Better Low Z driver stage design based on our Bootstrapped Gain Stage thread

[Helmholtz]

A 90° phase means an essentially reactive impedance. That's not typical for audio amplfier input or output impedance even including coupling capacitors. It also helps if the meter provides different measuring frequencies.
A good digital LCR meter allows to exactly measure the ESR of a cap. This would be a situation where phase is close to 90° (depending on frequency)

I mainly use this: https://www.peaktech.de/productdetai...tech-2170.html

[nickb]

A 90° phase means an essentially reactive impedance. That's not typical for audio amplfier input or output impedance even including coupling capacitors. It also helps if the meter provides different measuring frequencies.
A good digital LCR meter allows to exactly measure the ESR of a large capacitance ecap. This would be a situation where phase is close to 90°.

A good one will but the cheap ones don't. A 1% Z meter one won't measure ESR very accurately at all. I've seen those popular Peak ones display 200 ohms for ESR of a ceramic cap when it's really much less than 1 ohm. Not even in the ballpark. There's a a nice BK one that's 0.05% and goes up to 200Khz with graphing that looks just the ticket. I did wonder about getting a Red Pitaya board. With 14 bit ADCS and 40Mhz bandwidth you should be able to make a very decent instrument. I went off that idea as their own LCR meter is only 1% up to 1Mhz or so accurate and it made me wonder just how good the board really is.

[Helmholtz]

I did verify ESR measurements with my Peaktech using other methods (including a HP4194A Impedance analyzer) but I actually don't care if the result is 1R or 0.9R with a 32µ cap. Most important with low ESR is 4-wire measurement.

I don't see the relation to amp impedance measurements.

[SoulFetish]

whoa! I didn't even realize there were any replies. (for some reason, I didn't get any notification emails...)
I'm gonna have to circle back and catch up.
But real quick, ~ Helmholtz, I'm not trying to drive these grids into orbit. There is a cathode resistor, as well as large grid stoppers on the EL84s (51k right now). So, you might wonder if it's all even worth the trouble?
Honestly, I'm really not sure. I'm a bit discouraged about the amp lately. I might trash the entire project. I'm wondering if the HF response nickb mentioned is actually problematic in this design. I mean, the balanced portion of the amp is super quiet (as it should be). You almost can't tell that it's on, if I pull the preamp tubes and leave the PI, driver, and output tubes in. But It's so friggin loud, and I don't like the "touch" I'm getting on the high end of the frequency spectrum. It's way sensitive right now.

[SoulFetish]

Change in feedback is fairly small for Zsrc= 50k because of the 470K grid leak resistor. For infinity (quite useless) it's something like 3k for this cct with a 12AU7.

Another thought. The bias voltage is very uncertain as you rely on the tube to set it. It would be more predictable if you were to feed a reference to the grid. That would lower the input impedance and couple power supply noise. This first is not an issue an the second just requires attention to filtering.

These are two reasons I changed the biasing arrangement. As explained in the original post, the loading on the PI closely matches what would be seen in a conventional AC coupled output stage. The input impedance is set by the biasing resistors R1 in parallel with R2+R3. (I know you know this). So I chose resistor values to allow the range of bias voltages I wanted and would provide an input impedance of around 220k.
But power supply noise at the gate is not an issue in this case. ...Gate? (RG's got me thinking about FETs) *power supply noise at the grid is not an issue.
In a normal Uni-polar supply, ripple voltages attenuated by the resistor divider would be coupled in at the grid. However, this stage utilizes a bipolar supply, with the power +/- supply ripple being out of phase. The bias point is set so close to the center of the two supplies, that it any power supply ripple at the grid should be almost completely nulled. (Assuming well matched components.) Correct?

[SoulFetish]

FWIW, source or output impedance can be determined from the signal reduction caused by adding a load resistor.
Or more easily by using an LCR meter that provides AC resistance measurement (also works for input resistance). To block DC wire a cap in series with the meter.

I should be able to measure the actual output impedance using the amp as a test rig. I already have the fixed bias set up, and it shouldn't be too much to set it back up as a cathode biased/bootstrapped cathode follower.
If I feed the input of the cathode follower with a 1k source impedance from my signal generator, would that be adequate? Or should I use a high Z source impedance from the plate of the PI?

[Helmholtz]

If I feed the input of the cathode follower with a 1k source impedance from my signal generator, would that be adequate? Or should I use a high Z source impedance from the plate of the PI?

Ok, seems you want to try the "additional load" method to determine the cathode follower output impedance.
- First connect a dummy load to the amp's output. Or you may pull the power tubes.
- To not disturb relevant impedances feed 1kHz sine signal to the PI input. Keep signal level low to avoid any curve distortion.
- Measure signal level at the CF output : Vo
- Wire a 10µ cap in series with a 1k resistor ( = Rload) between CF output and ground and measure again : Vload

Calculate CF output impedance (Rout) from Rout = (Rload/Vload) * (Vo - Vload). Formula is derived using a bit of "Thevenin" and the fact that the current through both resistors is the same.
E.g. for Vload = 0.5 Vo, it follows Rout = Rload.

[SoulFetish]

Ok, seems you want to try the "additional load" method to determine the cathode follower output impedance.
- First connect a dummy load to the amp's output. Or you may pull the power tubes.
- To not disturb relevant impedances feed 1kHz sine signal to the PI input. Keep signal level low to avoid any curve distortion.
- Measure signal level at the CF output : Vo
- Wire a 10µ cap in series with a 1k resistor ( = Rload) between CF output and ground and measure again : Vload

Calculate CF output impedance (Rout) from Rout = (Rload/Vload) * (Vo - Vload). Formula is derived using a bit of "Thevenin" and the fact that the current through both resistors is the same.
E.g. for Vload = 0.5 Vo, it follows Rout = Rload.

Right

[Helmholtz]

??

[SoulFetish]

As in yes that sounds good

[Helmholtz]

A 1% Z meter one won't measure ESR very accurately at all. I've seen those popular Peak ones display 200 ohms for ESR of a ceramic cap when it's really much less than 1 ohm. Not even in the ballpark.

I admit that measuring the ESR of an ecap was no good example for a close to 90° impedance.
So I measured the ESR of a 10nF Styroflex cap as 0.02R (@100kHz). Meter was calibrated to 0.000R with shorted leads.
Then I wired a 10.0R resistor in series with the cap and measured again. Result: Rs/ESR = 10.06R.
That's good enough for me.

[Helmholtz]

BTW, the driving source impedance as seen from the power tube grids is grid stopper plus CF output resistance. Means that CF output impedance is rather insignificant with 51k grid stoppers.
It's this total resistance together with the EL84 input capacitance which matters for (open loop) HF response. The EL84 input capacitance I estimate to 22pF. This means a HF cutoff frequency of around 140kHz for this LP filter.
Means that limiting power stage HF response by driver impedance or grid stoppers isn't very effective as power pentode input capacitance (including Miller effect) is rather low.

[nickb]

I did verify ESR measurements with my Peaktech using other methods (including a HP4194A Impedance analyzer) but I actually don't care if the result is 1R or 0.9R with a 32µ cap. Most important with low ESR is 4-wire measurement.

I don't see the relation to amp impedance measurements.

A small error like that is of no consequence, but the one I saw where it read 200 ohms when it should have been less then 1 is a problem. No, it's not an issue for amp impedance measurements but that is just one case. it is crucial for component measurements and I don't want to spent my money on two instruments where one will do.

Sorry SF, got a little sidetracked here

[nickb]

any power supply ripple at the grid should be almost completely nulled. (Assuming well matched components.) Correct?

......Yes!

[SoulFetish]

BTW, the driving source impedance as seen from the power tube grids is grid stopper plus CF output resistance. Means that CF output impedance is rather insignificant with 51k grid stoppers.
It's this total resistance together with the EL84 input capacitance which matters for (open loop) HF response. The EL84 input capacitance I estimate to 22pF. This means a HF cutoff frequency of around 140kHz for this LP filter.
Means that limiting power stage HF response by driver impedance or grid stoppers isn't very effective as power pentode input capacitance (including Miller effect) is rather low.

of course. Im not using the grid stoppers as a LPF. Im using them to in conjunction with cathode degeneration as a means of clamping the the grids at 0V to prevent grid current and promote compression and clipping

but to your point, once i begin testing the phase response again in the feedback loop. More needs to be done if there is any inaudible oscillation or instability