Here is the *.asc file again:
john curl said:
Zero information in this response. I thought you wanted to help educate us. The patent number please? You brought up PIM in this thread, and there is nothing wrong with me discussing PIM, since that is indeed an Otala-related topic. You just don't like to have your assertions challenged. Give us some technical discussion, John. Why can't you tell us what it is your friend did or showed that made you make the assertions you did?
My MOSFET amplifier had less than 0.1ns of PIM (that's 100 ps), and it had a generous amount of NFB. That would normally be enough for people to infer that a generalized claim that large NFB and small OLBW cause PIM is wrong.
I'll say it once again. Matti deserves credit for specifying a measurement method for the distortion that he described. This is intellectually very honest and diligent. All I did was build an instrument to measure PIM in accordance with his method and showed my results. My results were in accordance with theory. That theory showing that an amplifier with higher NFB and lower OLBW need not have more PIM.
Matti's own measurement method disproved his assertions about PIM.
It is also worth noting that PIM that is created by the feedback process happens mainly because amplitude intermodulation distortion in the open loop response causes the closed loop bandwidth to vary dynamically in accordance with the amplitude intermodulation, causing the phase response to vary as expected. What this all means is that it is almost impossible for PIM to get generated in the absence of amplitude intermodulation distortion as well. Such amplitude intermodulation distortion will make itself known on many other conventional tests as well (including THD and CCIF IM). Indeed, in the latter, the only difference between PIM and IM is the relative phase of the upper and lower distortion sidebands. If you want to eliminate PIM, just make ordinary distortion very small (this is a good case for striving for low THD and CCIF).
PIM is also created in the total absence of NFB. If you have a no-NFB amp with a VAS that has non-linear collector-base capacitance effects you will get PIM right there.
Bob
john curl said:
Walt Jung, Seems to share that thought.
Last page read text:
http://waltjung.org/PDFs/WTnT_Op_Amp_Audio_3.pdf
john curl said:
If you were really concerned with the math of Barrie Gilbert, you would have done the math. When you posted this, it became obvious to me that you hadn't done the math at all. Doing the math of the Gilbert paper and not noticing the third harmonic is like driving down the street and not noticing what color the traffic lights are. It just doesn't happen. Your "concerns" with Gilbert's results are apparently limited to waving them around.
Did Barrie Gilbert retract his conclusions. If so, where? It would be unprofessional for Barrie Gilbert NOT to retract his conclusions, if he were found wrong. You have been listening to someone who doesn't like his results, but this is not disproof. You should know better than that. Now when it comes to my colleague. It is premature to give out any further information to ANYONE, as he asked me not to. I had no idea that this would be met with such hostility, especially on an OTALA thread. I thought that people here wanted to learn how to make better audio designs. Apparently, some are very satisfied with the status quo. So be it.
john curl said:
Actually, I was interested in his results enough to extend them to the case for which the input-output phase shift of the open-loop amp is arbitrary. His results are good as far as they go.
But what do they neglect? They neglect common-mode effects. This was first brought up by Pete B (PB2) in the Gilbert ("What's your reasoning...") thread several years ago. At first I ignored this, but then became more curious about it. The logical extension of Gilbert's analysis is to take an input stage consisting of typical bipolar transistor models, follow it with an ideal integrator and output stage (as Gilbert did), and look at the distortion of that.
What happens is that distortion due to common-mode effects of the input stage is an order of magnitude, at minimum, higher than that due to difference-mode effects for an amp with sufficient emitter degeneration in the front end (or FETs). The solution is to bootstrap the input stage as Scott did with the AD797. This distortion consists of two components - Early Effect and nonlinear Ccb/Cgd. Of these, the Ccb/Cgd term dominates. This is a pure PIM effect not considered in Gilbert's analysis, and is at least an order of magnitude higher than the difference-mode effect Gilbert considers.
The bottom line is that if you don't bootstrap the input stage, the whole TIM/PIM discussion becomes a complete joke because other sources of input stage distortion are dominating.
But hey, figuring this out requires SPICE, and the "learn and grow" advocates who also claim they knew everything there is to know about audio design 40 years ago can't accept that.
Interesting, but I doubt that you have thrown away the essence, even though you found other factors. My colleague's measurements, then, should not bother anyone, if and when they become available.
This might also prove that tubes are really much better, and solid state will always be a poor compromise.
This might also prove that tubes are really much better, and solid state will always be a poor compromise.
john curl said:
Again John I just don't know what you are talking about. I just read again from Andy's link, Barrie states repeatedly that he used a >1MHZ GBW< op-amp, that's circa 1967 technology, in order to walk through his equations. I'm afraid I have to agree with Andy, if you throw the AD829 (for instance) in the same boat you really aren't doing due diligence in understanding the analysis.
Hello John
If negative feedback or lots of it are the ills of all bad amplifier designs then how do you explain why the Halcro amplifiers always get exceptional reviews in Stereophile in terms of both their subjective and objective performances ?
Either negative feedback is always bad or it isn't ! You can't be half pregnant on this issue.
regards
Trev
If negative feedback or lots of it are the ills of all bad amplifier designs then how do you explain why the Halcro amplifiers always get exceptional reviews in Stereophile in terms of both their subjective and objective performances ?
Either negative feedback is always bad or it isn't ! You can't be half pregnant on this issue.
regards
Trev
I would like to point out a few facts that many may not realize. I hope it clarifies, rather than confuses the situation further.
Both Matti's amp and my JC-3, designed within a year or so of the AES paper, have 20KHz open loop bandwidth, with about 20dB of negative feedback. More feedback would be OK, IF the 20KHz open loop response is maintained.
Both amps also use a SUMMING INPUT in order to reduce common mode distortion. Please note this.
We came to understand, early, that common mode distortion compromised typical amp designs.
Practical considerations, later forced us to make non-inverting amps and this was made tolerable by cascoding the input stages, or equivalent techniques.
Both Matti's amp and my JC-3, designed within a year or so of the AES paper, have 20KHz open loop bandwidth, with about 20dB of negative feedback. More feedback would be OK, IF the 20KHz open loop response is maintained.
Both amps also use a SUMMING INPUT in order to reduce common mode distortion. Please note this.
We came to understand, early, that common mode distortion compromised typical amp designs.
Practical considerations, later forced us to make non-inverting amps and this was made tolerable by cascoding the input stages, or equivalent techniques.