Idea for a post-PI master

[potatofarmer]

Wish I'd drawn a picture of this, but basically I'm considering using a dual-ganged pot, wired up as rheostats between the output coupling caps and grid leak resistors. Full volume would be 0 ohms, and as you increase resistance you make a voltage divider.

I'm thinking about this because the existing (pre-phase inverter) master volume results in too much loss of bass. This rheostat approach will also change the frequency response of the amp, but I don't think it'll be as offensive a change.

Anyway, here are my thoughts so far:

• Don't have to rely on a pot for maintaining bias voltage
• If rheostat wiper loses contact, only effect is volume drop
• Linear taper pot at roughly 1x or 2x the resistance of the existing grid leak results in a nice audio(-ish) taper
• Only able to attenuate signal to ~30-50% depending on choice of pot. Not enough attenuation for bedroom use, but we get a nice big sweep over the range of attenuation that a MV is most useful for
• Rheostat doesn't change effective Rg1 for power tubes
• Rheostat increases load on PI at higher attenuation - fighting the attenuation - but in spreadsheet simulation, this effect seems negligible
• Bass response will increase as attenuation is increased
• Blocking distortion performance changes - longer times for bias excursion and recovery, lower excursion ratio. Less likely to overdrive the power tubes with an attenuated signal anyway

My questions are my usual: what am I missing, why hasn't this been done commercially, etc.

This seems to be more of an individualized solution, good for me and my 47k grid leaks, but probably not for what most people would want from a 'master volume.' I've disabled the global NFB loop so I'm not worried about interactions with that.

Original schematic here: http://www.thevintagesound.com/ffg/s..._135_schem.jpg

[Shadrock2]

I would say that yo get useful attenuation the "Rheostats" would have to be 2 - 10 times the value of the grid leak resistors. For most this would result in pots greater than 1 meg which would be hard to find. 47k grid leak sound pretty low to me. IMHO

Also you are right to think it will change frequency response but how much is an open question.

[potatofarmer]

I would say that yo get useful attenuation the "Rheostats" would have to be 2 - 10 times the value of the grid leak resistors. For most this would result in pots greater than 1 meg which would be hard to find.

That is a good point; I can always tack in resistors to see what a good value would be.

47k grid leak sound pretty low to me. IMHO

You'd think that, right? For thermal stability Rg1 should be max 100k, and since there are two tubes sharing the same grid leak resistor that means max 50k. Which would be a stiff load for an LTP, except Fender went with a 12AT7 with 47k plate resistors for a Zo in the 8.5k-ish territory. The Silverface phase inverter is one of the first things people "Blackface" but it's all-around better than the Blackface version IMO.

Also you are right to think it will change frequency response but how much is an open question.

Well, as far as the high end is concerned you're adding a big honking grid stopper, but even at 500k high-end rolloff doesn't occur until around 31 kHz. The low end is another story; I've cut the coupling caps back to 47nF for a -3dB point of 72Hz, but adding more load resistance means the -3dB point drops rapidly down to the 10 - 20 Hz region. Being a Twin though, that's around what it was with the original 100nF couplers.

The distortion performance is probably anybody's guess.

[Shadrock2]

"

For thermal stability Rg1 should be max 100k, and since there are two tubes sharing the same grid leak resistor that means max 50k. Which would be a stiff load for an LTP, except Fender went with a 12AT7 with 47k plate resistors for a Zo in the 8.5k-ish territory.

Oh yeah for two tubes... but you will notice that in the schematic the bottom pair of tubes uses a 68k grid leak. Plus there is a 10k balance pot that figures in also. This implies that the two ganged pots need to be different values. Maybe not though.

Well, as far as the high end is concerned you're adding a big honking grid stopper, but even at 500k high-end rolloff doesn't occur until around 31 kHz. The low end is another story; I've cut the coupling caps back to 47nF for a -3dB point of 72Hz, but adding more load resistance means the -3dB point drops rapidly down to the 10 - 20 Hz region. Being a Twin though, that's around what it was with the original 100nF couplers.

Also two tubes worth of capacitance. My GE data sheet for 6l6gc says 10uuf total input capacitance per tube. So 20uuf total right? Plus any for extra wiring. If I use my standard rule of thumb formula f = 1/(2 * pi * R * C) I get ... f = 16kHz for 500k grid stopper. This would still be useable though.
Don't really know how you calculate low f cutoff.
I am curious how this might work though.

In a slightly different direction... have you ever tried a rheostat in series with say a .1 mfd cap that connects between the grids of the LTP for a master volume?

[J M Fahey]

Rather than a rvariabl resistor in series which provides a very limiter range, wire them as conventional volume pots referred to ground, which can dial output power down to 0, far more useful.
The limited range of simple series resistors will be perceived by users as a still painfully loud amp, that's why it's not used commercially or in boutique amps.
You will need to add a couple extra coupling caps after them to avoid messing with bias, *or* refer them to bias voltage, to achieve the same.
The bias cap is audio ground so no problem with that.

[potatofarmer]

"

Oh yeah for two tubes... but you will notice that in the schematic the bottom pair of tubes uses a 68k grid leak. Plus there is a 10k balance pot that figures in also. This implies that the two ganged pots need to be different values. Maybe not though.

I replaced the 68k with a 47k.

Also two tubes worth of capacitance. My GE data sheet for 6l6gc says 10uuf total input capacitance per tube. So 20uuf total right? Plus any for extra wiring. If I use my standard rule of thumb formula f = 1/(2 * pi * R * C) I get ... f = 16kHz for 500k grid stopper. This would still be useable though.
Don't really know how you calculate low f cutoff.
I am curious how this might work though.

Ah, you got me there on doubling the input capacitance!

In a slightly different direction... have you ever tried a rheostat in series with say a .1 mfd cap that connects between the grids of the LTP for a master volume?

I haven't tried a "crossline" master, no. I've heard they can be somewhat finicky but I believe Matchless uses them with some success.

Rather than a rvariabl resistor in series which provides a very limiter range, wire them as conventional volume pots referred to ground, which can dial output power down to 0, far more useful.
The limited range of simple series resistors will be perceived by users as a still painfully loud amp, that's why it's not used commercially or in boutique amps.
You will need to add a couple extra coupling caps after them to avoid messing with bias, *or* refer them to bias voltage, to achieve the same.
The bias cap is audio ground so no problem with that.

Yeah, plotting the attenuation on a semi-log scale was what I was missing. I'll probably go with a Lar/Mar type, which is referenced to the bias supply (or in this case to each side of the bias balance pot) with a resistor between the wiper and the bottom of the pot to ensure you don't lose bias if the wiper lifts.

Oh well, thanks guys!