Partial FB: 2 SE + 1 PI = PP?

[mashaffer]

I have been looking at partial feedback in PP amplifiers and the increase in complexity and trickiness seems quite a bit greater. I have looked at the possibility of LTP ala Yves and Baby Huey and it seems that those were quite difficult to get right. I also considered Paraphase which has the advantage of symmetrical impedance and FB ratio but rather unsymmetrical HD spectra.

Then the thought occurred to me. Why not just use two SE amps with a PI as the front end? Lets say you used a simple cathodyne PI as the input stage of the amp and fed the output of each phase to what amounts to an RH84 amp with PP OPT across the output plates. Why would this not work?

[Sch3mat1c]

Way more parts, terrible efficiency (class A only) and high distortion. If you connect the speaker between outputs (i.e., "bridged" mode), you at least cancel the 2nd harmonic. In the more general case (including class AB, etc., which requires the biasing of a proper PP amp), then it is exactly an SE amp doubled up somewhere along the line.

Nothing hard about an LTP. Having any specific problems?

Tim

[mashaffer]

Sorry I wasn't clear. What I was essentially trying to describe is a normal PP output stage with a CC driver after the PI having partial feedback to the driver's plates. That way you can have the simplicity and balance of the cathodyne but a symmetrical impedance for the feedback to work into. No need for CCS, B-, or any of that.

Would a traditional resistive LTP provide symmetrical enough output impedance to keep the two feedback paths working together? I presume that you would need to have the tail long enough to use the same plate load on both halves.

BTW. I am looking at using low microphonics preamp tubes in this application so am considering 12AY7 if that seems like a reasonable choice. Outputs EL84 UL or pentode mode.

[revintage]

Hey Mike,

PP sounds like good idea. But as said so many, many times before, triode-drivers and anode to grid feedbacked output tubes(tetrode/pentode Schade) don´t mix well at all. Instead go for a cascode LTP and you´re there. Right done you will not need any GFB at all. If I have understood it right you are talking about a three-stage amp that probably will need over 400V B+ for the drivers to work properly. Go for a 2-stage with B- instead!

[mashaffer]

Thanks Lars. Do you think that triode driver in PP will be worse than triode driver in SE?

It seems like cascode LTP would require even higher voltage supply than simple LTP and add the issue of nasty distortion and clipping when swinging high voltages (i.e. pushed hard) unless I am missing something.

I just discovered that EF86 small signal pentodes are in current production. Maybe if triode LTP doesn't work out a LTP of EF86s might do the trick.

[revintage]

Hey Mike,
EF86 would be to wimpy. As there are few, if any, High Gm pentodes produced today, cascode is the alternative.

After my first post I modeled a PP by just turning the second upside down and adjusting a few things. Should work well. It could be refined using gyrators insted of anode resistors. Unfortunetly odd harmonics are dominating, although they will be very low, but is same with all PPs as even are cancelled out.

[Sch3mat1c]

Originally Posted by mashaffer

Sorry I wasn't clear. What I was essentially trying to describe is a normal PP output stage with a CC driver

Define: CC Driver. I don't get where CC comes into this..

after the PI having partial feedback to the driver's plates. That way you can have the simplicity and balance of the cathodyne but a symmetrical impedance for the feedback to work into. No need for CCS, B-, or any of that.

Well, you don't have any of the simplicity of the cathodyne if you have a driver after it...

Cathodyne has symmetrical output impedances. The voltages are equal as long as the loads are equal. In a PP amp with shunt feedback, they will be.

Would a traditional resistive LTP provide symmetrical enough output impedance to keep the two feedback paths working together?

Well, the feedback paths aren't really "working together", they work independently. That independence is fine with an LTP. LTP just steers current.

I presume that you would need to have the tail long enough to use the same plate load on both halves.

As long as the tail is more than 10/Gm ohms, and has enough voltage across it to accommodate the largest input swing, the LTP will work as well as anyone could care.

Typically, an LTP, "floating" at a DC bias, direct coupled to a preamp perhaps, will have its grids and cathodes in the 100V range, and maybe the plates at 200-300V, supplied from 300-400V. The LTP will probably be clipping with 10V input, so there isn't much required from the preamp, and there is plenty of compliance left in the tail resistor (10Vp-p out of 100VDC isn't much change in current). For a 12AU7, you might bias the stage at 5mA, so the tail will be around 100V/0.01A = 10kohms. Transconductance is around 2mmho, so the cathode impedance is about 500 ohms, or 1000 between cathodes. 10k > 10*500, so this tail resistor is big enough not to care.

Tim

[mashaffer]

Thanks for you work Lars. That is an interesting circuit but of course we are back to CCS or huge PS.

By wimpy pentode is it the low gm or low cathode current or both that is of concern? Something on the order of 6AU6 would have more stones right?

I may go see what tube models I have on the LTspice machine upstairs and try a few things.

[mashaffer]

Wow Tim, a lot to absorb there. Thank you. By CC driver I was thinking of the Williamson where the PI is buffered by push pull drivers. To my inexperienced eye the drivers look like common cathode gain stages sharing an un-bypassed cathode bias resistor.

Of course you are right that the cathodyne is impedance balanced but I presume the Z out is quite low and so the feedback would tend to disappear right? That is why I thought a driver would be necessary.

You information on LTP is quite useful. If I follow the math correctly then the required tail for a 12AY7 with 1700umho gm would be similar to the 12AU7 example you gave but the tail resistance requirement for the 12AT7 with gm between 4000 and 5000 would be even lower making it possibly the better choice (assuming microphonics are acceptable).

That might be worth trying as it keeps it simple and keeps that parts count down.

Am I following you correctly?

[tubelab.com]

Lets say you used a simple cathodyne PI as the input stage of the amp and fed the output of each phase to what amounts to an RH84 amp with PP OPT across the output plates. Why would this not work?

It works well. I have been doing this with Simple SE amps (12AT7 driving 6L6GC, EL34, or KT88). I use an external PI in front of the Simple SE board. Some tweaking of the cathode bias resistors, jumpering for pentode or UL operation, and a bit more drive allows a 2 X 8 watt class A SE amp to become a P-P 50 watt mono board.

I have been told that Schade type local feedback does not work with triode drivers. It does work, but not as well as it could with a pentode driver. Most of my P-P applications of the Simple SE board have used GNFB.