Optimizing RC time constants in the filter cap supply

[diagrammatiks]

Hi, I searched a bit and didn't seem to find much information concerning the optimization of recharge time in filter supply circuits.

I'll try to summarize what I've figured out.
- RC filters are used in the B+ supply to filter out ripple current.
- RC filters also decouple preamp stages from each other in order to prevent cross-talk, feedback, and oscillation.
- Under-filtering is bad from a "hi-fi" design stand point but contributes to the sound of vintage amps due to increased sag and looser bass response.
- Over-filtering can lead to blocking if the pulse current required for recharge is too large. Does this only apply to the first filter cap? I know it's especially dangerous when used with vacuum diodes. Can also lead to tighter bass, more "sterile" sound.
- Perfect filtering - where is this magical place?

Alright so I know valve wizard has the smoothing article up on his website that gives the calculations for an ideal filter cap size depending on the parameters of the decoupled stage...

However, it seems that the ideal filter cap size does not necessitate the "perfect" RC circuit for that particular stage.

Two questions come to mind here.
1. What are the limitations on the maximum number of stages that can be coupled together? It seems that the average or standard is 2 stages per rc ripple filter. Some amps couple up to 4 triode stages, although these stages seldom consist only of voltage amp stages.
2. Is there a necessary lower limit to the number of stages that must share a single RC ripple filter? For instance, I haven't seen a schematic that decouples 2 triodes from the same tube. Is this because a single dual triode such as the 12ax7 must be run a single B+ voltage?

Of course given the number of under filtered circuits that have been produced it seems that under filtering is generally not a technically worrying problem.

What if we went in the other direction...not just in the size of filtering but also in the number of ripple filter stages.

Suppose, for instance, that we have an amp with 4 triode stages derived from 2 12ax7s into a LTP derived from another 12ax7. What would be the most efficient way to filter these stages in terms of optimizing the time constant?

In Morgan Jones, "Valve Amplifiers", he talks about the utilization of series LC filters to take the place of 1 LC filter at the start of the B+. He states that by utilizing 2 LC filters in series, both filters could be smaller then the single filter need. The first would be slightly smaller, and the second would be several orders of magnitude smaller. Each dropping resistor between the chokes could also be reduced respectively. Jones seems to state that doing so would not only increase the ripple filtering capabilities of the input, but also reduce the time needed for a full recharge as well as lower the pulse intensity.

The question is...would this work for RC ripple currents as well?

Referring to the hypothetical preamp outlined above...suppose that...

The first tube is filtered by a 22uf cap and a 10k resistor. The second tube is filtered by a 22uf cap and a 10k resistor. The third tube is again filtered by a 22uf cap and a 10k resistor.

Is there a way of splitting up some of these filters to improve the efficiency of the supply?

For instance, could the first tube utilize a 20uf cap and an 8k resistor in series with a 2uf cap and a 2k resistor? Could the stage be connected to a series of 10 2uf/1k resistors in series?

What would be math necessary to figure out the compound RC time constant so that each stage could have optimized recharge time?

Thanks!

TLR - Rc time constants? compound decoupling? what maths? pleas halps. thnx.

Thanks!

[Ken Moon]

In Morgan Jones, "Valve Amplifiers", he talks about the utilization of series LC filters to take the place of 1 LC filter at the start of the B+. He states that by utilizing 2 LC filters in series, both filters could be smaller then the single filter need. The first would be slightly smaller, and the second would be several orders of magnitude smaller. Each dropping resistor between the chokes could also be reduced respectively. Jones seems to state that doing so would not only increase the ripple filtering capabilities of the input, but also reduce the time needed for a full recharge as well as lower the pulse intensity.

The question is...would this work for RC ripple currents as well?

This sounds like the CLCLC filter used on the Talon amp. It uses a 10uF reservoir cap, then a 9H choke, 40uF cap to ground, another 9H choke, and another 40uF cap to ground.

The idea is that the reservoir cap can be very small (I've used as low as 3.3uF), and this keeps the current pulses in that super-important (noise-wise) first current loop much smaller, and the recharge time faster, while the second LC filter prior to the power tube plates filters out most of the remaining ripple.

You can do the same thing with a CRCRC filter before the plates, but you'll have a much lower B+ at the plates than with the CLCLC (of course you can start with a higher HT to get around this).

Or, take a look at the thread I posted about making CLCLC-equivalent filters using cap multipliers - this looks like an interesting option to me.

[Steve Conner]

The resistors in the RC preamp filters have a second function: they drop voltage to get the preamp tubes down to their datasheet operating conditions. The resistor values are determined by the main B+ voltage and the required B+ for each stage, so the only variables are the number of RC stages and the capacitor size.

Yes, what Morgan Jones mentioned works for RC filters too, he actually covers the RC case in the book and gives design equations for it. It is a commonly used trick, but slightly disguised. Maybe the PI runs off the first filtered node, then the second RC filter stage is fed from the already filtered PI supply and so on. The first preamp tube is at the end of a 3-stage RCRCRC, which is the right thing as it's the most sensitive.

If the RC time constants are too short, the amp can motorboat because of low frequency feedback through the supply rail. You have accidentally reproduced the old Fender tremolo oscillator. I use something like 27 or 56k and 10uF most of the time. But be aware that motorboating can also be caused by a badly designed NFB loop in the power amp, or that can take it close to instability and bad supply rail decoupling gives it the extra push over the cliff.

If you want two triodes to share a supply node, then make it two consecutive stages. That way, any feedback through the rail is negative, so there is no risk of oscillation, at least of that stage. With two non-consecutive stages, or three stages, the feedback can be positive. Screw up your eyes, look at it sideways, and if you can see something that looks like the tremolo oscillator, go back to the drawing board.