We do not know that isolator solves it. You are assuming that. I think it is possible that modulated RF is either going through the signal line or picked up by the bare wires.l and demodulated in the amp to produce audio.

Note that I used solves in italic to note ground/common isolation is part of well known countermeasures (power splits, shielding, filtering, bypassing, balancing, etc ...) to the digital noise transmission issue. Sure, modulated EM is moving from the RPi to the amp and it's using the common power loop wires as the lower loss transmission line path for common/differential mode conducted (surrounds and follows the conductors) noise. Breaking the loop with a ground isolation or separate supplies moves the noise path to the higher loss air-gap radiated energy path.

https://www.tdk.com/en/tech-mag/noise/02

Noise, a form of electromagnetic energy, can be divided into conducted noise and radiated noise, depending on how they are transmitted. Conducted noise is noise that is transmitted along with signals through power supply lines, signal lines and trace patterns on printed circuit boards. Radiated noise is noise that travels through space and arrives as unwanted electromagnetic waves.
Conducted noise has a relatively clear path of intrusion, but it is difficult to identify because it appears similar to signals. In general, electrical signals are transmitted as patterns of voltage change, and noise becomes part of those patterns as it piggybacks on the signals.

 

Then that is the obvious solution, as we have already suggested.

Sorry, I did say in one of my replies that I need to run both from the same PSU, but I should've put that in the original post! I thought it was obvious (from the fact that I'd clearly already solved my problem if 2 PSUs were fine!).

 

Sorry, I did say in one of my replies that I need to run both from the same PSU, but I should've put that in the original post! I thought it was obvious (from the fact that I'd clearly already solved my problem if 2 PSUs were fine!).

IMO split power supplies is a superior solution but you can try filtering between devices. YMMV
https://www.murata.com/~/media/webrenewal/products/emc/emifil/knowhow/26to30.ashx

 

IMO split power supplies is a superior solution but you can try filtering between devices. YMMV
https://www.murata.com/~/media/webrenewal/products/emc/emifil/knowhow/26to30.ashx
View attachment 325266

This youtube video is exactly the same issue - sounds like he's coming to the same conclusions!

 

This youtube video is exactly the same issue - sounds like he's coming to the same conclusions!

Shocking!

As you can see the RPi design is well known for being problematic (with USB power related coupling being a common issue that may have been improved with the RPi4 and later revisions) IRT to circuit noise pollution when dealing with audio or evel low level DAQ measurements. The RPi, it all its forms, was not designed with low electrical noise in mind. It's a electrically noisy small general purpose computer that requires noise countermeasures (sometime several) for sensitive measurements.

I played around with the RPi a few years ago building and testing some Linux kernel driver hardware and software. You can get decent noise spec's with a little practical engineering to isolate and filter out RPi created noise.
https://forum.allaboutcircuits.com/threads/super-moon-shine.100322/post-898360
https://forum.allaboutcircuits.com/threads/super-moon-shine.100322/post-904573

 

I'm curious what may suggestion in #14 would reveal.