|
Unless there's some internal calibration path, it took some effort to get that square wave on the screen with no probes connected. I think that, or maybe the hammer, is what tips the Poe's Law scale to parody for me.
|
# ? Aug 15, 2022 15:24 |
|
|
# ? Jun 11, 2024 01:55 |
|
A friend just pointed out that the wrench is hex and the sockets are all square. Also the wrench isn't actually attached to the socket in front of it: This is like those old I Spy books
|
# ? Aug 15, 2022 15:41 |
|
Stack Machine posted:Unless there's some internal calibration path, it took some effort to get that square wave on the screen with no probes connected. I think that, or maybe the hammer, is what tips the Poe's Law scale to parody for me. Can you view saved waveforms on that model? That's what I assumed it was, but my scope is from the 80's, so I don't know about them fancy new fangled options.
|
# ? Aug 15, 2022 16:34 |
|
One Legged Ninja posted:Can you view saved waveforms on that model? That's what I assumed it was, but my scope is from the 80's, so I don't know about them fancy new fangled options. It's a DSO so my guess is the procedure was something like connect probe to calibrator, trigger scope, disconnect probe at the BNC end (but leave it clipped to the calibration output for a little extra humor), take photo. No clue about stored waveforms (other than the most recent one) but probably wasn't necessary here.
|
# ? Aug 15, 2022 16:39 |
|
Hot drat, everywhere is now out of my favorite under-appreciated micro family, the ATXmegas. I thought I was the only person in the world who used them judging by how unhelpful google results have always been. The only ones in stock at any supplier I actually trust are one or two BGA variants and some of the weird-rear end B models that have built-in LCD drivers instead of actually useful things.
|
# ? Aug 15, 2022 18:07 |
|
Shame Boy posted:Hot drat, everywhere is now out of my favorite under-appreciated micro family, the ATXmegas. I thought I was the only person in the world who used them judging by how unhelpful google results have always been. The only ones in stock at any supplier I actually trust are one or two BGA variants and some of the weird-rear end B models that have built-in LCD drivers instead of actually useful things. Sing it with me! https://www.youtube.com/watch?v=u1vuz8EtC9s It's a chip shortage! It's a chip shortage! Every day! In every way!
|
# ? Aug 15, 2022 18:17 |
|
Stack Machine posted:It's a DSO so my guess is the procedure was something like connect probe to calibrator, trigger scope, disconnect probe at the BNC end (but leave it clipped to the calibration output for a little extra humor), take photo. On modern digital scopes you can usually store/recall a trace as a reference waveform to either a USB stick or maybe a couple of internal memory slots. It can be pretty useful when you want to compare results from different tests before/after changing some component. The reference waveform is typically a different color than the normal traces, that scope seems to use yellow for channel 1 and the trace on the screen is yellow, so yeah it's just the last thing the scope triggered on.
|
# ? Aug 15, 2022 23:04 |
|
Rescue Toaster posted:Drown the fucker in flux. It's the only way to be sure. This goon knows what's up. I spent two summers doing a lot of SMD soldering, including plenty of itty-bitty ICs with lots and lots of pins. I'd do several large boards in a morning, then work on other stuff in the afternoon. Here's my suggestions: #1 most important: Use flux. Lots of flux. A flux pen is a start, but if you really want to do it right, get liquid flux and a syringe. You put just the gentlest pressure on the plunger and touch the (blunt) needle against the board and the flux just kind of wicks out. If you're installing some godawful 100-pin QFP chip, you really ought to be drag soldering. They make special tips for it, but you can do it just fine with a regular soldering tip too. Put flux on the pads before you set the chip down, then put more flux across the pins when it's in place. Tack down the corners, then get a little solder on your iron or on a corner pin and drag it across the pins on that side. It'll sound like the world's tiniest zipper being zipped. Because you're moving quickly, the component won't be damaged by the heat. Because you've used lots of flux, you're not likely to get any bridging between pins. If you do get a bridge between pins, you can usually just wipe your iron tip on the sponge, apply a little bit of flux to the bridge, then swipe the tip of the iron between the two pins, starting near the body of the chip and moving away briskly. Inspect with a microscope to make sure it's clear. Try a pointy soldering tip. The rounded/wedge-shaped ones are fine for through-hole, but for SMD I prefer a sharp tip. I would run the iron at 700F. Buy the skinniest solder you can find, because it gives better control over quantities. Use a microscope if you can, or at the very least get a pair of those magnifying goggles for soldering. Jeweler's loupes are not worth the hassle. When you're done, the board will be covered in flux. If it isn't, you didn't use enough flux. Run the tap water as hot as you can (we always kept the water heater turned all the way up in the office for this reason) and scrub that bastard with an old toothbrush. When it's clean, shake the water out and put it somewhere to dry. I found the work pretty pleasant once I got used to it. I can't stand paint-by-numbers, but I think the satisfaction I got from properly populating a surface-mount board is probably the same thing normal people get from paint-by-numbers. Best thing about SMD: no flipping the board, and no trimming of loving leads!
|
# ? Aug 17, 2022 01:09 |
|
This is true for through-hole as well, but for SMD work in particular you'll often be tempted to re-heat a joint without adding more solder, since the area is so small. Since you don't get more flux from the solder, you must always add extra flux from a little flux pen every single time you heat a joint. Yes, every time. Every time I forget or try to cheat it, I have to add flux and re-heat it a third time. Quick-chip has some nice gel fluxes with a little needle applicator that I use constantly.
|
# ? Aug 17, 2022 03:37 |
|
I got a Mesa 6i25 I/O card a little while ago but never got around to figuring out how to use it. The problem with it was that there are no drivers available for it for any OS, just a text file about what memory locations in it do/mean if you read/write them. I was worried I was going to have to write an actual device driver for it, but it turns out that Linux's generic PCI userspace I/O driver lets me just ram raw bytes into the memory locations. It looks like this awesome pointer hell but it works, and it actually works a lot like the GPIO modules in Microchip PIC micros too which is nice. code:
|
# ? Aug 18, 2022 09:36 |
|
BattleMaster posted:I got a Mesa 6i25 I/O card a little while ago but never got around to figuring out how to use it. The problem with it was that there are no drivers available for it for any OS, just a text file about what memory locations in it do/mean if you read/write them. I was worried I was going to have to write an actual device driver for it, but it turns out that Linux's generic PCI userspace I/O driver lets me just ram raw bytes into the memory locations. Thanks for posting this. I dont feel like such a bitch that I've always opted for hobbyist grade stuff like Arduinos or RPis for automation/machinery projects now. I even kinda like CircuitPython
|
# ? Aug 18, 2022 12:55 |
|
https://youtu.be/-hVZoaRTVaQ Now all I need to do is hook it up to a 3.5mm jack and I have a rudimentary square wave. Circuit is an astable 555, blatantly stolen from here. It’s not the most sophisticated or high-quality circuit, but I haven’t done any non-trivial circuit work since college physics, so I’m just happy it works at all. From here I need to figure out how to: - Adjust the frequency (aka period) of the wave with a potentiometer (plus another, less sensitive trimpot for fine tuning) - Adjust the pulse width of the wave with a potentiometer(?) - Add a 3.5mm jack that provides 1V per octave attenuation on the frequency - Derive a triangle wave, saw wave, and sine wave from the output of the 555 I uhhh don’t know how to do that. Y’all are better at electronics than me, any tips? csammis posted:Bit of a learning curve but Kicad 6 is good and free and possibly way overkill for what you want to do right now Work has KiCad 6 actually, I’ve used it and it’s not so bad. Still don’t really understand what I’m doing. I really wish there was a program out there that lets me: - Draw electrical circuits - Show electrical circuits on a breadboard (and, ideally, stripboard) - Add common electronic parts and ICs, with fully implemented specs (e.g. resistance, capacitance, pinouts, max current for LEDs) - Simulate circuit flow and find mistakes and poo poo that would blow everything up - Check resistor, capacitor, etc. values for a given input voltage - Provide arbitrary input voltage values so I can check against 5V, 9V, and 12V But KiCad and Fritzing are both missing one or more of these pieces, most notably a 12V power source as a drop-in simulateable part. Any advice on proving a circuit that uses this 555 virtually before moving to a physical breadboard? Pollyanna fucked around with this message at 15:49 on Aug 18, 2022 |
# ? Aug 18, 2022 15:36 |
|
The Falstad simulator does a lot of what you're asking for. Googling around about how 555s work and different configurations should get you up to speed on how they work. I don't have time to elaborate just this second, but the input side has a comparator (triggers if one voltage is less than another) and just waits for decay in voltage in the cap and then flips the input and charges it back up. With a good understanding of how they work, you can do all sorts of unusual stuff with 555s.
|
# ? Aug 18, 2022 17:06 |
|
By the way, if you use a mac you can try iCircuit, it's basically a way less jank Falstad (I'm pretty sure it's based on the falstad code or something even). It's also available on android and iOS if you want to use it on your phone for some reason. I used it a ton until I got better at the goddamn impenetrable bullshit of LTSpice and I still use it occasionally if I just want to play around with something quickly.
|
# ? Aug 18, 2022 17:51 |
|
ante posted:The Falstad simulator does a lot of what you're asking for. Good point - the change in charge in the capacitor over time would be a triangle wave, wouldn’t it? Maybe there’s a way to output it.
|
# ? Aug 18, 2022 19:36 |
|
Pollyanna posted:Good point - the change in charge in the capacitor over time would be a triangle wave, wouldn’t it? Maybe there’s a way to output it. It's a triangle-ish wave between 1/3 and 2/3 of vcc and yeah it's output on one of the pins all the time. But it's not really a very accurate triangle wave, because it's charging and discharging a capacitor, so you wind up with this: RC circuits (resistor/capacitor) charge and discharge exponentially like that so you'll only ever get a more or less triangle wave out of em', though if you just want to make triangle wave-y sounds that's probably fine. Shame Boy fucked around with this message at 19:42 on Aug 18, 2022 |
# ? Aug 18, 2022 19:39 |
|
Hhhhhuh, okay. Could still be an interesting wave to play with. Would be really nice if I could have something like a multimeter that lets me put both probes on a wire coming out of a pin, and measure the current value over time at that pin. So I might DIY an oscilloscope or something. Or shell out for an NTS-2…but lmao that thing is $230. Pollyanna fucked around with this message at 19:45 on Aug 18, 2022 |
# ? Aug 18, 2022 19:41 |
|
It's a logarithmic curve, but yeah You can get a triangle wave by (dis)charging a capacitor with a constant current, but that's a different topic entirely
|
# ? Aug 18, 2022 19:48 |
|
ante posted:It's a logarithmic curve, but yeah Logarithms are just sneaky exponents
|
# ? Aug 18, 2022 19:56 |
|
True
|
# ? Aug 18, 2022 20:12 |
|
BattleMaster posted:I got a Mesa 6i25 I/O card a little while ago but never got around to figuring out how to use it. The problem with it was that there are no drivers available for it for any OS, just a text file about what memory locations in it do/mean if you read/write them. I was worried I was going to have to write an actual device driver for it, but it turns out that Linux's generic PCI userspace I/O driver lets me just ram raw bytes into the memory locations. Did you measure how fast it can wiggle the pins? One of the most frustrating things about doing desktop GPIO through a shift register on the parallel port is it's too slow for DRAM or any CPU with dynamic logic.
|
# ? Aug 18, 2022 21:10 |
|
Stack Machine posted:Did you measure how fast it can wiggle the pins? One of the most frustrating things about doing desktop GPIO through a shift register on the parallel port is it's too slow for DRAM or any CPU with dynamic logic. I just went and did that. Using this loop to toggle pin 1 as fast as it can: code:
But I still get around 125-250 ns pulse lengths: Which is pretty alright. I assume the jitter is pegged CPU usage combined with background tasks, I/O access to other devices, and the fact that I'm not using an RT kernel so there are fewer guarantees on task scheduler timing. And what's better is that these things have modules for various things like stepper pulse generators, PWM generators, UARTs, counters, and others so you don't need to spend that much CPU time on certain things. The problem is that if you want to vary the number of these modules and what pins they're connected to, you have to make alterations to VHD code, recompile it, and reflash it. I'm not even sure how to get the VHD tools to do that yet, let alone how to actually do it because of the state of the documentation. Also, I still plan on building that Busyboard you sent me, but I'm still looknig for the parallel port connector. Digikey was out of the one you used or any other of the same type, so I need to go scrounging elsewhere. A while ago I built a parallel port to I2C interface using some resistors and a 74HC05 hex open drain inverter. I learned that in Linux you don't need to write to arbitrary I/O addresses but you can open up the parallel port device file and call ioctl functions on it to do stuff with it. Aside from not having to hope and pray you're writing to the right addresses, it obeys permissions so users in the lp group can do parallel port stuff without having to be root. I'm still trying to figure out if I can do something like that with the Mesa card and the uio_pci_generic driver. I was able to set a udev rule to make /dev/uio* files writeable by my new uio group, but the /sys/ tree files that let you do most of the work still are root only. BattleMaster fucked around with this message at 00:52 on Aug 19, 2022 |
# ? Aug 19, 2022 00:49 |
|
You'll probably get occasional blips of 10s of ms. Default settings for linux's default scheduler aims to wake processes that want to run no more than 7ms after they become runnable
|
# ? Aug 19, 2022 01:32 |
|
Foxfire_ posted:You'll probably get occasional blips of 10s of ms. Default settings for linux's default scheduler aims to wake processes that want to run no more than 7ms after they become runnable I was getting stuff like that with my I2C parallel port code: It uses nanosleep to ask for 25 microsecond sleeps, which are each supposed to be one quarter of a 100 microsecond clock cycle. Of course, Linux just takes these as suggestions because PC hardware and operating systems are not designed for any sort of timing guarantees. I wasn't watching my Mesa card's bit toggling for an extended period of time, but with no sleep calls it seemed to just constantly plug away without anything that bad happening. BattleMaster fucked around with this message at 02:44 on Aug 19, 2022 |
# ? Aug 19, 2022 02:40 |
|
If nothing else wants to run and you never block, the OS scheduler won't ever boot your process and you won't have to wait out scheduler latencies when unblocking. uio_pci_generic maps the card's physical addresses directly into the process's memory, so there's no context switch accessing them either. As soon as you add anything that blocks to your process or have other processes that want CPU, scheduler time will pop up again. For the ns-scale jitter you see now, I'd guess that's contention in the PCIe bus controller when something else is using it. No OS code runs beyond setting up page tables at the start. When you do an access, the CPU's MMU hardware is doing address translation and then it's going out as a real read/write that's claimed by the bus controller e: also, if that card can do DMA, using uio_pci_generic with it will bypass all security on the system. Anything that can get to the device can read/write all memory in the system. Foxfire_ fucked around with this message at 03:22 on Aug 19, 2022 |
# ? Aug 19, 2022 03:19 |
|
Yeah my parallel port code is oops! all system calls which definitely doesn't help. Each I2C byte is a couple dozen ioctls and nanosleeps. Regarding security, I guess in that case giving group access to uio stuff altogether would then be a bad idea and instead I should instead setuid executables so that only approved code can be executed by users. Not that I really intend to use this stuff in a security sensitive context - I just don't want to have to sudo everything - but I guess it would be nice to follow better practices.
|
# ? Aug 19, 2022 03:34 |
|
Is there a site that matches out-of-stock ICs with in-stock equivalents? I'm putting together an order for a BOM that includes eight of these op-amps, a "JFET Input Low Noise Op-Amp", but they're not available right now. I figured there might be a drop-in replacement like NE555/LM555, but I can't find any similar products that don't have some major difference like a different height (e.g. 1.57 mm vs 4.57mm) or a different Input Offset Voltage (e.g. 10 mV vs 3mV). Do I have no choice but to wait for or somehow find this exact chip, or should I feel free to use alternatives? If the latter, how do I know my choice is appropriate? EDIT: This is for a pre-existing PCB, so I can't just use a different IC and change up the wiring or something. Pollyanna fucked around with this message at 00:36 on Aug 20, 2022 |
# ? Aug 20, 2022 00:21 |
|
I'm confused by your mention of height. Essentially any PDIP-14 package will be physically identical, so if you can find one, go for it. In terms of functional replacements, op amps are tricky. If your use case is not that critical (ie. high quality audio) then it doesn't much matter. Find another one on Mouser in the same package and read the datasheet to make sure the pinout is the same (very very carefully) then it just should work. I think the TL074 is old enough that whoever specified it wasn't looking too closely at it's input offset bias or whatever
|
# ? Aug 20, 2022 00:36 |
|
If it helps, it's for this. Currently trying to do the same for its switches which are apparently hard to find. As far as I can tell TL074CN, TL074A, TL074BCN, TL074BCNE4, all these motherfuckers are exactly the same with the one difference of input offset bias. 10mV vs 6mV and even 3mV are all so small that they might as well be identical. I'll just order 8x TL074ACNE4 and be done with it, this is a dual slope generator for chrissakes it doesn't need to be perfect. What's the difference between all these anyway. ante posted:I'm confused by your mention of height. Essentially any PDIP-14 package will be physically identical, so if you can find one, go for it. I also don't know what the gently caress, but Mouser says the TL074CN is 3 mm "shorter" than its alternatives. I'm gonna guess it's bullshit and if it's not then maybe Befaco needs to fix its poo poo IDK.
|
# ? Aug 20, 2022 01:08 |
|
Almost nothing except simple passives will ever be an exact duplicate of another part. Even things like 555 knockoffs will be different from each in various subtle ways (e.g. a TI NE555 has an output rise time of 100ns and a Analog ICM7555 has a rise time of 40ns). Most likely, the design you're looking at didn't particularly care about most op-amp parameters and most things will work, but there's no way to say for sure without knowing what exactly it's doing in the circuit. e: Datasheet looks like all of those are exactly the same package Foxfire_ fucked around with this message at 01:23 on Aug 20, 2022 |
# ? Aug 20, 2022 01:19 |
|
Then I’m just gonna use whatever. I had a similar question for resistors - the BOM calls for 1/8 watt, and I have a large chunk of them already in 1/4 watt. I want to use up what I’ve got, but it looks like 1/4 watt is significantly bigger than 1/8 watt: and I can tell that the proportions on the PCB are a little off. But they’re getting soldered to the board anyway, so is it really that much of a problem? All it really means is that the leads look a little weird underfoot.
|
# ? Aug 20, 2022 01:49 |
|
Do you have any height limitations for your parts? You can mount the 1/4W vertically in the 1/8W positions if you have headspace, it'll likely be easier than putting them on flat.
|
# ? Aug 20, 2022 02:07 |
|
You've already found a very close replacement, but I wanted to note this just in case you or anyone else is in a similar situation without an easy replacement available. Those op amps are old, decent, and pretty well understood and show up in a lot of designs but not particularly spectacular. If you know why they were selected, finding newer replacements isn't terribly difficult. As JFET op amps they have very good input impedance and input offset and input bias currents. They have a particularly wide operating voltage range. It's not rail to rail but it's close, with under 1 volt of headroom on both rails.The input offset voltage is good to average depending on the variant, the gain-bandwidth product and slew rate is moderately fast, the input capacitance is decent and the noise is alright. Looking just through TI's catalog, I found that the OPA404, available as a DIP-14, beats the TL074 by every metric except having a smaller +/- 15 volt supply range but is oddly pricy for an older part (supply chain bullshit or maybe it was always a premium part?). If you know what you can compromise on, that selection will probably grow a lot bigger. For instance, if you need an exceedingly high input impedance but don't mind having much slower GBP and slew rate, and a lower +/- 8 volt supply range, the LMC660 has an incredible 2 femtoamp (!) input bias current while being available for only a few bucks. BattleMaster fucked around with this message at 02:36 on Aug 20, 2022 |
# ? Aug 20, 2022 02:33 |
|
I’m willing to bet that the Rampage is just a relatively old design, circa 2015-2016. Probably explains why the PCB uses 3.5mm jacks you can’t find anywhere anymore.PDP-1 posted:Do you have any height limitations for your parts? You can mount the 1/4W vertically in the 1/8W positions if you have headspace, it'll likely be easier than putting them on flat. Hard to tell without getting the whole thing together. It’s actually two PCBs, one of which attaches to the other in parallel via a pin header. BattleMaster posted:You've already found a very close replacement, but I wanted to note this just in case you or anyone else is in a similar situation without an easy replacement available. Those op amps are old, decent, and pretty well understood and show up in a lot of designs but not particularly spectacular. If you know why they were selected, finding newer replacements isn't terribly difficult. What I’m learning here is that 1. it probably doesn’t matter so use whatever 2. I should just design these boards myself so I don’t have to think about this poo poo.
|
# ? Aug 20, 2022 03:45 |
|
3. It's always faster to figure out someone else's design or cheaper to buy something else than design it from scratch (this takes a few years of hubris to learn)
|
# ? Aug 20, 2022 03:51 |
|
In your shoes I would just grab the TL074ACN's if they're in stock. They're guaranteed to work and it's not worth spending time on it to save pennies per unit at best.
|
# ? Aug 20, 2022 03:54 |
|
Pollyanna posted:What I’m learning here is that 1. it probably doesn’t matter so use whatever 2. I should just design these boards myself so I don’t have to think about this poo poo. I wasn't trying to say either of those things. Unless it's an incredibly undemanding application the part was probably picked for one or more of its characteristics and finding a replacement requires matching or exceeding those desirable characteristics. If you designed it yourself you'd probably have a better idea of why the parts were picked but if not, reading any available documentation, looking at the circuit, and thinking about the specific application you can take a guess at what characteristics are desirable and pick a replacement based on that. (or just look at parts and find one that is equal or better.)
|
# ? Aug 20, 2022 03:56 |
|
Fair enough, I’m just grumpy after a whole evening of tracking down parts on Mouser
|
# ? Aug 20, 2022 03:58 |
|
Pollyanna posted:I’m willing to bet that the Rampage is just a relatively old design, circa 2015-2016. Probably explains why the PCB uses 3.5mm jacks you can’t find anywhere anymore. If you design it yourself you have to think about even more poo poo, but also yeah you can put notes on it so you or the next person has an easier time selecting alternative parts
|
# ? Aug 20, 2022 04:20 |
|
|
# ? Jun 11, 2024 01:55 |
|
ante posted:3. It's always faster to figure out someone else's design or cheaper to buy something else than design it from scratch (this takes a few years of hubris to learn) Oh I can design a much better post than this, hold on
|
# ? Aug 20, 2022 18:10 |