Well, I spent the better of part of today rewiring and testing my PSUs. There’s no solution (yet), but I seem to be getting closer to the cause.
I started off by taking out the 2/3 PSUs, and properly wiring the remaining one with 1.5mm^2 wire. After hooking up the scope, this is what the 12V supply rail looks like:
As can be seen, there’s a ~30mV 50Hz ripple on the supply. There’s also an opposite ripple on the negative supply. I ran some tests by plugging in a sequencer and looking/listening to the output. What should be a steady DC voltage, has that same 30mV audible ripple on it. From the tests where I have connected only 1 PSU, 1 module, and where I’m listing to the signal through headphones, I have concluded that the problem is NOT one of ground loops. In this situation there are no ground loops present, still the hum persists.
Okay, so clearly the problem is a supply ripple, not a ground loop noise. So far, I have used only 1 of the routemasters AC inputs. Getting two AC inputs with opposite phase will cut the ripple by half, right? So, as my next step, I replaced my single-secondary transformer with a centre-tapped one. After wiring that up, this my supply rail:
Oh god, it got worse. Now the sample ripple is there, it just doubled in frequency. Well this is just very, very odd. Let’s look at the input of the regulators:
The problem lies not in a sagging input of the regulator. Instead, it’s somehow unable to deal with the peak of the input. Is it just too slow to respond to the transient here? I was very surprised to find this.
Just to verify, I wired up my previous PSU, a FC microbus, and found a fairly clean supply with ~2mV ripply. No audible hum on the DC output.
So it seems the problems lies within the regulators, or the surrounding components. I will go check my boards and components, maybe see if I can swap out the regulators. These are the exact ones I used:
From a quick look at the datasheets, I can’t see any reason why these would not be suitable. If I’m overlooking something, please tell me.
So yeah, no solution yet, but I seem to have found the problem.