[Animats]

You don't have to let the spikes from the switcher get very far.

Here's the schematic for a switcher I designed.[1] This is a strange application - USB power in, 120V out, to drive an antique Teletype machine. Without any filtering, there would be huge spikes in the DC across C1-C2. But it didn't take much filtering to fix that. There's a small ferrite bead at L2, and an RC filter at the snubber at R1-C7. The back to back Zeners are to absorb inductive kickback from the output electromagnet. That's the output side. On the input side, there's more noise suppression, to prevent injecting noise back into the USB power source, which is usually a laptop here. Note L1 and C12. Those are all tiny surface mount parts, total cost in quantity maybe US$0.20.

It's an exercise in LTSpice to get the values right and make the DC power smooth DC, in both voltage and current. This is well understood.

There are radio hams using this thing, and they report it's not blithering in the RF spectrum.

[1] https://github.com/John-Nagle/ttyloopdriver/blob/master/boar...

[SV_BubbleTime]

Oooh. I clicked through your PCB there. I would highly recommend you read a couple books on circuit design. You have a 2 layer PCB with no ground plane and you aren't routing ground, signals passing parallel with each other on adjacent layers, you have mixed analog and digital domain, massively long traces where they could be shortened with a small jumpers.

Worst offender is you aren't using a ground plane or routing a return path. You might be under the impression that your signal travels on the copper you routed for the signal - it does not. It travels mostly in a magnetic field between your copper signal and the closest signal of largest difference. Which in your case is only sometimes going to be your ground trace.

Short version... I would not use this as any sort of example for RF performance, at all anywhere, ever, and I'm being a nice as possible on that. I bet if you made a quick loop with an oscilloscope it would off the charts in reality. This would never pass FCC background.

EDIT: I see this was 7 years ago, but I would not use that as an example. At a very minimum if you are still making circuits... Watch every Phil's Lab video from 1 to 100. But somewhere in 50s is a good one on stack ups and signal returns.

EDIT2: While I'm picking you apart, which you implictitly asked for, your board is HUGE. So who cares how large L1 and C12 are? On that note, I could almost not find L1 at all, the schematic is a bit of a mess. KiCad is great and now allows for global and bussed component blocks I would recommend. Again, there is a Phil's lab video on that.

[Animats]

Nothing there is going faster than about 300KHz, so the signal paths are not a big issue. If things had to go faster, there would be a proper ground plane. The tiny area around the switching IC did require careful layout (it just follows the Linear Technology data sheet) but the rest doesn't matter much.

There is no one ground. The 120VDC side has an HV ground, and the 5VDC side has a ground. The important point is to have separation between the two.

[SV_BubbleTime]

> nothing faster than 300k

That is exactly one of the misunderstandings I’m trying to warn that you are missing.

I’m certain that your switching transition times are massively faster than the 300k period (each rise and fall time).

You can have split grounds, but you don’t have traces under your grounds so in many cases you aren’t even effectively using them. You are grounding a portion of every signal inside every other signal trace.

“The rest doesn’t matter much”… my man, I can’t explain in this form just how wrong you are.

It’s one thing to be wrong on a design from years ago. My old designs were worse. But it’s another thing to willfully ignore all the ways in which someone who knows better is trying to help you with.

Everything you think about that circuit is wrong in ways you aren’t getting - yet.

Try Phil’s lab videos for EMI.

[xeonmc]

off topic, but what do you think of this video by Asianometry talking about the intricacies of analog design for a laymen audience? https://www.youtube.com/watch?v=lNypq1XuZRo

[SV_BubbleTime]

I disliked that video very much. It was FAR too long, talked about things like parasitic capacitance long before even getting near the topic of the video.

Then out of no where it's talking about tantalum nitride...

No, I'm pretty sure that video was made by an AI.