Bias supply and tonal affects?

[Chuck H]

It's been proposed by at least one member here that using larger bias supply caps can improve low end. I could just experiment to find out, but I was just wondering if there is any solid theory to consider on the matter that makes the answer obvious.

My thinking is that there are a lot of peripheral considerations. The power tube bias resistors in series with the impedance of the bias supply does present a load. And the impedance of the bias supply @ frequency presents a coupling between the grids. So it seems to me that the loading would mean larger caps cut bass. But the coupling consideration means larger caps improve bass by decoupling the lower frequencies.?. Certainly there are specifics that are design dependent, but is it even worth consideration?

[soundguruman]

I would not say "improve..."
I would say "distort..."

A person with good musical hearing can quickly tell the difference.
A person who is tone deaf, will probably notice no difference.

But enlarging that filter cap can make the bass control more pronounced.
Maybe a "little too" pronounced.

Yes it's worth consideration.
The designers of Fender and Marshall amps (I would say) did consider it.
But the average tech who is not a musician, ignored it.

[J M Fahey]

A person with good musical hearing can quickly tell the difference.
A person who is tone deaf, will probably notice no difference.

Mmmmmmhhhhh, I'd love to know who's the High Judge who decides who's deaf and who's gifted

The designers of Fender and Marshall amps (I would say) did consider it.
But the average tech who is not a musician, ignored it.

Nice *theory*.
Pity neither Leo nor Jim could play guitar while ***many*** Techs do.

But now it's the Techs who are deaf.

Nice bit of knowledge, brought to you by Essss !! Geeee !!!! Emmmmmmmmmmm !!!!!!!!!!!!

No doubt constantly despising Techs will make you very popular around here

[soundguruman]

Like I said, the tone deaf hear no difference. (including you)

I do not give one crap about being "popular."

[Chuck H]

To get back on topic... It seems to me that the issue is dependent on the impedances in play. If the bias supply represents a high impedance it matters more. But is it the loading or the decoupling of phase cancellation that is most significant? Could it swing either way depending on the specific impedance of the bias supply as it relates to the tubes? And does it matter enough to consider, at least for bias supplies that have significantly high impedance like one might find in some Marshall amps or amps that use bias vary tremolos?

[Chuck H]

Like I said, the tone deaf hear no difference. (including you)

There was no need for Juan's attack. Of course there was no need for your instantly critical judgment of any individual either. I'm going to suggest that you both go hug a puppy or something.

[Enzo]

By bias resistor I assume you mean the grid returns - the 220k or whatever?

Those in series with the supply (and grid) form a load? Just what do they load? Normally the grid is not conducting current, so there is zero volts dropped across them. But if they are a load, then figure it out. It is one thing to say something is a load, but yet another thing to find that the difference is that with one cap the amp rolls off at 2Hz, but with another it rolls off at 3Hz. I doubt even those golden ears will not hear a difference between guitar amps that differ in bottom end by 1Hz.

[Chuck H]

Forgive me if my application of the nomenclature is incorrect. I've often seen such resistors in preamp circuits referred to as loads. If the concept of a load only applies to current and not voltage then I apologize for the inaccuracy. It's probably more accurate to note their roll as part of a voltage division with the driving amplifiers impedance being the series resistor. Which I'm certain is real. And there will then be AC dropped across that resistor. And looking into it I found that the filter/decoupling cap values used in many bias supplies actually don't go to the bottom of the audio range. I had always just assumed they did, but they don't. So I was wondering if there were any relevant considerations. We go to some lengths to learn about even small tonal manipulations. Little of it matters a lot but a lot of it might matter a little

[Dave H]

I’ve seen those resistors called both grid leaks and grid loads and they are the load for the driving stage. I think the usual 8u bias supply cap is big enough and making it bigger won’t audibly change the bass response. It would probably work fine without a capacitor as that point is a virtual ground anyway.

[Alan0354]

You talking about the cap on the -50V for grid bias of 6L6 type? I don't see it makes a dent of a difference in sound. It is for filtering the rectified noise from the transformer only. If you don't have enough then you have 60,120, 240 noise etc. going through, then you have hum problem. The filter cap goes through at least a 100K resistor before connecting to the grid, how can that has anything to do with the low end?.......Of cause if you talk about using a ridiculously small cap of a few hundred pF instead of the standard of over 50uF!!!

[R.G.]

Let's think about circuits for a minute.

We have a "T" circuit here. There is the leg of the T, which is a capacitor to ground. This cap is fed current/voltage from the rectifier network to keep it charged up to some voltage. The bias cap and its resistor/cap/diode/transformer network have some impedance which is calculable if you know the part values. In general, the cap will be picked for a value which holds ripple to quite small values compared to its DC voltage.

From that cap, grid leak resistors lead to the output tube grids in a fixed bias system (which is what I take as the subject here).

The grids have equal-and-opposite voltages forced on them by the phase inverter. The equal-and-opposite-ness may not be perfect, but has to be pretty good; maybe 10% different in some amps as a guess. The signal voltages come from a PI that may or may not have a low impedance. Often it is a differential stage with a 60K(ish) plate resistor in parallel with an 82K to 100K plate resistor. Sometimes it is the low(ish) impedance of the cathode of a split-load phase inverter.

The grids look like a very high resistance leakage to a negative voltage from electrons which hit them, and a capacitor to ground which is composed of the actual capacitances and the virtual ones from gain effects. These capacitances are quite small, as they're primary of concern in RF gain applications.

So a signal goes from the PI through the PI source impedance of maybe 10K (cathode output) to 50K (plate paralleled with a plate resistor) through a DC blocking signal cap of usually around 20nF, maybe more or less, and into the grid. The grid is so high impedance that it's negligible to the drive signal (by design), so most of the loading on the PI is the grid leak resistor unless you drive the grid positive.

The signal then is highpassed by the coupling cap from the PI, and divided by the ratio of the drive voltage at the PI and the grid leak resistor and whatever impedance it sees back at the cap/middle of the bias supply T.

To the extent that the PI signals really are equal-and-opposite at the grids and the grid resistors are equal, they cancel one another at the bias cap. The bias cap then doesn't matter. To the extent they are mismatched, they are attenuated by the ratio of the grid leak resistors and cap to ground. The lower the capacitor's impedance, the more they are attenuated to ground.

If the cap is not big enough, it doesn't attenuate very well at low frequencies, so any non-matching of the grid signals can appear as some cross coupling from grid to grid. But this cross coupling comes from the PI source impedance, through the grid leak resistance, is shunted to ground by the bias cap, and is attenuated again by the other grid leak resistor and its PI source impedance.

Just picking a typical bias cap value, 50uF, you get Zc = 1/(2*pi*F*C) = 1/(2*3.14*42*50E-6) = 75.8 ohms at 42 Hz. Grid leaks are on the order of 100K to 220K in general; the Twin Reverb I picked the 50uF from has 220K. So you get attenuation of the mismatches of 75.8/220,000 = 0.000344665, or a bit over 69db. There's another 6db or so from the back attenuation back into the other grid, and there's another 6db if you're only concerned with guitar stuff up at 82Hz.

Anyway, that's how I'd go about thinking about it.

You can see where this goes. Increasing the cap increases the attenuation between grids, by 6db each time you double it.

The eccentricities of the cap itself matter to the extent that they alter the attenuation ratio. High-ESR caps could make it worse (that is, less attenuation) by adding to the impedance of that cap. So an old, drying-cap with an ESR climbing up to a few hundred ohms might be detectable. Might. Probably the high ripple would be worse than the crossfeed.

[R.G.]

But enlarging that filter cap can make the bass control more pronounced.
Maybe a "little too" pronounced.

Care to explain how that works?

[Chuck H]

Thank you R.G. for the excellent rundown (as usual). One note would be that more typical bias supply caps for most amps are between 10uf and 25uf. But that still doesn't prove to matter much WRT decoupling as I've found.

I went ahead and ran a SPICE simulation. Noteworthy is that I didn't (don't have the technical expertise) program in all the anomalies and derates under specifics for the individual components. But let's ignore that for now. I ran a typical Marshall circuit with the dual 10uf's and a typical small AB763 circuit with 25uf. Replacing these values with 100uf for either circuit made no noteworthy difference in the audio spectrum. Litterally a twitch that was so tiny I couldn't visually determine it on screen. And only a couple of dB somewhere below 10Hz, as Enzo noted would be the case. No change at all to the knee of the coupling cap.

I'm ready to deem this myth busted.

[nickb]

<snip>
I'm ready to deem this myth busted.

Just to stir the pot a little consider what happens when you drive the power tubes grids hard, the positive swing gets clamped and the coupling capacitors charge given blocking distortion. In addition the average DC level on the grids goes down and so also the bias cap voltage go down, a little. Now the recovery from blocking it quite quick due to the small value coupling capacitors but the the big bias cap takes quite a bit longer to recover. The effect is not big and so I'd expect the audible effect to be quite small.

Also, the size of the cap does not reduce the magnitude of the voltage change, just the time constant.

[Alan0354]

Just to stir the pot a little consider what happens when you drive the power tubes grids hard, the positive swing gets clamped and the coupling capacitors charge given blocking distortion. In addition the average DC level on the grids goes down and so also the bias cap voltage go down, a little. Now the recovery from blocking it quite quick due to the small value coupling capacitors but the the big bias cap takes quite a bit longer to recover. The effect is not big and so I'd expect the audible effect to be quite small.

The bypass cap is connected to the transformer through a low value resistor, so the impedance on the cap side is very low. I don't think you can change the voltage on the cap much driving through the grid resistor that is over 100K.

I think the charging effect on the coupling cap from the PI to the grid can affect the sound a whole lot more.