Testing Global feedback stability and question about practical componsation for outputstage

[trobbins]

Sorry to drag you in to the quagmire. The X-Y plots can get pretty bizarre, especially when distortion from different sources (output stage valves, and especially output stage transformer) impart their own quirkiness.

Can you confirm you have disconnected all the high-freq roll-off circuits that are not of influence - the schematic in your first post shows a 200pF cap shunting the PI output, and you need to short out the grid-stoppers as they are likely an influence with miller capacitance. Also the dominant phase shift is likely to be from the plate resistance of the EL84 (which could be quite high in pentode mode) and the shunt capacitance of each half primary winding (ie. remove the MOV's or add capacitance to see how much of an influence that is).

[SoulFetish]

Sorry to drag you in to the quagmire. The X-Y plots can get pretty bizarre, especially when distortion from different sources (output stage valves, and especially output stage transformer) impart their own quirkiness.

No worries, Tim. it's a little slower going trying to learn something this way, so I appreciate the help.

Can you confirm you have disconnected all the high-freq roll-off circuits that are not of influence - the schematic in your first post shows a 200pF cap shunting the PI output, and you need to short out the grid-stoppers as they are likely an influence with miller capacitance. Also the dominant phase shift is likely to be from the plate resistance of the EL84 (which could be quite high in pentode mode) and the shunt capacitance of each half primary winding (ie. remove the MOV's or add capacitance to see how much of an influence that is).

I disconnected all the compensating components, as well as the MOVs.
Here is a schematic of the current test setup:

I was wondering about the large value grid-stoppers after testing it last night. I'll short those out and re-test.

[trobbins]

I have a VOX AC15 (sort of) clone on the bench at the moment, so looked up an STC 6BQ5 report for plate resistance Ra - it is stated as 38kohm for pentode and 1.5kohm for triode mode, so another test if you were keen would be to temporarily connect the EL84 screens to the anode (say through a 100 ohm stopper) and check the frequency response/phase shift using triode mode, as that may confirm that your phase shift gets pushed out by about a decade (assuming a simple RC filter response, although the actual output stage dominant filter would also include leakage inductance).

[SoulFetish]

After shorting the large grid stoppers, the first breakpoint in the open loop response moved up to around 15kHz with a shift in 45deg.

My next question is: Now what? do I close the loop and retest as is with all compensation removed? Or do I choose a suitable frequency to bring down the first breakpoint and choose the lag compensation?

[trobbins]

Perhaps try and get a few more measurements points in open-loop first - see if you can get a gain value for 90deg, 135deg, and then see how your brain copes when trying to record values if the phase starts getting less then more, and the gain also goes through ups and downs. This is where an automated Bode plotting test setup makes life so much easier.

Then yes, try and set up for some minor feedback like 3 or 6dB, and repeat the X-Y tests to see if gain can reduce below 0dB whilst phase doesn't go much below say 150 deg. And choose a simple compensation scheme and some nominal values and repeat. If you have some graph paper, you can try and follow some of the old articles that go through response plotting, and feedback compensation - that may see painful, but it does push you to realise what others have done decades ago. You may not really know if you are at a good spot without further trials, or just do some basic stability tests to give some confidence.