I'm for sure way to young to know about the former definition you mentioned. But my point is that there's nothing marketing related in the current definition. It's just a simple interpretation of a circuit theory formalism.
Actually there is nothing new about feeding back a divided voltage to a low open-loop impedance node. Most 1950's valve amplifiers do exactly that by feeding back to the cathode of the first valve. No-one called that current feedback back then, because current feedback only had the 'traditional' meaning that Jan explains in posts 15 and 18.
To make everything even more confusing, Nordholt reserves the term current feedback for amplifiers that take a sample of the output current and subtract that from the current of the input signal source. That is yet another way of using the term.
To make everything even more confusing, Nordholt reserves the term current feedback for amplifiers that take a sample of the output current and subtract that from the current of the input signal source. That is yet another way of using the term.
At some moment in history, someone started using the then well-defined term current feedback to mean something it had never meant before, thereby creating a lot of confusion. I guess Jan's point is that someone who is stupid enough to do that must have been a marketeer (-;
Well, up until a few weeks ago there was no confusion in the minds of most people as to what the two types of topology were, and certainly in the semi industry there is no confusion. CFA as used in hundreds of opamp data sheets over the last 30 odd years should attest to that. We seem to have gotten into a semantic argument here on DIY.
Firstly, rather than look at it from the input to output voltage view, let's look at what's happening with the feedback network. Here is how I understand it from an intuitive POV.
VFA
You provide a sample of the output voltage to the inverting input by dividing it down, and then feeding it into a diff pair, which in turns steers the LTP current into the TIS input through the LTP collector load resistors, or a mirror. If you keep the ratio of the feedback resistors constant, (within reason) you do not affect the rise fall time or the bandwidth of the amplifier. These are controlled by the LTP current and Cdom. Note, that if you change the feedback network ratio, you can affect the stability, but that is a different issue. Importantly, the maximum practical current into the TIS input stage is limited by the LTP current source.
CFA
In the CFA case, you also sample the output voltage, but the primary feedback quantity to the feedback node is current. Why? Because this feedback current, set of course by the value of the feedback resistor and its associated lower leg to ground, directly moderates the TIS drive via the emitters of the input buffer summing junction. There is no interposing LTP stage. If you change the resistance of the feedback network, but keep the ratio constant, you can affect the rise fall times and the loop bandwidth. Hence, on CFA's the feedback resistor is the recommended way to compensate for loop stability. And that is why is called a CFA. Note, that because of the use of low feedback resistor values, the drive into the TIS input stage can be much, much higher than a VFA - hence the great speed of CFA's. You could do this on a VFA, but the LTP current would have to be very high (100 mA) - a totally impractical proposition. VFA designers can get around this by going for MIC, so I am just pointing this out to try to highlight the differences.
The two compared:-
VFA- higher DC gain, lower offsets, lower noise; generally, slightly more complex (generally!). Bandwidth related to closed loop gain.
CFA - lower gain, wider bandwidth, higher offsets, faster rise fall times. Bandwidth, within reason, not related to closed loop gain.
That's how I understand it. I heard somewhere, IIRC, that Scott Wurcer rewrote the loop equations for CFA to express them more fully in current terms, but I may have got my wires crossed on that one.
Firstly, rather than look at it from the input to output voltage view, let's look at what's happening with the feedback network. Here is how I understand it from an intuitive POV.
VFA
You provide a sample of the output voltage to the inverting input by dividing it down, and then feeding it into a diff pair, which in turns steers the LTP current into the TIS input through the LTP collector load resistors, or a mirror. If you keep the ratio of the feedback resistors constant, (within reason) you do not affect the rise fall time or the bandwidth of the amplifier. These are controlled by the LTP current and Cdom. Note, that if you change the feedback network ratio, you can affect the stability, but that is a different issue. Importantly, the maximum practical current into the TIS input stage is limited by the LTP current source.
CFA
In the CFA case, you also sample the output voltage, but the primary feedback quantity to the feedback node is current. Why? Because this feedback current, set of course by the value of the feedback resistor and its associated lower leg to ground, directly moderates the TIS drive via the emitters of the input buffer summing junction. There is no interposing LTP stage. If you change the resistance of the feedback network, but keep the ratio constant, you can affect the rise fall times and the loop bandwidth. Hence, on CFA's the feedback resistor is the recommended way to compensate for loop stability. And that is why is called a CFA. Note, that because of the use of low feedback resistor values, the drive into the TIS input stage can be much, much higher than a VFA - hence the great speed of CFA's. You could do this on a VFA, but the LTP current would have to be very high (100 mA) - a totally impractical proposition. VFA designers can get around this by going for MIC, so I am just pointing this out to try to highlight the differences.
The two compared:-
VFA- higher DC gain, lower offsets, lower noise; generally, slightly more complex (generally!). Bandwidth related to closed loop gain.
CFA - lower gain, wider bandwidth, higher offsets, faster rise fall times. Bandwidth, within reason, not related to closed loop gain.
That's how I understand it. I heard somewhere, IIRC, that Scott Wurcer rewrote the loop equations for CFA to express them more fully in current terms, but I may have got my wires crossed on that one.
Last edited:
At some moment in history, someone started using the then well-defined term current feedback to mean something it had never meant before, thereby creating a lot of confusion. I guess Jan's point is that someone who is stupid enough to do that must have been a marketeer (-;
Nonsense. Stop hammering marketers.
Nonsense. Stop hammering marketers.
To me the fundamental difference, the 'difference that makes a difference' is the asymmetry of the input stage. VFB has symmetry between + and - inputs and CFB does not. There might though be exceptions to the asymmetry come to think of it - the LM3900 springs to mind
I suppose a CFB amp with both + and - inputs as current (low impedance) still counts as CFB.
To me the fundamental difference, the 'difference that makes a difference' is the asymmetry of the input stage. VFB has symmetry between + and - inputs and CFB does not. There might though be exceptions to the asymmetry come to think of it - the LM3900 springs to mind
This is one reason I started playing with a CM amplifier stage, to mitigate asymmetrys of the input transistors, straighten out the transfer per se.
Jan Didden wrote
"There might be a semantic issue here. Traditionally, the name 'current feedback' has been given to the kind of feedback that returns a sample of the output current. This could be for instance the voltage across a small resistor in series with the load. Current feedback thus makes the output impedance very high.
With the new so-called 'CFA' opamps that appears to have taken on a new meaning (blame it on the marketeers) namely: feedback to a low impedance node. Personally I think this is an ambiguous way to name it because, what's a low impedance node? The emitter impedance of a starved BJT can be pretty high, while 'voltage feedback' to an invering input fed from a low-impedance source may be pretty low impedance."
I would call that current sense mode feedback amplifier and indeed the output quantity controlled is the current. I do recall pointing this out somewhere else on the forum a few years ago.
Going back to VFA and CFA point a little earlier in this thread, note that they are both still voltage amplifiers - voltage in, voltage out. Might also be that's where the confusion is.
Note also that CFA opamps are not new. The first integrated circuit devices were produced by Elantec and Comlinear in the late 1980's and the CFA term was already used then to describe them. I recall reading that the Japanese were already using a similar topology in discrete designs earlier than this - but cannot be sure. When I was still doing linear design in the 1980's, the was a flurry of papers and articles about the subject - Sergio Franco wrote a very nice article in EDN that I read.
Wiki is your friend : http://en.wikipedia.org/wiki/Current-feedback_operational_amplifier
"There might be a semantic issue here. Traditionally, the name 'current feedback' has been given to the kind of feedback that returns a sample of the output current. This could be for instance the voltage across a small resistor in series with the load. Current feedback thus makes the output impedance very high.
With the new so-called 'CFA' opamps that appears to have taken on a new meaning (blame it on the marketeers) namely: feedback to a low impedance node. Personally I think this is an ambiguous way to name it because, what's a low impedance node? The emitter impedance of a starved BJT can be pretty high, while 'voltage feedback' to an invering input fed from a low-impedance source may be pretty low impedance."
I would call that current sense mode feedback amplifier and indeed the output quantity controlled is the current. I do recall pointing this out somewhere else on the forum a few years ago.
Going back to VFA and CFA point a little earlier in this thread, note that they are both still voltage amplifiers - voltage in, voltage out. Might also be that's where the confusion is.
Note also that CFA opamps are not new. The first integrated circuit devices were produced by Elantec and Comlinear in the late 1980's and the CFA term was already used then to describe them. I recall reading that the Japanese were already using a similar topology in discrete designs earlier than this - but cannot be sure. When I was still doing linear design in the 1980's, the was a flurry of papers and articles about the subject - Sergio Franco wrote a very nice article in EDN that I read.
Wiki is your friend : http://en.wikipedia.org/wiki/Current-feedback_operational_amplifier
Last edited:
One other point that should be emphasized is that because the VFA front end tail current is quite low, you will get a pole formed the the LTP current an the TIS input capacitance. Since the VFA open loop gain is much higher, and you generally have to consider the output stage pole, you would end up with excess phase and instability. The cure of course is to lower the TIS pole frequency with Cdom so you get 20 dB/ decade slope and a ULGF intercept below the output stage pole.
The CFA does not have this issue, because the current available to charge and discharge Cdom is usually very high, you you get a much wider bandwidth and slew rates.
In power amps, you normally have to apply
The CFA does not have this issue, because the current available to charge and discharge Cdom is usually very high, you you get a much wider bandwidth and slew rates.
In power amps, you normally have to apply
Indeed that is the 'traditional' definition in use from say early 20th century, because it controls the output current, while 'traditional' VFB controls the output voltage.
I believe it is awkward to tie the definition to the impedance of the summing point, because it immediately raises the question: what impedance is 'high', what is 'low' , where do you put the transition? Endless discussions.
Also not a good idea to tie it to a particular circuit topology like: 'if you feed it back to an LTP it is voltage feedback' - there are so many different circuit topologies now and in the future, endless discussions. Don't forget audio is only a tiny subset of the large control systems world where this is founded.
That's where the reference to 'marketing' came from: if a competent engineer, who was aware of the 'traditional' definitions, had marketed these new 'CFA' opamps, I doubt very much that he would be so ignorant to ignore 80 years of engineering and throw everything into kilter. It MUST have been a marketing guy
. That doesn't make him stupid, he's just doing his job, but he looks at these things differently than an engineer.jan
Last edited:
Jan,
You should know that most (90 % in my employer) marketers in the semi industry are EE's, many have MSc's and many also have PhD's. So lets get the competent engineer bit out of the way. Further, most semiconductor marketers work incredibly closely with the designers, and especially on high performance or system level products.
You don't seem to have been aware that CFA's have been around for 25 years or so and no one ignored 80 years of electrical engineering history in this CFA name thing.
Lighten up on the marketers man!
You should know that most (90 % in my employer) marketers in the semi industry are EE's, many have MSc's and many also have PhD's. So lets get the competent engineer bit out of the way. Further, most semiconductor marketers work incredibly closely with the designers, and especially on high performance or system level products.
You don't seem to have been aware that CFA's have been around for 25 years or so and no one ignored 80 years of electrical engineering history in this CFA name thing.
Lighten up on the marketers man!
Hey, I don't really care one way or another, I just wanted to flag that the rest of the world has a different definition.
Once you start to study feedback and control systems you'll be confronted with the 4 topology definitions:
shunt - series feedback;
shunt - shunt feedback;
series - series feedback;
series - shunt feedback
where shunt stands for 'voltage' and series stands for 'current'.
Sorry if I rubbed you the wrong way!
jan
Once you start to study feedback and control systems you'll be confronted with the 4 topology definitions:
shunt - series feedback;
shunt - shunt feedback;
series - series feedback;
series - shunt feedback
where shunt stands for 'voltage' and series stands for 'current'.
Sorry if I rubbed you the wrong way!
jan
Those so called CFA are genuine VFB topologies , the difference
with a differential is just buffering of the return signal , nothing
that can be qualified as different on a technical ground.
The alleged low impedance of the inverting input of thoses "CFA"
doesnt hold since the transistor that buffer the said input in a differential
works in common collector , hence its output impedance is very low ,
in fact it can be even lower than the one provided by the FB network
in said "CFAs"...
with a differential is just buffering of the return signal , nothing
that can be qualified as different on a technical ground.
The alleged low impedance of the inverting input of thoses "CFA"
doesnt hold since the transistor that buffer the said input in a differential
works in common collector , hence its output impedance is very low ,
in fact it can be even lower than the one provided by the FB network
in said "CFAs"...
Pioneer were using what we term CFB topology in their amps which were sold on the market back in 1978. By CFB topology I mean the use of the diamond buffer inputstage. The amp in question is the MZ1 and a matching preamp the CZ1. If you view schematics of these amps and preamps youll find that the inputstage and transimpedance stage is represented as a block diagram and everything else is shown. The obvious reason was to conceal the circuit. They had problems obtaining USA patents (I wonder why when the circuit was novel) but finally did manage in 1983 although their first request was entered in 1980. How Elantec and Comlinear were able to patent the exact same circuit just 3 -4 years later I dont understand. Either the examiners failed to find the prioir art or through some tecnicality. It was indeed the Japanese that came up with the circuit.
What is also interesting is that some time after the Pioneer came on the market Harmon Kardon also released a amp featuring such topology. Here no fully reliable schematics are available but it seems to be a very similar design to the pioneer. Its difficult to establish whether it was independently developed or developed from the pioneer. This Harmon Cardon is in many circles regarded as the best amp ever produced by them.
I think everyone including the members in The Self book thread are incorrectly characterizing why CFB topololy is called as such.
At college I learnt that the term comes from the act that the feedback from the output to the low impedance node tries to force the inverting input current to zero and hence the term curent (feedback).
A reference for this is "HIGH SPEED OPERATIONAL AMPLIFIERS by Walt Kester from Analog devices.
At college I learnt that the term comes from the act that the feedback from the output to the low impedance node tries to force the inverting input current to zero and hence the term curent (feedback).
A reference for this is "HIGH SPEED OPERATIONAL AMPLIFIERS by Walt Kester from Analog devices.
Whatever the impedance of the inverting input of an amplifier, the basic principle of feedback amplifiers is that the voltage at this inverting input follows the non-inverting input voltage within fractions of mV with BJTs. The difference between the AC voltages at the two inputs defines an output AC current according to a transconductance process usually expressed in mA/V.
I can't see any current feedback in the scheme where the inverting intput is a single device or a push-pull of transistors (curiously, it is only this last configuration which seems to have concentrated the appellation of CFA on itself).
Expressions VFA and CFA should apply to amplifiers where it is either the output voltage or the output current which is under control, and should not describe the way the feedback is achieved [if we admit that the feedback applied to the emitter(s) of the input BJT(s) is of current nature which is fundamentally wrong].
Semantically, the concepts can be somewhat clarified by the expressions VC or CC (Voltage or Current Controlled) and VS or CS (Voltage or Current Source) where the controlling value and the controlled output become explicit.
I can't see any current feedback in the scheme where the inverting intput is a single device or a push-pull of transistors (curiously, it is only this last configuration which seems to have concentrated the appellation of CFA on itself).
Expressions VFA and CFA should apply to amplifiers where it is either the output voltage or the output current which is under control, and should not describe the way the feedback is achieved [if we admit that the feedback applied to the emitter(s) of the input BJT(s) is of current nature which is fundamentally wrong].
Semantically, the concepts can be somewhat clarified by the expressions VC or CC (Voltage or Current Controlled) and VS or CS (Voltage or Current Source) where the controlling value and the controlled output become explicit.
No. it has always been called a current feedback amplier: series derived shunt applied negative feedback.
I.e. the output signal being sampled is a current which is then algebraically summed with the input signal which is also a current.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.