janneman said:
Actually, you can build bad feedback amps, bad zero-global-feedback amps, good feedback amps, good zero-global-feedback amps. So feedback can be good or bad depending on the way you implement it into the whole design. You gotta know what you're doing!
Jan Didden
Yes I agree but according to the anti-feedback brigade any feedback is too much feedback whilst others think that there is a moderate amount of feedback which yields the optimum performance but cannot explain why this is so.
I used the example of the Halcro amplifiers to dispel the myth that too much feedback always results in sub optimum performance from an amplifier
regards
Trev
Come on... Anti-feedback brigade use emitter (source, cathode) followers, but dare to say they don't use any feedback...
Look at the picture. The JFET input opamp (inverting connection - I have a symmetrical input transformer to reverse phase) has own NFB on AC for it's own CL gain about 26 dB up to about 100 KHz flat. Then a driver has own feedback, so it's gain is about 8. Both driver and opamp have a common feedback that makes driver's output resistance under my control. Then the whole thingy including class C outputs have a common feedback that makes switching on/off of output devices gradual and smooth. As the result, the amp with ompamp on input and class C output devices eats class AB amps on lunch. Some positive feedback allows to damp speaker coils on main resonant frequency extending flat response below. However, that does not mean other resonances of speakers and boxes should not be taken care of. It was designed for bass guitar, but people revealed that it sounded cleaner than best Hi-Fi amps known then. I did not believe, and did not design it for Hi-Fi, I was cheating with feedbacks and approximation, but few years later I've found that approximation by summing outputs of class A and class C amps was patented and called Current Dumping. In order to implement it properly I needed very fast driver and error amp.
Look at the picture. The JFET input opamp (inverting connection - I have a symmetrical input transformer to reverse phase) has own NFB on AC for it's own CL gain about 26 dB up to about 100 KHz flat. Then a driver has own feedback, so it's gain is about 8. Both driver and opamp have a common feedback that makes driver's output resistance under my control. Then the whole thingy including class C outputs have a common feedback that makes switching on/off of output devices gradual and smooth. As the result, the amp with ompamp on input and class C output devices eats class AB amps on lunch. Some positive feedback allows to damp speaker coils on main resonant frequency extending flat response below. However, that does not mean other resonances of speakers and boxes should not be taken care of. It was designed for bass guitar, but people revealed that it sounded cleaner than best Hi-Fi amps known then. I did not believe, and did not design it for Hi-Fi, I was cheating with feedbacks and approximation, but few years later I've found that approximation by summing outputs of class A and class C amps was patented and called Current Dumping. In order to implement it properly I needed very fast driver and error amp.
I feel that, at this point, a little OT is deserved, to reduce the tension 
It is my pleasure and feel honoured that you mention this...
In a couple of months I could be posting something newer and from a different répertoire ...
Hey, Wavebourn, is it Gemini or Scorpio? ...hmm I'll wait for your answer to the previous post to conclude
Regards,
M
It is my pleasure
and feel honoured that you mention this... In a couple of months I could be posting something newer and from a different répertoire ...
Hey, Wavebourn, is it Gemini or Scorpio? ...hmm I'll wait for your answer to the previous post to conclude
Regards,
M
Rafael.luc said:
Hi Rafael,
Thanks for bringing Walt’s paper to my attention. If I read him right, he is saying that wide open-loop bandwidth in an amplifier will reduce PIM. I’m afraid I have to disagree with him on this one. Many of the intuitive arguments about PIM can be seductively misleading, and many smart people can be led to the wrong conclusion if they don’t do the math or actual PIM measurements. Nevertheless, much of what Walt describes is correct.
Phase Intermodulation Distortion (PIM) is the change in phase shift through an amplifier that results from signal excursions.
First and foremost, it is important to recognize that if you start with an amplifier that has no PIM, but does have some ordinary nonlinear distortion, and then put feedback around it with the usual frequency compensation, PIM will indeed be created. Walt is correct in describing this conversion of some of the distortion from the amplitude domain to the phase domain.
That PIM exists is not a point of controversy. It does. That feedback placed around an otherwise PIM-free amplifier is not a point of controversy. It will. That high feedback and low open-loop bandwidth exacerbate PIM is the controversy. It does not.
As Walt pointed out, as signals stress the LTP input stage, its transconductance becomes smaller. This changes the open loop gain. When the input stage is followed by a Miller-compensated VAS (essentially an integrator), the gain crossover point of the negative feedback loop becomes smaller when the input stage is under signal stress. This is because the open loop gain is falling at 6 dB per octave and the whole open-loop gain curve is moving downward. As Walt correctly points out, the closed loop bandwidth of the amplifier then decreases correspondingly.
Since we have the closed loop pole moving in under signal stress, there results a bit more lagging phase shift through the amplifier, even in the audio band. The result is that the phase shift through the closed-loop amplifier changes with signal and we have PIM. So far so good.
The amount of lagging phase shift caused by a single high-frequency pole (the closed loop bandwidth frequency) at a lower frequency in the audio band is smaller, the higher the frequency of the pole as compared to the low-frequency target frequency. This means that, all else remaining equal, an amplifier with a higher closed loop bandwidth will experience less PIM. This is very different than saying that an amplifier with higher open loop bandwidth will experience less PIM.
The test for PIM proposed by Otala is analogous to the SMPTE IM test, where 60Hz and 7000Hz signals are mixed in a 4:1 ratio and applied to the amplifier. The 60Hz signal stresses the input stage (and other stages) and the resulting amplitude modulation on the 7kHz carrier is measured. In the PIM test, the phase modulation on the 7kHz carrier is measured instead.
Returning to what Walt said, signal stress on the input stage reduces transconductance, pulls in the closed loop pole, and causes increased in-band phase lag, thus PIM.
Where Walt appears to have made the wrong assumption is that he appears to believe that increased open loop bandwidth will reduce the influence of changing input stage gm on the closed loop gain crossover frequency. This is not true. Consider a conventional Miller-compensated amplifier. Input stage gm and the size of the Miller capacitor set the gain crossover frequency for a given closed loop gain. If you simply look at the calculation of gain crossover frequency as a function of input stage gm and Miller capacitor, you see that the open-loop bandwidth is not in the picture. Open-loop bandwidth does not play a role in determining the gain crossover frequency for a given nominal gain crossover frequency. This appears to be where Walt was mistaken.
There is a second piece to this, having to do with what causes input stage stress. As Walt properly points out, input stage stress is at the root of a lot of this. In an amplifier with low open loop bandwidth and high feedback factor at low frequencies, there is actually fairly low stress on the input stage at low frequencies. This is because the forward gain at, say 60 Hz, is very large. The error signal feeding the input stage will be the output signal amplitude divided by the forward gain, making it small.
When that same amplifier is made to have wide open-loop bandwidth of 20 kHz as described by Walt using VAS shunt resistors, all that does is reduce the forward gain of the amplifier at frequencies below 20 kHz.to be as low as that at 20 kHz. Put another way, it increases the stress on the input stage to be as large at low frequencies as it is at high frequencies. It doesn’t improve matters at high frequencies, it merely makes them just as bad at low frequencies. This can actually increase distortion and PIM. Indeed, at the 60 Hz PIM test signal frequency, the input stage undergoes more stress when the open-loop bandwidth is forced to be higher. It is extremely important in this context to recognize that the appropriate apples-apples comparison is with the gain crossover frequency of the amplifier held the same in both cases.
The final piece is that amplifiers with no negative feedback also have PIM. Because device parameters change with signal amplitude, the phase shift of such amplifiers will also be affected by signal amplitude excursions. In my AES paper on PIM (“Phase Intermodulation Distortion – Instrumentation and Measurement Results” available on my website at www.cordellaudio.com) I illustrated a 35W amplifier (figure 13) that could be configured with or without negative feedback. PIM was measured for both cases (Figures 14 and 15). For the case with no NFB, PIM measured 10ns over moderate power levels. For the case with negative feedback, PIM measured over three times lower, at about 3ns over the same range of power levels. The application of negative feedback actually reduced PIM, apparently by its reduction of the pre-existing PIM.
Cheers,
Bob
.
This does not help. Anyone want to describe the difference between global feedback from local feedback, to Wavebourn and Syn08? I give up.
By the way, the most common question I got from Dr Don Pederson was: "Do you know the answer, Mr. Curl?" because I had 5 years of recent design experience, before I took his course. 'Put that in your pipe, and smoke it', Syn08.
By the way, the most common question I got from Dr Don Pederson was: "Do you know the answer, Mr. Curl?" because I had 5 years of recent design experience, before I took his course. 'Put that in your pipe, and smoke it', Syn08.
Bob Cordell said:
Hi Bob,
Let's say you idealize the LTP so that it has no change in transconductance, but have Miller capacitance that is only the non-linear internal device capacitance, would you agree that such an amp would have PIM?
This is not to disagree with anything that you wrote above.
Pete B.
Use schematics, it is an international language. Draw 2 emitter followers, one of which is wildly oscillating. Does it have a global feedback? Or it has a local feedback? Or global feedback? Or both? Or none?
Or, may be you can show me the difference between somebody completely pregnant and partially pregnant, John?
Or, may be you can show me the difference between somebody completely pregnant and partially pregnant, John?
john curl said:
Can you please provide a definition of it and what criteria is used to distinguish one topology from another ??
Also you said that your amps have a 20KHz open loop bandwidth. Does this then mean that people with acute hearing above 20KHz will compromise their listening experience when using your amps ??
regards
Trev
Trevor, you do not understand, perhaps you are not technical. I don't know. However, we have found that the error overshoot that occurs in the input stage at frequencies above the open loop bandwidth cause both AM and FM distortion. It is the FM distortion that we are mostly concerned with. Even 1% 3'rd harmonic distortion (AM) is barely audible. I can prove this with ANY analog tape recording. As an analog tape recording engineer, I have made a number of master recorders and I know what to expect, even with the best tape, highest speeds, and widest tracks. 1% distortion is continually present in every master tape that is made. 10% on peaks. Therefore, .1% 3'rd harmonic distortion is almost meaningless me, however the FM component is something that feedback amps seem to make, and we have never bothered to completely measure it. This might be what the ear hears and is expecially sensitive to.
Now do you understand?
Now do you understand?
It depends on loudness and dynamics. It is barely audible on peaks, but sounds nasty on soft sounds, especially when they decay. The same tape recorder may be used for demonstration, but this time instead of clipping decrease HF bias level, then measure and listen.john curl said:
Also, feedback amps don't "seems to make", rather phase delayed, especially non-linearly phase-delayed feedback "seems not correct" them.
As far as the difference between global and local feedback, we separate them because usually local feedback has such a very high frequency rolloff, that we do not have to worry about excessive phase shift that will make the entire amplifier oscillate.
In the early days, we did not use global feedback, just local feedback. Then, we started to use global feedback for servos, and analog processes. It was limited in scope, mostly because of the low frequency phase shift due to transformers and coupling caps, to about 20dB, not a bad figure. Then IC op amps came on line and then the dreaded 'slew rate' became important. It wasn't really talked about before op amps were introduced. Along with slew rate limiting, came very low open loop bandwidth. At first, we didn't think much of it. Now, we consider it more important.
In my 40+years as a professional design engineer, only quibblers confused local feedback with global feedback, but they are present, here, so we must be more careful, or nothing useful can be said, without someone, making a point about it.
In the early days, we did not use global feedback, just local feedback. Then, we started to use global feedback for servos, and analog processes. It was limited in scope, mostly because of the low frequency phase shift due to transformers and coupling caps, to about 20dB, not a bad figure. Then IC op amps came on line and then the dreaded 'slew rate' became important. It wasn't really talked about before op amps were introduced. Along with slew rate limiting, came very low open loop bandwidth. At first, we didn't think much of it. Now, we consider it more important.
In my 40+years as a professional design engineer, only quibblers confused local feedback with global feedback, but they are present, here, so we must be more careful, or nothing useful can be said, without someone, making a point about it.