Power Supply Noise

It's important to get familiar with the audio cctry and it's performance requirements, so you can evaluate the extent of a potential threat. What's in it and what does it do (signal levels and functional blocks in the carved-in-stone linear lay-out)? Where does it get it's power, and what else is being powered by the linear section's source?

What kind of switcher? Off-line or post regulator? Strangely enough, coupling from 60Hz EMI filter components can be as severe in low level signal processing as any switching regulator interference.

Putting a power circuit on top of audio is pretty uniquely backwards - usually power circuitry gets buried due to it's relatively incompatible massiveness, thermal considerations ... and the obvious emissions threat.

Industry standard cabling conventions recognize the need to partition the I/O and power ports on the outside of any simple box assembly in a larger system (eg rack mounting). How these end up getting blended internally, at the design stage, is beyond me - but I've had the same issues plunked in my lap, after the fact.

Simply reorganizing the assigned real estate (...flipping the power assembly is an example) can add physical distance that wasn't obviously there to begin with. Adding physical plate barriers doesn't necessarily consume internal volume, either.

Count on real physical iterations and mandatory retesting of all audio functions for each, if the threat is real

RL

Shielded inductors, of course. High switching frequency, 200 KHz at least. Very tight layout loops to minimize currents making mag fields. Solid copper plane or multiple planes inside the board to block the ac magnetic fields. Ferrites/ceramic bypasses on power input.

The switching frequency won't be directly audible, but spikes can get rectified in opamp front-ends and have strange effects. Multiple switchers can beat.

Based on recent experience, be wary of synchronous switchers. Some can generate incredivle dV/dT spikes. Breadboarding might be a good idea here, to see how noise specific switchers might be.

Actually, this shouldn't be really hard with a few precautions.

John

Use shielded ones, preferably toroidal.

Yes - a ground plane mostly (but not entirely) attenuates 500kHz magnetic fields. And it will just about eliminate capacitive coupling. So try to arrange one so that it is in between the switcher transformer and the audio board.

Yes, the attenuation of a ground plane gets much better with higher frequency and it is much easier to filter any remainder.

Put the main inductors as far from the sensitive part of the audio board as possible - but keep away from the SMPS board edge since the fields will "fringe" around the edges.

Try to minimise the length of current loops that see a changing current, although you will probably find that the main problem is the switching inductor itself.

You can make a magnetic probe with a small loop shorting out the end of a piece of coax. Connect to a sensitive scope, can be set to 50 ohms. If you have a 1cm square loop and it picks up 1mV at the location of the audio board, you know the audio circuit will also pick up 1mV per square cm of any signal loop .

Similarly you can make a probe for electric fields by sticking a small piece of copper sheet or pcb material on the end of the centre conductor of a piece of coax (unshorted this time). Connect to sensitive scope (set to 1M impedance this time).

I found this quite educational, to get a "feel" for how the fields behave. The magnetic fields affect mainly low impedance circuits, the electric fields mainly high impedance.My application was not "audio" as such, but did involve low level signals in the audio range so perhaps similar.

In article , snipped-for-privacy@austininstruments.com says...>

Also the orientation of the power supply can matter. We have one that coupled into the audio transformers. Reversing it got the power supply magnetics further from the audio transformers with only a modification to the supply leads.

Right angle core to winding or or parallel winding to winding mounting of transformers can also make a lot of difference. Do it wrong and you're fluxed.

Best tip I've gotten in a month. Tried it out on a similar power supply and it became obvious where to put things in my design.

Use a switcher that is continuous conduction, not a pulse skipper. If you load is constant, there are ways to optimize the switcher. You could even spread the spectrum.