Valve Amp Power Supplies

[Cressy Snr]

Prompted by Simon's questions in Paul's Bates amp thread, I thought I'd start something in the beginners section because all this power supply stuff has left me wanting a better understanding.

Now the power supply below is the one I'm using in my 6V6 SE, E-Linear amp at the moment. It represents everything I know about power supplies Stop laughing!

Anyway what I show below is the quietest power supply I have ever built. It is completely silent...nothing, no hiss no hum, no mechanical noise, not even at switch-on. I mean what I say. You can physically sense the silence...literally. It allows the amplifier itself to sound superb.

Perhaps those who know a lot about these things could share their knowledge, post and illustrate how this supply could be developed from what it is below into something that is true state-of-the art. I for one would be eternally grateful.

So..what's wrong with this picture?

Screenshot 2022-01-23 at 16.58.31.png (63.7 KiB) Viewed 3068 times

[Nick]

What's missing is some description of the load. To the load the power supply will look like a 220uf cap. Once that cap starts to have the energy it contains reduced it will start to look like a voltage source with a 230r resistor in series. The AC voltage across the last cap will be a combination of a two stage RC filter of a 100hz as of a signal combined with the modulating effect of the load driven by the music signal. The transient response to a stepped load will be the result of the two rc stages.

I suspect though it doesn't show it that it's also supplying something earlier on before the RC stages in which case it gets more complex.

[Cressy Snr]

OK,
The output of the supply is 320V @ 86mA at idle, 43mA coming out of each branch of the HT. Because the supply feeds an e-linear amp with the load resistors to the input stages doubling as feedback resistors, then the HT is feeding both stages from the same point.
This means that there is nothing being supplied from anywhere except the two ends (HT right) (HT left)

[Nick]

Ok in that case the 440 ohm would make me think the supply will sag on load at lower frequencies. Spice would be the simple way to model the behaviour.

[Cressy Snr]

So RC filters can cause sag under load?

[Cressy Snr]

I ask this because Morgan Jones cites Scroggie (Cathode Ray) when talking about multi section power supply filters to drop noise way down. I mean why is there no consensus among engineers about this sort of thing. Yet multi section RC filters fook up the bass if that’s the case.

[Nick]

Of course once the cap starts to have a higher inductance than the 220r the 220r becomes the important part.

So at low frequencies or consistent power output the r becomes the source of power. It's not too bad. Let's say 5w. 5/350 is 16ma. Assume maybe 25% efficiency so that's 64ma of current. So that's 440*.064 say 30v so maybe 8% droop.

[Nick]

There is consistency. Maybe you are asking the wrong engineers. RC stages are fine for low current they work just fine. It's using them in a power stage that becomes a problem.

[Cressy Snr]

There is consistency. Maybe you are asking the wrong engineers. RC stages are fine for low current they work just fine. It's using them in a power stage that becomes a problem.

Ah!
You see that’s where the confusion was for me.
Power stages then: is it the case that resistors can cause voltage sag and big caps can slug the transient response?
Hmmm, maybe I need to back off on the filtration, go back to a basic CLC and take it from there.

[Cressy Snr]

A bit of feverish action later and I’ve reduced substantially, the amount of capacitance in the supply. The treble has speeded up, dynamics are sharper. The whole presentation is fresher and there is still no hum. Why I fill my PSUs with bloody capacitance, I don’t know. I mean it’s single-ended, not a class B push-pull. This thread is giving me one or two lightbulb moments.

Now I have 40u/4H/40u/ then splitting to a 220R/50u RC filter per channel. OK there are still resistors in there but I can hear in the sound that it is a step in the right direction.

[Nick]

Out of interest try shorting the 220r I can't see the point of any resistance you don't need in the supply to the output stage.

[simon]

Out of interest Steve, are you modelling these PSes in PSUD? It would be interesting to see what the model predicts compared to what you hear.

[Cressy Snr]

Out of interest try shorting the 220r I can't see the point of any resistance you don't need in the supply to the output stage.

I’ll have a look at that tomorrow Nick. Now you mention it there doesn’t seem to be any need for resistance at that point in the supply.

Out of interest Steve, are you modelling these PSes in PSUD? It would be interesting to see what the model predicts compared to what you hear.

At the moment I’m working empirically but I’ll get the supplies into PSUD: again tomorrow and see what gives there.

[pre65]

I must admit to liking PSUD, and if I can use it anyone can.

Just remember that accurate info in = accurate info out.

Some things, like capacitor ESR can be difficult to find though.

[Nick]

I cant think of a case where ESR is going to matter in a valve circuit. And TBH, if you need more that 5% accuracy then I would suggest you are designing too close to the edge.

Where PSUD can be useful is investigating the dynamic behavior of LC stages and getting a nice transient response. But it won't tell you what a good one is, you will have to decide that yourself by experiment.