CFA Topology Audio Amplifiers

Jump to Latest
#3,941
jcx said:
the "phase accumulation" from high gain at low frequency is actually rather weak when you understand the weighting factors

the Bode integral of feedback sums log(gain) over linear frequency

40 dB of "extra" loop gain DC to 20 kHz only "costs" as much as 1 dB of gain over a 1 MHz range

of course there is the practical added cost of the sloping gain over the necessary roll off frequency range - which shows the value of higher order compensation

another way to look at it is the phase from the zero of the open loop gain (the hf pole, veiwed from the right, as a zero cancelling the 90 degree integrating gain) - in the "high bandwidth" CFA example this is < 100 kHz - so the zero is adding < 5.7 degrees to the CFA "phase advantage" by 1 MHz

so the low frequency gain "Excess" of VFA is not really the 1st most important determining factor for stability at loop gain intercepts > 1MHz

the bigger cost in phase shift are excess poles within < 1 decade of frequency

the output Q are usually limiting

for the VFA/CFA difference the places to look are diff pair "added" Q, mirror, tail and VAS for the poles causing the loop gain droop beyond 5 MHz

properly applied small signal Q can easily have ft > 100 MHz, so "extra transistors in the path" aren't per se a big concern with ~ 10 MHz ops Q
Click to expand...

Good points jcx - this certainly makes sense if you look at the HF roll off beyond the ULGF as well. On a VFA it drops off rapidly and that must be due to the HF poles you mention. 'Excess phase' perhaps an unfortunate choice of phrase on my part.
 
#3,942
Bonsai said:
... this certainly makes sense if you look at the HF roll off beyond the ULGF as well. On a VFA it drops off rapidly and that must be due to the HF poles you mention. 'Excess phase' perhaps an unfortunate choice of phrase on my part.
Click to expand...

Yes. JCX's point more or less parallels what I said in my posts to you earlier, but more detailed.
I think Bob also made similar points so we have a consensus now.😉
It is much clearer now we have plots to examine and not just hazy verbal descriptions.
I didn't understand your use of "phase accumulation" because I visualise that phase accumulates as frequency increases.
But you could consider that it accumulates as frequency decreases towards the ULGF, the maths is quite symmetrical.

Best wishes
David
 
#3,943
Waly said:
And here is the explanation
... one step further and discuss exactly why the VAS output impedance behaves differently for the two topologies...
Click to expand...

Thank you for your explanation.
I have not checked details yet but it is consistent with my previously posted view that any practical differences between VFA and "CFA" response should be explainable in terms of the pole and zero placements.
And then, similarly, that the placement should be explainable in terms of the circuit immittances.
So I am keen to read more about the VAS output impedance and any other related topics.

Best wishes
David.
 
#3,944
Waly said:
...There is a frequency (here, around 3MHz) at which the two output impedances are equal, from there up, the CFA output impedance becomes lower than the VFA output impedance...
Click to expand...

I asked Andrew for the ASC because I wanted to check the effect of the different VAS feedback structures.
I plan to place the probe at the VAS because this reveals the effective total loop gain.
When the probe is placed like Andrew's, at the OPS, then the different feedback around the two VASs will be seen as different VAS output impedance.
This is classic Blackman's Theorem and is an example of what I call different mental models. Can be seen as an impedance difference or as a result of feedback.

Best wishes
David
 
#3,946
Bonsai said:
. your analysis up Dave... Hopefully, your results will allow us to peel the onion back a bit more.
Click to expand...

Yes, nice to have actual data, yours was educational. Lazy Cat claimed "back to square one" but I say on the contrary.
Could you please post your models?
I would like to make my results precisely consistent with yours, for obvious reasons.

Best wishes
David
 
Last edited:
#3,952
Waly said:
And here is the explanation.
Click to expand...
Bingo!
Waly said:
A lower VAS output impedance at HF, combined with a constant input capacitance of the output stage (not discussing yet the bias modulation effect in the OPS) renders a higher output stage pole. Has this anything to do with the global feedback loop? No.
Click to expand...
Oooppzz...?

The common limitation is that many good engineers cannot hear well. Listening ability will help I believe. How?

Build and listen some CFAs, can you hear differences that is unique or common to the CFAs? Correct, CFA can sound subjectively worse than VFA, and IMHO, the chance for sounding bad is higher than VFA. But can you hear something "unique" to these amps? (This special trait may trick people into thinking that these amps are better).

If you can hear this sound character or trait (sorry, I'm not going to explain), then it is hard not to think that it is the direct feedback loop that makes this difference.

Now, to directly refute the idea, build two amps (CFA and VFA) that will sound the same. May be you can, but I doubt it.

There's one more important thing than just knowing the special trait, it is how to build a good sounding one. Many here trying to go to that direction but many also going nowhere trying to figure out what is a CFA, a VFA with different first letter.
 
#3,953
Dave Zan said:
Have you done the simulations that Waly discusses?
If so what were your results?
Click to expand...
I already published a comparative study at the early begining of this thread:
http://www.esperado.fr/temp/VSSA/vssa-vs-vfa.html
Quite the same than Bonsai apart no diamond input and a landscape seen from a slightly different point of view ;-).
May-be you'll have less problems with my models ?
 
Last edited:
#3,954
Dave Zan said:
I receive a fatal error when I try to run your sim with the extracted "bip models" file...

Closer inspection also reveals dissimilar models in the OPS of VFA and "CFA"
One is QNJW0302g, other is QMJL1302a.
Click to expand...

Extracted just the needed models and was able to run the simulation.
I made the bias for the OPS identical, I don't know why it was different in Andrew's attachment.
Once I made the transistors in the VFA match the "CFA" and made the feedback network identical then the hi-frequency differences between them almost vanished.
The "CFA" has open loop 0 dB just above 14 MHz. The VFA just below 13 MHz.
Theory predicts that the OPS pole should be similar so VFA and "CFA" differences should be small.
The results are consistent with theory, in fact the difference is even less than I expected.

Best wishes
David.
 
Last edited:
#3,956
Esperado said:
I already published a comparative study
Click to expand...

A quick look shows results that are also consistent with theory.
Your test "CFA" has more gain than the comparison VFA practically everywhere so naturally distortion is lower.
Unusual for a "CFA" to have so much gain. Well worth my closer study to see how that is achieved.

Esperado said:
Now, time to build the two and listen ;-)
I bet differences will be much more than expected.
Click to expand...

I bet differences will be much less than expected, if we only listen i.e. blind.
In fact I will bet a meal in a Michelin starred restaurant that it's not detectable in a test that meets normal standards of peer review.
Where do you live in France? I should be over later in the year.

Best wishes
David
 
Last edited:
#3,958
Dave Zan said:
A quick look shows results that are also consistent with theory.
Your test "CFA" has more gain than the comparison VFA practically everywhere so naturally distortion is lower.
Unusual for a "CFA" to have so much gain. Well worth my closer study to see how that is achieved.
Click to expand...
The purpose of this comparison is to compare the two topologies everything equal.
Both versions have EXACTLY the same gain.
I design CFA from decades and had experimented a lot of various gains and topologies. Optimization is all the fun.
Dave Zan said:
I bet differences will be much less than expected, if we only listen i.e. blind.
Click to expand...
Please, do as i did, don't bet, listen 🙂
 
#3,960
Dave Zan said:
I bet differences will be much less than expected, if we only listen i.e. blind.

Best wishes
David
Click to expand...

If my CFA(or any other) can beat my TT VFA amp in blind listening test I could convert to the CFA fan, not that I think that VFA is better, just good amps not easy to distinguish in blind listening test.
Sooner or latter I have to build my CFA amp, just to many project waiting in my workshop.
BR Damir
 
You must log in or register to reply here.
Top Bottom