Then that buffer does the work of Merlin. Over a range of closed loop gains, the CFA has a constant bandwidth, while the VFA has only a constant gain-bandwidth product.
I built as similar a VFA and CFA as I could and showed the above in The Audio Voice 204: Getting to True Wireless Stereo... There Must Be a Better Way! Current Feedback: Fake News or the Real Deal? Your oversimplification ignores the demonstrable.
And even if it didn't, the above claim does nothing to deny the fact of current feedback, stated and ignored many times by you: Inverting input current equals ground current minus output current.
This is true of any amplifier. Using terms like "space-time" violation is not appropriate. There are situations where small signal analysis gives useful information and there are times when large signal analysis (usually a transient response simulation though there are other methods) is necessary.
I see a schematic of the "H-bridge" input op-amp has been posted and I suppose the arguments around "is it VFA or is it CFA" will start again. I see it as one of several ways of enhancing the slew rate of an amplifier with a constant GBW product accompanied by the usual degradation in DC specs (Aol, offset, drift, noise, etc.) due to the resistive degeneration.
Using "CFA" as a simple identifier for an amplifier where the closed-loop BW is relatively constant over a useful range of gains and has enhanced slew rate is innocent and conflating it into some kind of deception is pointless.
A clever guy calling on military contractors in the Loveland/Fort Collins area saw designers making discrete versions for large signal pulse applications because there were no commercial versions. The availability of good complementary bipolar devices led to the possibility of making dual supply versions of these that looked like "op-amps" but were different.
This was still cold war era military applications so the margins were huge and this was a very lucrative business. Harris had some very clever designers and could have made these years earlier on their DI process but didn't, that's history.
The topic is the basic principle involved in a CFA. People interested in the discussion know the effect of closing the loop in a CFA and in a VFA.Originally Posted by forr
It has been said many times : VFA input stage is only a CFA input with a buffer.
Then that buffer does the work of Merlin. Over a range of closed loop gains, the CFA has a constant bandwidth, while the VFA has only a constant gain-bandwidth product.
Why not consider "resistance feedback” or RFA?, if only to support the inclusivity of the other questionably esteemed member of Ohm’s law now huddling and sobbing all alone in the corner (... perhaps just a little sympathy to appease HR).
Figure 1 (sorry for the crude drawing) is the conventional schematic of a unity gain feedback amplifier. Figure 2 is the equivalent schematic of a CFA that includes the two resistors responsible for gain loss. R1 is the trans-resistance R1, and R2 is the feedback resistor. The resulting equation for gain can be approximated as:
Av = R1 / (R1 + R2)
The AD844 provides access to the Tz node, permitting the resistor R1 to be placed across its internal 3Meg Ohm trans-impedance. To verify the equation a 10K Ohm resistor was placed from this point (pin 5) to ground and a 10K Ohm resistor was placed in the feedback path with gain diminishing to 0.5. Similarly, a 30K Ohm in the feedback path results in a gain drop to 0.25.
This equation can be alternatively expressed in Ohm’s law whereupon R = V / I and becomes:
Av = (V1 / I 1) / (V1/I1 + V2/I2) where V1 is the voltage drop across R1 and I1 is the current passing through R1, etc. This suggests that voltages and currents are inter-dependant upon one another, with neither having sole dominion over Av. Yet this is also true for resistance.
The voltage gain Av is a unit-less number, hence the right side of the equation must also be unit-less. This also means that the formula can be equally expressed independently in terms of resistance, voltage or current. Hence Av can be expressed in terms of resistance, voltage or current respectively as:
Av = R1 / (R1 + R2) or:
Av = V1 / (V1 + V2) or:
Av = (1/I1) / (1/I1 + 1/I2)
In these equations the resistors are just that, voltages must appear across the resistors indicated and currents must be measured as passing through them to be valid in the equations of Av. In conclusion, none of these formulas have dominion or superiority over the others as to thereupon justify a CFA as being an RFA, VFA or CFA.
Figure 1 (sorry for the crude drawing) is the conventional schematic of a unity gain feedback amplifier. Figure 2 is the equivalent schematic of a CFA that includes the two resistors responsible for gain loss. R1 is the trans-resistance R1, and R2 is the feedback resistor. The resulting equation for gain can be approximated as:
Av = R1 / (R1 + R2)
The AD844 provides access to the Tz node, permitting the resistor R1 to be placed across its internal 3Meg Ohm trans-impedance. To verify the equation a 10K Ohm resistor was placed from this point (pin 5) to ground and a 10K Ohm resistor was placed in the feedback path with gain diminishing to 0.5. Similarly, a 30K Ohm in the feedback path results in a gain drop to 0.25.
This equation can be alternatively expressed in Ohm’s law whereupon R = V / I and becomes:
Av = (V1 / I 1) / (V1/I1 + V2/I2) where V1 is the voltage drop across R1 and I1 is the current passing through R1, etc. This suggests that voltages and currents are inter-dependant upon one another, with neither having sole dominion over Av. Yet this is also true for resistance.
The voltage gain Av is a unit-less number, hence the right side of the equation must also be unit-less. This also means that the formula can be equally expressed independently in terms of resistance, voltage or current. Hence Av can be expressed in terms of resistance, voltage or current respectively as:
Av = R1 / (R1 + R2) or:
Av = V1 / (V1 + V2) or:
Av = (1/I1) / (1/I1 + 1/I2)
In these equations the resistors are just that, voltages must appear across the resistors indicated and currents must be measured as passing through them to be valid in the equations of Av. In conclusion, none of these formulas have dominion or superiority over the others as to thereupon justify a CFA as being an RFA, VFA or CFA.
If you add a buffer to a one transistor CFA, topologically you get a VFA. So what?
If you add a buffer to a one transistor CFA, phenomenologically, you get a VFA, with the difference in "effect" that you acknowledge.
Therefore, a VFA is not just a CFA with a buffer.
Using same topology as i did 40+ years ago --- same results.... with reasonably matched compliments, Harmonics are < -120dbv. SR = 145v/usec.
Very simple and very effective... Current-Mode Amplifier.
App is suitable for low z loads, headphone amp, and as output stage for preamps, DAC etc.
View attachment HPA 6.pdf ***
THx-RNMarsh
Happy Thanksgiving day in USA
*** See Linear Audio pub for details
Very simple and very effective... Current-Mode Amplifier.
App is suitable for low z loads, headphone amp, and as output stage for preamps, DAC etc.
View attachment HPA 6.pdf ***
THx-RNMarsh
Happy Thanksgiving day in USA
*** See Linear Audio pub for details
Last edited:
The DC/LF gain equations (voltage or current output) are exactly the same for a CFA and VFA - so you will not see any differences there. Of course, the resistor values are much lower in a CFA to ensure correct operation because the -in is a current port.
Prof. Sergio Franco's article linked below explains why you cannot model a CFA as a VFA (see page 2) and why a CFA is not a VFA and was written as a rebuttal to Michael Kiwanuka's similar claim that CFA's were non-existent. I understand that a short while after this, Kiwanuka recanted.
https://www.edn.com/design/analog/4458753/1/In-defense-of-the-current-feedback-amplifier
Last edited:
The reasons of Michael Kiwanuka's temporary recantation were not clear. Some days later he came back to its initial point of view, at least, concerning the concept of current feedback.
I can't believe how recent the comments are on that article. Barry Harvey's comment echo's mine.
OMG
Give up.
The data and results here were explained by actual in-circuit measurements! Measurements which explained why the simplified/incomplete SIM was not accurate enough in fully showing how CFA works.
"Why the differences in the internal voltages and currents? Because the two-port representation of a negative-feedback system is not exact but is only approximate [3, 4], though the approximation is usually quite good, and in many cases, it is even exact. But, if we set up the CFA circuit in the lab, and measure the actual voltages and currents of its feedback network, we will get the data of the CFA circuit, not the data of the VFA circuit! Consequently, it is wrong to use the (imprecise) VFA model to invalidate the (precise) CFA model! The VFA model is an abstraction intended to facilitate the paper-and-pencil calculation of the loop gain T, but it fails to accurately account for the actual physical operation of the CFA (more on this as we move along)."
I have said this before, in my own apparently Cryptic ways, but this gets it all concisely.
Though I have MicroCap CAD/SIM ($$), why do you think i have so much invested in test equipment.... to see what is 'really' going on.
THx-RNMarsh
Last edited:
The attached contains an application that tests the depths of understanding of CFA amplifiers, and perhaps can inspire further thoughts on the advantages of limited gain open loop applications of CFA's, or perhaps then more correctly CMA's. It should be noted that the AD844 permits connections of resistors in parallel with its internal 3M Ohm trans-resistance at pin 5.
To begin with, if the inverting terminals are inputs where is the current coming from if it can only come from the mirror of itself?
To begin with, if the inverting terminals are inputs where is the current coming from if it can only come from the mirror of itself?
That's an easy one: ultimately, from the power supply. ;-)
BTW, I used the AD844 10 years ago in an experimental power amp with error correction but no global feedback loop.
The design has been written up for Elektor and AudioXpress/Audio Amateur and is available on my website:
Part 1: https://linearaudio.nl/sites/linearaudio.net/files/UK-1 2008040241.pdf
Part 2:https://linearaudio.nl/sites/linearaudio.net/files/UK-2 2008050441.pdf
The articles include a section on using the AD844 and taking advantage of its CFA structure and the fact that the 'Tz' pin is brought out to the package. That was key for my use.
The unconventional use doesn't make the circuit intuitive and may require actual attentive reading...
A few weeks ago I dusted the amp off for a demo in Northern France for a group around Jean Hiraga, and again it earned compliments for the neutral and effortless reproduction. I gave away lots of my amps but not this one ;-)
Jan
PS At the time I used the term 'current conveyor' for the AD844 analogous to the proposals in the Toumazou and Lidgey book I linked to earlier.
Last edited:
It looks very much like a differential amplifier, balanced voltage in, balanced voltage out.
The doubters will never become believers. I see ‘Moo Koo’ recanted his recant.
Happy Thanksgiving BTW 🙂
Few seem to think it doesn't exist, so I shouldn't worry about it 😉
When are we going to get thanksgiving over here, oh, hang on, we do, except it's called harvest festival cuz we are pagan after all.....
When are we going to get thanksgiving over here, oh, hang on, we do, except it's called harvest festival cuz we are pagan after all.....
“To begin with, if the inverting terminals are inputs where is the current coming from if it can only come from the mirror of itself?”
You’ve just pointed out an important point regarding CFA operation:
Cuurent comes OUT of the -input. Why? Because the current developed across the DB emitter degeneration resistor (be that discrete or integrated amplifer) has to raise the voltage at the emitter of the DB so that it equals the +input voltage.
The voltage at the emitter of the DB output transistor (which must be the same as the +input) is thus made up of two components: the feedback current via Rf flowing into Rg plus the current flowing out of the - input through Re.
The current comes in through the collector of transistor from the + or - supply as Jan has pointed out - just like a normal emitter follower. We normally refer to the current in the collector of the DB as the output current and the current from the -in as the input current so this terminology might be a bit confusing.
You’ve just pointed out an important point regarding CFA operation:
Cuurent comes OUT of the -input. Why? Because the current developed across the DB emitter degeneration resistor (be that discrete or integrated amplifer) has to raise the voltage at the emitter of the DB so that it equals the +input voltage.
The voltage at the emitter of the DB output transistor (which must be the same as the +input) is thus made up of two components: the feedback current via Rf flowing into Rg plus the current flowing out of the - input through Re.
The current comes in through the collector of transistor from the + or - supply as Jan has pointed out - just like a normal emitter follower. We normally refer to the current in the collector of the DB as the output current and the current from the -in as the input current so this terminology might be a bit confusing.
Last edited: