CFA Topology Audio Amplifiers

[ontoaba]

Why do you think the diagram you show in your post above is a CFA?

Please correct me, this topology classification may help anyone feels easier, which is CFA and not.
The single input (no.1) looks like a current feedback, is it classified as CFA?

 

[MiiB]

Singleton CFA with cherry compensation. Jfet input and Hawksford cascode. I have great expectations for this circuit and how it performs music.

 

[Esperado]

The single input (no.1) looks like a current feedback, is it classified as CFA?

The problem, with Bonsai classification, and if i understand well his point, is: when you higher feedback impedance, you are limiting current.
At witch moment it is no more a CFA ? ;-)
For me, feedback impedance is just a matter to fix the bandwidth, so, no.1 is CFA, on my point of view.

 

[Esperado]

My personal definition should be: It is CFA when feedback is applied, with no active device in between, to the input transistor's emitter or source, while input signal is applied at the base or gate of the same transistor. For the signal, this active device works as common emitter, while, for the feedback, it works as a common base.
Don't know if Mr. Marsh will agree.

 

[Edmond Stuart]

I agree.

 

[Petr1951]

ostripper
Part 2 of the article - The Class i low-distortion audio output stage (Part 2) | EDN

This output stage as well as the cascade of patent US4595883_Kazuaki_Nakayama works fine on resistive load constant resistance, but also does not work with the external load. Transfer coefficient is also highly dependent on the load, checked when the load changes from 4 to 8 ohms with step 2 ohms.

[/URL][/IMG]
http://radikal.ru/fp/11acaff24b1b4e22a34f603bd94cb1c7] [/URL]

regards
Petr

 

[Max Headroom]

Merci...

My personal definition should be: It is CFA when feedback is applied, with no active device in between, to the input transistor's emitter or source, while input signal is applied at the base or gate of the same transistor. For the signal, this active device works as common emitter, while, for the feedback, it works as a common base.

Thank you.

Dan.

 

[Bonsai]

The CFA will morph into a VFA. The point at which this happens is when the CLG and the -3dB CL BW become interdependent. This will happen at high CLG.

But, it's easier just to apply the two tests to determine if it's a CFA or not.

By these two tests, the circuit you posted earlier Esperado is not CFA - it's VFA because the peak current into the TISs is set by the LTP current sources.

 

[Waly]

Here is a short summary of CFA and VFA feedback modes.

The CFA will morph into a VFA.

It's good that you posted this document; finally, everybody can see what this is about.

Your notation of V and I doesn't make much sense. In any op amp, the current into the inverting and non-inverting inputs is zero (hence the "virtual ground" concept.

Now that you confirmed that the VFA and CFA feedback have the same topology, go no further than the first column in your table, the series-shunt feedback. Following the canonical formalism, in A/(1+AT) (the closed loop gain expression), A=Vo/Vi is the op amp open loop voltage gain and T=Rg/(Rg+Rf).

When calculating A, we need to consider (again, this is the canonical analysis, and make no mistake, there is no difference between the VFA and the CFA) the feedback network loading effects. The op amp output is loaded by Rf+Rg and the inverting input is loaded by Rg||Rf (that is, Rg in parallel with Rg). Now we can calculate A, which obviously depends on the op amp internals.

Now, if the open loop gain A does not depend on the inverting input loading Rf||Rg (like the case for a LTP input stage) we have a pure VFA. However, if due to the op amp internal topology, the open loop gain depends on Rg||Rf, the we have a CFA. Please note, the feedback loop has no bearing on this property. For all purposes of the feedback theory, the CFA is still a VFA, but with some special properties of the op amp.

About the gain-bandwidth properties of the CFA. It is wrong that the CFA does not follow the constant gain-bandwidth rule as the VFA. If we agree to call a CFA the op amp with the special properties, then we may note right away that this op amp follows precisely the constant GBW rule, for the open loop gain. When the loading of the inverting input changes, the open loop gain, and the open loop bandwidth product, is constant!

Now, can you see the fallacy? You call first a CFA some op amp which has a special property for the open loop gain, the you call again a CFA a circuit with closed loop (that doesn't follow the GBW constant rule, like a VFA).

You correctly noted that CFA's degenerate to VFAs for large closed loop gains (and my comment was that for all audio purposes, the so-called CFAs behave almost like pure VFAs).

I guess you can not better understand my CFA and VFA definitions: VFAs are "op amps" having the open loop gain independent on the inverting input loading, while CFAs are "op amps" having the open loop gain depending on the inverting input loading. Otherwise said, a CFA is an amplifier that has the open loop gain modulated by the feedback network (since Rg||Rf are both defining the feedback network).

Still with me? You will finally note that the slew rate is nowhere in this discussion. this is a small signal discussion, and slew rate is essentially a large signal property. You already got an example of a VFA with "current on demand" (the Stochino) having 300V/uS slew rate. Anyway, I think it's obvious that the large slew rate property has nothing to do with the CFA or VFA. You can have a high slew rate (or low slew rate, for what is worth) in both. In all cases, the slew rate is defined by charging the Miller (or equivalent) compensation capacitor. I would concede that CFAs have a native way to do this, while VFAs need extra tricks. But ultimately, do you think it makes any difference how this slew rate is created in the amplifier?

I disagree with your assessment that the VFA current at the TIS input is limited by the LTP current, why would you assume that all VFAs (defined as above) have a LTP input stage? As much as why all CFAs would have the current at the TIS input defined as Vo/Rf? You are here extrapolating the properties of a particular topology (for example, diamond buffer input stage plus a TIS) to all possible CFAs, which is incorrect. The same applies to the statement that CFAs have always one gain stage. Designing a 2 gain stage CFA is really trivial.

Now that I'm done, I can't stop thinking that I wasted another good chunk of electrons to post this. I'm pretty sure it won't persuade anybody to adjust their understanding of the so called CFAs. So be it.

 

[Bonsai]

Waly, this is you at your argumentative best.


The fact that the closed loop gain equation derivations between CFA and VFA are completely different also seems to have escaped you.


I am honestly not going to waste my time trying to convince you. Live in your CFA=VFA world, and continue to miss-apply feedback canonical forms to amplifier topology.

 

[Waly]

You've completely missed it.

I give up. I honestly give up trying to explain it to you.

Somehow, I don't believe you even read what I wrote. You are obviously a marketing person that can't make a difference between the feedback network properties and the intrinsic properties of the op amp topology. If you prefer to put everything conveniently under the same marketing label of "CFA", then so be it.

I give up myself explaining to you that your test and points, being tied to a very specific topology, are in general flat wrong.

 

[Edmond Stuart]

Hi Andrew,

The CoD adds only circa 15% more SR over and above the standard MC (250 V/us to the claimed 300 V/us).
[..]

Only 15%? It depends on how you look at it. Compared to LTP's without the cross-coupled C and R (200R and 1nF) and keeping the collector currents of the input trannies the same (i.e. 2mA, by lowering the tail currents from 12 to 4mA), then the CoD is much larger, say 100%.

Cheers, E.

 

[Waly]

The fact that the closed loop gain equation derivations between CFA and VFA are completely different also seems to have escaped you.

Beeeeep, wrong! From a feedback (type) perspective, the closed loop equations are exactly the same: A/(1+AT) at all frequecies where A=Vo/Vi is a voltage gain and T is Rg/(Rg+RF), for both what you call a "CFA" and a "VFA". The fact that for a particular "op amp" circuit intrinsic topology A (the open loop gain) is also a function of the inverting input loading Rf||Rg has no bearing with the feedback type. It's still voltage feedback and the amplifier is a VFA.

That's why I defined a "CFA" as an "op amp" with the special property of having the open loop gain depending on (or "modulated" by) the inverting input loading. Again, the open loop gain.

CFA is a confusing denomination, it could be named for example (certainly less confusing) LGM (Load Gain Modulated) amplifier. Apply voltage feedback to a LGM and you get your beloved "CFA".

Over and out.

 

[scott wurcer]

It's good that you posted this document; finally, everybody can see what this is about.

Your notation of V and I doesn't make much sense. In any op amp, the current into the inverting and non-inverting inputs is zero (hence the "virtual ground" concept.

You guys are still at it, amazing. The above statement is inaccurate, as I pointed out mikeks. All the current that charges the comp cap and makes the voltage at the gain node move MUST flow in the inverting input. Grounding the inverting input and looking at the open-loop transfer (rougly related to the gm and Cc in the same way as a VFA) is a relatively pointless construct. See any number of EE texts (such as Gray and Searl Electronic Principles, Physics, Models, and Circuits 1969 pages 660-661) for the extraction of a and f for the feedback equation (Ao/1+af) for the simple emitter feedback doublet. The feedback network is considered part of the forward transfer and the input is the voltage between the plus input and ground not the voltage between the base and emitter.

So maybe we should just stop calling it an op-amp.

EDIT - So I see we agree on something maybe, Things are the same but different, A contains terms from the feedback network in one case but not the other. So in one case the feedback network can modify the closed-loop BW. Think about an 8 legs where the user has no access to the frequency compensation, is this just a marketing trick or a useful differentiating feature to the user?

 

[Waly]

So maybe we should just stop calling it an op-amp.

EDIT - So I see we agree on something maybe, Things are the same but different, A contains terms from the feedback network in one case but not the other. So in one case the feedback network can modify the closed-loop BW. Think about an 8 legs where the user has no access to the frequency compensation, is this just a marketing trick or a useful differentiating feature to the user?

If you insist at looking at it as an "op amp" with current flowing (only) in the inverting input, then probably the "op amp" denomination is not entirely correct *). But I consider this view as uselessly clouding the core issue, there is no need to consider this current to explain the properties of the CFA and yes, we agree on your EDIT. That's enough to explain the CFA properties. I am not debating the CFA properties, only that, strictly from a feedback perspective, the "CFA" denomination is confusing and ultimately incorrect. "CFA" is related to a particular "op amp" topology and open loop property, and not to a particular feedback type/topology.

P.S. I'm not mikeks

P.P.S. If the current is flowing in the inverting input, and you consider the "op amp" as a black box, where is that current flowing out? Charge need to conserve, right?

 

[Krisfr]

Link Sausage and Pancakes are good for breakfast.
Link Sausage wrapped in a pancake is called a Pig in a blanket.
I like Pig n a blanket and I call it Pig n a blanket. I can order Pig n a blanket at the restaurant.
Maybe some people do not like me calling Link Sausage wrapped with Pancakes that.
It is okay.
But it is not a Pig in a Poke.
Can I have Pig n a blanket with some eggs over easy and a cup of tea too?
You Bloody well can. But not a vampire tea bag though.

 

[Bigun]

I like Toad in the Hole, it's pretty close to a Pig in a Blanket. I'd have to ask for the right one were I ordering at a restaurant in the UK or I might be mightily upset by what I received

Waly - I'm with you on the perception that CFA could become a marketing term more than a technical term but it still serves a purpose for talking about different topologies so long as we don't over-analyze it ...

 

[scott wurcer]

P.P.S. If the current is flowing in the inverting input, and you consider the "op amp" as a black box, where is that current flowing out? Charge need to conserve, right?

P.S. I hope not!

Current gets out of black boxes too. Put a 1000V/us ramp on the input with for instance 10pF comp cap. That's 10mA of current that has to flow in or out of the inverting input and the drop in any impedance from it to ground must be accounted for. I think your idea of "black box" is too narrow, if an amp delivers 1A to a load where does the current come from?

In the first case the compensation is usually a shunt C to ground the displacement current in C finds its way to ground and that loop has no problem with KVL, KCL, or charge conservation.

EDIT - BTW I agree on several other points. It is trivial to add a diamond buffer plus resistor to a "CFA" to get a slew on demand input. So say you pick a closed-loop gain and a resistor and comp cap for the same closed-loop BW. You are designing discrete amplifiers for a fixed application and you have access to everything. The resulting CFA vs VFA comparison would show little difference in performance in the direct signal path. I would look elsewhere, PSRR, CMRR, EMI for differences.

 

[Esperado]

By these two tests, the circuit you posted earlier Esperado is not CFA - it's VFA because the peak current into the TISs is set by the LTP current sources.

At witch moment did-you consider the AC current in the feedback resistance as negligible ?
Really, i prefer to describe the topology than to consider its suboptimal implementations.
In the same spirit, we can ruin all CFA's advantages with miller compensations.

am not debating the CFA properties, only that, strictly from a feedback perspective, the "CFA" denomination is confusing

It is just a semantic question. CFA is for "Controversial Feedback Amplifier".
And you have highly contributed to it.
One thing is for sure, the guy who, first, had this bad name idea has not contributed to the world's peace.

 

[PMA]

Useless debate, nothing else. You guys must have a lot of free time.