5d4 Floating Paraphase experimental results

[octal]

After the discussion I had with PDF64 in another thread about Paraphase inverters, I thought I'd try an experiment with my Fender 5D4 Super ish amp to check the output impedance of the PI circuit and the effect of bypassing the PI cathode resistors. The PI circuit is built stock- the only difference is the '50K' resistor on the schematic is 51K.

Output tubes were removed for this test.

I fed a 400Hz sinewave in and adjusted for 10Vac output on the 'phase' output of the PI . The 'antiphase' output measured 9.46V. (Conclusion: balance decent.)

I added 100K loads from each output tube control grid to ground. Outputs measured as follows: phase 7.5V, antiphase 6.3V. (Conclusion: output impedance of unbypassed cathode R antiphase tube higher than 'phase' output, but not as much as one would expect in a non-floating paraphase circuit.)

I then bypassed the cathode resistor on the bottom tube (the 'antiphase') one. The 100K loaded output voltage measured 7.8V. The 'phase' output was unchanged at 7.5V. (Conclusion: output impedances more similar now, better balance.)

I then checked the outputs with the extra 100K resistors removed AND the extra bypass cap on the cathode R of the antiphase tube installed. 10V on 'phase' and 10.7V on 'antiphase.' (Conclusion: balance a bit worse with both cathode resistors bypassed, but probably not of any consequence.)

I'd be interested to see if my interpretations of the data are correct, if anyone has any thoughts & insights on this circuit, etc.

Nathan

[pdf64]

Interesting stuff Nathan, thanks for reporting back.
A couple of thoughts; if we follow Merlin's convention, then the paraphase PI just refers to the stage with the inverting stage with nominal unity gain, that you are calling the 'antiphase', with the input to the paraphase being your 'phase'.
What effect does the 51k grid leak resistor have, eg what if it's swapped for 1M? Theoretically it shouldn't make any difference, but...?
The input to the paraphase already has a ~220k load due to the paraphase input impedance (consequently the output tubes wouldn't need a grid leak if cathode biased), so to be fair the paraphase output might be loaded with 220k also.
I'm really surprised that the cathode un/bypassing has such an affect, given that there's so much negative feedback - I'm wondering if it is because the signal level is fairly high; is it the same with a 1V signal? Pete.

[MWJB]

Swapping the 50K for 1M would increase the grid load values for the power tubes, as this resistor is situated between the 220K & 270K 6V6 grid loads & ground. A pot can be substituted for the 50K resistor to fine tune balance, or skew it as desired...Gerald Weber did this when developing the Kendrick Texas Crude Harp Amp.

[octal]

What effect does the 51k grid leak resistor have, eg what if it's swapped for 1M? Theoretically it shouldn't make any difference, but...?
The input to the paraphase already has a ~220k load due to the paraphase input impedance (consequently the output tubes wouldn't need a grid leak if cathode biased), so to be fair the paraphase output might be loaded with 220k also.
I'm really surprised that the cathode un/bypassing has such an affect, given that there's so much negative feedback - I'm wondering if it is because the signal level is fairly high; is it the same with a 1V signal? Pete.

Pete,
I thought my experimental results mostly verified your contention that the NFB through the virtual earth serves to lower the output impedance of the 'second' triode.
An unbypassed 12AX7 with 100k plate load would have about a 60K output impedance, so an extra 100K load to ground should have dragged the output voltage down by a higher percentage than it did, right?

I think a 10V swing is less than half the swing needed to achieve full power and is something like a fifth of the maximum possible clean swing of this stage (with no output tubes installed), so I don't think my 10V test level is too crazy. I don't think I'm going to have time to do a lot of additional measurements, though I am a bit curious about what varying the value of the 50K resistor might do. Does it only cause imbalance because of the unequal value of the 220 & 270K resistors, or does it cause imbalance because it limits the degree of NFB at the virtual earth?

Nathan

[MWJB]

If you short the 50K to ground totally, then the amp is SE, because you're no longer providing the out-of-phase signal to PI triode B. Ideally, with the full value forthe balancing resistor, balance should be adequate...as you reduce it's value, you attenuate signal to the OOP/bottom/antiphase power tube

[octal]

MWJB,

I get your point about the extreme case of lowering the 50K resistor to zero. As that resistor is lowered (within reason) you are lowering the grid drive to the 'bottom' tube, but you are also reducing the NFB going to the summing point of the three resistors, which tends to reduce the effect of the change.

Would it be safe to say that as the value of the 50K increases, the self balancing tendency of the floating paraphase increases, and as it is reduced it decreases?

Nathan

[MWJB]

Hi Nathan,

I've never really experimented with going larger than 50K (or 56K). My guess, & it is just that, is that there will be point that gives best balance, after which increasing/reducing will skew the balance, plus the larger you go the more aggressive the amp will become as grid load values for both 6V6s are increased.

As the signal to the OOP tube needs to be attenuated some, because it then goes through another stage of gain, there is the possibility that, rather than the "in-phase" tube seeing the bigger signal & the OOP tube seeing a smaller/equal one, the OOP tube could then be seeing the larger signal and the balance see-sawing the other way?

[pdf64]

'though I am a bit curious about what varying the value of the 50K resistor might do. Does it only cause imbalance because of the unequal value of the 220 & 270K resistors, or does it cause imbalance because it limits the degree of NFB at the virtual earth?'

My guess is that as it's a virtual earth point, then the 51k grid leak resistor could be any value from <10k to 3M3 without the circuit operation being affected. Though as Mark points out, above 50k then another resistor would have to be added, eg at the paraphase input to ground, to keep the 6V6 grid leak value within spec. Pete.

[octal]

I can try lowering it easily enough by tacking something in parallel. I'll see if I can try it later & report back.

Nathan

:LATER:

OK, I added a 15K in parallel with the 50K. The output of the second tube dropped to 6.5V (don't remember exact number- I wrote it down when I was downstairs, upstairs now, but it was 6 point something.) while the first tube's output stayed constant at 10V.

[pdf64]

Hmm - I thought that it might hold up better than that, but the open loop gain is only ~60 (or ~30 with unbypassed cathode), so can't expect ideal op-amp performance.
It does at least indicate that it's a floating paraphase, because despite the (open loop) gain falling to about 0.2 of it's previous value, the output only fell to 0.6. So the negative feedback is doing it's best to make up the loss.
Thanks for sharing the fun - Pete.