What's a reasonable amount of power-supply ripple?

[bst]

I'm in the very early stages of my first schematic design. It's a somewhat wacky idea to make use of all the orphan tubes I've accumulated with dual top caps. The rough outline comprises a single power supply, rectified by a 1641, feeding input/driver/splitter circuitry populated with 7193s, and a push-pull output stage with a single RK-34 in each channel, configured as a hum-bucking common-cathode.

It's all still a very rough sketch, but I've been playing with TubeCad and PSUD II to see if this project is actually feasible. Now to my specific question(s):

The ideal power supply would be a pure DC battery supplying unlimited current at an unvarying voltage. In the real world, however, I've seen commercial tube circuits from the 1950s and 1960s which have seemingly huge amounts of ac ripple on their B+ supplies. I've seen blanket statements by current tube designers that power supplies must not have more than a few millivolts (or even microvolts!) riding on the DC. Yet I've listened to some very clean-sounding older tube amps which turn out to have as much as two or three volts of AC imposed on the HV supply.

I've roughed out a simulation in PSUD II which gives me about 120mV of ripple on the push-pull output stage supply, and less than 1mV for the input and driver stages (screenshot below). Is this a workable configuration, or should I go back to the drawing board?

[Eli Duttman]

A PP amp can work well with relatively high ripple levels. OTOH, a SE amp needs VERY well filtered B+.

[DigitalJunkie]

In PP output stages,the ripple will cancel-out in the output transformer. Obviously lower ripple is a noble thing to shoot for,but I'd imagine that your numbers are probably pretty suitable.

[Yvesm]

When designing an whole amplifier you should consider how different functions will match / react together.
Before working at the PSU, it would be a good idea to check PSRR (power supply ripple rejection) of each stage to feed.

As was already said, a PP can accept considerable ripple (may be up to 5%) while not producing any hum the loudspeaker.
More: The gain of a pentode PP is amost unafected by variations / noise of may be 30% of the B+ applied to its anodes ... providing the screen grids and bias both remain stable and clean.

A contrario, a SE scheme with a triode does not accept more than a few millivolts before becoming noisy.

In short, acceptable noise from PSU depends on the topology you're planning to use and unless the values you used in your example were choosen according to the contain of your junk box, they are somewhat ... mmmh ... discutable, like f.e. the 225 Ohms choke. and the caps with so high ESR.

Yves.

[bst]

Yves,

Yes, this is a pure 'junk-box' amp. The power transformer is an old Stancor PC8412 purchased at a surplus store, the choke is an unidentified unit rescued from a piece of ancient test equipment, and all the capacitors are oil-filled motor-run types from yet another surplus dealer. The transformer and choke resistance are measured values, but the cap resistance was a 'worst-case' number I pulled out of the air.

TubeCad shows a PSRR for the output stage of almost 29 dB, and some of the possible input and driver configurations yield PSRRs of well over 30dB. I'll have to start refining my measurements and my models, but the preliminary results look promising enough to start breadboarding fairly soon.

[SY]

What Yves said. Start with your requirements (how much noise on the output), then see how that translates into allowable ripple for each stage. There is no one answer; if you want -100dB from a cascode, the requirements are far different than if you want -80dB from a push-pull grounded cathode stage.