Member
Joined 2009
Paid Member
I'd like to add my best wishes to Bob on his birthday - always enjoyed the wisdom shared on the forum by you 
Hi Edmond,
OK, so half the offset comes from the finite hFE of Q2 and Q3. When I try to estimate the various contributions, I actually find a larger contribution from R12 and R13 than from the hFE of Q2 and Q3:
IB,Q2=IB,Q3=140 uA/hFE~=1.4 uA assuming a typical hFE of 100
with a 1:1 ratio between Q5 and Q8 and between Q7 and Q9, this corresponds to 1.4 mV across R10 and 1.4 mV across R11.
The current through R12 and R13 must be about 7 or 8 uA before Q5, Q8, Q7 and Q9 start conducting significant current (VBE/100 kohm). Calculating back to an input offset voltage: 7 uA/1.5 mS=4.6666... mV (per side)
Regards,
Marcel
/OT
Hi Andrew,
Sorry for being off topic. Nevertheless I will reply to Marcel one more time.
Hi Marcel,
Thanks for pointing this out. After some fiddling, I got a stable sim without evil R12 and R13. Together with an increased beta to 3000, the circuit complies much better with the geometric mean rule. Now, the product (of currents) only varies +/- 6%. Please notice that this only applies to the simplified circuit. The real circuit with increased beta showed this effect far less pronounced.
Cheers, E.
Hi Andrew,
Sorry for being off topic. Nevertheless I will reply to Marcel one more time.
Hi Marcel,
Thanks for pointing this out. After some fiddling, I got a stable sim without evil R12 and R13. Together with an increased beta to 3000, the circuit complies much better with the geometric mean rule. Now, the product (of currents) only varies +/- 6%. Please notice that this only applies to the simplified circuit. The real circuit with increased beta showed this effect far less pronounced.
Cheers, E.
To get back on topic, I agree with what Jan wrote in post 15 of
http://www.diyaudio.com/forums/solid-state/208416-pros-cons-current-feedback-amplifier-2.html
That is, it would have been better if the people who introduced "current feedback op-amps" had devised a different name for them, as current feedback already meant something else. This doesn't make the technique any less (or more) useful, of course.
Best regards,
Marcel
Funny discussion and i understand why theres no such as agreement on this. as there's no such thing a true current feedback unless the output Voltage (feedback signal) is transformed into current with super high output impedance.
But some sort of intermediate solution where the current variation over a low value gnd resistor is sensed and used for feedback can yield fantastic results.
But some sort of intermediate solution where the current variation over a low value gnd resistor is sensed and used for feedback can yield fantastic results.
Don't confuse the inappropriate (because so controversial) name of "CFA" with the way so called amplifiers work, that had been described so many times in this thread. CFA describe here the way feedback signal is subtracted from input signal before classical amplification.
Last edited:
I'd like to add my best wishes to Bob on his birthday - always enjoyed the wisdom shared on the forum by you
Hi Bigun,
Thanks for the kind words. I have indeed enjoyed this birthday. Life is such a gift!
Cheers,
Bob
Esperado,
Sorry I fail to see the difference weather the feedback shunt network goes to the input node or to a low impedance node inside the circuit, only difference as I see it is that where the node is placed at the input you also deal a circuit with low input impedance. You get an IV-converter or transimpedance circuit.
Sorry I fail to see the difference weather the feedback shunt network goes to the input node or to a low impedance node inside the circuit, only difference as I see it is that where the node is placed at the input you also deal a circuit with low input impedance. You get an IV-converter or transimpedance circuit.
Last edited:
The issue is people confuse current controlled output with voltage controlled output with VFA and CFA.
They are 4 completely different things. And, as Esperado has mentioned, we've covered the first two and there are threads about them on the forum. This thread is about the feedback mode, and specifically CFA.
In the topologies discussed here, we sample the output voltage and the feedback is in the form of a current.
They are 4 completely different things. And, as Esperado has mentioned, we've covered the first two and there are threads about them on the forum. This thread is about the feedback mode, and specifically CFA.
In the topologies discussed here, we sample the output voltage and the feedback is in the form of a current.
May-i suggest you play with the two simulations files of the same VFA and CFA amplifier ?
http://www.esperado.fr/temp/VSSA/vssa-vs-vfa.html
Play with square waves on both topologies, and see what happens regarding voltages and currents in the input stages. Play with the feedback impedances too. I think you will understand why one is compressive, and the other expansive. Too, you will see the differences in bandwidths and slew rates and the influences of parasitic capacitances (one use emitter in base common, while the other base in emitter common topology for the feedback path).
I think it is a better way to see the whole story in once, than pontificating on the various aspects and try to correlate them with words.
Once everything will be clear, on the technical aspects, we can argue endlessly about the advantages of high slew rates and bandwidths. As far i'm in concern, it is useless, i have made my mind by myself based on my listening experiences 🙂
... Or wait for Bob's next publication, i'm sure he will resume all this this in the simple, clear and clever way is is used to. Just i hope this thread will give him the desire to compare the two topologies in real life, and correlate them with listening experiences... if it is not yet done. 🙂
Ultimately, this also applies a to VFA.😀
Remember, the IPS is a transconduction stage (voltage in, current out out ).
Cheers, E.
You mean *after* the feedback voltage had been transformed to current at the -input emitter of the LTP ? ;-)
The character of the signals has to be defined : voltage or current ?
In non-inverting amps, the input signal is in the form of voltage.
So, if we admit that the feedback is a current subtraction process, the input voltage has to be converted to a current first.
And as Edmond just said, this conversion is done by a transconductance stage...
EDIT after reading post #2374
The input transconductance stage is controlled by the difference of voltages between the input electrode (base) and the feedback node connected to the following electrode (emitter).
By the way, how is calculated the open loop gain of CFAs ?
Last edited:
Esperdo,this is also how i see CFA where the feedback is injected in the current path of the input device and not applied as a voltage signal on the inverted side of a LTP. But it's not pure as the current injected is the current function of a voltage AC shunt. (caps to Gnd). When you look at the node under the input stage weather its VFA or CFA then that node is voltage modulated with the input signal, but if you look at the currents they are shifted 180 degrees. so by enlarging the voltage-swing under the input device the currents are reduced by the current created (V/I) over the gnd resistor in the feedback shunt. In the VFA the backside transistor performs the V/I conversion and acts like a voltage/current buffer. I see the main difference in the two that the CFA does not have to deal with input capacitance of the backside inputtransisors which explains the better slev-rate and also better distortion figures in the treble.
Is the resistance in the emitter of the input device playing a double role :
- as a degenerative resistance giving local feedback
- as the "foot" resistance of the negative feedback network which divides the output voltage.
?
- as a degenerative resistance giving local feedback
- as the "foot" resistance of the negative feedback network which divides the output voltage.
?
(PS: I was just teasing a little)
lurking....
Excellent "take" on the better performance of CFA's.. (there , I admitted it. 😱).
Excellent thread , as well !
I see the same faster slew/uber low THD in non-miller compensated VFA's , but device Cob
is still the limiting factor. The CFA topology even eliminates this.
I use lower resistance values in my VFA NFB (+ no miller in the VAS) to offset this , but still fall short
of what a CFB setup can do.
Keep up the good work... ( and good explainations) 🙂 .
OS
Excellent "take" on the better performance of CFA's.. (there , I admitted it. 😱).
Excellent thread , as well !
I see the same faster slew/uber low THD in non-miller compensated VFA's , but device Cob
is still the limiting factor. The CFA topology even eliminates this.
I use lower resistance values in my VFA NFB (+ no miller in the VAS) to offset this , but still fall short
of what a CFB setup can do.
Keep up the good work... ( and good explainations) 🙂 .
OS