I often wonder, when reading all these posts like 'I build this CFA power amp and I can say it sounds much better than my CFA power amp', who actually DID do that?
jan
You can start with a badly designed VFA and then try an average
CFB , no doubt that it will sound better , moreover if it use good
devices while the former was stuck with some 2N3055 like OPS.
For a bit of history, as a salesman (I think) Carl Nelson founder of Comlinear saw the folks at HP Loveland Co. using the "CFA" circuit for high speed pulse processing, and saw dollar signs. At a given level of process speed the "CFA" type circuits continue to obtain better high speed pulse fidelity, DC performance is secondary. Their applicability to audio and low frequency instrumentation was never a factor.
Trying to apply some beginner logic...
So there is at least one advantage of a "CFA" front end. If you avoid using shunt compensation then you could keep the front end in class A with the safety net of the class B operation. This could then allow the use of smaller value shunt capacitor at the input and keep phase shift at 20Khz much lower.
A small advantage admittedly but along with keeping clipping behaviour under control you could end up with a better behaved amplifier under extreme conditions.
So there is at least one advantage of a "CFA" front end. If you avoid using shunt compensation then you could keep the front end in class A with the safety net of the class B operation. This could then allow the use of smaller value shunt capacitor at the input and keep phase shift at 20Khz much lower.
A small advantage admittedly but along with keeping clipping behaviour under control you could end up with a better behaved amplifier under extreme conditions.
Bonsai, how else would you explain the so-called "CFA's" propensity (with a complimentary push-pull input stage) to generate very high slew rates, well in excess of those supported by the stage's quiescent current?
In other words, where do think the relatively high currents (higher than the standing current of the input stage) come from to drive the compensation capacitor during slewing?
You run the front end 'rich'
You use a front end filter
Ccomp is small relative to VFA so if you do the above, no class AB
You can confirm this by simming it
Your observaions are right on the bulls-eye mark !!
[except being th only one who has deigned/built/listened to both topologies]
Thx-RNMarsh
You really haven't answered my question.
Nevertheless, how do you run the front end "rich" to begin with, and how do you run it "rich" enough to slew at 1KV/uS, for instance, without exceeding your front end's dissipation limits?
And, no, the shunt compensation capacitor is unlikely to be smaller than the minor loop feedback capacitor used in a VFA of the Thompson variety.
Thank you.
This shunt capacitor is not mandatory to compensate a CFA. As you already know you can use a form of miller compensation to compensate a CFA. Then you can compare the current output of a VFA front end with the current output of a CFA front end operating in class A. Whichever can produce the most current will win out in the slew rate war.
Comparing apples with apples is possible where the only difference in topology is the front end.
It would be good for CFA designs to get away from assuming shunt compensation is the only way forward.
This shunt capacitor is not mandatory to compensate a CFA. As you already know you can use a form of miller compensation to compensate a CFA. Then you can compare the current output of a VFA front end with the current output of a CFA front end operating in class A. Whichever can produce the most current will win out in the slew rate war.
Comparing apples with apples is possible where the only difference in topology is the front end.
It would be good for CFA designs to get away from assuming shunt compensation is the only way forward.
In a VFA you use a combination of tail current, degeneration and Cdom ( assuming MC) and a front end filter to ensure the LTP never overloads (analogous to class B operation when one of the Diff transistors is fully off and the other fully on).
Using a similar bag of tricks, you can also set the CFA front end to ensure it never overloads either.
Using a similar bag of tricks, you can also set the CFA front end to ensure it never overloads either.
The majority of so-called "CFAs" require shunt compensation because the input stage is coupled to the second by a current mirror, viz. you cannot apply Miller compensation.
While I can see a benefit to using the current mirror to isolate the loading from the front end. To then replace it with another load (shunt comp) seems to be at little counter productive. It does allow for a defined loading of the front end but at the apparent expense of precluding the use miller comp.
So I understand you can constraint your dislike to CFAs "coupled to the second by a current mirror". Perhaps a majority in ICs, but definitely not in discrete audio.
To add insult to injury, most of the CFA audio power amps are running at closed loop gains that makes the CFA effects ("open loop gain controlled by the feedback network") insignificant. For the same reason, even those CFAs with current mirror coupling can be in majority Miller compensated, as the Miller cap is to small to significantly impact the open loop gain, anywhere close to the ULGF. Therefore, your dislikes are a) on terminology and b) the typical CFA current on demand input stage (that can go Class B on fast transients, has poor PSRR, higher noise, etc...). As b) is easily tamed by limiting the rise time at the input and by running the input stage hot enough, the PSRR has also solutions, noise in power amps is not an issue, what remains is the CFA terminology. Which you have the right to deny as much as I have the right to accept and use when appropriate.
Good night.
I am with you. Being at the middle.
0.01% for THD margin and 0.02% for only H2. I feel that non clipping THD doesn't change the sound quality, its only added harmonics. It is feedback that change quality, but I am not saying feedback is bad thing, since my CFA amp using 80dB loop @20k and more than 110dB loop @1k.
You are not comparing its flexibility when improving feedback amplifier stability.
Looks like you know about phases.
Is there so called damping phase exsist as good and bad?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.