaps provide step down.Since the 'thing' is connected to AC the law states that it has to be insulated. So as lomg as you get the ecterior structure baxl together exactly as it os intended you should be safe from whatever bodges you put inside.
Still working on the SMPS.I am now validating the design by running all the math, SPICE and overthinking it:>picrelIs from the application note for ML4800 and compatible chips.I stumbled across this:The schematic, explanation and all makes sense.I's just AC line, rectumfryer and divider, if you ignore the caps / filter for the time being.The author always implies RMS when talking about line voltage. The author is talking 110V +/- some here.So wtf is it with the sqrt(2) ? The cap is only going to average out what it sees on the divider and the divider is applied to | sin(wt) * VRMS | so ARV. So why sqrt(2)? Show me the peak detector.
>>2967535Complaining to (You) always works. There it is:Pi/(2sqrt(2))=1,11They simple omitted or forgot about that later. And probably got away with it because THEY NEVER PROBABLY VALIDATED EVEN THEIR REFERENCE DESIGN and designed it for universal mains, with the extreme ends being far from what they have at the wall.I guess I have to go figure where excatly to put that VRMS voltage now.
You might not like it, but this is what (Im)proper rf shielding looks likeAll my hommies hate ignition coils
>want to make a pi hat>don't want to pull everything out and actually build itIs there a cure for this (other than amphetamines)?>>2967739Looks like it's ready for orbit 10/10
>>2967742>Is there a cure for thisStop thinking and start doing. You can think when you fuck something up. lmao
What is a quick and easy way - also not to computation intensive - to make a behavioural Vpulse in SPICE, I want to control tw inversely proportional to a voltage.
>>2967742Only took me 2 years to build a datalogger instead of paying $700 for one but I fuckin did itI posted before but engine noise was fucking up SD card access, so I made my own shielded cable and fuckin taped the whole box I made while I was at it
>>2967743It may be called manic depression, which would probably be my case:Never get out of bed, never do anything, daydream about everything you'd do, overanalyze it, make up excuses or find something or someone to blame for why you aren't doing it.Until the switch flips any you do everything, actually productive and fast in a mania for several days, until the switch flicks again.Fun fun.
>>2967744in the meantime ill go with triangular vpulse and comparator but there must be a better way.
>>2967749This might sound stupid to you, but start taking vitamins C & D, and get your heart rate up to 120bpm for at least 20 minutes a day. Fresh air and sunlight too when you get a nice day.
>>2967751and drink lots of water.
I want to use pic related PSU to power a dell optiplex 7050 in place of its usual PSU. Dell uses a proprietary 6 pin connector for the board power. According to https://www.dell.com/community/en/conversations/optiplex-desktops/optiplex-proprietary-8-pin-and-6-pin-pinout/647f8f68f4ccf8a8de077d92 pin 1 and 4 need to be +5V pin 2 and 3 need to be GND, and pins 5 and 6 need to be +12V.If I were to pull 12V off of one of the unused connectors from the PSU, take out a +12 rail, hook it up to a buck converter and regulate it down to +5V, could I then splice it together to the system power connector and make the thing work? I'm assuming the 5V lines aren't carrying much current as the system board does all the regulation to supply 5V and 3.3V.
>>2967755You should look at the current ratings.Last time I looked at a dell PSU the 12 V current was cosmetic with a pathetic output current for driving 2 or 3 legacy hard disks or CD drives.There is a reason dell has proprietary connectors, it’s not a scam for once.That means that most of the supply went into the 5V rail (the bean counters probably made this power strange calculus work).That said, with a 12 V PSU line and an efficient DC/DC converter, you should be able to get twice the current out of it for 5 V.Note that many PSUs just have all the 12 and 5 V wires tied to a single lug on the psu’s pcb.
>>2967751Funny enough, for a while now I am physically very active, outdoorsy alot and the 2+L or 3L water a day thing.Well sure this whole thing first appeared before that and it has improved by now, idk if it's this change or whatever else might have contributed to improving.It's just that there's no cure. Improve sure, but it leaves an eternal aftertaste and it's a very weird aftertaste. It's not like you can reverse that shit.>picrelYEAH BUDDY! Turns out a silly little PFC stage really makes a rectumfryer capacitor load look ohmic to the grid. Who would have known!>mfw so ohmić
>>2967766If I recall, the input current waveform with PFC should look like a sinusoid that's in-phase with the voltage waveform, but those green and tan waveforms don't look very sinusoidal so I'm guessing they're current out of the filter caps?
>>2967763>Last time I looked at a dell PSU the 12 V current was cosmeticThis one ONLY supplies 12 volts. It has 2 12V rails, one at 16 amps and the other at 16.5. It doesn't actually supply 5V really, its just a signal voltage that it generates. I think its the board telling the PSU to wake up and the PSU telling the board that its OK to supply power. The motherboard handles all the voltage regulation for 5v and 3.3v.
>>2967769It's a simulation I bodged fast to sanity check. The waveform is far from perfect, as for the control I just used a voltage controlled switch to compare a 100kHz triangle to the 50 Hz mains and adjusted the offset until it started to behave as expected. So the whole feedback is dumbed down to this one thing which can not work as the relations are not linear, right? As in 10% line voltage doesn't call for 10% duty cycle that high line voltage calls for. Also the load is constant.But IMO it's visible, especially when you compare it to the peak charging waveform of something uncorrected.Regarding inrush current limiting NTCs:They state a max power dissipation. In the low R state? I don't really expect it to get fried in the first few cycles if it then goes low ohm?
>>2967773>NTCsIntegrate their power dissipation to approximate their temperature, and ensure that it doesn't cross some maximum temperature threshold. I think spice can do that kind of integration.
>>2967742I think this is pretty normal for self-driven creatives. If you can’t structure yourself around deadlines or other people pushing one another forwards, you might just need a change of pace or perspective.I built a pair of ham radio antennas a month ago, both failed to work the way I wanted, so I’ve been spending the last month playing video games instead. War Thunder sim ground battles are actually kinda fun. But man, imagine if all the man-hours put into video games did something productive. Like building a pyramid or whatever. Anyhow, next month I’ll get back into it. I always have like 12 projects on the back-burner so I can always pick a different one. By that I mean one that doesn’t require any programming. Maybe I should get on that Claude shit.
>>2967774Well that is the obvious definitive way to do things.You'have to rely on a few assumptions or first measure the exact behaviour of the part in all sorts of ways.But generally good enough estimates exist. And IMO doing the measuring and math would be a bit too effort intensive for one rather irrelevant part.Which is why I assume the ratings mfgs give in their datasheet have some relation to the usual mode of employment: Put it in a mains line and run that into some sort of cap.Thinking about it I suspect you can approximate things by assuming it should behave in an exponential fashion, similar to an RC curve.
Is there anything wrong with torroidal ferrite transformers? Is it more than just the automated manufacturing difference?E(xx) cores be like:ExpensiveCoil former is difficult to findneed stupid clips tooMeanwhile chad torroids:Just wind your windings on that fucker. It's all self supporting.No further assembly required.Windings double as pins.Heck I might just put the E(xx) 47 or so compatible footprint on the board but I feel like I want to just get a big fat torroid and validate everything. I can still make the appropiate transformer later. And have a useful left over donut.
>>2967779It's not going to be linear though, resistance will be like a sigmoid curve. The peak power production for an ohmic load would be when the thermistor has half the total supply voltage across itself, so you can at least size the power rating of the thermistor based on having half the mains voltage across itself. That said, for a capacitive load (e.g. charging up the input caps before the SMPS's own load current begins) the total energy delivered to the series thermistor is resistance independant. You can just assume the whole 0.5*C*V^2 is dumped into the thermistor, then use its approximate heat capacity to figure out how hot it gets assuming it eats that all at once. If it's too much, then figure out how long to space it out by and size the resistance appropriately.>>2967784Winding toroids is ass. If you need to wind 20m of wire, then starting out will involve winding almost the full length of that wire through the toroid's hole for every single winding. Or maybe 50m if you start half and half, but that's not feasible if you have to wind more than one layer. 100 windings with 20m of wire would mean feeding a total of 1000m of wire through the middle of the core. If you're doing RF stuff, you've got like 10 windings or so, you're laughing. If you're doing SMPS stuff, if you're using a decent ferrite at 100kHz+ then it might take up to an hour or two. If you're winding a mains transformer, you're better off training a small third-world child to do it for you. The big mains toroids are also large enough that they need proper mounting with a bolt through the middle.EI cores can be as simple as winding wire onto a plastic bobbin (on a drill, lathe, etc.) with constant tension, then slipping that down onto the core once you're done.
>>2967776>WT>sim>groundGod oh god, I'm so sorry to hear that. I hope you get better soon. I too have so many ideas on the backburner and I've been spending time on vidya, albeit good ones (I've been clean from WT for years now). If I can muster the mental energy I might manage to make that analog PID demonstrator I've always wanted to do.
>>2967791Sim has less of a CAS problem and smaller battles, and forces you to use a different lineup every day. It's kinda comfy, until you get blapped by a false flag tank. I was clean for a good year or more, but I've just worked through a retail christmas so I deserve a break.I've got a bunch of rotary switches I should probably use to make a decade box myself. That and 5 assembled but untested DC-to-DC converter boards I should probably make a battery charger, solar charge controller, and benchtop power supply from. I expect at least 500W capability from my naive design. Disappointment is the spice of life.
>>2967795>I should probably use to make a decade box myselfreally, nigger?you dont have 7 dollars which will save you tons of work and money?
>>2967788yeah pretty much what I've said I believe.So suspicion confirmed, no downsides to toroids other than PITA.The last core I calculated this for, ended up being a 50:33 ETD49/25/16-3F3 at 100kHz. I'll run the numbers for a few of them fair-rite toroids mouser has. One of those on the list has a uh 4,5mH per sq turn. Another weighs 188g. Not gonna heat that up with what, 15W loss or so.>>2967808Use pots?
BTW chip for the SMPS needs barely any power. As usual it starts up via resistor. Afterwards some choose to feed it of an aux coil and regulator. Any reasons to do so other than the .5 % decreased standby power?I might end up giving the thing standalone gate drivers in case of later upgrades, si I might go auxiliary coil - isolated DC-DC - chip and drivers.
>>2967808I already have the switches, and resistors are cheap. As for effort, I’d rather spend a few hours making a decade box hats convenient to use, rather than that piece of shit that uses jumpers. Having higher watt capacity (especially for the low resistances) would be nice too.>>2967837Less heat production is also a benefit of not using a resistive regulator. Especially so when using 240V.
>>2967842heat <-> efficiency, the terms appear interchangeable to me unless making a toaster>>2967788calculating this for some toroids came out - again - at either 33 or 32 Np. Who would have guessed. I calculated the length of wire, neglecting the pitch, leads, etc. Just based on turns and circumference of the toroid cross section. It's only 2m for 33 turns so I'll got with a nice donut.I have found the difference tho. Sure the toroid offers nice window area but it is a much much bigger component really. If you think in rectangular footprints at least. ETD49 is - unsuprising, about 55*55. One of those donuts is 74mm outer dia. Guess I'll have to stand it up or leave alot of room if I want to design for ETD49 but go with the Dunkin' first.
>>2967842When I’m done prototyping my project… sometimes using decade boxes… I just stuff the decade boxes, solderless breadboards, etc. , into a big project box and “ship it”.Sometimes, if it needs a readout, I’ll hot glue my old multimeter into the box with a little cut-out window and buy a new one.
>>2967745I want to kill myselfAt the track with a live motor, and the fucker ain't loading from the SD card stillIt lists the tracks, then shits the bed trying to load it inMaybe I should disable the RPM interrupt while reading and writing....
>>2967776Funnily enough, I'm also putting off building antennas. Fucking around on the computer is just so much more consistently rewarding than doing things in real life that can fail and require cleanup. A terrible state of being, I know.
>>2967881Dev-boards are one thing. But deploying anything with solderless breadboards inside is just asking for trouble. At the very least use a breadboard-patterned proto-board.>>2967891If you can fail fast it’s fine, but when you put in tens of hours of effort only to be faced with hours more of troubleshooting, if not being slapped back to the drawing board, it’s demoralising.The resulting incentive is to subdivide projects into smaller steps that are easier to attempt and quantify. Often one of those steps is metrology itself, since one of the barriers to quick projects is being able to validate the intermediate steps. If you’re designing a 3D filament drier, you shouldn’t get to the point of having a drier with filament in it that ends up printing like shit, you should buy/make a thermo-hygrometer to properly validate your design before you proceed.Yeah so anyway I’m like 15 hours into a thermo-hygrometer that goes below 10%RH and shows absolute humidity and I hate it.
>>2967902>you should buy/make a thermo-hygrometer to properly validate your design before you proceed>Yeah so anyway I’m like 15 hours into a thermo-hygrometer that goes below 10%RH and shows absolute humidity and I hate it.kekI'm just plopping a GPS module on top of a Pi. The hardest part in my estimation is going to be installing a pin header socket upside down. I've already got it breadboarded and working... except when the worthless Chinese jumpers I'm using don't make contact... or when I forget what colors I used and connect the signals wrong and spend an hour troubleshooting it before pulling the scope out and realizing that I'm retarded. Which is why I want to make a hat (plus I want my breadboard back).
>>2967905I'm building it, boys. It's going to be less "hat" and more "bridge" because I've got a perfboard offcut that's about the right size and I don't want to do the upside down trick, but I'm building it.
>>2967916Imagine going back to 1987 and telling someone that you're building a hat for your Pi.
>>296792024 hours in our world would give anyone from 1987 permanent psychosis and/or brain damage from the sheer shock of everything.
>4 hours to do a <20 joint THT projectThis is why I was putting it off>>2967920An 80s nerd would get it if you explained that it was an expansion board for your computer. It's the part about receiving the time from military satellites with 10ns accuracy that would raise an eyebrow.
>>2967842oh well, when calculating the transformer I forgot that I even had put a half wave bias winding in my schematic. Doesn't matter there's space.So the way it is, there's R from the rectifier out to a cap, supposed to charge it until it comes on and the cap is supposed to hold above UVLO until the transformer is operational.Now I am wondering: I might aswell put a PFET in there, with a smaller resistor, and use, for example, the bias output to turn off the boot circuit. But the above is probably good enough if I R and C are large enough.Thoughts?
Correct me if I am wrong. I believe I can - assuming ideal behaviour only - simulate>picrela compensation network on the output of a VC-OP
>>2967960like so, right?
>>2967961>picrel is what I get of a 50R source, but ofc that doesnt make sense. But the plot makes sense.
>>2967952>added a blinky LED>cleaned up>finally done at T+6 hoursAt least it works
>>2967957Might mess with hiccough output fault state, if it turns off the output and there’s too much of a delay before the FET turns on again it could brownout the whole chip. Maybe switch it with the power-good pin directly? Or ditch the FET idea and use a discrete zener linear regulator with 300V pass transistor that outputs somewhat less than the feedback voltage. So the pass transistor’s output current I basically drops to zero when the feedback winding is outputting enough voltage. The quiescent current could be dropped by a factor of one hFE if using a single BJT, hFE squared if using a Darlington, or arbitrarily much if using a pass MOSFET. TL431s have too high of a quiescent current, but a zener is usable well into the uA. Just watch out for the Spiritio effect.
>>2967888>Maybe I should disable the RPM interrupt while reading and writing....seems reasonable, maybe out a physical switch on the antenna too
>>2967975Use a voltage divider to bring the "antenna" voltage down and feed it to an optoisolator, then to your i/o.
>>2967973AFAIK the chip has a mechanism for this, there's a connection to an external cap, fed from a current source and it's voltage is compared to a reference for 'soft starting' but honestly it's just a delay after which it goes full frequency full PWM. It's so the stage is set. When UVLO or any other fault happens that cap is discharged.This got me thinking about the only part of the schematic that isn't finished yet:Gate drive. Two switch forward. So a high and low side switch, both in sync (prefferably, so they share the switching load). The chip has 'integrated drivers' but those are very poor, maybe if you want to switch tiny FETs for a low power SMPS. So the obvious way is a push pull stage and a gate drive transformer. At least for the HS. Now I got plenty of single HS drivers. Naturally they do LS just fine.So I was thinking: Either have 2 bias windings or tap off one at 15V to 20V and use that for gate and chip, perhaps give the chip a zener if I end up with 20V. Use a LS and HS driver. Bootstrap HS.Problem is: I have no idea how sync the drivers would be in a noisy environment. Naturally on the bench they behave very identical. They usually have this delay to make sure a HIGH they're seing is a HIGH. The delay is the same for two.Stupid idea? Just chuck one driver in there and the gate transformer? Cant really think of another way to get two outputs from one such driver, one potentially several hundret volts higher.
>>2967975I tend to treat interrupts this way: Always disable all, identify states in which to allow and explicitly do so. The other way round always gets you.
>>2968026That gives me a pain in the ass idea. He could use an LED and a phototransistor to shoot the antenna signal through a TOSLINK cable. That would solve the potential problem of EMI coupling onto the signal wires, and also create a bunch of new problems that need to be solved.
>>2968029Why not use one gate drive transformer with two secondaries? That should keep them pretty-well in sync. I can’t see noise interfering with this topology even if you have two gate drive transformers. Do you lose anything much if they’re 1% out of sync?>>2968066Did you know that you can just buy toslink input and output connectors, with integrated optoelectronics? But using cheaper LED light-pipes are also an option.
>>2968066It's a good idea but the fiber needs to be insulated from heat and vibration, and the components would need to be sealed in a metal box. Maybe shielded coax that's grounded at the spark plug end and the tachometer box end? Also some passive RC filtration for HF beyond the desired range/RPM.
>>2968073or a ferrite bead/clamp.
>>2964845Sorry to bother, I need to get the transfer function of the circuit of the figure in the Vo/Vi form, but I get stuck and don't know how to relate Va with Vi in order to show the Vo variable and make the equation. i used block diagram to represent the circuit, I see two summing points, but IDK if this is right. Anon, could lend me a hand and help me to get the transfer function, please?thanks in advance Anon.
>>2967920The light bulb in my room has better specs than some 80's home computers.
>>2968077But can your light bulb run Pimp Wars?
>>2968072>Did you know that you can just buy toslink input and output connectors, with integrated optoelectronics?Chang doesn't appear to sell them to gweilos and Digikey/Mouser want $15 a piece. African American engineering is the only reasonable solution.>>2968073>the fiber needs to be insulated from heat and vibrationDoes it? I don't see why it would be any more prone to melting or rattling apart than whatever wire he's using.>the components would need to be sealed in a metal boxThey sort of are >>2967739
>>2968084Excessive vibration of fiber can cause spikes/errors. Cheap fiber is plastic so it should be kept away from the engine or insulated from heat so it doesn't deform.>They sort of areNo, that won't do. The circuit board(s) should be grounded inside the (metal) box, and the box should be grounded to the chassis. Adding a ground loop isolator might help too.
>>2968076Is that second amplifier meant to be an adder? It says VB on its non-inverting input, but it also has a ground symbol suggesting otherwise.
>>2968101I'm assuming it is an adder too, but I'm not sure, because of the same detail You have spotted.That's why with the blocks used two summing points. That and the fact I can't relate Vo with Vi baffles me.
>>2968106Well if it is two summers, I got:Vo/Vi = 1/(dRbR/(RaR) - bR/R - R/aR)If it’s just one, I got:Vo/Vi = 1/(-R/aR -bR/R -bR/aR)The transfer function for an unbalanced differential amplifier is:Vout = (V+ - V-)*Rp/Rs + V+And cascading two together, alongside that third stage gets a lot to cancel out. I got:Vo = aR/R*(Vb-Vi) - Vb*dR/RWhich is more elegant than I was expecting. Then just describe Vb in terms of Vo, substitute it in the above equation, and rearrange.
>>2968109First, thanks anon, very kind.>describe Vb in terms of VoBut how?, what's the trick to relate both?Bro, I'm ashamed to ask, but could you post the math that let you get Vo/Vi's please?I simply don't get where appear the 1/oversomething.I guess the cocient, but it does not matter what do I do, never reach the part where I should group the Vi's and simply make the fraction. I have almost a month, and never get it. It's driving me crazy.Thanks a lot again, anon.
>>2967390>>2967426>>2967469>>2967471Thanks for sharing your ideas, I will consider them if I end up prototyping that circuit. Pic related is another thing I've been working on recently. It switches the isolated output alternately at 24V and 36V. Breadboarded a similar circuit and managed to get it working. Although I used a pair of lab PSUs as power sources instead of flyback, boosters and all that stuff. However I'm not sure how I should power the optocoupler. Its supply should "ride on top" of the 15V drive rail. In my test circuit I bodged 3xAA batteries between its Vdd and (floating) GND pins. Worked like a charm, haha. Maybe the aux winding could be used to power the opto? The aux winding is currently unused because the flyback controller uses primary-side regulation. Does this make any sense? At this point I've absolutely no idea what else I should use instead of random 3xAA batteries.
>>2968072Yes yes that's exactly what I mean by gate drive transformer. You use one, with a HS and LS and drive that if a common driver.The only alternative I see is a driver for each ut maybe I should thibk about wether it indeed is possible to have one driver provising two sync outputs, one bootstrapped and the other not.>What use is itBoth FETs share stress IF they operate in sync. If one has already switched befoee the other even starts to that's not happening anymore. >>2968084 Damn would it look cool if they made opto isolators with a fiber link.>>2968109Why does the professor make everyone do stupid calculus instead of have them explain what is being done and why.>>2968129Mqn it's latw but thats fucked up isn't it? Your 24V is always on the output thanks to the parasitic diode. That could be what you want, 46V on the gate is a bit excessive and something is fucked with your PFet it's sort of in there the wrong way round. Your 'logic output' opto is referenced to 15V because you need more than? 20? to bias a Bjt base? I'm to. tired idk what you're doing.
>>2968129>common-drain MOSFET totem-poleWhy not use a half-bridge driver IC with shoot-through prevention like a normal person? I think you might be prone to exceeding the maximum Vgs on those FETs too, though it depends on the load.If you describe what you want the circuit to do, I can draw a better and simpler schematic for you, for free even.
>>2968141Ah ok ok I see the bjt is referenced to the same. Then that works. Power the opto with any 17,7 - 20,5V but why even insist on this very specific part?
>>2968115Ok define the point between aR and dR as Vx, and the point between R and R as Vy. Then the equations for each amplifier stage are:Vx = aR/R * (VB-Vi) +VB [unbalanced differential amplifier]Vy = R/dR * (VB-Vx) +VB [assuming it's another unbalanced differential amplifier]VO = -dR/R * Vy [normal inverting amplifier]VA = -bR/R * VO [another normal inverting amplifier]VB = VA [buffer, cR and Rf do nothing to an ideal op-amp, same for last R, literal schizo circuit]Then you feed them together:VO = -dR/R * (R/dR * (VB-Vx) +VB) = -dR/R * VB + Vx - VB = -dR/R * VB + (aR/R * (VB-Vi) +VB) - VB = -dR/R * VB + aR/R * (VB-Vi)VB = -bR/R * VOVO = -dR/R * (-bR/R * VO) + aR/R * ((-bR/R * VO)-Vi)Then rearrange that to get an equation for VO:VO = dRbR/R^2 * VO - bRaR/R^2 * VO - aR/R * ViVO*(1 - dRbR/R^2 + bRaR/R^2) = - aR/R * ViVO/Vi = - (aR/R)/(1 - dRbR/R^2 + bRaR/R^2) VO/Vi = 1/(-R/aR + dRbR/RaR - bR/R) now if aR is actually a*R and not just a variable name, you can simplify further:VO/Vi = 1/(-1/a + db/a - b)VO/Vi = a/(-1 + db - ab)This is why I prefer balanced differential amplifiers, you know, the ones with 4 resistors and the A*(v+ - v-) transfer function.
>>2968141New computers are going to come with RGB optoisolators since no one can afford RAM anymore>>2968147https://www.youtube.com/watch?v=gSFd_2oJgak
>>2968147Thanks, I'm studying the setup, still don't get why don't get Vo and ViI really appreciate your kindness and effort to write it down. This serve to me.
>>2968154I'm thinking about that now. As long as you work without rabdom access you can use any delay line, including fiberoptic, for memory.
>>2968147notation on the Rs is really weird.
>>2968206I meant like instead of fancy RGB RAM you're going to have a single 8 gig stick of DDR3 that you had to suck a dude off for, but the optoisolators in your power supply are going to cycle through funny colors so you can maintain some semblance of gamer cred
>>2968089didn't consider grounding out the shield cause I'm dumb, the copper wrapped around a wire for my shielded antenna, is grounded out and works well, unlike like the not grounded box...I'm visiting folks for new years, so deal with a pos phone screenshot.You can see how the tach is connected, capacitive/inductive loads on each end. the weird sinking signal it was generating, and now, after adding a 74hc14, the beautiful 3-5mS square wave that comes out (just did this last night, only bench testing yet with a drill)Heat ironically is the least of our concerns here, wire dangles on top of the motor, never really even seen a sense wire melt before lolAdding the 74hc14 last night gave me a DRASTICALLY cleaner signal to filter and work with, I keep forgetting to bring my scope to the track to see how that live signal looks...And the schematic I ripped from an Amazon tachometer with that same contactless principle as all the other commercial dataloggers use (at least from the user perspective)The bench top test doesn't generate emi like the running motor does, had this fucker at 3,000rpm and it read the SD card just fineIt's a really interesting phenomenon
>>2968232Wrong imageBut interesting That little noisy signal, and this "big sink" signal, came off the same circut, sometimes it was clean sometimes it looked like noise.Last night with the 74hc14 it was pure and clean every time I tested, which gives me drastically more hope
>>2967975Or use polling, and every ++counter % 16 == 0 do a display update or something.
>>2968233Schmidt trigger solved all problemsAll of the mAnd it now makes square waves
>>2968296Nice job, Anon. Record your schematic and share it with your buddies.
Ok time to get off my ass and do a simple project. With no feature-creep. Just an audio mixer for my desktop, tablet/phone, and ham radio, so just a summing amplifier and three volume knobs. Microphones can be done separately, no need for trrs. The only question is power, I'll likely be using TL072s so highish rails are probably going to be required, none of this 5v single rail rubbish. Ikinda want to use the +/-12V rails inside the computer, maybe I should use one of my expansion card slots with a DIN connector with those rails on it, plus fuses of course. Seems like the 24-pin connector to the mobo is the only one with -12V on it, and none of the little headers on my mobo will output the -12V rail from what I can see. It's an MSI B650-A Wifi.Should I use high-pass output caps too? I think a master-volume is probably unneccesary, but I could add it. Especially if I add clipping diodes to protect my ears, so I can vary the maximum output level between earphones.
How to I splice wires without them breaking off where the cable starts? I had put in shrinkwrap for strain relief but to no avail.
>>2968327use thicker heatshrink that actually prevents flexing, the double-layer glue-lined stuff usually works wellotherwise you can use something like silicone to provide a controlled amount of flex, or something harder to stop it flexing at all between the solder and where the insulation starts
>>2968327butt connector + crimp
>>2968327NASA splice
>>2968327Gob hot glue all over it
>>2968330> picrelImagine living in a world where you have to fake a picture of a solder joint with mspaint
>>2968330>NASA spliceactually called Western Union spliceit was documented 43 years before NASA even existed
https://www.youtube.com/watch?v=eeCY8pvld8ghow to learn optimal ground wiring like no bus wire across the pots, separation between ground points using the chassis built-in resistance at 3:30, safety ground having its own connection at 4:10
>>2968330lineman's splice
>>2968326Wait before that I’ll make a doohickey to prevent my appleshit charging above 80%, because I am not paying for another battery. A USB plug and socket with a MOSFET and current-sense resistor in between, when it goes below 0.5A I latch the FET off, requiring a button to enable again. I could use a microcontroller, to periodically enable the power for a short amount of time to see what the load is, but then my phone would vibrate each time. I could tell it to turn on again after 12 hours though. Maybe I can tell whether it’s plugged in via the USB data pins? I’ll give that a try.
>>2968304The "antenna" is a AAA positive terminal bladeYou could relocate the "sensor" by wrapping a wire in copper tape and grounding it out, works really nicely, kind of like a janky coaxial cableConnect the wire to the antenna spot, and the blade on the other end of the wire
>>2968546Thanks, Anon. Mighty White of ya.
>>2968567I got into racing and got pissed off karting dataloggers are $600-800Fuck you open sourced everythingHackTheTeack.netGithub links at the bottom for the timer library, the whole datalogger, and the data viewerOnly took me 2 and a half years of my free time but fuck you I did itNow adding a water temp/egt sensor is 2ez and the next step along with a bigger screen$60 ish in parts, all on amazon, no readme for the case files yet, been running it for 2 years now without the tachometer, works damned well
>>2968574That's the spirit. The time you spent gaining electronics knowledge will reduce the time spent on future projects as well as the cost. Sharing your project makes it harder for those niche producers to rip people off. I don't have a problem with people making money, but $800 for a tachometer is justification for a beheading.
Broke out my /ohm/ setup again to finally try to breadboard an analog PID controller so I could better understand it and hopefully learn something more along the way.As a process I want to have a small motor turn a pot, like in a demo I saw on yt, but I need to drive the motor. I got a small cheap brushed DC motor from my scrap pile that would do the job, and I thought about using a 555 to generate a PWM signal to control the speed (the idea was to use the 555 to make a sawtooth to feed in a comparator alongside the PID V_out to get the PWM, probably there are better or more elegant ways tho), but it then dawned on me that this wouldn't be able to drive the motor backwards even with a split power supply. Please /ohm/, I need pointers now.
>>2968631Use a stepper motor (for a rotary pot) or an actuator (for a slide pot).
>>2968633>actuatorWorm gear drive
>>2968633The mechanical implementation isn't a problem, since I'm a mech eng. I don't have steppers at hand and besides, this is one of those "learn something along the way" occasion.
>>2968640H-bridge motor driver.
>>2968642What I needed. Thank you anon.
About the SMPS. So I have decided to slap a 1P2S transformer for the 2 synchronous HS & LS switches and a random driver I have on hand for the gates.As to not stall this I will just measure the cross section of a few torroids I have on hand and characterize their frequency response as good as I can with a gen and scope.Does one simply add primary turns to a gate drive transformer until L is sufficient to keep I below the limit of the driver at the highest on period to keep the driver happy? Is it advisable to provide at least a pad for some R so one can limit I?Just slapping more turns would keep the flux in check so I guess I will end up doing it that way.
>>2968646No problem, Anon. Happy New Year.
>>2968655Oh I forgot: Wouldn't it make sense to capacitively couple the primary to GND, as to avoid driving the fucker into saturation? Especially since I see a future update where duty can be > .5 but I believe such duty cycles would mean I need to provide symmetric drive voltage to the driver so the transformer can be pulled the other way and reset.
>>2968141>>2968143Thanks, good point on the gate voltage. Maybe I'll try adding some zener Vgs limiters. I just used the stuff I found from our lab. Some random 60V N-/P-FETs and transistors.>>2968142Yes I really would like to use e.g. ucc21220 or something similar that would allow using only N-FETs.
>>2968666>only N-FETs literally me.But if you do this you have to jump through so much hoops and accept that sometimes, for various reasons, your design ends up worse.But N-Fet only drivers are simple. So why not cook one up?Also, regarding your schematic, I looked at it a bit longer and not sleepy rn:Why 24V flyback and boost to 36V ? Are you hoping this arrangement results in less ripple than a flyback and lin?Why drop 9 fucking volts on a lin?Why drop 24V to 15V only to boost it to 46V?Q4 and Q5 bases are at 31 V or 4 V with respect to Q5 C. How is that even supposed to be an output stage? What are N and P FET supposed to do when the gate voltage asymptotically approaches 46V or 19V?The longer I look the worse it gets. Monke on a typewriter.>only N-FETs>uses P-FETsplease don't tell me I wasted 5 minutes of my sleep on a schizophrenic.>picrelfor example is an isolated N-FET only driver that I made a while ago for a very specific purpose. Aka it's severely limited in some aspects, especially towards high frequency. This means if your gate drive needs only 9A or less and there are no other special requirements you'd almost always be better advised to just buy an integrated isolated gate driver.
>>2968631Going full-analogue, you’ll likely want to produce your PWM signal by feeding a triangle wave and a variable analogue voltage into a comparator. For bipolar PWM into an H-bridge motor driver (e.g. L298N), you’ll want two offset triangle waves, with a single analogue voltage getting compared to both. For example, one triangle wave from ~0V to +5V, and one from -5V to ~0V. Depending on the input topology of the motor driver you may need a logic gate or two between these two comparator outputs and the motor driver. To make a triangle wave, you can approximate it with an RC oscillator’s RC node, especially if the amplitude is low compared to the drive voltage.Otherwise, just feeding a single PWM signal into an H-bridge’s direction input might work. 555s make for a lousy voltage-controlled PWM though.Also brushed motors are fast, consider a brushed gearmotor.>>2968666What are you trying to do, swap between a 24V and a 36V supply for some sort of load? How many amps? Does it ever need to be able to shut off completely?
This has me confused all of a sudden:It's common knowledge you need X and Y capacitors to filter L to N or L/N to PE. So it's pretty common to put those as needed in your input filter stage and reference everything after the rectifier to it's negative terminal.Now I see PC PSUs and other galvanically isolated PSUs where the output is definately PE referenced.Would this not technically violate whatever regulation it is that mandates only class Y capacitors to PE?I mean sure, there is a whole series of elements, seperating L or N from PE, especially the transformer.But what then would the acceptable number of series components, that are galvanically isolating and not class Y caps, connecting L to PE, be? Or does the transformer get a rating like a class Y cap enabling you to PE reference the output?
>>2968678>two offset triangle wavesI don't see why they would want that, when using a DC motor and a L98N, which - AFAIK - behaves much like a L293D.>single PWM into direction inputI believe in case of the L293D it would at least not latch so it could work but most direction inputs are more like a 'reverse' input so the drive would be really noisy, always on but switching between directions at variable duty. But why not just feed a single signal into two of the comparators? Like I said: Why have two offset signals to begin with, we're not driving steppers, are we?>555s make lousy voltage-controlled PWM thoughKinda, depending on the variantn and circuit they do really well and you get real sharp edges on the output, stupidly high speed and good linear response over wide range.But I kinda agree, especially when something isn't for production, and just go with a 2 comparator package, one generates the triangle wave, the other compares it and provides feedback for the first.I suspect the entire PID functionality can be put into the 2nd comparators input, using only 2 OPs.>Picrel was a window comparator with hysteresis signalling a LM293D to drive a DC motor so not the same but: The LM293D module only needed 2 inputs for 2 directions and standstill.
>>2968673Sorry error:The first comparator provides its own feedback. That way you get a fixed frequency triangle oscillator. The second one then compares it.If you don't want to be as fast or have really fast OPs then you can use 3 OPs, where first provides a square wave, going through a pot to set the frequency to the second, that integrates it, feeds back to. the first and a third, that compares it to a volgage. So the first two create a really nice adjustable frequency triangle and the third the signal.Might want to replace no. 1 and 3 with a comparator since that is really all they do.
>>2968678>PWM and analogue voltageYup, my initial idea exactly.>bipolar PWMCool, didn't know you could do that, I'll look into it. My initial idea is to get the output voltage from the PID to direct the H-bridge, then just bias the triangle wave so it's zero centered and compare that to get the PWM to then feed into the H-bridge.>lousy>consider gearmotorYeah, I'm not going for high end here, and whatever the speed I can just reduce it afterwards, I'll print some pulleys, it will be fine. It just needs to work, it's just an educational project for me, baby steps basically. Once I'm satisfied I'll tear it down. Plan B is to use a resistor and a thermocouple in case controlling the motor turns out too hard, it's simpler although a slower process.Pic related my small "portable" setup.
>>2968684>I don't see why they would want thatPic related, so that you get PWM to go clockwise when the output of the PID loop is above 0V, and PWM to go counter-clockwise when the PID output is below 0V.>most direction inputs are more like a 'reverse' inputYes. I don't know how noisy it would actually be though. Motors are inductive loads with extra momentum, they smooth out high frequencies by their nature. By pulling hard up and down you'd be basically using it like the inductive element in a buck converter. The current waveform would be like a triangle-wave. With digital logic you could either pull hard high and low by PWMing the direction inputs, or instead PWMing the enable pin from the full-wave rectified PID output amplitude and using the sign of the PID output to switch direction. There's lots of ways of doing it. If you do let the motor go open-circuit instead of short-circuit/reverse, don't forget the freewheel diodes.>I suspect the entire PID functionality can be put into the 2nd comparators input, using only 2 OPsYou can do P, I, and D with a single op-amp, but you can't adjust the terms independantly with three potentiometers. So a quad op-amp (P, I, D, and a summing amp) in the more textbook way is probably what the anon wants to go for. Especially for a demonstration like this if he wants the circuitry to be easily understood.
>>2968697>my small "portable" setupi take it the baby penis is the load for the motor, and the project is some kind of prostate ticklera praiseworthy project suited to every red-blooded gooner on 4chan, and a brilliant candidate for a Kickstarter
Caracas is getting lit up by US forces right now. Portland Andy is restreaming several feeds. Will they announce the death of Maduro tomorrow?
>>2968697> portable setupI use this plano tacklebox for portable stuff and components I need for field service.You don’t need all that shit for portable setups…For example, you only need an IRFZ44N mosfet.Only a few zener diodes, you can customize the voltage after the fact with a divider.You need a lm358 op amp drawer. That’s it.Don’t keep your equipment in a wooden box, keep them on wire racks/shelves for air flow and get a mini desktop air filter if dust is a problem.Also, quit using the european symbol for resistors, it’s provably and objectively wrong, and learn the SI prefixes, such as kΩ because that’s standard.Get rid of that sewing mat — it’s a youtube meme.I don’t see any pink foam in those trays… use that for static sensitive components, or anything that fits nicely in there just as a precaution (e.g. cmos)
>>2968752>tacklebox for portable stuff and components i have fond memories of being on my knees for 10 minutes on a busy subway platform picking up and sorting components after my box fell and flew open after hitting my knee
>>2968752Believer it or not there's a reason behind everything:>it's portable in the sense that I don't have a dedicated bench at home>it's fully enclosed because it sits in a corner for extended periods of time and I want to keep the dust out>European symbol is for autistic reasons, as they're easier to draw consistently, but I use normal symbols otherwise>the mat is to protect the wooden table beneath>I don't have static sensitive components
I'm at the D part, I've tried a few functions and the derivative looks good every time, but I got all this ringing at the output whenever there's a discontinuity. Is this due to all the parasitic capacitance of the breadboard? Is there something I can do to filter this out without fucking up the output?When I get a real system hooked up it shouldn't be an issue anyway, as I expect the error to be a C1 function at the very least, but I'd like to know how to deal with this anyways.
>>2968735The triangle waves you've drawn - while it's arguably an offset - are 180 out of phase. I really don't see the point of generating two offset triangles. Many ways to skin this cat but all the ways I came up with seem less tricky to get right than this.>buck converter>triangle waveWell no shit, provided you have freewheeling. You still end up with the issue where you're running a current for no reason, even if you get the motor to act like a mechanical panel instrument.>sign of the PID outputNo shit again, that's probably what anyone would do. Unless they have very specific reasons to jump through extra hoops, so why 2 offset triangles?>Lot's of ways of doing it.I agree. Some are just more accessible and probably provide better performance for general applications. Especially since there is less that must be controlled.>Freewheel diodesWell there's the buck thing again. I don't think those drivers shoot through, they sure have a Tdead on the output drivers. I guess the condition that needs freeqheeling diodes is best desvribed as 'high Z' output.>P I and D in a single OPThat sounds about right, so you just make an integrator with pot thats also a differentiator with pot and has another R pot for the proportional term? Guess that works.>>2968742What would the error input to the loop be?A microphone maybe?>>2968752You carry 'components' - other than maybe rectifiers, caps and installation hardware, for 'field service'? Precisely you carry MOSFETs, OPs and Zeners? What do you work as it sounds odd and optimistic. >>2968764>I don't have sensitive componentsYou do. It just happens very very rarely with common, generally robust cocomponents. But when it happens it sucks ass. I don't do ESD either tho. >>2968776You're expecting an OP to just not tingle shit and feed back whatever the result is? I mean thats pretty much the thing an OP does. Got to fix your amplifier.Also first rule of electronics:Cant make oscillators, unless not trying.
>>2968776You need to limit the high-frequency gain somehow, like a small value resistor in series with your capacitor. Just like with an integrator, there are some generally accepted extra components to make a differentiator behave less ideal but work more reliably.Post schematic.>>2968808> I really don't see the point of generating two offset trianglesYeah, easiest way is probably one 5V to -5V triangle oscillator, with a pair of resistor bias circuits to push it higher and lower. It just looks nicer drawn that way.>provided you have freewheelingDon’t need it if the motor always has a low-impedance connection to one rail or the other (well as you say later there is dead-time). It’s synchronous buck. Or a class-D amp.>running a current for no reasonIt’s got current running regardless, you can’t really stop current running through a motor each PWM cycle because it’s inductive.> so why 2 offset trianglesProbably less components. You can passively create the two triangles as I described above, but getting a PWM duty-cycle proportional to the amplitude of a bipolar analogue feedback voltage seems like it will take at least an op-amp or two. But I might be overlooking something obvious.
>>2968826>Just like with an integrator, there are some generally accepted extra components to make a differentiator behave less ideal but work more reliably.Yeah, not knowing these things is why I miss not having a formal education nor work experience in electronics, but then again it's a hobby, it's just play for me.>Post schematic.It's more or less pic related, it's off this page:https://control.com/textbook/closed-loop-control/analog-electronic-pid-controllers/
>>2968826>you can't really stop currentthats besides the point. If the error is 0 for some time you want your motors movement to stop. Sure, you can buzz it back and forth at a frequency where the movements amplitude is below your allowable tolerance or the motors mechanical hysteresis swallows it entirely. That is runbing current through it for no reason. Irrelevant but ugly, if the above conditions apply and, your energy budget is irrelevant. Bad if the conditions above or energy budget are not met.>Low impedance connection to one rail or the other. Erm no? Firstly it matters which rail, in respect to the direction current was last induced. Secondly if the other rail is high Z and has no diode (parasitic like in MOSFET or discrete) there will be a voltage spike regardless,albeit it may be negative. Lastly the resulting harmonic reset of the inductance is ugly and brings about uncertainty for the next on cycle.>2 trianglesI might draw something up tonight after other work becomes exhausting. After all the effort posting about it might aswell be put into providing examples.>>2968853Again: What is the error input? A miv strapped to the users face?
>>2968860>what is the error inputIn my current circuit PV is given by an arbitrary signal generator, so I could feed functions and see the differentiator and integrator doing their thing: ramps, trapeze, sine, log, all have pretty easy to recognize derivatives and integrals. . This will change once I hook up a (buffered) potentiometer to give me my PV. The setpoint is just a buffered trimpot.
>>2968853Yeah, small resistor (1-100Ω probably, try 10Ω) in series with Cder, and a small capacitor (10p-10nF probably, try 100pF) in parallel with Rder.A large value resistor in parallel with Cint would be good too, but it should be at least 100 times larger than Rint, so unless you change the size of Cint and Rint to be lower impedance with the same RC constant, you’ll have to give it a miss.>>2968860>That is runbing current through it for no reasonIt’s not really putting much current at all, at least with a decent PWM frequency. You give the motor +12V and the current starts ramping up, but it gets to like +2mA before you flip the voltage to -12V and the ramp goes the other way to maybe -2mA, and back and forth like that. Not like you’re dumping significant heat into your driver or your motor when in such an idle state. I am assuming it acts like an inductive load with series resistance, that isn’t entirely true, but I think it holds in the motionless case.
>>2968878>not significantLike I said: Depends. It may be insignificant if your power budget allows for it. It depends on motor inductance too. But I am still convinced it is ugly in any case, when the output is zero you want your motor to be either free or 'locked'. And you still get the resonant reset. Unless you now want to add caps to clamp it to something reasonable - which may fuck your frequency response - it may damage things. And lastly it is still a non issue since the whole premise is away of doing things that might only be applicable if you have very specific reasons for doing so.Non issue.Ugly.More trouble.
>>2968680I want to add another question to that:Is it accptable to have a PE copper pour in a production design and is there regulation regarding what then the isolation distance would have to be?
>>2968899Designing for electrical safety classification is probably not something you'll get expertise for here. I don't know if there's a specific distance requirement (my guess is probably not), rather your appliance must pass earth leakage (and earth resistance to chassis if applicable) tests. Y caps can't be too large or else the leakage current will be too high.Also some class-2 appliances (appliances without a protective earth connection) will connect the live and neutral lines to the LV side via Y caps, even for USB chargers where the metal USB bit can be touched by a human hand. This nominally puts it at a Vac/2 potential, but it's high impedance so it doesn't matter.
Who's the fuckin genius who came up with this garbage?
>>2968972More than one genius came up with that. lmao
>>2968972burgers
>>2968903I just find it odd, you know. They dont say 'your cap must pass this and that test, not burn and not short' the regulations say you need a specific cap. But when it comes to the board n stuff, you can probably mount that cap on a super burny phenolic card, with a PE copper pour on the bottom - which is a cap besides other things - and be fine regulations wise.>AnywaysAnyways, started to layout the SMPS. First iteration. Very unhappy. Why the EMI section be sticking out? Found chinese non descript EMI modules, don't need to pass EMI anyways just want to play nice and Chang makes that shit For less than I can get the components.Is the output ground really convoluted? Yes. Yes it is. But then again the output traces gt some copper mass to them.
>>2969014How much isolation distance you guys figure one can realistically get from a TO-220 package by staggering Pin 2 / Drain / Collector without being ridiculous?
>>2969032> isolationIt’s not about isolation, it’s about the trace and pad layout on the PCB. Usually seen in > 3 pin devices like amplifiers where they want beefy copper.Ultimately, you can just remote mount the device to the case and use shielded wires that run to the PCB if you wanted to.
>>2969032I've decided I will bend it out by 2 rows. That should, assuming circular pads D, give an isolation distance of 5sqrt(2) - D, while source and gate remain isolated by 5-D mm. So in high side applications that's the weak point. But meh.
>>2969033>It's not about isolation. >Presents TO220-5 I have not claimed staggered legs like you find on many dense power packages were solely about isolation.Often enough it do be that way tho. Check HV stuff with TO220-3 or 247-3 components. You'll very often find staggered drain, perhaps with shrink tube on it, while the layout isn't tight at all.GEEEE I wonder why that be.
>>2968678>swap between a 24V and a 36V supply Pretty much this. Load is about 0.3 A(max) but connected via long cables. No need to shut off completely (at least 24V should be supplied to the load at any time).>>2968673Thx for your good points:>Why 24V flyback and boost to 36V Just figured it would be easier to find a suitable transformer for 24->24 than 24->36. >Why drop 9 fucking volts on a lin>Why drop 24V to 15V only to boost it to 46VYes it's pig disgusting. U2 supplies U4 because the ap3012 booster IC we had at our lab has 16V max. Vin. Figured I _might_ be able to get away with it as the currents would be rather low for supplying the drive rails. Maybe I'll check another part.>What are N and P FET supposed to do when the gate voltage asymptotically approaches 46V or 19V?When Vg= 46V the N-FET (upper) should supply 36V to the loadWhen Vg=19V the P-FET (lower) should supply 24V to the load Are the DIP-8 parts on your board optoisolators or something else? You use dual 2N7000s for driving each N-FET, interesting.
>>2969040Well that was somewhat short sighted
I hate the browser nuking comments...>>2969048>Dual 2N700 for FET drivingAnd you think I'll just pull them down again via resistor, large enough to limit current to <1A or whatever dual 2N7000s do? And then wait for VGS to fall? And the transformer to reset through body diodes somehow? Also I'd need yet another voltage for that stage gate, or shenahigans with a charge pump, or live with the exponential approach of VGS to VGS_max which then also sadly has to be <20V, as the 2N700 have to allow VGS_max + Vth on the gate?Nope those are mundane integrated gate drivers. Got tons, few cents a pop, the output stage is usually ~2 or 4 A or so. Most of the stuff on the board is a 'bit odd choice' since it's stuff I have laying around and fits the bill. The rectumfryer is D20XB80 for example. >Swap between two suppliesJust use one FF, Q goes to one MOSFET, Qbar to the other, add RGate and parallel Diode to achieve slow on fast off. If you insist on high side N_MOSFETs provision ONE DCDC, to provide ONE gate voltage. Your current's are low, so no need for high VGS, 46V GND referenced or 10V referenced to 36V, is 22V on the 24V one. This technically violates VGS_max of most datasheets, so check your transconductance curve if you can go even lower on the 36V one or clamp with a zener that is appropiately sized for RGate and the voltage.Regarding datasheet VGS: Those usually are very conservative, some datasheets will explicitly allow static 30V VGS, it's not unheard of MOSFETs and IGBTs to handle dynamic 80V on the gate, especially older ones.
>>2969048>Pretty much this>0.3AIf voltage dropout doesn’t matter, a diode from +24V to the output, and a PNP BJT (or equivalent sziklai pair) from +36V to the output, switched directly by a phototransistor optocoupler if isolation is needed.If voltage dropout does matter, then a PFET from +36V to the output, and an NFET from 24V to the output using the 36V as its maximum gate voltage. Both with 12-15V zener diodes limiting the gate to source voltage.An even smarter method would just be to change the flyback converter’s feedback, such that it outputs either 36V or 24V depending on whether a logic-level FET switches on and adds another resistor to the feedback loop.>>2969050That’s what soldermask is for.
>>2969067>SoldermaskSure the recommendations for isolation distances make a distinctions between soldermask and not.But I disagree with the statement that it was 'what soldermask is for', the primary purpose of which is to prevent wetting with solder, especially in automated soldering and providing corrosion resistance to bare copper.Either way I found a better layout where the high amp component now moved north of the FET, this allows me to run a much narrower trace there and I'll next see if I can make space between FETs to get all those signals through.>FiltersSo I am iterating over the output filter. Now commonly, with filters, the order of elements matters and noise must not be allowed to get past elements. It's why for example bypass / decoupling caps can be made entirely ineffective, by for example placing them after the pin, as seen from VCC. Is this much of an issue with things like SMPS output filters?Once I add copper pours to keep the real part of the impedance down I will inevitable get into situations where filter elements end up being connected on one plane, so that one could not reverse the exact schematic from it. The positive of the electrolytic and the preceeding film caps for example. I guess it'll be fine and the smaller caps will still provide the far lowest impedance path for noise...
>>2969070Got the thing alot smaller but maybe there's room still.Since I'll buy the EMI filter anyways I will bodge that on. Without that section sticking out its 180mm across.I'm wondering if I should change THT current shunts and a Zener for 1206, since it might save more space and I'm sitting on a bunch, up to 1,5W.Meanwhile I also wonder if replacing some SMDs that connect logic to power (sensing dividers and shunts, HS switch source) for THT ones will provide the isolation I should have there. Logic will be isolated from output.
>>2968330Wouldn't this still cause strain at the point where the solder ends? Using stranded wire btw, not single core.
>>2969126It's fine in tension. The less flexible the wires are, the less of an issue the stress concentration at the edges of the solder joint will be when bent side-to-side. If it's solid-core, no problem. If it's 60s multistranded, that's just like 7 cores total, probably still fine if it's not like 28awg. But if it's modern fine stranded wire, that's where you get your problems even with 14awg. Thick heat-shrink makes for a decent sideways strain reliever, while the solder holds better in tension than a butt crimp.
>>2969097Gonna carve any isolation slots? Check creepage guidelines for various industries. I'd put a slot between the opto at least. And you're probably familiar with the concept of a PCB spark-gap too, in addition to a MOV.
>>2969128>SlitsI'm comprmising my own work in a weird way here. I fist started out designing for at home production capabilities. Meaning no slits. Soon it became more and more unrealistic to fab at home. So now it's half baked.>creepageI am mostly concerned about through the air but it's a fair point, also creepage can quickly spiral into more.>Industries I'm designing this for personal home use but you're right and I don't want to look like a doofus. Also there might be dust and there will be water spray. Now that being said:The insulation distance I was talking about and worrying about would be the ones that end up frying the logic, or something, on the primary side.Thy isolation distance is limited by the opto and sits at 5mm. Something must go terribly terribly wrong elsewhere first for it to become an issue but 5mm is 5mm.The runner up is a spot with 8mm. And it's a way more realistic spot. The PFC output could jump there through the heatsink or various <5mm gaps on the board and get to the output.But c'mon: 8mm ? I doubt it. You're more likely to kill yourself holding either output terminal. I guess at 90V it's kinda possible to goof.>picrelPicrel is how it currently sits. I need something I can send out for production. Since the EMI arrives as a module I simply chopped it. That way the board space is way less.
>>2969135fuck
>>2969135Yeah it’s not flashing over that 5mm mark, at least unless you have a stray EDM inductive voltage spike, which this topology shouldn’t have if I understand. But with dust and such around, creepage is definitely a concern, unless you feel like glooping the transistors in resin. To make slots you can just drill subsequent holes and file them together. Or even make two seperate boards and run wires between the two. No reason you have to stick to the optocoupler’s default footprint. I was designing an ultra-low common-mode leakage power supply where the primary and secondary were on opposite ends of a toroid and with a light-pipe for optical feedback isolation. Or maybe I used a feedback winding and no opto at all.>>2969136You’re either using KiCAD on a low resolution screen or you have the same issue of it rendering at quarter resolution as I do. You do use Mitxela’s curvy/melted traces plugin, right?
>>2969141>inductive spikeI assume every active component in that circuit will break down long before an inductive spike reaches a voltage sufficient to jump the 5mm when clean and standard ambient conditions. I also believe all of the inductances in the circuit will have a hard time reaching the current needed to give the associated stray capacitance that hard of a stuffing.>slot manuallyYeah I usually wire the jewelers saw through. Usually when I break something, throw a fit, slap it on a bare piece of PCB and bodge it in instead.>picrelresulted from a particularly temperamental reaction. The bodge worked well.>>2969141KiCADYes always.>Low residk its 3430x1440 on that screen if that helps>pluginno idea man might aswell ask your ex about loyality.>No need to stick to the footprint. There's only so and so much you can stretch the legs. Even then, the bare pins do not promise the same isolation as the plastic package does and there's no feasible way of modifying them accordingly, just get a wider opto when needed. I'll call it a non issue tho. This is more than enough for rectified mains, especially on stuff someone builds and then uses themselves.
So I was happy enough to add the pours and while I didn't think I'd do it at first I ended up adding some via stitching, it just doesn't feel right.I like the output stuff much better already, none of the traces is thinner than 5mm at any point.I guess it's time for Chang to do his thang.I guess
>>2969043Case in point. Consider this primary MOSFET in a flyback. The designer would have no issues just extending the track for the drain 5 more mm. Nothing in the way there. Meanwhile bending the pin is just one extra assembly step, even if they come prebent, someone did the bending and someone pays for the extra step. Why would they?Isolation.
I'm experimenting with interleaving the output of multiple cheap DACs to get high(er) throughput than any one alone.Necessarily, the outputs will need to be multiplexed. In terms of output linearity and frequency performance, what would you more seasoned engineers choose:>raw-dog it using plain transistors as basic analog switchesor>use a dedicated analog switch ICor>something else/I'm baldI'm leaning towards the anal switch just because it already has linearity compensation and likely has better performance from using integrated electronics. Honestly I'm not sure just how much the frequency response of the analog switch matters, given that the output of the DAC (for a given output period) is technically DC. I think the rise/settling time would be the primary concern.
>>2969126The only goal is for the joint to be as strong as the wire itself. The western union splice is stronger than the wire it joins per NASA's testing
>>2969172this seems very silly compared to just getting a better DAC. is the price jump really that high?
>>2969174I could, but this is mostly for my own edification. I'm autistic and I want to see how far I can push 48/50MHz DACs and ADCs, for fun. This is usually the highest speed ADC/DACs you can get that work over SPI. It's also a good learning project for me for my FPGA development skills. I've brought it up a few times in past threads.
>>2969174>>2969176I also just like coming up with stupid solutions for nonexistent problems. You learn a lot that way.
>>2969176>>2969177fair enoughone stupid thing you could do is just tie all the DACs together via resistors, then compensate for the resulting averaging in software
>>2969179This is the solution.A lot of these DACs are fronted with simple transistor amplifiers that can operate in the MHz with a high impedance input. In osciloscopes.Switching seems like an absolute nightmare with settling time, noise, speed issues, etc.
>>2969206>>2969179>>2969177>>2969176Idk you guys, I only have actual in depth experience with delta sigma DACs from work, other than that every DAC I used for an application just worked for me after I chose it accordingly.The plan doesn't sound convincing. I don't expect Anon to achieve his goal of increasing throughput like that. Maybe draw a concept diagram for all outputs and inputs. If it's promising simulate the thing.
Pls can you guys review my shit? I'll start with the control section. Some values TBD but the math on most of it is done and the math is never the final thing anyways.
>>2969234Skipping the input filter as I'm using a module.Note: I_AC is for voltage sensing. It is only that the error amplifier takes current as it's input.
>>2969236Yes, I want that resistor directly at the gate.
>>2969237C4 will be any >400V PP cap.
>>2969238It's not going to be FR207. I forgot what I chose to go there but non-issue.
>>2969239Note the ground for the actual output being isolated.The bias winding is physically on the primary side.
>>2969240It's all PP caps and one electrolytic. PE on the output goes to case, Just in cas I want to grab it again for whatever reason.
>>2969241Most of the values here will hve to be determined. I'll just slap the poles of the filters somewhere where it makes sense. R across opto is in case the bias current for TL431 and opto dont agree.
>>2969242It's a 90V 5A single output SMPS, two switch forward.The datasheet mentions 'current mode' and 'voltage mode' - I assume they mean CC or CV operation, where in CC the primary current is the error input and in CV the output voltage is. Right? Can someone specifically look over the bias transformer and startup resistor, I am not sure if that is the best way to do it. It's a resistor that provides Vcc from the rectified stage (I believe taking it from the PFC has pros and cons) and a zener to clamp it. The bias winding conducts in half cycles, I imagine once it provides the zener current the startup resistor current should drop significantly.I am considering a few things: A pot to adjust current limit.A pot to adjust output voltage.A switch that fools the current limiter.A relay shunted by a precharge resistor and a few comparators that comes up when everything else is okay.A diode across the PFC inductor, so the cap charges to 325V before the PFC comes up but without potentially leaving residual current in the inductor.I bodged an LED in after the fact. it's on Vcc. Maybe there's better options, maybe I should just use one for rectifier out, one for PFC out, one for the startup current, Vcc, soft start capacitor, and output?Placed a pad in case I want to capacitively couple the GDT, do I want to?All diodes and switches but the output rectifier will share an heatsink, most are insulated packages.
>>2969236>calling the output of a mains rectifier GNDIt irks me.>>2969239Shouldn’t D4 go to above the current sense resistor, not below it? Just seems kinda sensible, though I doubt it matters.>>2969243>The datasheet mentions 'current mode' and 'voltage mode' - I assume they mean CC or CV operation, where in CC the primary current is the error input and in CV the output voltage is. Right?Yes, it’s about the feedback method being voltage or current. Current-mode seems to be better except for requiring compensation networks at high duty-cycles.Instead of pots, I’d just stick with THT resistor footprints that are tolerably easy to swap out. If in doubt, add extra footprints to your layout.
>>2969264>GNDIt is not called GND, the rectifier output is offset by <2V.But it hurts that you're saying this, since I removed all the ***_N nets and added GNDs just for you guys. What would you call it? 'GND', which is on the other side of the current shunt. >D4If I put it on the node I_pwm, it would mean that the xformer reset goes through R26, while it might not be much and realistically not matter, it could limit duty cycle, put extra stress on R26 and damage the IC, as the IC expects positive input and is probably rated for Vcc-0.3 or something. >current mode voltage modeI suspected so. I find it well confusing. Some application note presented two circuits, one labelled 'current mode' and the other 'voltage mode'. They were almost entirely the same. Meanwhile I am thinking:Wether it's in current or voltage mode merely depends on how you set the R26 in relation to the output voltage divider. This determines which, for a given load, becomes limiting first. Sure the extreme cases of low and high load impedance will (should) always result in the 'other' limit kicking in.>current mode seems to be betterI somehow feel like we both aren't on the same page, I'm probably missing something.The way I get it is that in current mode during normal operation (design load) the error input for current is always the one limiting the duty cycle, meanwhile the error inout for voltage is wide open. That gives a constant current source like for LED drivers. The voltage feedback will only catch it and prevent destruction if your load went open. On the other hand in voltage mode I'd expect reversed roles: The I_PWM kicks in at 1V or so, I believe. So 10A here, and only saves the device if something goes wrong (output short). But during normal operation the opto feedback governs duty cycle.-> The only change youd need is the current shunt or am extra divider and / or the output divider / reference voltage. The only optional change I see... (cont.)
>>2969270... is one where eunning in current mode allows you to maybe simplify the output voltage feedback. No linearity needed. Output too high - > jam the signal to the high rail and the thing will limit / wait for the output cap to self discharge or a load to be connected.
>>2969270Current-mode switching means a cycle-by-cycle current feedback. Switch turns on, current through inductor ramps up, switch turns off once this current gets too high, repeat. The current setpoint is a function of the error voltage, I think, but by setting a maximum value to the instantaneous switching current you also (roughly) limit output current like a CC supply. Instead of a conventional oscillator, current-mode converters often just have a dead-time before turning on again. A voltage-mode converter instead just has a regular constant-frequency PWM oscillator where the duty-cycle is changed by output voltage feedback directly, which is slower to react to dynamic loads since it’s not synchronously reacting each cycle, but it’s better for edge-cases.
>>2969295https://www.onsemi.com/download/data-sheet/pdf/fan4800cu-d.pdfHere's the datasheet, without the discussion is no use.It clearly shows, when I_LIMIT exceeds 1V it cuts the output. This makes sense, regardless of the mode of operation it is at least used to protect the hardware. There is a simple RC filter but the time constant is 0.2 uS in my current schematic. Switching frequency is only 100 kHz. So the filter is only for noise and shows instantanous current.And I finally got how it works by looking closely again:In voltage mode V_REF and R_RAMP generate the ramp using C_RAMP. This ramp is compared to get the output signal and that's it. This works if you assume that primary current was precisely a function of duty cycle.Now in current mode it is not V_REF that generates the ramp but it is V_I_LIMIT or the current in the primary. Meaning that if less current than expected was seen on the primary in any given cycle the ramp will have a flatter slope and the comparator will cut the output later. That way the cycle is extended to compensate for the lack of current in the cycle.Odd: Remove the load and less current is seen on the primary in the following cycle. This would mean we now get a longer cycle despite needing a shorter one, until the output voltage rises to the point where FBPWM says 'nuh-uh'.I guess you sort of get a faster response at the expense of some overshoot or something.Thanks for the help. I'll leave it as is, where V_REF feeds the ramp.
>>2969297I should add I have done some more guesstimation and calculation on the startup behaviour. But there's lot's of 'WHEN?'s that remain dubious.I gave my soft start capacitor a 15 mS timing. Meanwhile the PFC output capacitor completes 5 tau in 1.3 uS if the EMI filter lets it. Sounds like plenty of room.It's only that, IIRC, the step response of an LC circuit reaches twice the step voltage. And the diode following the PFC inductor should lock that in the PFC output cap.Ofc, this again is assuming steady 325V after the rectifier.Not sure if the funny EMI board has inrush current limiting. Guess I'll slap a limiter ... and I'll first start it up in series with a lamp or something anyways.
>>2969173i thought the issue was the strain like at 13:30 instead of the wire bending in a wide radius the strain is concentrated because of the rigid jointhttps://www.youtube.com/watch?v=kRQYUim554M
>>2969430the splice is for electrical cables, not audio cables (probably for that reason.) the NASA testing was for performance under load/tensile strength and not for flexibility
>>2969270>What would you call it?>>2969264
Really, none of you was gonna say something about Q1?
>>2969511Fuck this all 3 FETs for PFC and PWM were GSD instead of GDS all the time.
>>2969511The bent legs are kinda silly I guess if there's already ground between the legs, but it still makes for slightly better creepage resistance. I wasn't looking that closely at your layout, now that I am I'd suggest that you delete those superfluous ground traces and let the thermal reliefs carry the current.Also MELT THAT SHIT
>>2969529No. I get furious when curvy / round tracea for no reason, especially when it's cargo cultish as in 'it just better HF performance, somehoooow' like no, like always: It depends. What do you want from your trace? If you eanted it to behave ideal and if anything maybe like a LP, you could do the math for double 45 bends that act as a LP. And so on.No curves here. Love my good old 45s.>Thermal viasThere are none. Which do you mean? Under heatsink? They're primarily to ground and mount heatsink.>Better creepageYes>Ground betweenYes, if something happens it's most likely to blow the fuse. But I'm pretty chill now. It's only 400V maybe it's time I cut the crap about isolstion. I redid the entire section earlier and a few other 'improvements'. Also ordered mouser, some stuff on backorder so chang still has plenty of time to get an order for the boards.Anyone want to add artwork?
>>2969529Oh now I get you. You mean the spokes in thermal relief.No. Countless of ideas about 'good practice' but this is how I learned to do it and the reasoning is sensible:-You know you eliminated a connection from ratsnest, with or without fills.But importantly:-You forced yourseld to consider and evaluate possible return current paths.You can make a board where a global GND fill takes care of all your connections, but if you have traces running left and right top and bottom your current path will be long, irregular, squeeze through tight spots, share copper with various sources of noise etc.Better to explicitly define it. You can always use a 10 mil trace and leave in a 45 so it gets absorbed in a thermal reliefs spoke anyways if you want.
>>2969242Ieeeek...I just realised this wont work. Another lucky save I guess... TL431 widthstands 37V max... says the datasheet. Here it would have to put up with 90.I think this can be fixed by making another divider, for the cathode and then populating R6, to bypass enough current for the divider.
>>2969582Maybe it's easiest to just slap a resistor common to the reference divider and the anode to ground, which will drop the remaining 60V or more.
thoughts on this layout? I have an LTC6655 as a voltage reference for an ADC (LTC2380-24). the 6655's load regulation is 3 ppm/mA and the VOUT_S leg on the LTC6555 and the reference input on the 2380 both sink 2 mA. without doing anything that would throw the reference off by 4 mA * 3 ppm/mA * 4.096 V ≃ 48 µV. the MAX4238 in there has a Vos of 2 µV, which should be a huge benefit. or is it?the second thing is the wide copper area around pins 7-8 on the 2380 (top right corner). I have in mind to glob a bunch of solder on there. I reckon a 0.5 mm layer should be good enough. the fill is 1.5 mm wide and about 2 mm long, so the voltage drop should only be around 90 nVthe caps are 47 µF X7R
>>2969621schematic
>>2969623oh and the VOCM divider would draw another 2 mA. it getting its own amp is maybe excessive but you never know. this is high precision stuff
>>2969621I heard you should usually tie the unused pins to something lest they become antennas, so etimes it buried in the notes for the devices.
>>2969637that can be an issue yes. luckily this is all running at rather sedate speeds
>>2969554>it just better HF performance, somehoooowOh not at all, the guy who made that plugin is certainly of the opinion that it doesn't matter for shit, and I believe the same. Maybe if you're doing GHz circuitry you need to care about it, but if you're just melting traces instead of routing proper transmission lines at those frequencies you'll have a tough time getting anything to work. No that plugin is purely aesthetic.>artworkGot anything in mind? How about a classic LeMans car?>>2969555I guess it's still good practice, but it just looks kinda goofy. If you make it in a 45 even without thinning it out, it should be fine.>>2969582Ah yeah, TL431s can be a pain like that. I wish someone made a higher voltage version. You can always put a zener in parallel with them, with the right R5 and R6 value that wouldn't be an issue, besides maybe some heat production. Or you could use a proper op-amp with its own little zener supply regulator.>>2969621>>2969623Shouldn't pin 2 of the MAX4238s be connected to GNDA? Or have you just not routed that yet? A full plane of GNDA beneath this analogue sections is probably a good idea.
>>2969647>Shouldn't pin 2 of the MAX4238s be connected to GNDA? Or have you just not routed that yet?yeah I hadn't gotten there yet>A full plane of GNDA beneath this analogue sections is probably a good idea.I'm trying to keep it star grounded as much as possible. a ground pour would prevent that
>>2969652Pretty sure last thread someone was heavily implying that star grounding was inferior to other methods. I found an article that may or may not have been the one he linked, the archive isn't loading for me:https://www.allpcb.com/allelectrohub/analog-pcb-grounding-the-star-ground-myth-debunkedI'm on the fence about it, in your case I suspect it might be a good idea. I'd consider making a ground pour and just carving thin keepout zones to make it into a very thick-limbed star where needed. If EMI and loop area isn't an issue, I wouldn't be too concerned, but with regards to the digital side of your DAC I'd definitely want to ensure proper grounding and low loop area for the fast signal traces.What's the poject?
>>2969654it's a thing for reading Pt100 sensors with high accuracy, <1ppm/sqrt(Month). I've already proved that an earlier design works, so I'm implementing a bunch of improvements on this second version>https://www.allpcb.com/allelectrohub/analog-pcb-grounding-the-star-ground-myth-debunkedinterestingI think the most important thing is to keep the digital signals away from the analog ground plane (if I decide on a ground plane)also the analog and digital digital sides have completely separate supplies, with optocouplers to exchange signals. I could in principle run the analog side off of some batteries. the entire thing is double ovenized - one oven for the analog side, and another oven for the entire device
>>2969655>also the analog and digital digital sides have completely separate suppliesI guess the DAC will be the only digital stuff going on in the analog side? Looks like the DAC has a seperate digital and analog VDD pin, but a single ground pin. I'd want to put a choke on the VDD pin, if not more filtration, instead of routing it directly to the +5VA that you feed to the reference and op-amps. I also see you haven't used chopper amps, which might give you less offset voltage drift, but I'm not exactly an expert. Maybe a 2k resistor from U33's output to GND would ensure it's better matched to U26.>double ovenisedOh man you're really pulling out all the stops. Watch out for capacitor microphonics too. Are you using an ultra-low common-mode noise power supply for the analog side? A (resonant) DC-to-DC converter with coaxial) windings well spaced out on a toroid might be the way for that.>I could in principle run the analog side off of some batteriesHaving battery backup so it remains heated during power-cuts might be worth doing one way or another, if you want to stop the PPMs from leaking out.
>>2969657I'm feeding the ADC as per the datasheet. RC low-pass filters essentially. I'm not worried about high frequency noise because that gets averaged out. sufficiently fast drift on the reference is one thing I'm trying to watch out for. we suspect what's limiting us is actually drift inherent in the sensor. the way to fix that is probably to reduce the drive current>Watch out for capacitor microphonics toonah too high frequency to matter>Having battery backup so it remains heated during power-cuts might be worth doing one way or another, if you want to stop the PPMs from leaking out.so far power outages haven't been an issue. what has been an issue is the human caretaker deciding to increase the lab temperature from 25°C to 30°C, which caused the oven inside to fail to maintain a nice 40°C and all those PPMs to go on strike. I'll be sure to increase the oven temperature to 50°C or so this time aroundone thing I'd like to try is put a peltier in there instead. that'll have to be saved for version 3
>>2969657oh wait I'm confusing the PGA with the ADC. yeah maybe I should add a little bit more filtering. but also more parts is more things to behave strangely. like giving off popcorn noise
>>2969661Yeah peltier ovenising seems like a nice and elegant method, just takes a bidirectional DC-DC converter and you can actively cool or heat your oven. This also allows you to pick a lower temperature, which is probably better for noise? Hey, maybe to set the oven temperature as accurately as possible, you need some sort of precision temperature sensor, like a PT100. Oh but now we need some way of precisely digitising the temperature of that PT100 without it drifting...
Has anyone seen these anywhere other than Jaycar? They’re 4-pin signal connectors that lock together with Anderson PP15/30/45 connectors. I’m considering using them for a mixed-signal gender less connector for high-power DC power negotiation, but I’d prefer if I were able to buy the connectors from more than one vendor, considering they’re a discontinued part at Jaycar.
>>2969647>ArtworkI currently have freddy killowatt doing the stabbin on the output>looks goofyI've made a few designs for radio comms and opened alot of the big ass dollar equipment competitors sold. Thw things you see. But: It's informed decisions that do work. No one there gives a fuck about how it looks if it means a potentially better noise figure, or message error rate. But if the MER sucks everyone gives a fuck.>TL431I thought more about it, came up with some additional options and also looked at what others doing:The divider options have their own problems, you can make them better at the expense of efficiency (so additional optocoupler mA don't matter).Zener on top:Seems to be popular. I think you still will have to check which leaks more, Z or TL, else the whole idea of the zener doing something might go out of the window. Might help the leaking with a shunt.Series Z to GND. Looks like no one does that. Idk why everyone wants their TL anode at GND? Same with the dividers. No one does that.I was also thinking of cascoding a FET. If I give the FET gate < V_TL_max + Vth, it pinches the channel before the TL dies. And leaks more or less again.>Z in parallelWas unironically my first idea. And I immediately came to hate it. That's a can of worms I wont dig into unless I have to. Everything is suddenly fucked.>CarNo this SMPS isn't to support LGBT.>>2969623Always consider if you can delay something like shutdown_bar from Vcc so as to avoid abnormal operation. Give them individual time constants if you imagine an order of startup. >LTC2380it's clearly polluting your GND. Didn't check datasheet but it's obvious:It has digital. So it will need to be on GND / DGND at least somewhere. It's mixed signal. So it needs to be on both and the DS probably gives a layout example for seperation line. Right side of symbol is probably all DGND. >PGA280Same issue. DGND->GNDA.(...)
>>2969725(...) Filters poles of the filters on VREF and other V lines appear to be placed suspiciously low. Idk for sure either at what they pull back up but maybe do your math and check again.>>2969621I think it's okay but my professor and boss would've given me flak for the traces not fanning out early at that one SOIC, for the Vout copper fill, for traces that abruptly change dimensions and thus impedance, traces that come onto pads at 45, traces that merge 45, and corners or bends that are so close together the bends merge into one ill defined thing or a 90.>Inb4 wizardryIt depends on your application and figures od merrit.Make examples and chuck them on the VNA.Or get a tinfoil hat and scream the smith chart wasn't real and VNAs are big antennas marketing tool that just reads out a hidden chip.
>>2969725>LTC2380As I mentioned earlier, it only has one ground pin, I'd consider getting a different chip if just for that reason. But he says it's going to an optocoupler for all the digital comms anyhow, so it's not like there's much digital shit capable of giving noise. The DAC is what I'm worried about, I assume the SPI comms to the PGA are low enough speed it won't be an issue. But he also said that high frequency noise isn't really an issue, he'll be averaging values anyhow. And I think star-grounding will probably prevent digital current-noise on the GND rail interfering with anything, if there is any synchronous sampling noise that could cause systematic error.
If mouser says mfg lead time. is 20 weeks and stock is zerodo they really mean to imply O'll be waiting 20 weeks?REEEEEEEEEEE THE WEST HAS TRULY FALLENI'm going to. duplicate my order on lcsc with lots of chink parts.
So everyone seems to be doing this, more or less. I don't plan on reinventing the wheel and there's probably good reason for everyone else to do this.Now:All 3 elements, optocoupler, Z and TL431 leak. current and also require a certain minimum current for stable, normal, linearish... operation.So I have added R1, R2 and R3 to shunt whatever current other elements may need in excess of their parallel element past it.It should follow that at least 1 R is not needed and can be left open, as one can expect one element to require the highest minimum current.Am I right so far?
>>2969751Makes sense to me. I'd guess that the zener has the least leakage and the TL431 has the most, so R1 probably isn't needed. You could probably combine R2 and R3 into a single resistor.Also I'd still consider a zener (or TVS diode) across U9 just in case a transient comes along. If the parallel diode's breakdown voltage is 10V higher rated than whatever the TL431 will ever see, the leakage should be arbitrarily small. Don't suppose there should be any capacitors here, should there?
>>2969754I checked the following so far:Max rating on the zener is tight, see here for the 62V one.https://www.vishay.com/docs/86337/bzt52-g.pdfDon't want to take that thing past 5mA.Meanwhile this thing here:https://www.vishay.com/docs/83666/sfh610a.pdfYou probably want to design nominal I_F around 5 mA. So perhaps shunt something past the zener but that gives you 90V on the TL431 again, unless a lower impedance path to GND is provided, like another shunt.This fucker here:https://www.ti.com/lit/ds/symlink/tl431.pdfwont get hurt by any of this if it isn't overvoltage.So perhaps:Opto - Zener - ReferenceShunt some opto current past everything else, like 1 mA or so.That way the zener sees 4mA by the time the opto does 5 and the zener gets to protect the TL431.Next also shunt just uA or sot past the TL431, just in case the zener leaks more. Vishay doesn't say a thing about leakiness.Another solution:TL431 says recommended cathode current is greater than 1mA.After which it should be fairly linear.Maybe it's possible to design for 1 mA at 30V output, 2 mA at 60V, 3mA at 90V and 4 at 120.This should cover the whole dynamic range and also ensure the zener is zenering before the output is enough to damage the TL431.I even have a suspicion this is exactly how many designs I see work:They will use R, Z, TL and the opto somewhere in there. No shunts. I suspect, knowingly or not, their TL is passing enough current to make the zener work before the TL is damaged. But those designs are like 48V so that's not a suprise.I shall sim and test this. Sim models are probably crap.
>>2969757>Sim models are probably crapAll I remember is that LTspice has some built-in models for voltage references very similar to the TL431. And they don't work at all.
>>2969758So I built and tested this and am now more confused.1K (later 560R) to opto anode, cathode to 10V zener (I am voltagelet PSU rn) to tl431k (recycled) cathode, anode to GND, tl431 ref to cathode, so its self referenced so should stay at 2.5V. Output side common emitter with some uh... 3k or so R.So on the primary side I expect a voltage drop of Vz + Vf + 2.5 = 14V or so and a current of (Vin - 14)/1k.>Voltage on lowas expected low Vref, high collector out.>Voltage increaseas expected Vref goes up too and at some point output reacts but it's not fair, as I'm using 1 voltage for both...>More voltageVref is stuck at 1.6 V or so.>20VVref is 1.6Vwell thats a nice regulator but why isnt it 2.5V?
>>2969760forgot
>>2969760Uuh, I've seen some TL431 clones that aren't actually 2.5V, so maybe try one that isn't salvaged?
>>2969763>try one that isn't salvagedI ran out. New ones coming in... 20 Weeks when the fucking ferrite factoria delivery the ferrite to mouser heh.I still like the cascode thing. Just have a BJT or FET take the brunt of the blocking voltage. Idl how it could go wrong but why no one does it.
>>2969766regulashiuns just finei think
>>2969766Seems good. Just tune the photodiode bypass resistor for the right amount of bias current and you’re fine. Or you could go extra funky and put a “constant-current diode” in parallel with the photodiode, to put 1mA through the TL431 regardless of the output voltage.
>>2969666lolbut yeah a Peltier would be really nice. one issue is frost. unless one sets the oven to 25°C or so which isn't as exciting as being able to do -50°C>>2969725>it's clearly polluting your GND. Didn't check datasheet but it's obvious:it may be helpful to know that I don't do SPI while converting. I strobe the CNV pin 256 times exactly, at 2 MHz, so as to get one 24-bit sample at a time at 7.8 ksps, then read it. so keeping CNV from interfering is the main concern>>PGA280>Same issue. DGND->GNDA.I only switch it like once per second, and I have deliberate dead time to let things settle when I do>>2969726I'm modulating the sensor at like 10 Hz so RF concerns are moot. except for the digital maybe, but again only the CNV pin is of much concern. thermal is slowthe biggest concern we've had testing an earlier version of this thing isn't any of the analog stuff, but the oven interfering with the measurement. I pulse density modulate the heating resistors at a leisurly pace, and it's possible to see them turning on and off in the data. we "fixed" this by discarding measurements when the heater is on. the optocouplers are mainly for this reason, because the heaters sit on the purely digital side of things. they poke into the thermal baffle but are otherwise entirely separate from the analog boarda shot of the whole board might help. the analog board is mechanically separated so as to prevent thermal stress from the heaters making their way in. it should also make the board stiffer (higher relative thickness)blue is the signal path, red is the CNV pin. it'll fuck off to one of the optosthere's more optos than I'll probably need. a future revision might remove some and further shrink the analog sidethe idea for having cutouts comes from the LTC6655 datasheetthe point of the cutouts on the far left/right is to have the analog board somewhat sprung. again the point is stress relief
>>2969770I just realized I should also try to drive the heaters with a constant current rather than pulsing them
>>2969769its more realistic like this, as the gate voltage wont be independent from. the ref voltage, unless I use a reg.It's easy to see that like this I will be exceeding most FETs gate breakdown voltage.But if I gave it a 10:1 divider I believe it will still work over a reasonable range and never exceed the VGS.I think it also does not suffer from. startup issues. At low output voltage the FET passes no current and does not signal the PWM to put the brakes on. GOD DAMN IT NEED A 90V SUPPLY TO TEST CIRCUITS FOR 90V SUPPLY.I think I'll end up slapping pads for all versions or making an extension connector so I can cut the feedback stuff and redesign a small pcb to plug in.
>>2969770>I dont SPI whenthose chips have clocks and the clocks are running.Perhaps the current is lower while the output drivers arent doing anything at clk frequency but digital things are noisy unless the clk is off. Idk if it is. Is it?Also most DAC and ADV are some delta sigma stuff and similar. Again: Clock, while sampling. Should have a ground where you can dump clock stuff.>it's not causing problemsthen it's good.>the oven is the problemyou could ramp it up and let it fly during sampling and correct in software for the down slope?But I would just do what most OCXO do:A beefy beefy lin reg. For the oven. Just dump the heat you don't need into a heatsink at the other end of the board.
>>2969774>those chips have clocks and the clocks are running.yes the conversion runs on a clock. what am I going to do about that? not convert? the clock is internal to the chip. I trust that Linear have done their job and the one GND is fine. the input is differential>you could ramp it up and let it fly during sampling and correct in software for the down slope?I'd rather have it steady state. but yeah I should probably go for a constant current op amp circuit and stick the MOSFET in there and use its waste heat alsowe plan on measuring ppms while deliberately sweeping the oven temp up/down, with the Pt100 in a gallium cell, to correct for any remaining thermal effects. I'm hoping to get down into the 100s of ppbs
>>2969776>what am I going to do about that?Ground digital ground to dedicated digital ground. I know this sounds obvious so I am. assuming I am missing something.>constant current op amp circuitNah, you want to and AFAIK do regulate your oven. So sensor, feedback compensation, linear regulator. But everything else spunds reasonable, only that if it was my job i'd use a resistive heater and burn the part that is not needed at any point in time in a transistor with enough thermal isolation from the oven.You can make pretty nice single digit R resistors that are very well distributed by putting a meander on PCB. You can either stick one to your oven or use the smave you have in the oven to kind of distribute your resistor acros the oven.
>>2969763verified this TL431 works. It does, Vref is 2.5V in the most simple configuration.I suspect it must have been the zener, rated at 1W it might behave funky at lowish currents. It's meant to get 11V fown to 10 at 1A, or 20V down to 10 at 0.1 A.Then again I did put 20V on the thing and didn't get what I wanted to see. 100mA should have been more than enough.Anyways MOSFET source follower thing next.
>>2969778>Ground digital ground to dedicated digital ground. while the 2380 only have one ground, there might be some sense in having a ground tie and separate grounds for it and the rest of the analog side. in effect I'd then have three grounds:>digital ground on the digital board>digital ground on the analog board>analog groundthen tie the grounds on the analog board at the power connector in the lower left or close to itI'll see what I can come up with>Nah, you want to and AFAIK do regulate your ovenI mean controlling the set point from software with say an RC filtered PWM signal -> voltage -> op amp -> MOSFET -> sense resistor -> feedback on op amp>distribute your resistor acros the ovenI tried something similar to this with an earlier design but it was kind of a hassle
>>2969783>softwaremaybe thats just me because I hate software and I used to work with guys who do all the software so me and a colleague could do all the HW>inb4 I do appreciate that things are alot of effort and dollars to do can be done in 5 minutes in uC and softwareThat being saidFirst thing I'd do was slap purely analog OP and BJT or power OP on there and compensate loop with passives. Filtering issues other than setting your poles and zeros.
>>2969773Yay! We got regulation.The TL431 is set at 15VThe scope shows crossing from 14.99 to 15.99V. The yellow trace shows the opto common emitter output (it is clamped by zener to 5V3 as to make it immune to effects of varying the single input voltage). The green trace shows the voltage across opto diode and TL431 - > so across TL431 is ~1.3V below this.I believe this shows the thing works as far as:The TL431 will not see the full voltage.The thing provides feedback.It begs the question: Why would TL431 be needed anymore, at all. I believe a zener in it's place would do the job but then again the whole thing could be just enough zeners.I will put this in the schematic and board and add some additional footprints for the alternative solutions and all the current shunting and stuff.
Guess EDM and SMPS anpn is getting bammed again and will ragequit again very soon. It was nice while it lasted.
>>2966970>posting aspergers.Pic rel is autism. https://elm-chan.org/
>>2969803indescribably sovlful
>digikey, farnell and mouser all have my IP and throw anti-bot shit in my facewhat are some suppliers that might actually want my money?
>>2969833LCSCmaybe ArrowRS's website is kinda ass, though still better than LCSC's
Tachometer guy, last circut worked flawlessly on all the small engines/generators I had at the house but not the race motor, screen read fine but SD card still corruptingI finally gave in and bought a scope, now I gotta learn how this fucker works
>>2969851the solution was to use Chromium. maybe it's just that I need to upgrade Firefox
>>2969873or you can just tell sites that youre using chromehttps://addons.mozilla.org/en-US/firefox/addon/uaswitcher/
>>2969881or you can just change your DNS to Cloudflare, Google, NextDNS, etc., and flush your DNS cache.
>>2969871When it comes to EMI, how you route your scope probe makes a big difference, and measuring common-mode noise can be downright difficult if the scope itself is grounded. Have fun.You can make magnetic field probes out of thin coaxial wire though, that might be worth trying.>>2969873Maybe instead it was Ublock Origin interfering with first-party spyware like it's supposed to, which you can't really do anymore on Chromium-based browsers.
>>2969903> how you route your scope probe makes a big differenceI thought coax 50 ohm probe cables were essentially shielded so no issue ?
>>2969968I mean the antenna wrapped around the spark lead.
NEW BREAD>>2970026>>2970026>>2970026MIGRATE
