PSU problem - Murphy is an ***hole!

[Rundmaus]

Hi there,

just finished a test setup for a power supply, schematic as follows:


In PSUD2 the circuit simulates quite well, about 11mV ripple at "A" and an acceptable step response:


A simulation in LTSpice went quite weird - the ripple is still the same, but the output oscillates wildly at low frequencies about 1-5Hz:



Guess which one Mr. Murphy has chosen to be closest to reality?

Ideas anyone?

Greetings,
Andreas

[Salas]

Many times in passively filtered PSUs with tube rectification on preamp circuits especially, VLF ''pumping'' occurs on woofer speakers cones. I would take the LT Spice hint seriously.

[godfrey]

I guess you need to tame the LC resonances. In your first schematic I would try adding 200 ohms resistance in series with L22. Or maybe some resistance in series with both coils. Your second circuit shows 87 ohms with each coil. Maybe that's why it behaves better.

The other damping option is to put resistors (maybe 100 ohms?) in series with each of C21, C23, C27 and C29. That will give more ripple but less voltage drop.

[tinitus]

shouldn't there be a cap before the first choke ?

[Salas]

There is choke input persuasion also.

[Irakli]

I would not call 0.3Vp-p oscillations "Wild" for 400V power supply.

Also resonance frequency for LC filters is around 6.4 Hz. As it was already recommended, try to damp them.

[Rundmaus]

Hi there,

yes, trying to damp the resonances will probably be the next step.

Still I wonder about the strange irregular shape of the oscillation. I ran a FFT on the simulated data and it shows no predominant frequency apart from the 100Hz + harmonics. The large oscillation "above" the ripple seems to be completely random.

Would one expect that?

The difference between the two simulations seems to be the ESR of the caps - PSUD2 regards them to be 2 ohms per default. Current production electrolytics probably have milliohms. I will try adding series resistances and see what happens.

Other ideas?

Greetings from Germany,
Andreas

@godfrey: The chokes have 87Ohm DC resistance in both cases, although not stated explicitly...

[DF96]

You have a 2-cell lumped approximation to an unterminated transmission line, with a characteristic impedance of 204 ohms and cutoff around 6.5Hz. It is shorted at one end by the rectifiers and open-circuit at the other end, so will ring like a bell. Two things to try: change one of the L's or C's so they are not matched, or add a snubber across one of the caps (say, 100 ohms in series with 50-100uF?).

[Rundmaus]

Originally Posted by DF96

You have a 2-cell lumped approximation to an unterminated transmission line, with a characteristic impedance of 204 ohms and cutoff around 6.5Hz. It is shorted at one end by the rectifiers and open-circuit at the other end, so will ring like a bell.

Hi DF96,

I am not sure I fully understand your point. The other end of the filter chain was terminated by the load resistance (ca. 4,5kOhm). This is not well-matched, but should not act as an open end either.

What do you think?

Greetings,
Andreas

[DF96]

Your PSUD2 simulation shows a fixed current load, not a resistance. A fixed current load has infinite resistance, so to AC looks like an open circuit. In any case, 4.5k is almost open circuit when compared with 204ohms.

You will be getting some damping from the choke resistance and the rectifiers. I think the basic problem is that by 'traditional' valve standards the PSU has too low impedance for the applied load. This seems to be a common problem nowadays, as big caps are available. In the olden days you might have had 16uF or 32uF caps with a 10-15H choke. This would have a higher resonance frequency, but would be higher impedance so better damped by the load.