Platform forum

Grounding, noise, clock signals...

mhx , 07-25-2020, 08:15 AM

This is my first post here and I hope I'm not asking too much. I'm an electrical engineer, but have spent most of my life writing software (from embedded stuff up to services spanning thousands of machines). I could always hack up some (mostly digital) electronics, but never did much analog or mixed-signal stuff. I've never done any PCBs up until a few months ago. I'm now working on filling these gaps and one project I'm doing is a low noise, split rail (+/- 15V) power supply using SEPIC converters and post-regulating their outputs with linear regulators.

I did a few SPICE simulations of the circuit itself, and I'm now working on the second version of the board after identifying several issues with the first version. I've watched almost all of Robert's absolutely awesome videos, and with every video, I've found so many things to improve on my layout.

The thing is, just like with reading books, if you're applying the freshly learnt knowledge to your own project, there's always this moment when you stumble across questions that haven't been answered. Or, you don't even realize that there could potentially be an issue.

So, with my power supply project, I've got a few concrete questions:
  1. Clock distribution. One of the problems with the first version was that the DC/DC converters were running asynchronously and were interfering with each other, causing instabilities. I've verified that the first version runs fine if I provide the DC/DC converters with an external clock from my signal generator. Now the new version has a clock generator on board and I need to route the clock signal to the DC/DC ICs. As it's a two-layer board, the clock needs to run on one of the outer layers. I've tried as much as possible to avoid layer changes and keep the traces short. For the 5V output on the right of the board, I preferred a slightly longer track over a layer change, for the +/15V outputs, I had to jump to the bottom layer very briefly. My questions:
    1. Does my layout look reasonable?
    2. Are there reasons why you would probably have done it differently, and how?
  2. Grounding/noise. I obviously want the +/-15V outputs to be as noise free as possible. I've tried to follow the application notes for the LT3045/LT3094 with regards to guarding the SET pins, but that's surely just one of many ways that noise can get to the output. I've also tried to keep the DC/DC ICs and the clock generation as far physically away as possible. The thing that I still find it hard to fully wrap my head around is how to ground everything properly. I started off with a solid ground plane, but then thought that it's probably a bad idea to have the return currents from the linear regulators flow across the DC/DC or clock sections. So what I did now was to have a separate ground area for the linear regulators and connect both ground areas together at the VIN header. My questions:
    1. Again, is my thinking correct here?
    2. Would it make sense to further separate the grounds for both linear regulators?
Last but not least, I'd very much appreciate any feedback on the layout. It feels to me like this is the best I'm currently able to do with my very limited experience, but I'm sure there's a ton of things that could actually be improved.

The full project is here: https://easyeda.com/mhx77/smps-test-v2

The PCB (directly linked above) is "v2-work-in-progress".

Many thanks,
robertferanec , 07-27-2020, 02:33 AM
@mhx that is a very specific project. Maybe someone with a lot of experience in power supply design can help?
mhx , 07-27-2020, 03:50 AM
@robertferanec thanks for getting back. I know this is quite niche and I'll probably just have to trial-and-error my way through. 😄
qdrives , 07-30-2020, 12:43 PM
Hello @mhx I had a quick look at your design.
A few simple comment to start with:
1) If you want low noise, do it on a 4 or more layer board. Layer 2 is only ground
2) Do not route track below inductors, beads and transformers.
3) Use via stiching all over to connect grounds.
mhx , 07-31-2020, 08:57 AM
Thank you @qdrives! Doing a 4 layer board (S-G-P-S) is definitely something I want to try going forward. 2) makes sense and should be much easier to avoid when going 4-layer. 3) I've already tried to do that around the linear regulators, but I'll make a note to do even more vias in the future. 😁
Use our interactive Discord forum to reply or ask new questions.
Discord invite
Discord forum link (after invitation)

Didn't find what you were looking for?