Bypassing B+ in PP amp

[oshifis]

I think the variation of B+ due to change of load in a push-pull amplifier is common-mode, so it has less effect than, say, on a SE. My question is: how important is bypassing the plate voltage filter electrolytic capacitor with a smaller foil capacitor? Will it improve anything? What is the recommeded value and type?

[SY]

Yes, it's worthwhile. Unless your amp is strictly in class A, the total required current will change on a dynamic basis.

[Shoog]

I found that even in pure class A it made a significant difference to the overall sound. More top end detail.

Shoog

[oshifis]

I have 4 x 220uF / 450V at 415V B+, running triode-connected 807s (12.5 W output power @ 1kHz/1%, no NFB). I have some 47nF Russian teflon capacitors around, or some WIMA 1uF/630V polystyrenes. Will any of them improve the sound, or should I go for some polypropylenes? What value, 100nF, 1uF ?

[Shoog]

I advise that any of those should deliver so just swap them around and see what you think. Start with the teflons.


Shoog

[richwalters]

Having just completed the Leviathan p-p amp of the century, this project used output trannies from a different vendor, custom based on an 18 sect Williamson design. So now I have two monster amps of sim design using same tube types but different output trannies to compare, and is noticeable on audible tests.
The Sowter transformer sounds smoother and the Majestic definitely brighter with better imaging. Initially, I blamed the KT88's, til I swapped for 6550 and no difference. Both amps have sim upper freq resp and both UL output stages require snubbing to clean up ripple to create a sensible looking square wave. The Sowter tranny required harder snubbing due to 14 sect design and although responded well at 10Khz, waveform slope slewing is high. So leakage Cap within windings is higher than Majestic tranny.
Snubbing consumes power, and at say 10Khz the tubes will require more drive current to deal with the increased reactance of parasitic components at that frequency. So in effect with all the reactances the equivalent impedance of the output stage drops at high audio freq forcing more of that freq ripple appear on the B+. This is confirmed by THD measurements for p-p amps at 10KHz at the same power for 1Khz is roughly 10-15X the thd figure for 1Khz.The ESR of the smoothing cap has to deal with the harmonics and figures begin to stack up.
I regulary use 1uF low loss polyprop from ICW which I am quite sure plays quite an important part in the sound definition. The arguement I make is valid especially when parallel p-p stages are used, as when power tips into class B, peak currents get high, i.e
100W parallel UL p-p in AB can reach over 0.5A so cap ESR becomes important.
Those members with thd analysers, graphical FFT methods or even the HP 331A can easily illustrate the complex harmonic problem.

Those might say there ain't much music around at 10Khz. Yes there is. To reproduce a 10Khz square wave,(trumpet and other blast instruments can produce quite steep sided harmonics) the amp must have a passband of min 30Khz and things get worse higher. So again we are back to where we have started, the quality of the output transformer. With p-p, without decent amp loop gain and global nfb, the fidelity war is completely lost. Some members have successfully used rogue mains transformers and others for output stages, but in reality there is no consistence with leakage parasitics, hardly capable of being optimised for fidelity, so the bandwith is poor.

richj

[Shoog]

I am encountering similar issues with my current amp. I was getting very sharp overshoot on the leading edge of a square wave at all frequencies. This also showed up as frequency peaking at about 50khz. This is all using main toroidal's as outputs. I have introduced a primary side zobel of 1.5Kohm and 0.047uF which has helped a huge amount. Of course you are right that zobel effect bandwidth - but I'm happy that I still have good response past 30khz.
That sort of spiking is very audible and not very pleasant to boot.

In the case of mains toroidals its all about parasitic capacitance and nothing to do with leakage inductance. I have used the same transformers as outputs on other projects - but this amp is more of a current drive amp and I think this is what is causing the overshoot combined with the interwinding capacitance.


Shoog