Big high uf!

[R.G.]

[...] Really, I think power supplies are WAY more complicated and subtle than most of us give them credit for. In particular, chokes seem like a good idea at first, but if you're not careful you can end up with a PS that rings at some audible frequency. Since the ringing is not related to the music, it comes through as, well, non-musical. And, in most cases, nobody ever sees the ringing when they hook up a scope because we look at the B+ while the amp is idling, or maybe with a steady input signal, but we don't excite the ringing while we're looking, so we don't even know that it's there.

Anyway, the choke leads to a supply that does not respond quickly to changes in load and actually leads to an unclean sound. A resistor works better. A resistor and choke in series can work, but you need to tune the series resistance to get the damping right. When have you ever heard of anyone doing that?

That used to require paper-and-pencil math, knowing equations, and such. I got extruded through several semesters of math to do just that kind of thing. And I still prefer circuit simulators. They could make me do the equations, they couldn't make me like it.

One thing that stuck with me from designing switching power supplies is that any time you have an inductor and a capacitor, you have the potential for ringing. And since all capacitors have some inductance, they all have a self-resonant frequency (that's the minimum-impedance point for every cap, where Xl and Xc are equal, and it just looks like ESR). Again, the RF guys are hyper sensitive to this issue. They have to worry about the difference in self-inductance between the same copper area being round versus flattened rectangular.

So we always have the problem. What do we do about it? We could discuss it to death until we feel better, I guess. I just feel better with some numbers, even numbers picked to reasonably bound the problem if I don't know the exact inductances and capacitances. I can whip 20H and 47uF into a circuit simulator and get back a plot of its resonance, and then put resistor values in series (and parallel!) to damp the resonance. I can even whip in an estimated load, which also damps an LC resonance. Designing damping to keep the output filter from ringing in the face of a varying load/damping was touchy at times.

I can also step sideways, and put simple regulators on various stages. This actually works pretty well.

One way to see what's happening is to, as you note, hook up a scope to the power supply nodes and run a sine sweep into the inputs, looking for any sign of increased AC at the nodes, and - with modern scopes - run the FFT and see where spikes rise that are non-harmonically related to the power line. This used to be the stuff of dreams or well-equipped corporate labs, but my $250 Picoscope does it pretty well.

I always liked numbers better than "here be dragons". Personal limitation, I know.