|
ryanrs posted:You've also neatly explained most of the reasons I only use 32-bit ARM microcontrollers these days. Absolutely nothing I build is cost-sensitive enough to deal with that kind of 8-bit aggravation. My first MCU family was AVR, like ATTiny44, ATMega48, etc. They had a lot of things going for them which made them excellent learning platforms: 1. Powerful yet manageable instruction set (much more powerful than a PIC). I started off in assembly and don't regret it. 2. Most instructions take 1-2 clocks, maybe 4 max for certain branch/return instructions. ISR latency is ~7-10 clock cycles IIRC. At 16MHz you could get quite a bit done. 3. Great community support. This was in the days before vendors started pushing their own support forums on you. AVRfreaks was excellent and still exists today (I think). 4. The IDE, AVR Studio. Completely free, and by the time I started up it was drat stable and robust. v4.3 I think. A lot of engineers never upgraded from that because later versions (which became Atmel studio) jumped to VS and that made everything a mess for several years. The best thing about their IDE is that the vast majority of the AVR devices had full simulator support. At the time, I actually thought that was normal for all MCUs to have a simulator (MPLap offered simulators for PICs, I think). You could literally run your code without connecting to an actual target device via a debugger. You could step the thing not just by instruction, but by CPU clock cycle. You could set data/program breakpoints anywhere, with no limits. It's difficult to explain how valuable this was, and how much I wish I had such a thing for the devices I use now. It means you can perfectly profile execution time of any code without modifying it. The peripherals were also modelled as well (more on this later). When I reached the limits of those AVR chips and had to find something more powerful, I was sad to find that none of them had anywhere close to the level of simulation I was accustomed to. But if anything, having a cycle-level simulator is even more useful with fancier CPUs with deeper pipelines. For example, for some families of devices, whenever the debugger halts the cpu (like when hitting a breakpoint or single-stepping) on a real target device, the pipeline is flushed. This means whatever whatever instructions were in the pipeline need to be re-loaded from the start, meaning they will execute later than if you hadn't halted the cpu. Overall, the more sophisticated your cpu, the more difficult it is to observe what the cpu is doing without the debugger. For simulating ARM cpus there aren't many (free options). Keil MDK is the main option I see from an established vendor, but the free version is pretty limited. Haven't tried it myself though, would be interested to hear anyone else's experience. And AFAIK none of the ARM simulators offer any real peripheral support. Not a big deal if the peripherals are implemented and documented competently. If not, then... poo poo. In the C2000 devices I'm currently using, the I2C peripheral is loving miserable, and they offer almost no support via example projects. For example, when handling a repeated start condition as a slave, there are up to five status bits that may be set/cleared by hardware, all within the span of a few I2C clock periods. They all can fire interrupts, so it's critical to know not only which ones fire, but also the order in which they fire. But sometimes the difference is one or two CPU clock cycles. When stepping via the debugger, you're actually jumping by several clock cycles at a time, so you will usually (but not always!) miss this. A couple of these bits can actually be set and cleared by hardware within those gaps, meaning you you can't tell in debug view that they were set at all. In the end it can work, but it's agony to get there. ANIME AKBAR fucked around with this message at 17:32 on Oct 23, 2022 |
#
?
Oct 23, 2022 17:29
|
|
|
|
| # ? May 9, 2024 14:10 |
|
|
I also started in a similar low level manner, with PICs, and I agree that that has made me better about understanding everything that's going on. But man, it really does take ten times the time to write anything performant for 8 bit micros. The powerful new stuff, you don't even need to think about, you just throw a bunch of C++ at, and it runs, fast. It's definitely frustrating that the lazy way is "best" for some definition of best. It's the same argument for everything being written in JavaScript / Node / React, I guess, and I do hate that
|
|
#
?
Oct 24, 2022 02:27
|
|
|
It's all just volume right though? For smaller quantities your design and develop time is a larger percentage, so just throw an esp32 or rp2040 or stm32 if you can get it, who cares about the price. I was designing some prototypes for work at like qty 20 and frustrated with the cost of some available parts until I was reminded if I spent even one day looking for alternatives and updating the design my billable rate would exceed the difference.
|
|
#
?
Oct 24, 2022 06:03
|
|
|
Engineering costs dominate small runs, I remind myself, as I submit another $2,500 Digikey order.
|
|
#
?
Oct 24, 2022 07:27
|
|
|
Rescue Toaster posted:It's all just volume right though? For smaller quantities your design and develop time is a larger percentage, so just throw an esp32 or rp2040 or stm32 if you can get it, who cares about the price. I was designing some prototypes for work at like qty 20 and frustrated with the cost of some available parts until I was reminded if I spent even one day looking for alternatives and updating the design my billable rate would exceed the difference. I use heaps of parts from Wurth even though they're way more expensive. Wurth parts all have extremely high quality schematic symbols and pcb footprints though. It saves me so much time that I calculated it once and it would take hundreds of thousands of units to break even on the time (and therefore money). This is at work of course, where my time has an actual monetary value. When you're a hobbyist, spending time is half the point so it's kind of free.
|
|
#
?
Oct 24, 2022 11:41
|
|
|
which one of you runs https://www.leadedsolder.com ?
|
|
#
?
Oct 24, 2022 12:21
|
|
|
Splode posted:I use heaps of parts from Wurth even though they're way more expensive. Wurth parts all have extremely high quality schematic symbols and pcb footprints though. It saves me so much time that I calculated it once and it would take hundreds of thousands of units to break even on the time (and therefore money). Yeah a few places have systems to download footprints in some standard format and convert it to whatever cad software. More and more are supporting kicad too. Nothing better than not having to worry about loving up a footprint. I had to do some weird rear end 9-pin dsubs that were flipped upside down so one of the middle pins was shifted between the rows. I probably would have missed it honestly, but the footprint I got from them made it obvious what was going on. Yeah for hobbyists, I still think my time is the most valuable thing, it's just a question of where I'm spending it. I do mostly software at work so I don't actually normally want to gently caress around with it as much at home, I use a lot of micropython at home because that's not the part I want to spend time on and I can just spend an extra couple dollars to get a raspberry pi pico carrier for a RP2040 and use that, or whatever. That said the recent videos I've been watching about building CPUs from scratch have made me interested in building something and writing my own assembler and compiler and such. Haven't done anything like that since college. I seem to have lost my 'Build your own Z80 computer' book that I was keeping for sentimental reasons...
|
|
#
?
Oct 24, 2022 17:38
|
|
|
I would blow Dane Cook posted:which one of you runs https://www.leadedsolder.com ? I believe that would be Seat Safety Switch.
|
|
#
?
Oct 24, 2022 19:19
|
|
|
If you have two 12V power supplies, can you connect the NEG_2 terminal to POS_1 and get 24v between NEG_1 and POS_2?
|
|
#
?
Oct 24, 2022 22:18
|
|
|
Paul MaudDib posted:If you have two 12V power supplies, can you connect the NEG_2 terminal to POS_1 and get 24v between NEG_1 and POS_2? You sure can, just make sure that the power supplies aren't both grounded on their outputs
|
|
#
?
Oct 24, 2022 22:23
|
|
|
Splode posted:You sure can, just make sure that the power supplies aren't both grounded on their outputs Normal industrial supplies with screw terminals usually don't so you can series connect them. Not relevant for 24 V but there is usually a soft limit on the number of supplies you can stack like that, normally +/- 250 VDC or so relative to chassis. This is limited by the Y-capacitor voltage ratings I assume. Rescue Toaster posted:But I had way more problems hand soldering a FFC connector because the legs just took NO solder, so drag soldering it created like 20 bridges even with tons of flux, the legs really sort of 'scraped' the solder off the hoof tip and then it had nowhere to go. 0.5 mm pitch FFC connectors are miserable to solder, I've found buying brand name ones from Digi-Key is very slightly better since the solderability of the leads is better. AliExpress types require a decent flux to be applied, but the construction of the leads means liquid flux just disappears somewhere inside it instead of soaking the leads. Flux gel might work better.
|
|
#
?
Oct 25, 2022 08:24
|
|
|
mobby_6kl posted:I've been looking at the internals of my cheap Chinese laptop and it's pretty interesting what's left inside The plate is pretty thin so has gently caress all thermal mass and heats up quickly and takes a long time to cool down in testing. This film has to work more as an insulator rather than help with emissivity , right? I thought it would be completely negligible anyway but plugging some approximate numbers into this here calculator, enabling emissivity lowers the temperature by like 30 degrees, which seems ridiculous. 0.07  0.7  https://www.heatsinkcalculator.com/flat-plate-heat-sink-calculator.html But looking at the calculated examples here, they do have an example at 5W and natural convection, and radiation is up to 30% of heat transfer  https://www.heatsinkcalculator.com/blog/the-importance-of-radiation-in-heat-sink-design/ The best thing would be to test it of course but that film is one-way only, I won't be able to stick it back on once removed 
|
|
#
?
Oct 25, 2022 12:41
|
|
|
mobby_6kl posted:One more laptop thing. That copper thing is the heatspreader that does the passive cooling and is covered by a black film glued to the top, which I'm lifting a little here. Sometimes those are thin vapor chambers and not just plates, but it's hard to tell from your image or from not knowing what's underneath the plastic. Is there a heatpipe out to any fins or fan assemblies? The plastic might also be preventing whatever it sandwiches to from shorting on the copper.
|
|
#
?
Oct 25, 2022 13:07
|
|
|
Rexxed posted:Sometimes those are thin vapor chambers and not just plates, but it's hard to tell from your image or from not knowing what's underneath the plastic. Is there a heatpipe out to any fins or fan assemblies? The plastic might also be preventing whatever it sandwiches to from shorting on the copper. Seems very unlikely, I haven't tried actually removing it but I seriously doubt there are any heat pipes or vapor chambers there. It's just an Atom cpu so they didn't bother with anything fancy I think. The surface shouldn't be in contact with anything either, it's just the back cover ~1mm opposite of it.  Also they kindly left enough empty space to stick a 15Wh phone cell, which would increase the battery capacity by 50% lol.
|
|
#
?
Oct 25, 2022 15:03
|
|
|
Leo's bag of tricks has a nice new video about getting started with getting into electronics. In case we currently have beginners reading the thread: https://www.youtube.com/watch?v=IU9CpR9EeA4
|
|
#
?
Oct 25, 2022 16:22
|
|
|
longview posted:And if it's PC/server power supplies they probably do have grounding to chassis ground on the minus side. Neat, was thinking of daisychaining astron 12v supplies. I don’t have an immediate application, was kinda just wondering in passing if it was possible.
|
|
#
?
Oct 25, 2022 16:43
|
|
|
longview posted:And if it's PC/server power supplies they probably do have grounding to chassis ground on the minus side. Another thing to consider when stacking is noise performance can get weird sometimes especially if you're stacking to create a +/- supply with ground in the middle. Just something to watch for. These FFC's were molex from digikey, so I wasn't expecting much trouble. I use chip-quick gel flux but it's really only gel for application, it flows pretty quick once hot. I've seen some of the guys doing repairs and poo poo on youtube have some flux that seems to stay put much better while heating, but of course it's a nightmare to clean it looks like and I like my PCB's spotless.
|
|
#
?
Oct 25, 2022 17:32
|
|
|
I bought a small battery monitor for a 48v battery + inverter setup that I have. The shunt that came with the monitor appears to have a slot that was cut somewhat haphazardly, about halfway across the whole width of the blade in the middle. The listing on ebay shows a couple small notches instead: https://www.ebay.com/itm/125301026178 Does this sound normal for a shunt? Is there a good way to test for accuracy with a multimeter?
|
|
#
?
Oct 25, 2022 18:46
|
|
|
Yep, they calibrate each one at the factory. In the case of very, uh, budget products, that might be a random warehouse, with an angle grinder. But I wouldn't be that worried
|
|
#
?
Oct 25, 2022 18:53
|
|
|
And if the notch is deep enough it also becomes a bonus fuse! Only way to test a current shunt is to apply a current, you'd need a (preferably actually DC, not some LED driver thing) constant current power supply and a known good current meter in series. Run the current through the shunt and you can measure the voltage across it. Or just hook up the meter and check what it says.
|
|
#
?
Oct 25, 2022 19:02
|
|
|
Bench multimeter in 4-wire resistance mode.
|
|
#
?
Oct 26, 2022 22:45
|
|
|
Little while ago I asked for some advice on servos and arduinos for my wife's Halloween costume. Thought some of you might want see how it's coming along. Still not quite done yet, more painting and stuff, but looks really good so far. https://www.youtube.com/watch?v=zPjBmnWhDgo And here's the completed one https://www.youtube.com/watch?v=uYNungujZR0 A Proper Uppercut fucked around with this message at 01:31 on Oct 28, 2022 |
|
#
?
Oct 27, 2022 03:18
|
|
|
A Proper Uppercut posted:Little while ago I asked for some advice on servos and arduinos for my wife's Halloween costume. Thought some of you might want see how it's coming along. Still not quite done yet, more painting and stuff, but looks really good so far. That's both really cool and incredibly unnerving, great job!
|
|
#
?
Oct 28, 2022 19:44
|
|
|
So question: I've used these 74AHC1G14's to both add some nice Schmitt trigger hysteresis filtering to inputs and simultaneously level shift from 5V to 3.3V by putting a 10K resistor on the input to limit the current. The datasheet even explicitly says it's fine: quote:The input and output voltage ratings may be exceeded if the input and output current ratings are observed. However, in a recent design, I consolidated two of them into one of these guys without really thinking about it cuz I had it handy, and looking at that datasheet there's no similar note. TI's version of the same part does mention this: quote:The input negative-voltage and output voltage ratings may be exceeded if the input and output current ratings are observed. but 1) that's TI, not the one I have, and 2) it seems to be specifically talking about only the negative voltage rating. I think it will be fine, since basically you're just letting the built-in TVS diodes sink the tiny bit of current needed to drag the input down to 3.3V, but I don't actually know and the datasheet has me second-guessing that now. If I put 5V via a 10K resistor into this thing, do you think it's going to destroy itself? More importantly considering how annoyingly hard it is to get new ones these days, is it going to let the 5V go straight through to the raspberry pi at the other end? Shame Boy fucked around with this message at 20:04 on Oct 28, 2022 |
|
#
?
Oct 28, 2022 20:02
|
|
|
Shame Boy posted:... Shame Boy posted:More importantly considering how annoyingly hard it is to get new ones these days, is it going to let the 5V go straight through to the raspberry pi at the other end? Probably not. The worst case I can think of for damaging other things is if there's a diode back to the 3.3V supply, so you're pulling up on that supply through a diode and a 10k resistor, which gets you like 100uA pulling up on the 3.3V supply. Assuming the load on your 3.3V rail is more than that, the regulator will just adjust for it and you'll never notice. I guess if the load ever gets too light there may be nothing around to sink that current and the 3.3V rail could start creeping up. The easy/kind of shoddy way to guarantee this will never happen is to have a 10k resistor from the 3.3V rail to ground to guarantee a 330uA minimum load. I'm not really able to think of a plausible way your 5V could bypass the supply rail and get to the output driver, since the output driver is going to be a set of MOSFETs with their gates on earlier stages and their drains on the output pin. You can't pull the DC voltage of the drain of your MOSFET down below its source with the gate, so as long as the supply is OK the output should be OK, unless it's melted itself to a puddle of conductive slag (again, unlikely), then all bets are off.
|
|
#
?
Oct 28, 2022 21:22
|
|
|
Why not just build a proper level shifter with two resistors and a mosfet?
|
|
#
?
Oct 28, 2022 21:38
|
|
|
Or feed the 5v into a resistor divider to the IC's input
|
|
#
?
Oct 28, 2022 21:43
|
|
|
ante posted:Or feed the 5v into a resistor divider to the IC's input This works if the IC only draws a small amount of current, but if it draws enough that the voltage at the IC pin drops below the logical high level, it won't work. Best to use an actual level shifter.
|
|
#
?
Oct 28, 2022 21:46
|
|
|
Cojawfee posted:Why not just build a proper level shifter with two resistors and a mosfet? If you're asking why I didn't do that, it's cuz I already needed the Schmitt trigger anyway and I've used that first chip to do exactly this sort of thing tons of times in the past and it's always worked fantastic  Hell I think I got that trick from an app note for it at one point even.If you're asking why not add that now, it's because the boards are already made and in my hands 
Shame Boy fucked around with this message at 22:12 on Oct 28, 2022 |
|
#
?
Oct 28, 2022 22:10
|
|
|
Stack Machine posted:It's probably fine, but if it's not in the datasheet it's not guaranteed to work. If this is something you're selling I wouldn't do it, just because you wouldn't be effectively covering your rear end, but if it's just for a personal thing, go nuts. My gut feeling of the risk is 99% works fine, 0.9% ESD protection causes weird behavior on pin, 0.1% you cause it to latch up in some bad way and it somehow shorts power to ground and destroys the part by discharging the decoupling capacitor into it. Yeah it's just a personal thing so no big deal, like I said my only real concern is frying the RPi since I haven't seen one for sale at not-scalper prices in over a year. But it sounds like that's not really likely to happen (I'm sure the thing will have enough load on the 3.3V line at all times to sap away any leakage) so cool, thanks.
|
|
#
?
Oct 28, 2022 22:16
|
|
|
Cojawfee posted:This works if the IC only draws a small amount of current, but if it draws enough that the voltage at the IC pin drops below the logical high level, it won't work. Best to use an actual level shifter. How much current do you think a CMOS input pin draws
|
|
#
?
Oct 28, 2022 23:42
|
|
|
ante posted:How much current do you think a CMOS input pin draws How thick is a gate oxide? But yeah, for a pure input a resistive divider is reasonable. As long as the divider impedance is a relatively low compared to the input impedance, it won’t matter much. You can figure out a reasonable input impedance approximation by taking the input voltage divided by leakage current spec at said voltage. Resistive dividers are typically a larger problem for higher speed signals because the resistive divider and input capacitance forms a low pass filter which slows down edges in the time domain.
|
|
#
?
Oct 29, 2022 08:10
|
|
|
Any diode fans here? what could I use instead of a 1SS184 that's actually in stock at mouser?
|
|
#
?
Oct 29, 2022 09:24
|
|
|
SnoPuppy posted:How thick is a gate oxide? Most of the pin current is leakage from ESD devices and body diodes. Those schmitt trigger inverters have max pin currents up to around 1uA in the data sheet and that's orders of magnitude more than MOS gate currents. (It will be nowhere near 1uA at room temperature for most units.) Leakage currents are so much higher than gate currents that it leads to some real consequenses in semiconductor manufacturing. There are "antenna" design rules that are all about keeping the gate from accumulating enough charge before all the layers are manufactured by either having a large enough gate to share the charge that might accumulate or having sources/drains on the same net (without hopping up to higher metal that hasn't been added yet!) to drain it down with body leakage. SnoPuppy posted:But yeah, for a pure input a resistive divider is reasonable. As long as the divider impedance is a relatively low compared to the input impedance, it won’t matter much. You can figure out a reasonable input impedance approximation by taking the input voltage divided by leakage current spec at said voltage. If you ever find yourself in this situation, throw a capacitive divider with the same ratio in parallel with your resistive one. That will handle the high frequencies and the resistors will hold it at DC. Stack Machine fucked around with this message at 14:40 on Oct 31, 2022 |
|
#
?
Oct 29, 2022 13:36
|
|
|
Shame Boy posted:So question: I've used these 74AHC1G14's to both add some nice Schmitt trigger hysteresis filtering to inputs and simultaneously level shift from 5V to 3.3V by putting a 10K resistor on the input to limit the current. The datasheet even explicitly says it's fine: In my professional opinion, Yolo
|
|
#
?
Oct 29, 2022 14:12
|
|
|
I would blow Dane Cook posted:Any diode fans here? what could I use instead of a 1SS184 that's actually in stock at mouser? Mouser option: https://www.mouser.com/ProductDetail/ROHM-Semiconductor/BAV70HYFHT116?qs=QNEnbhJQKvZ5nyvoPGp0dg%3D%3D Digikey option: https://www.digikey.com/en/products/detail/panjit-international-inc/MMBD6100-R1-00001/14660895 The mouser one has slightly higher forward voltage drop, but I think within the worst case of the 1SS184. One thing to keep in mind is SC-59 is basically the same as SOT-23, so you might have more options than you realized. I think they might technically be like a fraction of a millimeter off in the leg length or something but nothing that would be a problem hand-soldering a replacement. Most important parameter is obviously reverse voltage, and forward current (want to exceed those). Then try to make sure forward voltage at your typical current isn't too much higher if the circuit depends on it (hard to know without knowing the circuit). Then recovery time is usually the next factor to consider, again depends on the circuit if that matters. Since the 1SS184 is specifically billed as a 'ultra high speed switching application diode' I figured it was important to keep the reverse recovery and capacitance in line.
|
|
#
?
Oct 29, 2022 14:48
|
|
|
Holy crap I love these gears: https://www.youtube.com/watch?v=QrkiJZKJfpY
|
|
#
?
Oct 29, 2022 23:11
|
|
|
I saw that too and thought they would be fun to play with until I saw how much they cost.
|
|
#
?
Oct 30, 2022 00:32
|
|
|
So far the Schmitt trigger level shifty thing is working fine without exploding so that's good. Had a different problem though, a part of the circuit is getting enabled even when it shouldn't. I checked the power supply enable circuit, since the LDO was supposed to have a built-in pull-down, and everything seemed fine. Then I noticed the actual value of the built-in pull-down: 4 megohms. Gee, thanks. I don't know if this is typical, but from other chips I'm used to built-in pulldowns being more like 50-100K so I guess jokes on me for not checking. Leakage from some diodes in another part of the circuit is almost certainly enough to raise the voltage above the 1.2V high level threshold if the pulldown is freakin' 4M. Oh well, at least I think it'll be easy to bodge in a resistor...
|
|
#
?
Nov 3, 2022 17:32
|
|
|
|
| # ? May 9, 2024 14:10 |
|
|
Check out this blank-rear end wall. I wanna fill it up, and a thought that came to mind was a bunch of LED RGB fuckery. Anyone know of a good LED wall art project that isn’t too simple, but still doable for a hobbyist? Size is not a problem! I found this, which looks cool: https://youtu.be/g6n8XLmZ__I But ideally it wouldn’t be hooked up to my vital signs.
|
|
#
?
Nov 6, 2022 00:57
|
|
