At the point of this discussion, may-be it can be interesting to remember the patented schematic of a CFA amp, published by Analog Device in1989 to demonstrate CFA, made by one of their consultants, Mark Alexander in the AN-211 application note..
He use a SSM2131 OPA as input device and the feedback is applied to the output of the OPA. A current mirror sense the two rails of the OPA to provide the push and pull signals to the following stages.
In HF, when the OPA reach its open loop limit, its efficiency diminish, so, it ask more current for the same level. A kind of auto pre-compensation of the hf losses of the following stages.
On my side, i do not bother with current vs voltages point of view, because V=RI.
Just i see the input stage difference this way: In a LTP, we are obliged to set inverting and non inverting impedances equal. For DC offset balance as well as distortion cancellation. It means the inverting input present the minimal 10K impedance you need for your input signal.
And the base of the - input transistor , working is common emitter mode present a high value parasitic capacitance.
In CFA, while the input signal is applied this way, the feedback is applied to the emitter. This means it is * common base* for the feedback.
With both a low impedance and a lower parasitic capacitance, giving the feedback node a much higher cutoff frequency.
I hope those will not confuse with the point Richard has put his camera, but provide an other angle, for a better 3D understanding of the "expansive" behavior.
He use a SSM2131 OPA as input device and the feedback is applied to the output of the OPA. A current mirror sense the two rails of the OPA to provide the push and pull signals to the following stages.
In HF, when the OPA reach its open loop limit, its efficiency diminish, so, it ask more current for the same level. A kind of auto pre-compensation of the hf losses of the following stages.
On my side, i do not bother with current vs voltages point of view, because V=RI.
Just i see the input stage difference this way: In a LTP, we are obliged to set inverting and non inverting impedances equal. For DC offset balance as well as distortion cancellation. It means the inverting input present the minimal 10K impedance you need for your input signal.
And the base of the - input transistor , working is common emitter mode present a high value parasitic capacitance.
In CFA, while the input signal is applied this way, the feedback is applied to the emitter. This means it is * common base* for the feedback.
With both a low impedance and a lower parasitic capacitance, giving the feedback node a much higher cutoff frequency.
I hope those will not confuse with the point Richard has put his camera, but provide an other angle, for a better 3D understanding of the "expansive" behavior.
Last edited:
Wahab, I think Richard is just pointing out that this topology/technology is capable of very high performance at GHz frequencies. So, yes, not really applicable to audio.
However, for audio discrete designs (lets leave out IC's for now because IC designers have a few things available to them that discrete designers don't), this topology easily achieves 15 ppm at 20 kHz at high powers. With a bit of effort, 5 ppm. You cannot get this figures with an MC VFA and have to resort to TPC, TMC or other advanced comp techniques.
There is no mania about CFA, it's been around for years and there are already many fine audio amps out there using this topology, so I struggle to see the problem that you and a few others have with it.
Chill man. It's only an amplifier. If you don't like it, pretend its not there and carry on building VFA.
However, for audio discrete designs (lets leave out IC's for now because IC designers have a few things available to them that discrete designers don't), this topology easily achieves 15 ppm at 20 kHz at high powers. With a bit of effort, 5 ppm. You cannot get this figures with an MC VFA and have to resort to TPC, TMC or other advanced comp techniques.
There is no mania about CFA, it's been around for years and there are already many fine audio amps out there using this topology, so I struggle to see the problem that you and a few others have with it.
Chill man. It's only an amplifier. If you don't like it, pretend its not there and carry on building VFA.
Member
Joined 2009
Paid Member
I agree with this. But interestingly, there is no requirement to balance the impedances at the input to an LTP. We do it for convenience, more of a 'it's easier' but you could if you wanted use dissimilar impedances and deal with the dc-offset just as with a singleton input.
I believe one commercial amp, I read about, NAD maybe, used a LTP where one of the devices was a CFP and the other not. There's also the Rush Cascode input (also NAD I think) which is a VFA input but is not symmetrical and has the same DC offset issue as a singleton.
p.s., it's not V=RI, everyone knows it actually V=IR
Hi, begun.it's not V=RI, everyone knows it actually V=IR
Not everyone.
V=RI is the way we had been learned in Europa.
We too are using strange measurements units, like meters, degrees centigrade, liters or Kilos
Last edited:
Now Richard can go on with explanation why in the end high bandwidth can influence our perception although we do not actually hear those very high frequencies. We want to hear how current mode, that at the end will become current feedback amp, achieves that. We have intuition that the same thing that provides high bandwidth, provides stability, and eventually better perception in the audio band. We also have intuition that this behavior is highly desirable for audio power amps, in spite of the fact that we do not actually hear those very high frequencies.
We shall also keep an eye on VFB thread because we are not limited to one topology, just curious to understand why we have a certain (very consistent) perception.
We shall also keep an eye on VFB thread because we are not limited to one topology, just curious to understand why we have a certain (very consistent) perception.
Then, i believe you will loose all the advantages of the LTP (including distortion cancellation) for no advantages, because you'll never get the low parasitic capacitance of the emitter.
We will have to compare the behavior of an amplifier using inverting input as signal input, where feedback is mixed passively (means in an ideal way) with the signal BEFORE the input stage. The amp performance in bandwidth will be limited by those capacitances as well.
The trend though I notice is that the CFA proponents prophesize the topology as the best thing since sliced bread and are pushing it hard, while on the other hand VFA guys have Always been just satisfied with their topology, tinkering and improving to get the best results without attempting to downplay CFA, unlike what the CFA guys do.Chill man. It's only an amplifier. If you don't like it, pretend its not there and carry on building VFA.
That's been a reason amongst others why this topic is surrounded with quite some heat. =)
I think its best advantage will show in high level amps at higher audio freqs. RNMarsh
Richard, I think that this is a very good distinction to make. It seems that many times I see the arguments against the use of the cfa amplifier circuit the comparisons are made on the basis of class-A operations. This seems disingenuous to me in so many ways. Unless you are using high efficiency speakers, ie horn loaded types or few professional large drivers how often is this really the case? I do not think most amplifiers in most home systems are running on the first watt or there about with common lower efficiency speakers. This is where I think that the topology differences seem to be important, where we are really driving an average speaker load at a higher power output. If I want to listen to a smaller sized speaker at a higher spl level this will require much more power under commonly much harder to drive impedance/reactance loading.
Thank you for keeping it simple in the beginning and I hope that this thread does stay on track and not become another argument over semantics.
Richard, I think that this is a very good distinction to make. It seems that many times I see the arguments against the use of the cfa amplifier circuit the comparisons are made on the basis of class-A operations. This seems disingenuous to me in so many ways. Unless you are using high efficiency speakers, ie horn loaded types or few professional large drivers how often is this really the case? I do not think most amplifiers in most home systems are running on the first watt or there about with common lower efficiency speakers. This is where I think that the topology differences seem to be important, where we are really driving an average speaker load at a higher power output. If I want to listen to a smaller sized speaker at a higher spl level this will require much more power under commonly much harder to drive impedance/reactance loading.
Thank you for keeping it simple in the beginning and I hope that this thread does stay on track and not become another argument over semantics.
buzzforb,
Perhaps it doesn't at all, but at the same time it seems so often that topologies are compared as the class-A against most amplifiers operating in both class-AB or even class-B and anything that operates out of this A range is discounted as not worth discussion. So I guess that as far as cfa or vfa that is out of context to this discussion.
Perhaps it doesn't at all, but at the same time it seems so often that topologies are compared as the class-A against most amplifiers operating in both class-AB or even class-B and anything that operates out of this A range is discounted as not worth discussion. So I guess that as far as cfa or vfa that is out of context to this discussion.
Bob C touched on something which to me was a very valid point... and that is that the CFA architecture being discussed is I think symetrical push pull which is just one particular form. If you are trying to understand differences in sound quality you should also look at more traditional single ended CFB designs that hark back to the very early days of SS amplifiers, and why, whatever technical shortcomings they may have, can achieve the sonic appeal that they do. Everything seems based on visual symmetry these days, not just circuitry but PCB layouts too.
Just a sayin'
+1 !
One thing i loved in his approach was the very simple and lazy way it cancel the two main evils of CFA due to the input device toplogy: PSSR and distortion, using an operational low distortion and good PSSR amplifier.
While this amp measured amazingly good for the time and sounded very nice (i had two samples evaluation boards from AD and tried-it at the time), i wonder if this idea would be able to break again the wall today, using some well known no distortion guru Wurcer's baby ?
I think that, in an LTP, (emitters of same polarity coupled) each transistor can be seen as working in the three configurations all together : EC, BC, CC. A picture of considerable help is to consider that a single transistor works as a differential device and its behaviour mainly depends on the value of the potential difference between its base and its emitter : it is voltage driven by Vbe.
The parasitic capacitances should be defined by their value and their effects described. "High" alone is not informative and comparisons are to be made with a push-pull input.
Given the linearity advantage of the cfa input stage in isolation, and coupling to a class A output stage with its reduced need for loop gain, the match may be a positive tradeoff.
thanks
-Antonio
About datasheets, a remark of old doddering.
Il loved so much those BIG data books, from Analog Device, PMI, Harris etc...
It was like a Christmas gift, an Alibaba cavern...
With each OPA, the schematic you can know at first sight if it was sympathetic or not. Full of application notes giving-you ideas...
Nowadays, they look more like a Marketting commercial paper, simplified schematics (When you are lucky to find one), incomplete measurements etc...
Had they turned paranoiac to not provide any information that can be used by some competitor, or lazy ?
And all your preferred integrated circuits, killed by programed obsolescence !!! A grave yard.
Il loved so much those BIG data books, from Analog Device, PMI, Harris etc...
It was like a Christmas gift, an Alibaba cavern...
With each OPA, the schematic you can know at first sight if it was sympathetic or not. Full of application notes giving-you ideas...
Nowadays, they look more like a Marketting commercial paper, simplified schematics (When you are lucky to find one), incomplete measurements etc...
Had they turned paranoiac to not provide any information that can be used by some competitor, or lazy ?
And all your preferred integrated circuits, killed by programed obsolescence !!! A grave yard.
Last edited:
Here are some of the things about CFA I would like to see discussed and resolved on this thread:-
1. What are the key topological elements that allow the LG -3 dB bandwidth to be extended
2. Explore in depth how to maximize the loop gain (and by implication, further reductions in distortion)
3. What are the fundamental limits to open loop linearity in the CFA, assuming standard EF3 output stage
4. Are comp techniques like TPC, TMC applicable to CFA (I suspect the answer to this is not straightforward)
5. Statements have been made that at high closed loop gains, CFA's degenerate into compromised VFA's. Explore this concept and provide a basic explanation (a few equations will not go amiss here)
6. Explore further the applicability AFEC distortion reduction and its limits (initial LTspice investigations seem to indicate there is a clear limit as to where an how it can provide distortion reduction in CFA amplifiers)
7. Are there any techniques that offer the possibility of significantly improving CFA PSRR other than brute force cap multipliers or AFEC.
The focus for all these questions is discrete CFA audio power amplifiers.
1. What are the key topological elements that allow the LG -3 dB bandwidth to be extended
2. Explore in depth how to maximize the loop gain (and by implication, further reductions in distortion)
3. What are the fundamental limits to open loop linearity in the CFA, assuming standard EF3 output stage
4. Are comp techniques like TPC, TMC applicable to CFA (I suspect the answer to this is not straightforward)
5. Statements have been made that at high closed loop gains, CFA's degenerate into compromised VFA's. Explore this concept and provide a basic explanation (a few equations will not go amiss here)
6. Explore further the applicability AFEC distortion reduction and its limits (initial LTspice investigations seem to indicate there is a clear limit as to where an how it can provide distortion reduction in CFA amplifiers)
7. Are there any techniques that offer the possibility of significantly improving CFA PSRR other than brute force cap multipliers or AFEC.
The focus for all these questions is discrete CFA audio power amplifiers.
Last edited:
Agree.
When I lived in the USA, one year (1986), I went to a semiconductor show and most of the vendors were giving data books and applications manuals away. I walked out of the show with books stacked high in my arms. I think I made about three trips to cart them out and for the second two trips, everyone was looking at me. They were probably thinking 'what's that crazy guy doing'. I still have a lot of those books in storage (National, Siliconix, PMI, BB, ADI - IIRC TI were still mainly digital then . . . And of zero interest to me ).
You can buy the LT apps manuals on Amazon - a lot of stuff written by Jim Williams (RIP) which I love because it was so down to earth and practical. I designed a 1ppm linear 20 bit D-S A-D based on some of the stuff I learned in those books.
Now, I am a marketing guy