Can I double up the voltage doubler caps?

[Deke609]

Can I put two parallel 220uF/350V caps in place of each single 220uF/350V voltage doubler caps?  It looks like a cheap and easy way of cutting ripple voltage in half.

The only issue I can foresee is that it increases the initial current draw through the doubler. PSUD shows a roughly 40% increase in current at the initial spike that peaks at 3 ms after turn on.  But thereafter, the parallel doubler configuration appears to behave identically to the stock configuration. I figure two caps in parallel can easily handle an 40% increase over the current through a single cap.  Perhaps I'd need a higher current rated slo-blo fuse?  Are there other considerations or potential problems?

cheers and MTIA, Derek

[Paul Birkeland]

I mean, you could.  Can you hear the power supply ripple? 

[Deke609]

Nope - or at least I can't hear any hum.  Might it make a subtle audible difference? Dunno.  But since it's easy and cheap, why not try to get the stuff that you don't want to produce signal to be as quiet as reasonably possible?  For my Kaiju rebuild, I will be experimenting with a solid state HV regulator [edit: so that the B+ feeding the output stage is also regulated]. The designer reported a max of 300 uV output noise, but almost all of that was residual ripple passed through by the reg.  Looking at his output noise plots, by my eyeball estimate the intrinsic noise output of the reg, when mains ripple is removed, is about 3 uV.  So I am aiming to have minuscule ripple going into the reg via a CLCLC filter comprising doubler, 10H/200mA choke, 200 uF high ripple voltage rated film cap (huge! 25 cm long!), 10H/200mA choke, 1500 uF film cap (also huge, but mostly diameter-wise). Overkill? Absolutely.  Fun factor: off the charts.  Plus, I think the giant aluminum can film caps will look cool on an all aluminum chassis alongside the new big black MM output transformers.

cheers and thanks, Derek

[Paul Birkeland]

Nope - or at least I can't hear any hum.  Might it make a subtle audible difference? Dunno.  But since it's easy and cheap, why not try to get the stuff that you don't want to produce signal to be as quiet as reasonably possible?

Mostly because a 300B is only so quiet, so if you chase the power supply ripple down to 1 pico volt, but the noise floor of the 300B itself is 10,000 times higher, what have you accomplished?

[Deke609]

Fair point. Hadn't considered the intrinsic noise of the tubes. It's still fun though.  And I wonder whether some 300B's are quieter than others? I use EML 300Bs, but also have JJ's and the stock Sovteks. I only use the Sovteks for testing new builds/mods, and the JJ's have just sat in a drawer since I got the EMLs.  I'll have to try the amp with all 3 types.  Just for fun.

cheers, Derek

[Paul Birkeland]

But also consider how much noise you can hear.  From my experience, anything below about 200uV RMS is going to be tough to hear.

[Deke609]

That's a helpful data point to have. Thanks. In my case, super low voltage ripple going into the reg should help me test whether any audible differences of the reg are truly the result of regulation and not just (or in part) ripple rejection.  It's hard to tell from the PSUD plot grids which only go down to 1 mV, but I think the new CLCLC filter should get me down in 10s of uV ripple, maybe even less. So I am confident that I can rule out ripple as a factor when I compare the CLCLC alone against the CLCLC with HV reg.

cheers, Derek

[Paul Birkeland]

PSUD will show you 1mV peak-to-peak.

In the output stage, the power supply noise is divided across the parallel feed output stage, then stepped down through the output transformer according to the impedance you have set at the output.

[Deke609]

In the output stage, the power supply noise is divided across the parallel feed output stage  ...

Trying to understand this. Do you mean that the impedance of the parafeed cap in series with the primary of the OPT forms a voltage divider? THat makes sens to me, but I don't see how it reduce noise without also reducing (intended) signal.

then stepped down through the output transformer according to the impedance you have set at the output.

Oh, right! That's a huge step-down. But at this point any noise is part of the signal, right? If so, I would have thought that the OPT isn't doing any noise reduction separate and apart from the step down of the signal. The entire signal/noise mix is stepped down in proportion to primary:secondary ratio.

Doesn't any noise reduction need to take place before B+ reaches the plate of the output tube (in convectional flow terms).

cheers and thanks, Derek

[Paul Birkeland]

Trying to understand this. Do you mean that the impedance of the parafeed cap in series with the primary of the OPT forms a voltage divider? THat makes sens to me, but I don't see how it reduce noise without also reducing (intended) signal.

Is the signal present on the B+ line or the plate of the tube?

Oh, right! That's a huge step-down. But at this point any noise is part of the signal, right? If so, I would have thought that the OPT isn't doing any noise reduction separate and apart from the step down of the signal. The entire signal/noise mix is stepped down in proportion to primary:secondary ratio.

Sure, a few mV of noise in the presence of hundreds of signal volts.

[Deke609]

Is the signal present on the B+ line or the plate of the tube?

I would have thought both. Isn't the plate just the endpoint of the B+ wiring?  So any variance away from true DC in the B+ line appears at the plate as a varying voltage potential (noise signal).  And when music signal passes to the plate from the cathode according to grid voltage changes, the music signal and noise signal sum.  [Edit: or rather, the variances in plate voltage cause variances in the music signal appearing at the plate] Do I have this wrong?

cheers and thanks, Derek

[Paul Birkeland]

I would have thought both. Isn't the plate just the endpoint of the B+ wiring? 

When we refer to "B+", it's presumed that there is a big ass cap or some kind of regulator there to present a very low impedance to AC.

So any variance away from true DC in the B+ line appears at the plate as a varying voltage potential (noise signal).

Think about the loop that this noise current needs to follow, then you'll see how a parallel feed output stage has some advantages in terms of noise reduction and you can start to quantify the divider values at 120Hz. 

[Deke609]

OK. Caught my mistake (or at least one of them). For the last while I've only been thinking of capacitors in terms of reactance and so started to think of them only as frequency dependent resistors, forgetting that, unlike resistors, they absorb and release charge when put in parallel with the load - and in doing so effectively regulate voltage applied to the load.

Many thanks again PB for responding to my misunderstanding-fueled questions.

cheers, Derek