Paraphase inverter question

[Helmholtz]

It is not, take a look at the pics in post 95

I was speaking of Dave's sim results in post #98 where he varied PI output caps.

[catalin gramada]

Thanks for You good intentions but in two hours have to go at work, I.m very sorry must I leave You. Thanks for support gents. Have a nice week!

[Dave H]

I don't see much change in the differential signal.

True, the autoranging fooled me. I just noticed the increased phase shift and the differential signal now being larger than the grid to ground signal. There goes another theory. I'm still thinking

[Dave H]

I see a lot of pictures from the oscilloscope, various numeric values from the screen, who it mean nothing.
We just don't know what we want with a bunch of data.
I am afraid that we will stray because proposals and results do not indicate constructive debate.

Let's go back to the beginning

https://music-electronics-forum.com/showthread.php?t=49645&p=537990&viewfull=1#post537990

and to define what we want from thread.

You may well be right

[catalin gramada]

Latest news.
Before leaving the house I still found some time to modify the bias circuit as in picture below, as it was suggested:



Now the amp is absolutely quiet. Zero noise, none.
I renounced at dual rail, meant half resistance, low current, low ripple, and extra 100uF was added directly in the wiper. I.m more than satisfied. Thanks.

[Dave H]

Why did you put in an independent bias.

To use unmatched tubes?

[Helmholtz]

The PI outputs are low impedance compared to the 220k resistors from grids to bias supply so there's a potential divider which reduces the ripple at the grids when the PI is connected.

I tend to think that this is the actual explanation of the problem. The different PI plate impedances differently load the bias ripple voltages at the grids and thus cause a differential ripple signal. This would also explain that phase shift/difference changes with lower cap values.
To verify that this is the only/dominating effect one could disconnect the 47nF cap at the second PI triode's grid and watch for any changes.

Did you determine PI output impedances by sim?

[vintagekiki]

To use unmatched tubes?

Schematic is good while new and matched tubes in the amplifier.
Matched tubes is a relative term, with limits of +/- 5%, beyond those limits tubes is unmatched.
Let's not forget that power from amplifier declines with the square tubes emission. 90% emission from tubes gives 81% power on amplifier.

Since schematic in the bridge, any deviation from the matched leads to a bridge imbalance, which manifests as a quiet hum from the speaker.
During their work, tubes in amplifier wear unevenly, so asymmetry (unmatched) can be and to 20-30%. In order to balance the output tubes, different bias voltage is supplied.

With different bias voltage (different work point), tubes is static balanced, have the same idle current, however what happens when from phase inverter brings for example 20VAC signal to the output tubes.

For the same AC voltage from the phase inverter the output tubes are not same excited, so one side of the PP is higher open, and the other side PP is less open.
Further, everything is clear, I don't have to write.



After all, what used and unmatched tubes looking for in amplifier?

[pdf64]

The second triode inverting amplifier will have much lower output impedance than the first, so the bias ripple present at its ac node will be (much?) more attenuated. So the bias ripple will no longer be common mode.
To exasperate that, the inverting amp will flip the polarity of the bias ripple at the first plate, so the bias ripple will become a differential rather than common mode signal.

[Helmholtz]

Yeah, that's what I am saying.

[Dave H]

To verify that this is the only/dominating effect one could disconnect the 47nF cap at the second PI triode's grid and watch for any changes.

Did you determine PI output impedances by sim?

I don't think there will be any differential ripple signal with that cap removed. It removes the second triode's NFB making its output impedance the same as the first triode and the second triode no longer inverts the output of the first triode.

I didn't measure output impedances by sim. I just thought the first triode's output impedance would be about 35k (for a 12AX7) and for the second triode it would be about 35k divided by its loop gain.

[Helmholtz]

I don't think there will be any differential ripple signal with that cap removed. It removes the second triode's NFB making its output impedance the same as the first triode.

Good point, thanks.

I didn't measure output impedances by sim. I just thought the first triode's output impedance would be about 35k (for a 12AX7) and for the second triode it would be about 35k divided by its loop gain.

Just curious, because Zollner states loaded output impedances of 33k for the first triode and 12k for the second triode of the paraphase in a Fender Deluxe 5D3, which indicates only a moderate 1:3 ratio.

[Dave H]

Just curious, because Zollner states loaded output impedances of 33k for the first triode and 12k for the second triode of the paraphase in a Fender Deluxe 5D3, which indicates only a moderate 1:3 ratio.

5D3 has the second triode's cathode unbypassed. I think that halves open loop gain and doubles output impedance. Would that account for its high output impedance?

There's also a 56k from the "virtual earth" to ground and it's not a true virtual earth. Could that increase output impedance somehow?

[Helmholtz]

5D3 has the second triode's cathode unbypassed. I think that halves the open loop gain and doubles the output impedance. Would that account for its high output impedance?

Yes, there are certainly significant differences, including a much lower plate supply voltage (and it's a 12AX7 not a 12AT7). But I doubt if these can account for a difference in output impedance of a factor of 20 or so.

There's also a 56k from the "virtual earth" to ground and it's not a true virtual earth. Could that affect output impedance?

Rather not, as signal current through the resistor (from "virtual ground" to actual ground) should be negligible.


If you could tell me the output signal drops (from sim) at the PI plates with additional 225K AC loads, I volunteer to calculate the output impedances.

[catalin gramada]

Can be introduced in simulation a build-out resistor to balance the output impedance as M Blencowe suggested, please?

http://deewm.com/Webcache/The_Valve_..._Inverter.html

I thinking 10k will be pretty much on spot