CFA Topology Audio Amplifiers

[FdW]

My TL072 is fed from the cap multipliers , to the locally decoupled IPS board
.. then to the 12V zeners /47u caps with "to the pin" .1uf caps on the
IC. The 2-47u's and .1u's are within mm's of the IC.

As I am using a slower IC , this would not be enough ??
(PS components circled)
OS

It may be enough, but I think it is better to do a bit more (as advised by some of the IC suppliers). It is (for instance) a good idea to 'short' HF from rail to rail and not to shunt it into ground.

 

[JOHN BALI]

It may be enough, but I think it is better to do a bit more (as advised by some of the IC suppliers). It is (for instance) a good idea to 'short' HF from rail to rail and not to shunt it into ground.

So we only need 1 cap, no need 2 more that shunt to ground?
or 3 caps in total 😕

Btw, why not use LF412 for servo instead of TL072...

 

[Esperado]

It may be enough, but I think it is better to do a bit more (as advised by some of the IC suppliers). It is (for instance) a good idea to 'short' HF from rail to rail and not to shunt it into ground.

One is better for common mode, the other for differential. I believe both are better.
But decoupling to ground is a requisite.

 

[Waly]

but some commercial arguments he likes to push, that i consider like snake oil.

Well, I would think that, in audio, the snake oil has to be branded and subjectively evaluated as well. It seems like every audio expert has his own snake oil, and then any other brand is snake oil. Your snake oil makes no exception.

 

[Petr1951]

Esperado (3815)
It is one of the questions witch remain unanswered in this thread, as nobody had the occasion to compare some complicated/sophisticated CFA designs with simplest optimized ones, VSSA like.
One thing i would dream we can organize, at the end.


Your colleague in the past engaged in professional sound, also have long been searching circuit solutions for the most natural sound.
He recently admitted that the best for getting a good sound is the amplifier without overall negative feedback.
The figure below shows the VAS out stage with a negative output impedance.

[/URL][/IMG]
http://radikal.ru/fp/05afa9d4ab854f768fae358dd0de6a88][/URL]

[/URL][/IMG]
http://radikal.ru/fp/b5e3cca231ed45da9650d46d43561ad0][/URL]
best regards
Petr

 

[Waly]

However there is another factor that is being overlooked, I doubt that simulation will demonstrate the effect thoroughly.

I don't see any reason why simulation would not be able to demonstrate this effect, if it exists.

You are making an extraordinary statement, so you'd better come up with solid arguments about. Unfortunately, so far, your descriptive explanation doesn't make any sense, and simply starting with "Bob I agree with you" doesn't help. Mr. Cordell was talking about a lead-lag compensation at the VAS output. As I mentioned myself, that's a Miller loop compensation aide and has nothing to do with the shunt dominant pole that you are wrongly promoting as the holy grail for CFAs.

You already got the explanation why the shunt compensation is suboptimal, all is needed is you to read it and digest it.

Now go ahead and start ranting.

 

[Edmond Stuart]

So we only need 1 cap, no need 2 more that shunt to ground?
or 3 caps in total 😕

Of course you need more caps, at least a cap between gnd and the supply rail to which the internal compensation cap is referred. In most cases, this is the negative supply rail.

Btw, why not use LF412 for servo instead of TL072...

The LF412 is much better, lower input offset voltage.

 

[FdW]

So we only need 1 cap, no need 2 more that shunt to ground?
or 3 caps in total 😕

Btw, why not use LF412 for servo instead of TL072...

If, even, a picture was not enough to show that, then I rest my case.

 

[john curl]

I have designed CFA amps in the past, both with bipolar and fet inputs. I found that they are too thermally 'messy' to use normally, BUT they are potentially better than a typical differential pair input due to the feedback loop feeds closer to the very input, than a typical differential pair.
I have avoided contributing here, to keep my potential for added controversy down, and the last time I tried to convey my experience with CFA amps to Richard Marsh on the phone, he hung up on me. Best I avoid further problems.

 

[Waly]

the last time I tried to convey my experience with CFA amps to Richard Marsh on the phone, he hung up on me.

 

[Esperado]

I have designed CFA amps in the past, both with bipolar and fet inputs. I found that they are too thermally 'messy' to use normally,

Hi, John. Nice to see you here.
Did-you refer to offset, or Quiescents ? What kind of OPS do you used ? Laterals or BJT ?

The LF412 is much better, lower input offset voltage.

the requisit is, indeed, high impedance (to can use film caps, means FET input), low offset (it would be stupid to try to cure offset with something witch can bring its own) and, if possible, low noise.

 

[Bob Cordell]

Is that what you found from reading the entire works? I am not sure about that.

IMO - Esp interesting to me are digital processes and what they do to phase and its audible.


THx-RNMarsh

Yes, anyone who knows about loudspeakers, loudspeaker crossovers, and the Linkwitz Riley crossover knows that it is an all-pass.

Cheers,
Bob

 

[Esperado]

the best for getting a good sound is the amplifier without overall negative feedback.

On a pure intellectual way, it seems logical to do not ask a stage to cancel distortions of an other one.
I have, personally, no experience on "no GFB" with power amps. Neither on negative impedance. No instability ? No added distortion to the speaker coil, due to Eddy currents ?
Thank for sharing.

 

[Bonsai]

TL71/72 for servo should be ok. No need for expensive opamps.

 

[31697B]

Yes, anyone who knows about loudspeakers, loudspeaker crossovers, and the Linkwitz Riley crossover knows that it is an all-pass.

Cheers,
Bob

Hmmm. Thats not what I meant. But, the work referred studies the 'phase spectrum' and finds many conditions which are audible. L-R crossover might fall into the aubible affects. If so, circuit designers can use the info contained in the article to design better crossovers -- ones with minimum added audible artifacts/affects. Seems we often solve a problem only to create another.

-RM

 

[31697B]

I have designed CFA amps in the past, both with bipolar and fet inputs. I found that they are too thermally 'messy' to use normally, BUT they are potentially better than a typical differential pair input due to the feedback loop feeds closer to the very input, than a typical differential pair.
I have avoided contributing here, to keep my potential for added controversy down, and the last time I tried to convey my experience with CFA amps to Richard Marsh on the phone, he hung up on me. Best I avoid further problems.

John, you must have me confused with someone else who hung up on you.

No one here has thought of thermal issues as a problem with CFA. They seem quite stable that way. No more or less than VFA. Many OEM use CFA for decades without thermal issues. Strange thing to say about CFA.... can you elaborate? Or, was it just your particular implementation?

Thx-RNMarsh

 

[MiiB]

I have several posts ago shoved a CFA front-end where the current is thermally compensated by a CCS, which has opposite tempco of the input pair. (The four transistor glued together) If currents then are mirrored into the VAS. Then there's no drift at all. I have built several and seen it by practical implementation. The first versions I built was terrible as I had to run everything very low in order to have room for the drift. The latter cures it.

 

[31697B]

For the many intellectuals here --- more reading material for you:

Texas Instruments AN840. www.ti.com/lit/an/snoa247b/snoa247b.pdf

Similar work can be found in the Microelectronics Journal.... such as Journal 30 (1999) 841-849. "New Precise SPICE macromodels for the current-feedback operational amplifier."

More views and description for understanding the subtle and not so subtle operation of CFA.


Thx-RNMarsh

 

[ostripper]

TL71/72 for servo should be ok. No need for expensive opamps.

But , with many being pin compatible with the TL .. one still has a choice.
The dual package is the way , as SO many NE5532's exist in the waste stream.
Fixed the Caps !! (closer Vcc-gnd's + /+vcc to -Vcc 3 cap)

OS

 

[john curl]

Well I was getting somewhat 'high handed' Richard and it was not the only time that you hung up on me. I haven't called you since, you might notice. It's OK.