"Best sounding" is subjective, so anything goes.
Likely the correct answer is "any, all or none".
Just read the review of your two integrated amps in Stereo magazine. They have different distortion spectra and the reviewer takes about sound differences too. The class A version has a bit more distortion and as far as i can tell it is third. He claims that the class A version sounds more creamy and sweet. Sorry, that may be not the exact meaning of the german words but i can not find a better translation at the moment.
That's an excellent article, but it doesn't address the topic at hand. I'm apparently making a poor presentation of my topic, so I'm try again in a different way:
The classic feedback model with an integrating loop around various gains and summers has a characteristic complex function of its own. See Baxandall, or recently Putzeys. This new complex function only applies to an integrating loop, and my contention is that a single degenerated stage doesn't meet the integrating loop criteria.
This can be tested (in the real world - thank you very much) and I'm working on something to satisfy my own curiosity. Others, for example Nelson Pass and BV in this thread have shown a path in the model world, but since this is a question of the model itself, only the real world will settle it for me. The actual numbers are small, but not beyond the gear here at the shack.
Thanks,
Chris
Hi Scott,
Not quite. You have been told that it is possible to make very low distortion output stages (I should note that EF2 is not low distortion and thus essentially counts as suboptimal design) and that if this is done, the VAS Stage often has more distortion.
With enough effort (or lack thereof) it is very easy to design a very poor linearity output stage, but given the current state of the art, why would anyone do such a thing.
John in the 1970's can be easily forgiven for his EF2 stage, (though he could have added a pair of J-Fet buffers before the EF2 to make it by far less problematic), plus it is largely Class A which reduces the issues somewhat.
But now is more than 3 decades later and with the work of Self, Cordell, Locanti and others there little excuse to continue such suboptimal practice. In fact, replacing the Drivers in an EF2 with low capaitance Mosfet's can bring dramatic improvements, something which I believe JC does these days.
Ciao T
Not quite. You have been told that it is possible to make very low distortion output stages (I should note that EF2 is not low distortion and thus essentially counts as suboptimal design) and that if this is done, the VAS Stage often has more distortion.
With enough effort (or lack thereof) it is very easy to design a very poor linearity output stage, but given the current state of the art, why would anyone do such a thing.
John in the 1970's can be easily forgiven for his EF2 stage, (though he could have added a pair of J-Fet buffers before the EF2 to make it by far less problematic), plus it is largely Class A which reduces the issues somewhat.
But now is more than 3 decades later and with the work of Self, Cordell, Locanti and others there little excuse to continue such suboptimal practice. In fact, replacing the Drivers in an EF2 with low capaitance Mosfet's can bring dramatic improvements, something which I believe JC does these days.
Ciao T
Hi,
There is no "patent recipy".
I would suggest that it is best to start using devices and topologies that have the best "inherent" linearity and then to seek to improve the linearity first.
Ideally we have the output stage the dominant source of distortion and apply just enough loop feedback around it to reduce this distortion to what we find acceptable.
My definition of "acceptable" is at least 10dB below the distortion of Speakers and similar in nature, if we do not actively exploit distortion cancellation between amplifier and speaker.
Now I know no dynamic, electrostatic or planar speaker that has much less than 0.1% THD (either 2nd or 3rd HD dominant) at 1 Watt input and non that have much below 1% at rated power.
So I would suggest that "acceptable performance" is one that produces no more than 0.3% 2nd HD and 0.1% 3rd HD at rated power below clipping and less than 0.03% 2nd HD and 0.01% 3rd HD at 1 Watt, with the distortion below 1 Watt behaving monotonic (distortion falls with level) and preferably also above 1 Watt.
It should be noted that well designed Class A output stages easily match the above figures without NFB, while optimised Class AB output stages only exceed the above by less than 10dB or so.
I think global feedback is perfectly acceptable, where care has been taken for the amplifying circuit to not suffer bandwidth modulation with signal levels (degeneration and local feedback can be used to produce such results) and the necessary stability requirements etc. have been fulfilled and where thermal memory distortion inside the loop is reliably attenuated by the feedback loop (and not adds increased problems).
Degeneration has limits, it is my experience that beyond a certain point degeneration is less effective in linearising circuits than looped feedback, plus it invariably raises the output impedance, which can impact open loop bandwidth (though not slew rate).
Global looped feedback has more potential problems, but it is probably much easier to design (it's design effort approaches "brainless" these days) and at least in measured terms it can hide a multitude of sin's, including excessive VAS distortion etc., so people feel they do not need to address them locally. It is an easy option to good specs. The results can sound good, however IME the results of carefully optimising the whole circuit and making do with less global lopped NFB sound better and measure reasonably similar.
Ciao T
There is no "patent recipy".
I would suggest that it is best to start using devices and topologies that have the best "inherent" linearity and then to seek to improve the linearity first.
Ideally we have the output stage the dominant source of distortion and apply just enough loop feedback around it to reduce this distortion to what we find acceptable.
My definition of "acceptable" is at least 10dB below the distortion of Speakers and similar in nature, if we do not actively exploit distortion cancellation between amplifier and speaker.
Now I know no dynamic, electrostatic or planar speaker that has much less than 0.1% THD (either 2nd or 3rd HD dominant) at 1 Watt input and non that have much below 1% at rated power.
So I would suggest that "acceptable performance" is one that produces no more than 0.3% 2nd HD and 0.1% 3rd HD at rated power below clipping and less than 0.03% 2nd HD and 0.01% 3rd HD at 1 Watt, with the distortion below 1 Watt behaving monotonic (distortion falls with level) and preferably also above 1 Watt.
It should be noted that well designed Class A output stages easily match the above figures without NFB, while optimised Class AB output stages only exceed the above by less than 10dB or so.
I think global feedback is perfectly acceptable, where care has been taken for the amplifying circuit to not suffer bandwidth modulation with signal levels (degeneration and local feedback can be used to produce such results) and the necessary stability requirements etc. have been fulfilled and where thermal memory distortion inside the loop is reliably attenuated by the feedback loop (and not adds increased problems).
Degeneration has limits, it is my experience that beyond a certain point degeneration is less effective in linearising circuits than looped feedback, plus it invariably raises the output impedance, which can impact open loop bandwidth (though not slew rate).
Global looped feedback has more potential problems, but it is probably much easier to design (it's design effort approaches "brainless" these days) and at least in measured terms it can hide a multitude of sin's, including excessive VAS distortion etc., so people feel they do not need to address them locally. It is an easy option to good specs. The results can sound good, however IME the results of carefully optimising the whole circuit and making do with less global lopped NFB sound better and measure reasonably similar.
Ciao T
Thorsten, I would like to point out that over the decades, my design philosophy has 'evolved'. At the very beginning, starting in 1967, I was NOT against using all the feedback that I could muster, without oscillation. In fact, complementary differential input stage gave a 6 dB improvement in this direction.
However, when Otala came on the scene, in 1970, I decided to give HIS ideas a good shot, and I did it for the Grateful Dead modules for the Wall of Sound. It WORKED FOR THEM, and independently, Mark Levinson heard the difference compared to the best IC's that he could get at the time, so he wanted me to design modules in that direction, which became the Levinson JC-2, perhaps my most successful design. Then, in the same spirit, of wide open-loop bandwidth, high slew rate, and low global feedback, the JC-3 was designed the way that it was. However, today I would design it somewhat differently, BUT one of MY original guidelines was to use paralleled circuits but to simplify the series circuits. So, I tend to avoid extra devices, IF it is at all possible to avoid them. What I usually do, instead is to use relatively low resistive loading on the input stage, and increase the current of the second stage to have its percentage change in current to be small for the job that it has to do. So, I might use 10-50ma of standing drive current to keep it linear.
In any case, adding extra cascodes or darlington followers is something that I try to avoid, if I can. I 'could' get better specs. if I added more of these components, but it would not necessarily sound better, in fact, it might sound worse.
However, when Otala came on the scene, in 1970, I decided to give HIS ideas a good shot, and I did it for the Grateful Dead modules for the Wall of Sound. It WORKED FOR THEM, and independently, Mark Levinson heard the difference compared to the best IC's that he could get at the time, so he wanted me to design modules in that direction, which became the Levinson JC-2, perhaps my most successful design. Then, in the same spirit, of wide open-loop bandwidth, high slew rate, and low global feedback, the JC-3 was designed the way that it was. However, today I would design it somewhat differently, BUT one of MY original guidelines was to use paralleled circuits but to simplify the series circuits. So, I tend to avoid extra devices, IF it is at all possible to avoid them. What I usually do, instead is to use relatively low resistive loading on the input stage, and increase the current of the second stage to have its percentage change in current to be small for the job that it has to do. So, I might use 10-50ma of standing drive current to keep it linear.
In any case, adding extra cascodes or darlington followers is something that I try to avoid, if I can. I 'could' get better specs. if I added more of these components, but it would not necessarily sound better, in fact, it might sound worse.
Scott,
What was the output device bias in your sim, did you use the same values as for JC's design? I don't think so.
If you want to show what happens in JC's Amp I think you need to fully replicate the design and then analyse it. Anything else may be construed as deliberately producing results to prove your point by altering the experiment.
Clearly you do. Does that make it true?
I think I stand with "folksy" inner loop and Walt as well John (and others, going back at least to Kiebert, Walker, Gow et al in the late 40's and 50's of the last century).
With respect, Scott, given that you earn your living by messing every day with designing stuff that relies on global feedback loops exclusively, it is not unusual for you to view things exclusively from that angle. However I think there is no need to try to declare this method to be the only true way and to try to defend it all costs. You may gain more by looking closer at what is really going on then by attempting to debunk things.
Ciao T
What was the output device bias in your sim, did you use the same values as for JC's design? I don't think so.
If you want to show what happens in JC's Amp I think you need to fully replicate the design and then analyse it. Anything else may be construed as deliberately producing results to prove your point by altering the experiment.
Clearly you do. Does that make it true?
I think I stand with "folksy" inner loop and Walt as well John (and others, going back at least to Kiebert, Walker, Gow et al in the late 40's and 50's of the last century).
With respect, Scott, given that you earn your living by messing every day with designing stuff that relies on global feedback loops exclusively, it is not unusual for you to view things exclusively from that angle. However I think there is no need to try to declare this method to be the only true way and to try to defend it all costs. You may gain more by looking closer at what is really going on then by attempting to debunk things.
Ciao T
if you are going to stand with Walt on this issue you have some walking to do - both he and Marshall Leach initially embraced Otala's prescription of "flat open loop gain" and after publishing designs incorporating that idea later came to the conclusion that flat open loop bandwidth was not required for low PIM - both concluded that you can design amps with large feedback factors, sloping open loop gain and still avoid/have reduced PIM
JCX, I think you are confusing TIM with PIM. It is true that Walt, Marshall, and I stopped thinking that extremely high open loop bandwidth was necessary for TIM, and we argued with Matti Otala about it, back in the late 70's. Then, Matti wrote a completely mathematical paper on PIM, and it was NOT put into the JAES, because Lipshitz et al did not want it there. Then, Cordell made some measurements that seemed to say that PIM was not very important, and we left the matter alone for about 15 years (a long time to be sure).
Then Barrie Gilbert came up with his paper on OP Amp Myths, and showed that PIM could indeed exist even with an otherwise ideal (distortion wise) toplogy like the 741 and its slightly improved versions, such as the 4558, and 4562 (as they use the same topology) and he published it, I believe, in house, at ADI. Walt Jung picked up on it, wrote an article about this paper in 'Electronics Design', I believe about 15+ years since Matti's paper was first given. He sent a copy to me, and he and I tried to get Barrie's paper popularized and available. Barrie apparently didn't care one way or another. I suspect that it is still available on Walt's website, and I started to cite it here.
I find nothing intrinsically wrong with Barrie's paper, and then Ron Quan did his paper on PIM, given last year at AES. The only criticism of Ron's paper that I know of, so far, is that he did not publish any modern circuits. Well, that does not mean that the problem has gone away, and Charles and I are going to push for what we find works in audio design, even if we can't 'prove' it to virtually everyone.
Then Barrie Gilbert came up with his paper on OP Amp Myths, and showed that PIM could indeed exist even with an otherwise ideal (distortion wise) toplogy like the 741 and its slightly improved versions, such as the 4558, and 4562 (as they use the same topology) and he published it, I believe, in house, at ADI. Walt Jung picked up on it, wrote an article about this paper in 'Electronics Design', I believe about 15+ years since Matti's paper was first given. He sent a copy to me, and he and I tried to get Barrie's paper popularized and available. Barrie apparently didn't care one way or another. I suspect that it is still available on Walt's website, and I started to cite it here.
I find nothing intrinsically wrong with Barrie's paper, and then Ron Quan did his paper on PIM, given last year at AES. The only criticism of Ron's paper that I know of, so far, is that he did not publish any modern circuits. Well, that does not mean that the problem has gone away, and Charles and I are going to push for what we find works in audio design, even if we can't 'prove' it to virtually everyone.
Last edited:
Thanks for the references.
The link given for Barrie's paper is usually cited as www.edtn.com/analog/barrie1.htm, Mar 9, 1998, but that site is dead and forwards to an empty page on EE Times. Searching the site for Barrie Gilbert or "Op-amp Myths" seems to give too few or too many hits, depending upon the date range selected.
The closest I could get was with the WayBackMachine:
Analog Avenue:Op amp Myths by Barrie Gilbert (3/9/98)
The link given for Barrie's paper is usually cited as www.edtn.com/analog/barrie1.htm, Mar 9, 1998, but that site is dead and forwards to an empty page on EE Times. Searching the site for Barrie Gilbert or "Op-amp Myths" seems to give too few or too many hits, depending upon the date range selected.
The closest I could get was with the WayBackMachine:
Analog Avenue:Op amp Myths by Barrie Gilbert (3/9/98)
Yes, I read the article. As far as I remember, he defined distortion as anything in the output which is not merely the input multiplied by a constant. It can be as simple or as complex as you like. NFB will reduce it, while complicating it due to reacting with itself. I think his work is only incomplete in the sense that it assumes that every stage has an algebraic transfer function combined with a filter. This close enough to the truth for most practical purposes. Remember, we are trying to make audio amps, not invent pathological circuits which deliberately break reasonable assumptions in the maths.simon7000 said:
Sorry, as you feared, I don't understand this sentence. The article I linked to does address the issue, because it shows by calculation and measurement that degeneration suffers from exactly the same 'feedback mixing' effect as is seen in other forms of feedback. Feedback theory does not care about the nature of the feedback, it simply says take a sample of the output signal (current, voltage, oil temperature, steam pressure) and by some means subtract it from the input signal. Just because we normally apply this to voltage sampling does not mean that it doesn't apply equally well to current sampling, as in degeneration. That is the beauty of maths.Chris Hornbeck said:
The fact that a stable loop must have "integration" in it pertains to the fact that loop gain goes through 0dB before phase goes round the clock, not that loop gain must necessarily go all the way to infinity at DC. In fact practical integrators always level off at some point where they reach their DC gain. With a degenerated transistor this just happens at a high frequency so for most of the band it appears almost like a P controller with no integration. In reality though there are NO true P controllers. Every physical has a bandwidth limitation i.e. some HF poles. Witness the fact that some follower circuits oscillate quite happily. Those that don't will upon close inspection turn out to have a dominant pole. It just happens to be real and well in the megahertzes. And yes, this loop has a transfer function (or complex function as you call it) just like any frequency dependent system, and it's written mathematically as a horizontal line with a polynomial below it and a shorter one above it.
Back to hiding, Jan Didden asked me to chip in so I hope this is more or less the answer he was looking for.
Last edited:
Thank you for the link.
What is this "a rather new alternative to precise low-distortion, wideband, voltage-mode amplification" ?
Regards
George
PS Barrie Gilbert's article continues here:
http://web.archive.org/web/20010810144803/http://www.edtn.com/analog/barrie3.htm
http://web.archive.org/web/20010810144803/http://www.edtn.com/analog/barrie4.htm
Last edited:
I wish to correct an oversight in my previous input. I meant to say the 4558, 4559, 4560 series of IC op amps,that so far as I know, are still very popular for mid fi. These are problematic, but the 4562 appears to be comparable to the best of today's op amps, and apparently it was deliberately designed to be so. However, PIM distortion may still present, in the 4562, but to a much smaller amount. What the ear detects, I do not know, but it might well be more sensitive to PIM than most of us realize.
I knew this post would come. Yawn, you couldn't be more wrong. In the first place I clearly stated that I just closed the loop with NO LOAD on the output stage save the 80k feedback so it wouldn't taint the results. What is shown is solely (very closely so) the distortion of the VAS.
I will state, to be clear, the local feedback (2-1M resistors) shown in that circuit do nothing useful.
I will state it again degeneration creates a fundamentally different voltage to current transfer (lower and more linear) and arguing as to if it is feedback or not is beside the point. In the JC-3 VAS the transconductance of the input pair is not changed by the feedback so current demaded by the output (including the VAS's own output impedance) causes exactly THE SAME ERROR VOLTAGE AT THE INPUT. jcx said essentially the same thing.
As for the last comment, I expected better.
- Status
- Not open for further replies.