motorboating ?

[shooter]

Took this from a resent post;

If one is too small, there may be coupling through the power supply of like-phase signals. That leads to oscillation.

(by HBP)

Is there a *rule of thumb* value?, or maybe a ratio Cap to R?
I'm getting ready to put my PS rail together, I don't need to drop much per tap.  I'm gator-clipping each R, and *simulating* Current load with a dummyload R (.08A @315 ~~4k@26W)

Thanks

 

[PRR]

Figure the bass response of the "forward amplifier", and the "bass response" going back along the B+ filter chain.

The filter chain must be tuned significantly lower than the amplifier.

[shooter]

OK, so I have 247uF total(100, 100, 47), I *choose* 10Hz, used Xc = 1/2piFC and come up with about 64 ohms, which seems like a reasonable value, but where do the R's tween the caps come into the math?

[HotBluePlates]

Got a schematic?

Don't total up individual parts. You need to calculate bass roll-off from R-C at each step through the circuit. Let's do a pretend example.

Let's say you have a 12AX7 gain stage with a 100kΩ plate load, a 0.022µF coupling cap and a 1MΩ resistor after the cap from the next-stage's grid to ground. Our 12AX7 stage has a 1.5kΩ cathode resistor and a 25µF cathode bypass cap. The power supply node for this stage is a 16µF filter cap, which has an upstream 10kΩ decoupling resistor running to the filter cap from the next-higher supply voltage node. Oh wait... this is a simplified version of the Vibrato channel's 2nd gain stage in the AB763 Deluxe Reverb.

1. Calculate bass roll-off "going forward" of the audio circuit.

First place to look would be the cathode circuit of the gain stage. Simplify the calculation to assume the cathode resistor is the total resistance to be considered (the actual effective resistance is somewhat lower).
Use 1/(2*π*R*C) to find the -3dB point ("C" is in farads, so you'll need to add decimal places).
F_-3dB = 1/(2*π*R*C) = 1/(2*π*1500Ω*0.000025F) = ~4Hz

Do the same R-C calculation for the coupling cap. However, "R" is all resistance between each of the cap's leads. That's a 1MΩ to ground on one leg, and a 100kΩ to a.c. ground, so the 2 are "in series" as far as the cap is concerned. (100kΩ is an approximation; the output impedance of the 12AX7 is the real number to use, which includes the tube's apparent internal resistance in parallel with the 100kΩ, for a lower total. But this moves the roll-off up a little, so our approximation is safe.)
F_-3dB = 1/(2*π*R*C) = 1/(2*π*1.1MΩ*0.000000022F) = ~7Hz

So for audio, the coupling cap roll-off dominates somewhat (which is typical unless you fiddling bypass caps to trim bass).

2. Calculate the bass roll-off "going backward" through the power supply.

Let's keep things simple and consider just this stage's filter cap, and the decoupling resistor going backwards to the next-higher supply voltage. So 16µF and 10kΩ

F_-3dB = 1/(2*π*R*C) = 1/(2*π*10kΩ*0.000016F) = ~1Hz.

So the power supply R-C sets a bass roll-off 1/7th that of the dominant bass roll-off of the audio path in our example. I would have guessed 10:1 would be a good ratio to meet PRR's suggested "significantly lower" recommendation. If you make C bigger or R bigger in the power supply, the roll-off moves lower.

It's a good idea for a preamp to have supply node voltages staggered such that the R is measured at least in kΩ's or tens-of-kΩ's. That way, C doesn't have to be very big, and you have reasonably-good assurance without checking that the supply's bass roll-off is well below the audio path roll-offs.

[shooter]

Wow, I did digest it, mostly.  My brain kept saying; This looks like an endless FOR-NEXT loop!
change coupling cap, change PS, change bypass, change PS, change PS, change........

Thanks for the time, I'll start cranking the initial broadside-of-the-barn numbers.

[HotBluePlates]

... My brain kept saying; This looks like an endless FOR-NEXT loop! ...

Yes!! Welcome to amp circuit design!

An amp circuit is an empirical thing, which often owes good function to the balance of different elements. It's usually an iterative process, where you often start with one factor concrete & fixed ("I'll use triode X in a gain stage"), then assume a few other factors to get started ("supply volts will be roughly Y" & "I'll need Z volts of peak output signal, starting with W volts of input signal, and need controls A & B along the way"), and then you do a trial design and evaluate it.

And when something evaluates poorer than you'd like, you decide what needs to change and you start over again with one or more pieces more-solidified.

Somewhere along the line, you'll probably find out that one or more assumptions you made (supply voltage X or plate load Y) won't easily produce the result you want, so you change and start over (especially in the case of the supply voltage example).

I just went through some of that when I decided to build a PRR 1/3w amp, and decided I had extra room on the chassis I'll use so I might as well add a 2nd, triode-based, preamp. I'm gonna use some 6BC4 triodes, because I have a bunch on hand, they have a nice middle-of the road Amplification Factor, and although current-hungry, I wanted to hear whether a high-Gm triode was worth putting up with for early preamp stages (the high current/Gm supposedly offers reduced-noise performance).

I've got what I think is a workable design, though this thread reminded me to go back and triple-check my power supply rail.

I might build the thing and decide it doesn't sound good, or doesn't offer me any practical advantage over the ubiquitous 12AX7, or that the 6BC4 is best saved for a case where I need really low output impedances (directly from the tube or after being further lowered by a transformer).
My only useful bit of advice is always start with the output stage-to-speaker part of the amp, and select a power transformer which meets the needs of the output stage. There are only so many voltages offered in power transformers, and only so many output transformers available (unless you roll your own), and the requirements of the output stage will set what drive is required of the preamp, as well as tell you how much supply voltage is left over to be used by the preamp.

While getting in the car & just riding can be fun, if you want to get somewhere you have to decide where you're going first. Then you can figure out how to make your way there.

[shooter]

always start with the output stage-to-speaker part of the amp

I've learned much here, and surprisingly my builds are better

This current build HiFish6V6, *I think* it's workable, my main concern, the PT is 260vac 100mA, unloaded *1.414
gives me 367, shooting for 315 plates......without introducing *unseen* problems like motorboating, which I was clueless about (I know what it is), till I read your comment.  So math is free, and knowledge is invaluable .  Whether it works practically

[shooter]

Ok Excel did my math, see attached.

To me, the 1st node, V1, *seems* ok, V2 not so much.  I can change the PS decoupling R easy, and get that number down.

wasn't sure how to *model* the PA so left it for now.  Is one node more *susceptible* than another, like noise vs gain?

[HotBluePlates]

... To me, the 1st node, V1, *seems* ok, V2 not so much.  I can change the PS decoupling R easy, and get that number down. ...

Why don't you reduce the value of the coupling cap between V2A & V2B and/or reduce the value of the grid reference resistor on V2B? That would raise the bass roll-off there, since I don't think anyone will miss the 3Hz sounds they're not hearing anyway...

I suggested lowering the grid reference resistor of V2B because I assume you can't easily reduce the output impedance of V2A...

[shooter]

I changed - on paper - the decoupling R's for the pre-nodes, got the *looking back* values down to .68Hz.  I believe I won't sacrifice much plate volts in the pre since I'm just trying to get 20-25Vpp out. 

I'm hunting up info on modeling the PS nodes and stage impedances, so as I change X I can see how Y gets effected, without solder

[PRR]

RDH3 suggests (IIRC) allowing 20%-30% drop in each B+ filter stage.

This makes sense. Each stage will have a gain more than 2. Therefore the stage before it "can" work on 30% lower supply voltage.

In traditional straight-ahead amplification we know what we start with and every stage makes it bigger. Conversely the filtering must be best at the input stage and can be less for other stages working at higher level.

It maybe doesn't work for guitar amps. We have a very wide range of input levels. We will sometimes have HIGH level in the first stage, then knock it down with volume and tone pots.

Still in all there is NO excuse for taking less than 5% drop at any point in the B+ filtering, and 10% drop is often a better plan.

(G2 node is a trade-off. 5% drop at idle may be 20% drop at FULL output, which may shave clean power output. Fortunately G2 does not have to be a lot cleaner than plate. Values around 1K are go-to.)

With resistors picked the caps could be computed. However it is silly to compute a shopping-list like 0.73uFd, 2.34uFd, 18.9uFd. Up-size them all and buy a baggie-full for a bulk price. 20uFd and 40uFd will usually cover all cases.

[shooter]

thanks guys for getting me 2 more re-working brain-cells, they have been dormant to long.