John, I think this is complete hogwash. I am not a pHd and can design circles around many of them, yet I respect them and would love to have gotten one myself (so I could teach at a place like Stanford). At the same time, I have rubbed elbows with many people with Masters and Bachelors that had little creativity.
Cheers,
Bob
Hi John,
I don't recall anybody saying the 741 was clipping. However, it was perilously close to, if not into, slew rate limiting in that test. Slew rate limiting, of course, is a form of internal clipping.
2F2-8F1 would be much less likely to show up because it is an even-order product, and the nonlineaities that lead to feeback-generated PIM seem mostly to be odd order (think input differential pair). Nonlinearities that lead to open-loop PIM may be more likely to be even-order, such as junction capacitance nonlineairity.
Cheers,
Bob
But dangerous. Some bozo will sooner or later use this for his claims
about better sounding power rectifiers.
rotfl, Gerhard
PIM in class D Amplifiers
Here's a thought for the day to put PIM in perspective. Let me know if you agree with me.
PIM in Class D amplifiers.
Class D amplifiers necessarily have a series inductor in the output filter. A typical value is 10-20 uH. These are often realized with a powdered core. It is generally accepted that distortion in the output filter is a problem for class D amplifiers because of this when the feedback loop is not closed around the output filter (which is hard to do because of stability issues).
Any nonlinearity in the inductor usually manifests itself as current-dependent inductance. Thus, we have a pole moving back and forth with the signal current, causing in-band phase shift changes. Voila, PIM.
Note that this is a version of PIM that is not dependent on being created by negative feedback - or even by any active devices.
Suppose the filter is 3 dB down at 100 kHz. If the inductance varies by 10% as a function of signal, the PIM created will be ten times worse than the PIM created by a feedback amplifier whose open loop gain moves around by 10% with signal if that amplifier has a NFB gain crossover frequency of 1 MHz.
(yes, this is VERY back of the envelope).
Cheers,
Bob
Here's a thought for the day to put PIM in perspective. Let me know if you agree with me.
PIM in Class D amplifiers.
Class D amplifiers necessarily have a series inductor in the output filter. A typical value is 10-20 uH. These are often realized with a powdered core. It is generally accepted that distortion in the output filter is a problem for class D amplifiers because of this when the feedback loop is not closed around the output filter (which is hard to do because of stability issues).
Any nonlinearity in the inductor usually manifests itself as current-dependent inductance. Thus, we have a pole moving back and forth with the signal current, causing in-band phase shift changes. Voila, PIM.
Note that this is a version of PIM that is not dependent on being created by negative feedback - or even by any active devices.
Suppose the filter is 3 dB down at 100 kHz. If the inductance varies by 10% as a function of signal, the PIM created will be ten times worse than the PIM created by a feedback amplifier whose open loop gain moves around by 10% with signal if that amplifier has a NFB gain crossover frequency of 1 MHz.
(yes, this is VERY back of the envelope).
Cheers,
Bob
Well, here we are again. Thanks again, Gerard for doing a creditable re-creation of the test. It looks like multiple tests, like we originally did in the laboratory with incrementally increasing and decreasing output levels, is in order. This should clear up the matter. Again, I regret that my original tests were destroyed in the firestorm, or I could have looked at them, and added further info, one way or another.
It should be remembered that if the numbers don't fit, then maybe they are derived from the wrong equations.
Personally, this situation reminds me of something that happened around WW1, when Einstein's theory was being discredited. Apparently, some great German physicists could emulate Einstein's predictions by asserting the presence of a planet in our Earth orbit, of a certain size, and 180 degrees opposite the Earth so that it was always hidden by the Sun. Some people will go all out to discredit something that shakes their 'world view'.
It should be remembered that if the numbers don't fit, then maybe they are derived from the wrong equations.
Personally, this situation reminds me of something that happened around WW1, when Einstein's theory was being discredited. Apparently, some great German physicists could emulate Einstein's predictions by asserting the presence of a planet in our Earth orbit, of a certain size, and 180 degrees opposite the Earth so that it was always hidden by the Sun. Some people will go all out to discredit something that shakes their 'world view'.
What shakes my world view, is creating of non-harmonic content from 2 periodic signals, when no other source of the signal present. As I said before, if some LF instabilities due to kind of thermal feedback with delays present, there will be a hidden signal playing the role, but non-linear stable system can't create in-harmonic content from 2 distorted and intermodulating periodic signals.
But you said before that you believe that a negative feedback around a thingy that itself creates PIM helps to decrease it, so may we assume that a class D amp with an additional feedback loop from speaker out will have less of PIM caused by filter cores?
multitone IMD tests could use incremental frequency and/or amplitude changes of the input tones to help identify the order of the IMD products as long as the map is relatively smooth and continuous (not guaranteed)
would 192K soundcard test tone generation be acceptable? – I really think 1st order 100KHz filtered square wave is a little excessive for a audio amplifier test – know anyone using 1st order roll-off microphones?
would 192K soundcard test tone generation be acceptable? – I really think 1st order 100KHz filtered square wave is a little excessive for a audio amplifier test – know anyone using 1st order roll-off microphones?
But you said before that you believe that a negative feedback around a thingy that itself creates PIM helps to decrease it, so may we assume that a class D amp with an additional feedback loop from speaker out will have less of PIM caused by filter cores?
the Class D output filter corner frequency severely limits the available bandwidth for global feedback around the filter (Bode Integrals anyone?) - but using BJ Lurie's ideas a higher order loop gain shape could allow some audio frequency distortion reduction - including the PIM components from the inductor that fall in the region of negative feedback - assuming (2nd order?) output filter corner frequency more than an order of magnitude higher than the region of functional negative feedback (audio range negative feeback limited to a fraction Fc/20KHz ratio and impacted by filter order)
http://books.google.com/books?id=-g...er&source=gbs_navlinks_s#v=onepage&q=&f=false
(the luriecontrol website seems to be undergoing modification)
Last edited:
That's exactly what I did in post 1032. I moved one tone a few Hz and the
IMD product moved n times as much, according to the order of its dependence
on this tone.
A sound card would be ok. Maybe the square wave should be clipped and then
low pass filtered again to make its risetime independent of the sound card.
Gerhard
But you said before that you believe that a negative feedback around a thingy that itself creates PIM helps to decrease it, so may we assume that a class D amp with an additional feedback loop from speaker out will have less of PIM caused by filter cores?
Yes, that is true, but I pointed out parenthetically that closing the NFB loop around the output filter can be difficult because of the lagging phase shift introduced by that filter.
Unless I am missing something, getting the output filter inside the NFB loop while retaining any meaningful amount of NFB at 20 kHz is very difficult.
Another very good reason to try to get the output filter inside the NFB loop is damping factor. A 20 uH inductor in the filter will reduce the damping factor at 20 kHz to about 3. I believe that having a decent damping factor out to 20 kHz is desirable (although at that frequency a DF of 20 is probably adequate).
Cheers,
Bob
But you said before that you believe that a negative feedback around a thingy that itself creates PIM helps to decrease it, so may we assume that a class D amp with an additional feedback loop from speaker out will have less of PIM caused by filter cores?
Yes. I've seen quite some time ago a (pretty indigest) Signetics App Note related to negative feedback in class D amps, it should still be available on the Signetics web site.
I think this is not much different from a VAS stage that modulates its
Ccb depending on the momentary Collector voltage.
Whether it can generate anything but normal harmonics and mixing products
depends upon whether we are willing to accept that amplifier/excitation system as cycle-stationary.
For a spectrum analyzer it is for sure, so I don't think that we can find anything.
Maybe if we could do FFTs over partial cycles we could see some momentary change,
but we could not express this in Hz because the Fourier transform
is simply not defined for anything shorter than an eternity. Perhaps we need
a Laplace analyzer. ;-)
I don't like math.
Gerhard
Last edited:
...now assume we used an air core inductor, so no PIM generated by it, then applied a negative feedback around this filter.
What should happen, if the entire amp around which we applied the FB is not perfectly linear?
First of all it’s good to see that this thread has moved in to this millennium.
Discussing the future in high-end power amps (aka Class D)
Bob
Yes of course PIM is a problem if you include the output filter in the closed loop, it is obvious, and as you say it, we would like to include the output filter in the closed loop to get a better damping factor.
As I see it that is the same problem and the same wishes we have with all amps, no matter what classes we are talking about.
Show me one amp no matter what class with no (nada) PIM and a good damping factor.
OTOH I don’t think it’s fair to generalize all so called class D amps into one group and say that they have this and that problem. Maybe there are just as many different ways to design a (so called) Class D amp as there are ways to design a (so called) Class A, Class AB or Class B amp?
What do you think guys?
Cheers
Stinius
Scott, I have redesigned the amp so just throw away the schematics I sent you. BTW have a nice trip.
Discussing the future in high-end power amps (aka Class D)
Bob
Yes of course PIM is a problem if you include the output filter in the closed loop, it is obvious, and as you say it, we would like to include the output filter in the closed loop to get a better damping factor.
As I see it that is the same problem and the same wishes we have with all amps, no matter what classes we are talking about.
Show me one amp no matter what class with no (nada) PIM and a good damping factor.
OTOH I don’t think it’s fair to generalize all so called class D amps into one group and say that they have this and that problem. Maybe there are just as many different ways to design a (so called) Class D amp as there are ways to design a (so called) Class A, Class AB or Class B amp?
What do you think guys?
Cheers
Stinius
Scott, I have redesigned the amp so just throw away the schematics I sent you. BTW have a nice trip.
I started building one for you, but you've stopped answering e-mails as soon as I said that you have to share at least some expenses.
This happens a lot in my world, also, Wavebourn.
For everyone else, I don't know what to say about the inductance variation. It can be very important, I'm sure.
However, I am in a situation where I can't address this topic here.
I will try to continue on a separate topic, once one comes up that I can contribute to.
It is interesting that I brought up this topic with one of my associates, and he said 'What inductor is actually linear? Apparently he has looked deeper than most, but I brushed it off, saying that we have to use 'something' and air coil looks pretty good to me, compared to many other types.
For everyone else, I don't know what to say about the inductance variation. It can be very important, I'm sure.
However, I am in a situation where I can't address this topic here.
I will try to continue on a separate topic, once one comes up that I can contribute to.
It is interesting that I brought up this topic with one of my associates, and he said 'What inductor is actually linear? Apparently he has looked deeper than most, but I brushed it off, saying that we have to use 'something' and air coil looks pretty good to me, compared to many other types.
Last edited:
- Status
- Not open for further replies.