A few quotes from Barrie Gilbert's article:
'That is, there has to be an additional quadrature component of RELATIVE magnitude.'
'Assume again the the signal frequency is 1/100th the unity gain frequency, that E=2V and G=10, all of which represent quite moderate conditions for an op amp. Then, the actual phase angle is 6.01 'degrees', by E =5, it has increased to 9.14 'degrees'. This is not what we would expect from a LINEAR AMPLIFIER, whose phase should be quite independent of amplitude. In certain applications, the excess phase and its variation with signal level will be very troublesome.'
Also, please get Walt's review of Barrie's work on his website, under Op Amp Audio, ' Walt's tools and tips' 'Realizing High Performance: Bandwidth Limitations'
What do the rest of you think?
'That is, there has to be an additional quadrature component of RELATIVE magnitude.'
'Assume again the the signal frequency is 1/100th the unity gain frequency, that E=2V and G=10, all of which represent quite moderate conditions for an op amp. Then, the actual phase angle is 6.01 'degrees', by E =5, it has increased to 9.14 'degrees'. This is not what we would expect from a LINEAR AMPLIFIER, whose phase should be quite independent of amplitude. In certain applications, the excess phase and its variation with signal level will be very troublesome.'
Also, please get Walt's review of Barrie's work on his website, under Op Amp Audio, ' Walt's tools and tips' 'Realizing High Performance: Bandwidth Limitations'
What do the rest of you think?
NO, PMA, I am talking about a general problem, ignored by just about everyone. Of course, we can MINIMIZE PIM, but is it good enough? I don't know for sure, myself.
For the record, I had 4 separate designs on display at the Munich Show. They are contemporary products, so where do you get off, insisting that I live in the past? I TALK ABOUT the past, I design today with the latest and the best.
For the record, I had 4 separate designs on display at the Munich Show. They are contemporary products, so where do you get off, insisting that I live in the past? I TALK ABOUT the past, I design today with the latest and the best.
Last edited:
That is where the op amps used are questionable.
It is important to NOTE a potential distortion source. Usually, discrete designers strive to minimize PIM, even if they don't believe it is important. However, commercial manufacturers often save a few cents by using the cheapest IC op amp that will spec out for them. This is the greater problem. The more subtle problem is that the mere existence of a 90 degree slope in the open loop gain within the usually working bandwidth means that there is a potential for PIM being generated, perhaps a tiny amount and fleeting in character, or perhaps, subjectively, the most important added distortion that the ear can detect as: something wrong, something imitation, or what we usually call 'solid state sound'. Serious designers strive to eliminate it.
It is important to NOTE a potential distortion source. Usually, discrete designers strive to minimize PIM, even if they don't believe it is important. However, commercial manufacturers often save a few cents by using the cheapest IC op amp that will spec out for them. This is the greater problem. The more subtle problem is that the mere existence of a 90 degree slope in the open loop gain within the usually working bandwidth means that there is a potential for PIM being generated, perhaps a tiny amount and fleeting in character, or perhaps, subjectively, the most important added distortion that the ear can detect as: something wrong, something imitation, or what we usually call 'solid state sound'. Serious designers strive to eliminate it.
Yes please read again, >> UNITY GAIN FREQUENCY<< which has nothing at all to do with the open-loop gain low frequency breakpoint.
I'll need to preface this by saying that I may not be in the best of mind and/or spirit. Headache and quite tired today, maybe just too old for all this bickering, perhaps? But a couple of things occur to me.
Jan's quote from the sid/tim AES paper needs to be taken in context. I think it is still accurate, insofar as all those measurements way back in the 1970's, i.e., that context was in terms non-linear distortions, NOT phase modulation.
I think John got a bit carried away in the attribution below. Such a simple statement is just that, too simplistic. But then, so are also many of what is often posted here, all op amps stink, discrete is automatically wonderful, etc. I never been able to understand why more people can't work towards examining the engineering why behind the objections...
But onto the Q of PIM, etc. Even the phase modulation mechanism has multiple degrees of dimension. The instantaneous gm variation in BJT vs FETs, for one. This is a giant one for understanding and is illustrated by the references on my web site, namely http://waltjung.org/PDFs/WTnT_Op_Amp_Audio_3.pdf for one, and also http://waltjung.org/PDFs/WTnT_Op_Amp_Audio_4.pdf for another. The latter article shows how PIM can be minimized via multiple feedback loops around synthesized op amp stages, whereby the open loop BW is directly under user control.
Hope this is helpful. As for the nature of my personal feelings about PIM, etc., yes, it is important. But, I'd rather not be quoted on such things, without prior review of all the background.
And, I'd also caution everyone to watch out for chasing any holy grail, you just might be disappointed.
wj
Jan's quote from the sid/tim AES paper needs to be taken in context. I think it is still accurate, insofar as all those measurements way back in the 1970's, i.e., that context was in terms non-linear distortions, NOT phase modulation.
I think John got a bit carried away in the attribution below. Such a simple statement is just that, too simplistic. But then, so are also many of what is often posted here, all op amps stink, discrete is automatically wonderful, etc. I never been able to understand why more people can't work towards examining the engineering why behind the objections...
But onto the Q of PIM, etc. Even the phase modulation mechanism has multiple degrees of dimension. The instantaneous gm variation in BJT vs FETs, for one. This is a giant one for understanding and is illustrated by the references on my web site, namely http://waltjung.org/PDFs/WTnT_Op_Amp_Audio_3.pdf for one, and also http://waltjung.org/PDFs/WTnT_Op_Amp_Audio_4.pdf for another. The latter article shows how PIM can be minimized via multiple feedback loops around synthesized op amp stages, whereby the open loop BW is directly under user control.
Hope this is helpful. As for the nature of my personal feelings about PIM, etc., yes, it is important. But, I'd rather not be quoted on such things, without prior review of all the background.
And, I'd also caution everyone to watch out for chasing any holy grail, you just might be disappointed.
wj
I apologize to Walt for getting him involved in this 'discussion' but I want to also thank him for clueing me into Barrie Gilbert's article's more than a decade ago, and for his own summary of Barrie's work in the cite above.
It does get old, after awhile, and if each and every one of you has found an IC op amp that you find just great for your application, then go for it.
I will continue to 'muddle along' with my discrete designs when I can, and I will also select IC's when it is useful and necessary to do so. I sure would like a really quiet AD825, that would be VERY useful. I wonder why they made it so noisy?
It does get old, after awhile, and if each and every one of you has found an IC op amp that you find just great for your application, then go for it.
I will continue to 'muddle along' with my discrete designs when I can, and I will also select IC's when it is useful and necessary to do so. I sure would like a really quiet AD825, that would be VERY useful. I wonder why they made it so noisy?
Yes, good audio reproduction is a series of 'non-events' that when taken into consideration, build to a 'major event'. The CTC Blowtorch was a 'major event' for a select group of audiophiles, and it is composed of a series of 'non-events' that came together. That is the 'point' of this thread.
It is quite noisy at 12nV per root Hz. But I think this is a nice device to put after a volume control. No coupling caps. I need to try it.
I guess that it is time to move on. What else could be important? I suspect that EMI proofing could be one of the next most important factors. For example, the CTC Blowtorch goes through: low leakage power transformer, high speed diodes, floating input cap, common mode choke, and ground setting caps, these three making a pi network, a traditional 317,337 regulator stage, all cap bypassed, shunt regulator, and finally a series mos fet based open loop multiplier for EVERY gain block, except for the phono stage which uses jfets instead of mosfets in the final regulator. (And people wonder why it costs so much?)
Overkill? I think not.
Overkill? I think not.
Hi Walt, nice to see you here!
Noted. I was more trying to highlight your research on the link between OL BW and good audio reproduction. Fully agree with you that such a direct link is non-existing.
Wrt PIM, I believe Pavel's tests are interesting. If we take as a realistic value the 5mV pk-pk linear range for the input LTP, in a 100W amp (80V pk-pk Vout) at up to 20kHz I get a minimum required OL gain at 20kHz of (80/5)*1000 ~ 84dB. That seems excessive to me, even if that linear range can be doubled probably with nfb.
My conclusion would be that the avoidance of PIM depends on available OL gain at a particular freq and not on OL BW. Also, nfb or the absence of it does make no difference, as the OL gain and linearity of the basic amplifier circuit does not change with the application or not of nfb.
Is this correct? That 84dB above still bothers me; any comments on that?
jan didden
Noted. I was more trying to highlight your research on the link between OL BW and good audio reproduction. Fully agree with you that such a direct link is non-existing.
Wrt PIM, I believe Pavel's tests are interesting. If we take as a realistic value the 5mV pk-pk linear range for the input LTP, in a 100W amp (80V pk-pk Vout) at up to 20kHz I get a minimum required OL gain at 20kHz of (80/5)*1000 ~ 84dB. That seems excessive to me, even if that linear range can be doubled probably with nfb.
My conclusion would be that the avoidance of PIM depends on available OL gain at a particular freq and not on OL BW. Also, nfb or the absence of it does make no difference, as the OL gain and linearity of the basic amplifier circuit does not change with the application or not of nfb.
Is this correct? That 84dB above still bothers me; any comments on that?
jan didden
Last edited:
OK, but still. Even with an 80mV linear input range you'd need 60dB OL gain at 20kHz to avoid gross non-linearity and thus PIM. Is that realistic? Or are we chasing a ghost? If we exceed the linear range by some amount, how much PIM is actually generated, does anyone have any numbers for that? Walt? Bob?
jan
jan
assuming several mV of error V is the same as assuming limited loop gain, high feedback amplifiers can keep diff input V sub mV to beyond 20 KHz
this is a powerful means of linearizing the input stage which for bjt diff pair has predominantly 3rd order nolinearity - distortion to signal ratio reduces as diff input V**2 == 10x higher loop gain gives 100x less distoriton from input tanh gm
this is a powerful means of linearizing the input stage which for bjt diff pair has predominantly 3rd order nolinearity - distortion to signal ratio reduces as diff input V**2 == 10x higher loop gain gives 100x less distoriton from input tanh gm
Last edited:
Yes but only if the amp itself has sufficient OL gain at that point, right? Suppose the amp has 30dB OL gain at 20kHz, you can put in all the nfb you want but for full power (80V pk-pk) you do need 80/30dB ~ 2.7V differential input signal at the LTP. I don't see that you can get around that.
So, are you saying that a good power amp has 80dB or more OL gain at 20kHz?
jan
I have built amps with 80 dB and more loop gain http://www.diyaudio.com/forums/soli...ain-composite-op-amp-circuits.html#post512806
the 1st post shows a 2-pole compensated composit amp with 80 dB gain @ 20 KHz
2-pole comepensation even allows high flat audio loop gain http://www.diyaudio.com/forums/soli...erview-negative-feedback-205.html#post1428480
which shows a direct comparison with huge IMD reduction over the normal single pole, "high bandwidth" compensation advocated here
the 1st post shows a 2-pole compensated composit amp with 80 dB gain @ 20 KHz
2-pole comepensation even allows high flat audio loop gain http://www.diyaudio.com/forums/soli...erview-negative-feedback-205.html#post1428480
which shows a direct comparison with huge IMD reduction over the normal single pole, "high bandwidth" compensation advocated here
It is not advocated, I just wanted to show how it works. And as there were concerns about Gm variations with amplitude, I did show it as well. Remember, that I started with analysis of input LTP without feedback and such analysis is valid for ANY compensation scheme.
Regarding your 2-pole compensation, how about transient response. Is there a fast rise followed by slow, long time settling? How about settling time at 1% and 0.1%
- Status
- Not open for further replies.