Your opinions are sought on Audio Power Amplifier Design: 6th Edition. Douglas Self

[DouglasSelf]

But otoh it is an extremely linear structure, keep it in class A and you have one hell of a driver stage.

Jan

Hi Jan.

It struck me that using a CFP compound driver stage for an EF output might be a good notion, but in sim I could find no improvement in either crossover distortion or Large Signal Nonlinearity. (LSN) What circuit configuration did you have in mind?

 

[jan.didden]

Here's one possibility (CFP is bootstrapped).
Not my own circuit to be honest.
But I believe the QUAD 303 was similar, no?

Jan

 

[DouglasSelf]

Here's one possibility (CFP is bootstrapped).
Not my own circuit to be honest.
But I believe the QUAD 303 was similar, no?

Jan

My version had CFP drivers driving the output EFs directly. I would call your circuit CFP pre-drivers.

The 303 output was different- output triples with local NFB round all 3. (APAD6 p247) Not easy to keep stable, with modern devices, anyway.

 

[manso]

Hi Jan.

It struck me that using a CFP compound driver stage for an EF output might be a good notion, but in sim I could find no improvement in either crossover distortion or Large Signal Nonlinearity. (LSN) What circuit configuration did you have in mind?

Youre sure about this ???

See some Lin designs like LK280, these use cfp driver. See what happens at high frequencies.

 

[Bigun]

I used CFP driver for single EF in my TGM3 - essentially a Triple but I thought at the time it would give less phase shift than a triple EF and more efficient use of power rail headroom?

 

[manso]

Contrary to Doug I find improvement in LSN and in case one uses a setup such as Lin amps the LSN is apreciable better especially at higher frequencies.

 

[Dave Zan]

... a CFP compound driver ... for an EF output ... but in sim I could find no improvement ...

Theory implies this should work well but when I tried it I also found no improvement at first.
The theory is pretty well based so I was quite surprised and curious.
I did some study and research to find out what factor(s) I had overlooked.
Turns out the output impedance of what drives the OPS is a major complication.
Easy to have zero in a Spice simulation but it varies in a complicated way in a real "VAS".
This also interacts with the output base stopper resistors in a non-linear way, to further complicate matters.
In the middle of another article for Linear Audio on this.

See some Lin designs like LK280, these use cfp driver.

CFP has the potential to be excellent but it is not simple to beat an EF.
Thanks for the reference, I am curious to see their solution.

Best wishes
David

 

[Dave Zan]

I used CFP driver ... less phase shift than a triple EF and more efficient use of power rail headroom?

It is an attractive possibility, isn't it!
Did you compare with triple EF to see how it worked out in your case?

Best wishes
David

 

[dadod]

My version had CFP drivers driving the output EFs directly. I would call your circuit CFP pre-drivers.

The 303 output was different- output triples with local NFB round all 3. (APAD6 p247) Not easy to keep stable, with modern devices, anyway.

There is a thread with that type of OPS http://www.diyaudio.com/forums/soli...-stage-audio-power-amplifier.html#post1349109, and also a group buy thread.

 

[vynuhl.addict]

Harmon Kardon used the CFP driver followed by an EF in a few of their older amplifiers. Transistor choices are finicky, and with poor choices can lead to much more phase shift than a triple. Having the negative loop of the CFP within the global negative loop requires some more care in compensation, when I tried this arrangement it would only remain stable with a slower slave device and 10r from collector of the slave to the emitter of the pre driver.


Colin

 

[jan.didden]

Turns out the output impedance of what drives the OPS is a major complication.
Easy to have zero in a Spice simulation but it varies in a complicated way in a real "VAS".
This also interacts with the output base stopper resistors in a non-linear way, to further complicate matters.
In the middle of another article for Linear Audio on this.

Best wishes
David

Hi David,

That surprises me, I was under the impression that the CFP as a (pre)driver in class-A presented a nice linear load impedance to the VAS/TIS. Was this with the CFP in class-A?

Jan

 

[jan.didden]

My version had CFP drivers driving the output EFs directly. I would call your circuit CFP pre-drivers.

The 303 output was different- output triples with local NFB round all 3. (APAD6 p247) Not easy to keep stable, with modern devices, anyway.

Yes indeed.

 

[Dave Zan]

That surprises me, I was under the impression that the CFP as a (pre)driver in class-A presented a nice linear load impedance to the VAS/TIS. Was this with the CFP in class-A?

It surprised me too, that's why it's so educational!
I think of it as the source impedance that the OPS looks back into, rather than the load impedance presented to the VAS/TIS.
The CFP was in class-A.

More tomorrow, it's Friday with a friend now.

Best wishes
David

 

[Bonsai]

I suspect that the CFP stage Zo is ok up to a point and then increases rapidly with frequency as the EF OPS Zi drops off at HF and the result is a very difficult animal to stabilize. This might explain the fact that the practical results don't quite meet expectation.

 

[Lazy Cat]

I suspect that the CFP stage Zo is ok up to a point and then increases rapidly with frequency as the EF OPS Zi drops off at HF and the result is a very difficult animal to stabilize. This might explain the fact that the practical results don't quite meet expectation.

BIGBT has both characters in balance, inherently stable, very easy to work with, very easy to drive, made a lot of amps with this output.

 

[Bigun]

It is an attractive possibility, isn't it!
Did you compare with triple EF to see how it worked out in your case?

Best wishes
David

No comparison was made, but I remember it (TGM3) sounded very good. I had a simple singleton input, simple VAS and cfp driving the EF. I can't remember why I didn't take this design to a final pcb, I may have to revisit that one day.

 

[Dave Zan]

... I built the probe, using a commercially available pulse transformer... Works fine for 1KHz up to 6-7MHz.

Does your probe do the correction for non-ideal injection point impedance?
Either the Middlebrook 1975 version or the later version.

Best wishes
David

 

[Dave Zan]

... Tain/dual injection is a pain ... stepped traces which makes it very difficult to move, copy, duplicate in a side-by side comparison

It helps to put the equations into the Plot Def. file but it's still inconvenient.
It would be easier to make stepped comparisons if the test could be done in one step.
I think this should be possible with simultaneous measurement of the injected current flow both forward and backwards.
But if it was easy I expect someone would have implemented it.
Perhaps people are just blocked by Middlebrook's technique, but that was intended to minimize sensitivity to experimental inaccuracy.
What do you think?

Best wishes
David

 

[AndrewT]

.............See some Lin designs like LK280, these use cfp driver. See what happens at high frequencies.

Lk280 is Linn manufacture.
Lin is different.

 

[yldouright]

Since a lot the heavy weights from the CFA vs. VFA thread are here, I thought posting this link discussing audio voicing would get pragmatic input. The discussion there ties in this thread with the CFA vs. VFA thread.