Mostly neither do I. The whole thing Bonsai? Just kidding about the first part.
https://www.theabsolutesound.com/articles/nelson-pass-four-decades-of-innovation/
I have a couple of his amplifier designs by Nelson lying around, an Adcom GFA-555 and a Nakamichi PA-5 I believe
Miller capacitance relates to devices. It isn't true that the Tiger amp handled the intrinsic Miller capacitances of those old transistors badly in its design.
Bandwidth, frequency response/ and slew rates relate to the source impedance driving the Miller capacitance and the extent it requires driving. The source impedance of the CFA input (internal to the Tiger), at the bases of Q10/Q11 is 1K (R15) to ground. As Bonsai pointed out earlier "the primary compensation is very heavy-handed VAS shunt loading via R15 and C3 which both Self and Cordell talk about avoiding." However back then it was done to handle and drive the extent of Miller capacitance present in those devices. Looked at from another perspective, corrective feedback signals imposed on the inverting terminal (input emitters of Q10/Q11) require something to work against at the base, being R15 at high frequencies, otherwise the Miller capacitance takes over to limit the response in the loop.,
The other factor to bear in mind is that for full output the input network of the CFA (Q10/Q11) moves only 1/3 the amplitude, reducing the effective Miller capacitance to 1/3 of what it would normally be if otherwise required to follow the full swing output.
P.S. - I believe I added a couple of compensation capacitors for stability... no issues for me to find or fix
https://www.theabsolutesound.com/articles/nelson-pass-four-decades-of-innovation/
I have a couple of his amplifier designs by Nelson lying around, an Adcom GFA-555 and a Nakamichi PA-5 I believe
http://espacenomeutente.altervista.org/wp-content/uploads/2020/04/RE_Mar-Apr1973.pdf
Miller capacitance relates to devices. It isn't true that the Tiger amp handled the intrinsic Miller capacitances of those old transistors badly in its design.
Bandwidth, frequency response/ and slew rates relate to the source impedance driving the Miller capacitance and the extent it requires driving. The source impedance of the CFA input (internal to the Tiger), at the bases of Q10/Q11 is 1K (R15) to ground. As Bonsai pointed out earlier "the primary compensation is very heavy-handed VAS shunt loading via R15 and C3 which both Self and Cordell talk about avoiding." However back then it was done to handle and drive the extent of Miller capacitance present in those devices. Looked at from another perspective, corrective feedback signals imposed on the inverting terminal (input emitters of Q10/Q11) require something to work against at the base, being R15 at high frequencies, otherwise the Miller capacitance takes over to limit the response in the loop.,
The other factor to bear in mind is that for full output the input network of the CFA (Q10/Q11) moves only 1/3 the amplitude, reducing the effective Miller capacitance to 1/3 of what it would normally be if otherwise required to follow the full swing output.
P.S. - I believe I added a couple of compensation capacitors for stability... no issues for me to find or fix
Last edited:
The amp has two feedback networks connected to ground. The output stage is depicted in figure 2, returned to ground with input emitters connected in the feedback, forming the CFA nature of this output. It should be noted that R26 and R27 in the collectors of the input transistors in Figure 2 seem in all likelihood for compensation using the Miller capacitance. I may have added capacitors between the collector and base of these transistors for stability purposes, perhaps in not having the identical transistors indictated.
I built numerous of these amps using my own devices/circuit boards, etc., with variant compensation for stability. I used them for driving two 6-8" Philips motional feedback sub arrays that pushed the peak resonance to 10 Hz. As they only went to 120Hz the smooth highs couldn't be heard, though high frequency noise could be heard as the feedback went up. However when you play 30-40Hz notes they go much more silent to the 2nd and 3rd harmonics as expected.
http://espacenomeutente.altervista.org/wp-content/uploads/2020/04/RE_Mar-Apr1973.pdf
I built numerous of these amps using my own devices/circuit boards, etc., with variant compensation for stability. I used them for driving two 6-8" Philips motional feedback sub arrays that pushed the peak resonance to 10 Hz. As they only went to 120Hz the smooth highs couldn't be heard, though high frequency noise could be heard as the feedback went up. However when you play 30-40Hz notes they go much more silent to the 2nd and 3rd harmonics as expected.
http://espacenomeutente.altervista.org/wp-content/uploads/2020/04/RE_Mar-Apr1973.pdf
Last edited:
It’s just an output stage configured with amplification. If you read Self (can’t recall is Cordell touched on it), there are a few derivatives but all basically provide some OPS gain.
A CFA has a diamond buffer IPS with a low z inverting input and a high z non-inv input, a TIS (aka VAS) or TAS (aka current mirror) 2nd stage and then a buffer OPS. Importantly, the output voltage and the feedback resistor set the front end gm. The whole lot together make a CFA.
A CFA has a diamond buffer IPS with a low z inverting input and a high z non-inv input, a TIS (aka VAS) or TAS (aka current mirror) 2nd stage and then a buffer OPS. Importantly, the output voltage and the feedback resistor set the front end gm. The whole lot together make a CFA.
As a contributor to the seeming vitriolic debate between identifying devices as CFA's and VFA's awhile ago now at the DiyAudio, there seems only one individual that came up with what I thought was a reasonable solution. This involved imposing stimuli to a device and dependant upon the result determine if a device can be characterized as a CFA or VFA. Of course this would require some consensus to be agreed upon as reflecting the threshold of the turnover.
To be more "on topic" specific to capacitors/capacitance I am currently designing an I/V using the LT1206 CFA. It supports varying the quiescent current to higher values as seeming used of advantage in the AD811, a CFA that appears gaining some favour. This allows for more control over the slew rate and high frequency performance, being useful in stabilizing the drive into higher capacitance loads, specifically interconnect cables, etc.
https://www.analog.com/media/en/technical-documentation/data-sheets/1206fb.pdf
Looking first at page 13, it shows a couple of examples a "Precision ×10 Hi Current Amplifier" and "Low Noise ×10 Buffered Line Driver" that both show a dual operational amplifier network with "an output stage configured with amplification", using an CFA LT1206. Being used as an "output stage" does not deny the LT1206 is still a CFA, notwithstanding what Self and Cordell might otherwise have concluded.
Looking at page 8 shows the simplified schematic of the LT1206. The schematic does not contain a current mirror, rather instead contains a current magnifier of unknown multiplication, being dependant upon two resistors of undisclosed value, the resistor in the collector of Q2 and the resistor in the emitter of Q6. The base/emitter drop across Q5 approximately equals the reverse base/emitter drop across Q6, meaning that the volt drop across the two resistors is approx. equal. It isn't necessary that these values be equal, and are unlikely to be so. The implication is that CFA designations are not reliant upon the currents being equally mirrored, rather can be magnified by the hfe of bipolar transistor as well.
As far as the diamond buffer goes (as being conditional upon its use to be identifiable as a CFA), does Pedja's implementation of an I/V using an AD844 CFA without implementing that buffer turn it into a VFA?
To be more "on topic" specific to capacitors/capacitance I am currently designing an I/V using the LT1206 CFA. It supports varying the quiescent current to higher values as seeming used of advantage in the AD811, a CFA that appears gaining some favour. This allows for more control over the slew rate and high frequency performance, being useful in stabilizing the drive into higher capacitance loads, specifically interconnect cables, etc.
https://www.analog.com/media/en/technical-documentation/data-sheets/1206fb.pdf
Looking first at page 13, it shows a couple of examples a "Precision ×10 Hi Current Amplifier" and "Low Noise ×10 Buffered Line Driver" that both show a dual operational amplifier network with "an output stage configured with amplification", using an CFA LT1206. Being used as an "output stage" does not deny the LT1206 is still a CFA, notwithstanding what Self and Cordell might otherwise have concluded.
Looking at page 8 shows the simplified schematic of the LT1206. The schematic does not contain a current mirror, rather instead contains a current magnifier of unknown multiplication, being dependant upon two resistors of undisclosed value, the resistor in the collector of Q2 and the resistor in the emitter of Q6. The base/emitter drop across Q5 approximately equals the reverse base/emitter drop across Q6, meaning that the volt drop across the two resistors is approx. equal. It isn't necessary that these values be equal, and are unlikely to be so. The implication is that CFA designations are not reliant upon the currents being equally mirrored, rather can be magnified by the hfe of bipolar transistor as well.
As far as the diamond buffer goes (as being conditional upon its use to be identifiable as a CFA), does Pedja's implementation of an I/V using an AD844 CFA without implementing that buffer turn it into a VFA?
There are numerous ways to configure the 2nd stage (same applies to VFA). However, the majority use either a current mirror (typically IC types although the 1206 is a bit different) or a more conventional VAS (discrete power amps especially ). The 1206 is a bit unusual, but I think that’s primarily so it can drive very capacitive loads Looking at the feed forward/bootstrap circuit around Q15 and 16.
Sorry, I am not familiar with Pedja’s work.
The best and most complete description of CFA operation is Prof Sergio Franco’s here https://www.edn.com/in-defense-of-the-current-feedback-amplifier/#:~:text=The%20current%2Dfeedback%20amplifier%20(CFA,CFA's%20application%20in%20myriad%20systems.
Sorry, I am not familiar with Pedja’s work.
The best and most complete description of CFA operation is Prof Sergio Franco’s here https://www.edn.com/in-defense-of-the-current-feedback-amplifier/#:~:text=The%20current%2Dfeedback%20amplifier%20(CFA,CFA's%20application%20in%20myriad%20systems.
Last edited:
Is it possible to have current feedback without having a whole amplifier? Obviously there can be local feedback in an amplifier. Is it by definition only voltage feedback? IOW is it that feedback is always applied at the base of a transistor instead of at the emitter, so a feedback signal is always subject to miller effect? Is that the view? If so, seems a little fishy.
Sure... as long as we can agree that the network that is "not a CFA" works just "like a CFA"... and I will still be calling a CFA
Last edited:
My rule of thumbs goes like this: In a CFA amplifier or section of an amplifier the feedback network's impedance affects the open-loop gain, and the voltage divider ratio applies a closed-loop gain on that.
That usually means that the feedback is a emitter or source to which the impedance adds as local degeneration.
I’ve been doing a lot of rooting around on electrolyte conducticity - unfortunately most of the papers are behind paywalls. The behavior of electrolytes with frequency, current and concentration looks to be complex. It’s probably electrolyte behaviour at LF due to low current flow that causes distortion. There’s a lot of research going on into electrolys from what I can tell because of battery development and, separately, in the biological field. The plot below is for an electrolyte over 10^2 to 10^8 Hz - I’m looking for plots of frequency and current vs conductivity linearity. Anyway, digging continues.
No need for a scientific study. Just think logical. Take a pure 1 kHz sine signal, listen to it. Fade in a 2 kHz sine signal. At some level of that added "K2" you may notice it, still the frequency is near, the brain accepts it as somehow correllated to the basic signal. Substitute the 2 kHz with 5 kHz. You will notice it at a lower level. The brain detects / discriminates the more distant frequency much easier and questions the correlation. This could initially fool the listener with more rich sound but in the long term results in fatigue, irritation. Also if there was another tone with basic 5 kHz, or near, it will smear that. So the music gets smeared by THD. The sound of instruments with low natural harmonic content, that have a pure tone like often the case with piano, will be degraded most. THD lowers the contrast between tones that naturally have either low or high harmonic content. So if a bunch of people who do not listen extensively, on first impression rate a high THD amp better than a low THD amp, its quiet understandable. After listening to different recordings over a longer period of time it becomes clear that the high THD amp has a sound signature. One fine example is the Musical Fidelity A1/X.
Not forget that THD not only depends on frequency but also on power. Amps should not have higher THD at low power than on higher power. That could be a plus for a simple single ended Pass amp over a Krell or whatever. Same with DACs.
The Pass amps tend to have high-ish THD overall.
Most amps will have lower THD at lower power. The only reason I don't say "all amps will..." here is that some amps do show a sweet spot at moderate output power and then rising THD towards higher output power. But the THD for a, say, 125 W rated amp will tend to be lower at 1 W than at 100 W.
I agree with your comments around 'richness' and listening fatigue. Great summary.
Tom