Audibility of PIM
Michael,
Your point is well-taken, and we must never under-estimate the ear's ability to hear something. There are really two points in this discussion. The first is how much PIM the ear can hear. The second is what factors in amplifier design influence how much PIM is made. Although I personally find it unlikely that the ear can hear PIM on the order of a few nanoseconds, my focus is mostly on the second question.
I believe that PIM, like any other distortion, should be made as small as possible (as long as doing so does not compromize other aspects of the amplifier). I believe that PIM, like other distortions, is just one more symptom of the underlying nonlinearity of the amplifier. I believe that PIM is extremely unlikely to exist in the absence of other measurable distortions. Finally, my measurement experience indicates that, in the real world with real audio power amplifiers, the application of negative feedback actually reduces the total amount of measured PIM. Moreover, high values of negative feedback and low values of open-loop bandwidth do not exacerbate PIM at all. One more case where feedback gets an underserved bad rap. As with anything, feedback is a tool that should be used responsibly, and a tool that can be abused by the ignorant or foolhardy.
Finally, let me end with a question: we often hear people talk about time smear in some part of the reproduction chain. Do we think they are talking about PIM? If so, then time smear would be measurable. For example, I believe I could use my PIM analyzer to measure the PIM of a loudspeaker. Just a thought...
Bob
Michael,
Your point is well-taken, and we must never under-estimate the ear's ability to hear something. There are really two points in this discussion. The first is how much PIM the ear can hear. The second is what factors in amplifier design influence how much PIM is made. Although I personally find it unlikely that the ear can hear PIM on the order of a few nanoseconds, my focus is mostly on the second question.
I believe that PIM, like any other distortion, should be made as small as possible (as long as doing so does not compromize other aspects of the amplifier). I believe that PIM, like other distortions, is just one more symptom of the underlying nonlinearity of the amplifier. I believe that PIM is extremely unlikely to exist in the absence of other measurable distortions. Finally, my measurement experience indicates that, in the real world with real audio power amplifiers, the application of negative feedback actually reduces the total amount of measured PIM. Moreover, high values of negative feedback and low values of open-loop bandwidth do not exacerbate PIM at all. One more case where feedback gets an underserved bad rap. As with anything, feedback is a tool that should be used responsibly, and a tool that can be abused by the ignorant or foolhardy.
Finally, let me end with a question: we often hear people talk about time smear in some part of the reproduction chain. Do we think they are talking about PIM? If so, then time smear would be measurable. For example, I believe I could use my PIM analyzer to measure the PIM of a loudspeaker. Just a thought...
Bob
Please Bob, don't 'second guess' me as well. A small number, to me, is still a real number. I tend to prefer 'reasonable worst case' rather than average conditions.
Most of my designs are little different from yours, except perhaps greater in power output. I have found that open loop design to sound better to my ears. At least, with regard to preamps. Your results may vary.
Most of my designs are little different from yours, except perhaps greater in power output. I have found that open loop design to sound better to my ears. At least, with regard to preamps. Your results may vary.
.
I imagine many people's ears would say the same thing. It doesn't surprise me at all that a circuit designed using an OL methodology will sound worse when a control loop is applied. The two systems are significantly different. Control system design is really complex in this application and needs a different skillset. For example, the argument that a CL amp will necessarily sound better with wide OL bandwidth and low OL gain rather than low OL bandwidth and high OL gain is not correct - this is a misplaced OL mindset argument.
It isn't that different. I have designed both high negative feedback and no loop feedback preamps in the last 5 years with essentially the same parts and topology. It seems pointless to add a feedback loop to a line preamp, IF you do it with a lot of care.
I don't know what to say about power amps, except that low feedback usually means high open loop bandwidth, and I think that is the most important factor.
While Charles, Nelson, and I tend to make open loop designs on occasion, this does not mean that Matti Otala ever said that negative feedback was forbidden in a quality design. I have one of his power amps and it has a negative feedback loop. The amp that he made for HK in 1980 had loop feedback, so I find that a lot of criticism of Matti is not based on much evidence.
What Matti promoted was high open loop bandwidth and high slew rate. This approach has been successful for me, ever since I first made the JC-2 line amps in 1973.
Our original work on TIM did not 'prove' that high open loop bandwidth was desirable or necessary. PIM tends to show that high open loop bandwidth is desirable, and maybe even necessary in the best designs.
I don't know what to say about power amps, except that low feedback usually means high open loop bandwidth, and I think that is the most important factor.
While Charles, Nelson, and I tend to make open loop designs on occasion, this does not mean that Matti Otala ever said that negative feedback was forbidden in a quality design. I have one of his power amps and it has a negative feedback loop. The amp that he made for HK in 1980 had loop feedback, so I find that a lot of criticism of Matti is not based on much evidence.
What Matti promoted was high open loop bandwidth and high slew rate. This approach has been successful for me, ever since I first made the JC-2 line amps in 1973.
Our original work on TIM did not 'prove' that high open loop bandwidth was desirable or necessary. PIM tends to show that high open loop bandwidth is desirable, and maybe even necessary in the best designs.
Re: Audibility of PIM
Comparison of Nonlinear Distortion Measurement Methods By Richar Cabot (Audio Precision). The most stringent test: 20KHz THD.
Rodolfo
Bob Cordell said:
Comparison of Nonlinear Distortion Measurement Methods By Richar Cabot (Audio Precision). The most stringent test: 20KHz THD.
Rodolfo
traderbam said:
6db is a nice number.
That smallish feedback is what you get by simply closing the fedbackloop around a no-fb amp without increasing openloopgain.
Otherwise i do not see any sense in your assumption.
A low performance OL amp with a feedback loop put around it will never/unlikely make for a high performance CL amp.
Another question to you guys:
Is it necessary to make a poweramp unitygain stable ?
An amp not beeing unitygain stable will amplify errors in the frequency range from where amp reaches 180° OL-phaseshift to when it finally lowers OL-gain below 1. (having positive feedback in this range)
Mike
john curl said:
You're being cryptic, John. I don't know what you mean by "second guessing".
What I WAS doing was stating that your reliance on Barrie Gilbert's article as a credible basis for your PIM position is misplaced. When you drop names and cite others' work as supporting your position, you should assume that people will go read that work and call you to task if that work indeed does not support your position as you assert it does.
This audio business is very subjective, and there is a lot that we still don't know. I have absolutely no problem with you preferring the sound of a design without negative feedback, or a design with wide open-loop bandwidth. It is possible that such a design is better and we just don't yet know why. I call that the "X" factor, and am comfortable with that. I don't know why my 35 wpc KT88 tube amp sounds so damn good, but I don't lose sleep over it.
What I DO have a concern with is when people say something is better for a particular reason, come up with some hypothesis that supports that position, but then get upset when the hypothesis is tested by others and is found to have no basis. You say over and over again that wide open loop bandwidth is better because it reduces PIM, but you get upset when others show measured results that show that not to be the case. Once you state a measurable, testable hypothesis for something, you have put a stake in the ground, and should expect that others will test it. Don't get upset just because we don't take your word for it.
You'll get no argument from me regarding design for worst case, and design on the basis that something may be too small to hear. Why else did I do an amplifier with less tha 0.001% THD-20, 300 V/us slew rate in a 50-watt amp, and less than 100 picoseconds of PIM? I may have a hard time believing that a few ns of PIM is audible, but that certainly doesn't make me complacent about it.
Bob
http://www.diyaudio.com/forums/showthread.php?postid=483904#post483904
http://www.diyaudio.com/forums/showthread.php?s=&postid=494035&highlight=#post494035
http://www.diyaudio.com/forums/showthread.php?s=&postid=481861&highlight=#post481861
http://www.diyaudio.com/forums/showthread.php?s=&postid=421693&highlight=#post421693
http://www.diyaudio.com/forums/showthread.php?s=&postid=494035&highlight=#post494035
http://www.diyaudio.com/forums/showthread.php?s=&postid=481861&highlight=#post481861
http://www.diyaudio.com/forums/showthread.php?s=&postid=421693&highlight=#post421693
It's possible to modify Gilbert's assumption of a quadrature phase relationship between the op-amp's output voltage and the input diff amp's output current. One can assume a more general case where the open-loop gain is a low-pass filter rather than an integrator. Suppose you consider the extreme case of very wide open-loop bandwidth, where the phase shift from differential input to output is zero. One finds when doing the KVL equation around the input loop in this modified version of Gilbert's analysis that there is no AM-to-PM conversion in the closed-loop amplifier at all.
But achieving this condition carries a high cost - namely throwing away a lot of open-loop gain to get there. Then for a fixed output voltage amplitude, the amplitude of the differential input voltage is much higher than the integrator case. Even though the AM-to-PM approaches zero as the input-output phase shift approaches zero, the AM-to-AM and the overall distortion get much larger because of all the open-loop gain that's thrown away to get there. So the assertion that this is a subjective improvement relies on the assumption that AM-to-PM distortion is orders of magnitude more objectionable than AM-to-AM. I'm not aware of any studies that have been done to compare relative audibility of the two types of distortion.
But achieving this condition carries a high cost - namely throwing away a lot of open-loop gain to get there. Then for a fixed output voltage amplitude, the amplitude of the differential input voltage is much higher than the integrator case. Even though the AM-to-PM approaches zero as the input-output phase shift approaches zero, the AM-to-AM and the overall distortion get much larger because of all the open-loop gain that's thrown away to get there. So the assertion that this is a subjective improvement relies on the assumption that AM-to-PM distortion is orders of magnitude more objectionable than AM-to-AM. I'm not aware of any studies that have been done to compare relative audibility of the two types of distortion.