Switched power supplies are notorious for causing EMI; why didn't he test for it? Frankly, it sounds like he just copied the circuit from a datasheet and called it a day.
How would you suppose he tests for EMI? He says he is using mostly power supply modules which will have some thought already put into EMI suppression. Beyond that, if the system works why should he worry about EMI? To take any quantitative measurements would require a chamber and calibrated instruments. This is akin to asking why a hobby software project doesn't have full coverage unit testing.
>He says he is using mostly power supply modules which will have some thought already put into EMI suppression.
He is not. MAX776 are controllers, and older ones at that. These require some skill to apply effectively. I wouldn't do it this way, I'd look at the LMR36xxx series (and friends, you'd want the 60V switches here) sync bucks for this. Hot loop on die = goodbye EMI!
If a software engineer said: this Rust code leaks a bit, but it shouldn't be too bad, you only have to restart it once a day, then what would you think of it?
so glad to hear none of the code you’re working on has ever leaked, even while it’s in development and being worked on. what the hell is wrong with you?
The point is not whether my code has leaked or not, but whether I was even aware that leakages are a possibility. I'd be happy if someone pointed it out, even if not entirely in the correct tone.
You’re pointing out an extremely common issue with switch mode power supplies. It’s so common that it comes off like you’re calling the author an idiot. It’s not helpful —- you’re 100% not at all being helpful.
The way to be helpful would be to LOOK at the layout, component choices, and design, and suggest changes, but you’ve done nothing of the sort.
> This is the initial commit, revision 1, and has a few issues, but it works and the computer boots! It will no doubt be obvious that I am not an engineer, and I certainly don't specialise in SMPS circuit design. I have used modular power options where possible and have tried to follow the recommended design for the -22V SMPS circuit as closely as possible. Please feel free to improve on this design. [emphasis mine]
More critical than EMI, methinks, is something he takes pains to point out:
> There is NO output protection (I'm not sure if there ever will be), so if you build one of these then make sure to check the voltage rails before using it.
Sounds like a nice, humble attitude. Despite the humility, the end result is a project that meets his needs and provides useful details for others to use or build upon.
> Switched power supplies are notorious for causing EMI; why didn't he test for it?
The necessary equipment - anechoic chamber, proper antenna and scope - goes into the six figures. Way out of reach for hobbyists and most if not all maker associations.
Designing switched mode supplies is hard. I've just barely started playing with them, but it's really tough choosing component values that don't ring like a bell, a loud bell, at pretty high frequencies. I've found these two series of web pages [0] [1] to be very useful, even for an electronics neophyte such as myself. Also, for simulations Xyce [2] handles the large current transients much better than Ngspice [3] (the default simulator for KiCad), and the Xyce docs gave me some really good guidance on how to tweak the settings.
One trick I stumbled on was to run the input voltage as high as feasable; for the part I'm trying to use the source/drain effective source/drain capacitance drops dramatically as the voltage goes up. It's the MOSFET capacitances that are causing me the most grief.
And testing for EMI seems to be a black art. Hack-a-Day has listed some do-it-yourself projects for probes and testing ([4] [5] ...), but haven't gotten that far yet.
>it's really tough choosing component values that don't ring like a bell, a loud bell, at pretty high frequencies
That doesn't sound right to me. It's not really about the component values. It's about the parasitics and the edge rates that can ring those parasitics. The main LC elements are not usually the problem (especially at high frequencies).
> for simulations
You cannot simulate any SMPS for EMC reasons without including the layout. Full stop, do not pass Go, etc. You can simulate for functionality and you can avoid some well-known traps, but you'll never be able to be confident without the layout included. Never.
> run the input voltage as high as feasable
Usually you want your duty cycle around 50%. Higher input voltages will push that down, which is usually a bad thing and requires special measures to compensate. And switch capacitances are rarely a major problem below line voltages, so you don't get much return on it either.
What, exactly, are you trying to do?
(Disclaimer: professional EE but not a power guy. I do little SMPS stuff up to 24V all the time though, especially low-noise analog supplies.)
I'm just getting into analog electronics. Nominally, I'm trying to synthesize 90VAC @ 30mA (telephone ring signal) using buck/boost from a "reasonable" input voltage with reasonable efficiency. Not anywhere close to worrying about interference, but seeing simulated 3A swings at 1MHz is not encouraging... A big chunk of that is due to gate/src & src/drain capacitance. I'm finding that jacking up the inductance leads to really high average currents, cutting the inductance means really small damping resistors and high current again, shrinking the output capacitance helps the ringing and quickly gets the voltage where I need it, but now the ripple is horrible... fun times!
