Friedman/Suhr PPIMV

[physics]

Hi!



I was wondering if anyone had any thoughts on the "new" master volume that Friedman and Suhr amps seem to be using, as seen on the Plex and SL68 amps. Bias grid-leaks are replaced with 2M2 resistors, and each resistor gets one gang of a dual gang pot (each gang wired as a rheostat) in parallel with it. As you turn down the pot, you reduce the grid-leak value and load down the phase inverter, or such is my current understanding.

I'll post a sketch after some sleep.

Thanks,
          physics

Edit: sketch
   

[K O'Connor]

Hi Guys

The 2M2 is a safety resistor to guarantee a connection between the bias supply and the tube grid in case of pot wiper discontinuity. Of couarse, if the the pot track is simply wired in parallel with the 2M2 and wiper linked to the top end, then the 2M2 is redundant.

As Physics (and physics) points out, dialing the pot to zero will tie the coupling cap to the bias supply and load the splitter while reducing the drive into the tubes to zero. This overall effect is the same as with the crossline MV inasmuch as drive to the tubes is reduced while the loading on the splitter is increased. In both cases the differential gain of the common Schmitt splitter is reduced and its distortion may rise. When feedback is applied to the power amp, as is the case with high-power amps, it can be difficult to retain a clean sound when the MV is reduced.

[physics]

Thanks for the info Kevin!

Edit: Kevin pointed out the circuit description that follows is confusing. It was meant to describe one of the DC-coupled PPIMV's in TUT4, not the F/S PPIMV. I'm leaving the description as-is to not mess with history/remove context for post #4.

The F/S PPIMV looks similar to the design you put in TUT4, with pot wiper to top of grid leak, the bottom of the pot to -Vb, and the top of the pot to the PI coupling cap. That one reduces the grid-leak to zero as well, but it keeps a relatively high impedance shown to the PI. It also has actual voltage-divider action happening with the pot instead of loading down the PI, if I understand it correctly. Would there be any reason to prefer the TUT4 implementation over the F/S implementation?

Thanks,
physics

[K O'Connor]

Hi Guys

I just lost my entire post...

Physics added the sketch to post-1 after my comments were posted. The sketch shows rheostat-wired pots, contrary to his description in post-3, which describes a standard DC-coupled
post-PI-MV.

For the two standard DC-coupled post-PI-MVs, the version with the wiper as the signal input loads the splitter, where the version with the wiper output does not. Neither effect the bias voltage to the tube. Any splitter that DOES change grid-bias voltage should be avoided.

The AC-coupled post-PI-MV can be wired in the same two ways as above, but is ground-referenced instead of referenced to B-. The two forms behave the same as their DC counterparts.

The design in the sketch has pot-X tied to B- if CW is supposed to be the loudest setting. The 2M2s are redundant.

[physics]

In retrospect, I see how post #3 was confusing. Sorry. To clarify, it was intended to describe one of the DC-coupled designs in TUT4. Here is my description reworked:

The F/S PPIMV looks similar to the design you put in TUT4, which has the pot wiper to top of grid leak, the bottom of the pot to -Vb, and the top of the pot to the PI coupling cap. Both your design and the F/S design are strapped across the grid-leaks.

Thanks for the additional info in post #4 Kevin. To your comment about changing bias voltage being a bad idea, am I correct in saying that the F/S design messes with bias voltage since it varies the grid-leak values, and thus it's not ideal?

Thanks,
physics

[K O'Connor]

Hi Guys

There is no MV design that I know of that will upset the bias condition of the tubes, assuming the tubes are not exhibiting high grid leakage currents - definitely a tube to be discarded anyway.

have fun

[physics]

Got it, thanks! I'll likely try at least one of these PPIMV in my Marshall 4104 as a stop-gap until I can do power scaling, and will post clips for reference.

[K O'Connor]

Hi Guys

With respect to the drawing in post-1, where the grid-leaks of the output tubes are replaced by a dual pot wired as rheostats to load the splitter, that is actually what happens.

Note that the cross-line MV does exactly the same thing and is used in Vox amps that have cathode-biased output stages. For a fixed-biased amplifier, it is best to AC couple the MV.

In a guitar amp the usual splitter is a stacked Schmitt, or simply a differential amplifier. The gain of the splitter depends on the sum of the plate resistors and the value of the shared cathode resistor. The basic differential form has a medium to high-value cathode resistor which behaves as a constant-current source. Its action is more ideal as the value is made higher and the voltage across it is made higher. This ideal form requires a reference voltage for the grids, which can take many circuit forms.

The stacked-Schmitt has a split cathode resistor with a small portion on top of the remainder. The split-point in Rk is used as a voltage reference for the grids, so in this way the voltage reference is stacked on top of the constant-current source. In most guitar amps, this entire circuit is further stacked on top of the feedback network for the power amp. The feedback signal is capacitively coupled from this second stack point up to one of the grids.

In the stacked-Schmitt, the current-source resistor still effects the balance of operation and it works better with more voltage across it. Were you to short it out, the balance of the output signals would suffer greatly. The upper portion of Rk above the voltage reference is what the tube "sees" as Rk and this gives this variation of a diff-amp its high gain. With typical 82k + 100k plate values and Rk=470, the gain is limited by the mu of the tube. Based on the visible circuit values (as a first approximation of gain), gain would be 182k / 470 = 387. Were the tubes replaced with transistors, this gain is easily achieved, but a 12AX7 has mu=100 and therefore the circuit's maximum gain is 100.

The external loads for the two signal outputs effect the overall gain. If we think of the 100k side and that the tube sees 100k of load, then add a 1M in parallel, the load is now about 90k. If we place 220k in parallel with 100k to represent a typical grid-leak for an output tube, the visible load is 68k. In differential terms, the two plate loads in series is now 136k and the small Rk portion of 470-ohms is unchanged, and a theoretical gain of almost 300.

The crossline MV is not ground-referenced; rather, it is floating, but it is in parallel with the sum of the plate resistors for the diff-splitter. The pot is wired as a rheostat, which means one end is tied to the wiper, or that only one end and the wiper are connected to the circuit. In either wiring, the net resistance the pot adds to the circuit ranges from zero-ohms at one end, to the full pot value at the other end, over the sweep of the pot. If you short the plates of the splitter together, there is cancellation and no output signal. if we did this directly there would be scratchy noise from the pot as it has DC current through it. To cure this we add a cap in series with the pot, or place it on the output side of the coupling caps.

With the modern balance values of 91k and 100k, Rk=470 and a 12AT7, Va=540V, we can get 100+100Vpk output with 6Vpk of input, giving a differential gain of 33.3. This is with 100k loads after the coupling caps. Adding 500k across the signal lines reduces the output slightly to 97+97Vpk and A=32. A 100k cross-line load reduces output to 65+65Vpk A=22. At 10k load, output is 12+12Vpk A=4. At 10-ohms, the signals overlap completely and there is still an asymmetric signal of +1v4 -5v fed to both lines. Surprisingly, at zero-ohms the same signal is present.

Note that with a less extreme input signal, the output signals will all be proportionately lower and the quietest output will be quieter.

If we replace the crossline-MV with the dual-pot rheostat loads, we see identical results down to 5k+5k loading. At 5+5-ohms we have symmetric signals at 12+12mVpk. At one-ohms total load the signal is at 1.2mV per line, still symmetrical, and at zero ohms there is zero output.

So, the dual-rheostat-MV works much better than the traditional crossline-MV (no surprise) while using a "rheostat" approach instead of a voltage divider approach, as all other MVs are.

Like all circuits that use dual-pots, the tracking of the pot sections is quite terrible and the ideal performance we see here is not likely to be attained.

[physics]

Thanks for the analysis Kevin!

[nauta]

Hey eh

Hasn't that MV been a round for a long time already? i thought I saw it in some ken Fisher notes way back

Rock on