Let me try to rephrase that.
Your argument is of the form, "Under the conditions X, which are assumed to be true, then Y will occur".
But that was not the question. The question was of the form, "What are the conditions under which X is true?". Re-asserting X does not lead to any new information.
Ultimately, it has to get down to fundamental mathematical relationships. Here's an example. If the 7th harmonic of the input distortion assuming distortionless output is a certain value, is it valid to assume the same output 7th harmonic given an undistorted input? If not, is there some modification to the conditions under which it will be true? Is it a really good approximation for the second and third harmonics, but lousy for others? It's not at all clear to me what the restrictions are.
Your argument is of the form, "Under the conditions X, which are assumed to be true, then Y will occur".
But that was not the question. The question was of the form, "What are the conditions under which X is true?". Re-asserting X does not lead to any new information.
Ultimately, it has to get down to fundamental mathematical relationships. Here's an example. If the 7th harmonic of the input distortion assuming distortionless output is a certain value, is it valid to assume the same output 7th harmonic given an undistorted input? If not, is there some modification to the conditions under which it will be true? Is it a really good approximation for the second and third harmonics, but lousy for others? It's not at all clear to me what the restrictions are.
I think that it is important to see where audio design has come over the last 40 years. Forty years ago, if you were sitting in an EE classroom, you might hear the professor point out that it is the OUTPUT stage that works the hardest, and therefore generates the MOST distortion. The input stage has relatively little to do, compared to the output, so it may be unwise to resistively DEGENERATE the input stage, because it will REDUCE the total open loop gain of the amplifier, and the loss of total feedback around the amp will increase the final distortion. It also might make the amp more noisy.
Let's look at a good example of this: The GAS AMPZILLA power amp of 1974 or so.
If you look at the design, it is very much what you see here today. Complementary differential transistor input stage, second gain stage, and complementary Darlington output stage. (basically).
This is almost exactly the same input stage that I made for my own amp in 1968, EXCEPT that I used factory MATCHED input pairs (duals) instead of individual devices, and all that I had to do was to match the beta of the complementary pairs, as they were already matched N-N and P-P.
What did GAS do? Well, Jim B. added series emitter resistors to each device, effectively DEGENERATING the input differences, reducing the offset, and making sure that each device runs at about the same current. Good idea! However, this reduced the AC gain of the amp and therefore there was less feedback to reduce the distortion of the output stage. So, individual electrolytic caps were added across each emitter resistor the get back the AC gain, yet the DC gain (and offset) was reduced. Good idea? Seemed a good idea at the time. BUT, the problem is that both TIM and PIM were significantly increased. Yet, on an IM or standard harmonic distortion measurement, it would have looked better. Is this a good tradeoff? I don't think so. Now, what if you removed the caps? Then the slew rate could be increased, the input noise would go up slightly, and 4 caps would have been saved from the BOM, but the xover distortion would probably be too high and JB would have to raise the idle current and then increase the size of the heatsink. Sort of a no-win situation, except that is how we would approach it today.
Let's look at a good example of this: The GAS AMPZILLA power amp of 1974 or so.
If you look at the design, it is very much what you see here today. Complementary differential transistor input stage, second gain stage, and complementary Darlington output stage. (basically).
This is almost exactly the same input stage that I made for my own amp in 1968, EXCEPT that I used factory MATCHED input pairs (duals) instead of individual devices, and all that I had to do was to match the beta of the complementary pairs, as they were already matched N-N and P-P.
What did GAS do? Well, Jim B. added series emitter resistors to each device, effectively DEGENERATING the input differences, reducing the offset, and making sure that each device runs at about the same current. Good idea! However, this reduced the AC gain of the amp and therefore there was less feedback to reduce the distortion of the output stage. So, individual electrolytic caps were added across each emitter resistor the get back the AC gain, yet the DC gain (and offset) was reduced. Good idea? Seemed a good idea at the time. BUT, the problem is that both TIM and PIM were significantly increased. Yet, on an IM or standard harmonic distortion measurement, it would have looked better. Is this a good tradeoff? I don't think so. Now, what if you removed the caps? Then the slew rate could be increased, the input noise would go up slightly, and 4 caps would have been saved from the BOM, but the xover distortion would probably be too high and JB would have to raise the idle current and then increase the size of the heatsink. Sort of a no-win situation, except that is how we would approach it today.
John, this is somewhat a simplification but to the huge class of amplifiers that basicly look like op-amps on steroids there are a limited set of parameters to trade off. I like the inductors across the degeneration like the JE990. Bob Pease sent me a private message stating that he proposed that at Teledyne in the 70's and they laughed at him.
As an asside, I used Jim's Sumo 9 as prior art just recently in a patent suit discovery phase.
As an asside, I used Jim's Sumo 9 as prior art just recently in a patent suit discovery phase.
Well that's sort of not fair since I've worked with Barrie for 32 years and he has his way of looking at things and everyone makes allowances at this level. In the end Barrie will not make a fuzzy or sloppy mistake in an attempt to prove a point or promote one approach over an other. I remember easedropping on an email trail between him and Charles Hansen concerning killing the open loop-gain to make it constant over the audio bandwidth. I personally think this has no merit but Barrie invested a serious amount of time to walk Charles around to seeing his point of view.
Look at slide #13 and #27
http://www.hypex.nl/docs/Bruno Masterclass/slides.htm
#37 about inverting amps
http://www.hypex.nl/docs/Bruno Masterclass/slides.htm
#37 about inverting amps
andy_c said:
Andy,
OK I see what you mean. The answer to those questions is not in Baxandall's article, of course. He never went into those details at all; his reasoning was a way to get to some simplifications that would be close enough to be useful for his purpose. I'm sure that people like you and Brian can wield some heavy math tools around to show that his reasoning would fall apart under this and that condition. In the mean time, he has used his tool and helped himself and lots of people to better understand some things. Things that are not 100% correct under any and all circumstances, but worhwhile nevertheless.
I'm not defending Peter Baxandall in any way, mind you (as if he needs that!). It's just that he and some others have a way to cut through the chase by knowing what to take in, what to ignore. I personally like that, it's a very efficient way to get close to your goal, and if you really need to be much closer, you can breake out the heavvy math.
Jan Didden
SY said:
Sorry SY missed that post. If you remember, the amp that didn't show the effect was a tube amp with a combination of positive inner loop feedback and negative global loop feedback. (The article describing that amp is on my website in My Library, it's by John Miller). Hardly an aples-to-apples comparison to Baxandalls' single FET stage. I also qualified the issue by saying that the Baxandall chart was an example of unexpected behaviour *for this particular circuit* and I also mentioned that Bob's research showed that it didn't occur in other cases.
Jan Didden
Jan,
Imagine you have 3 linear, forward path gain blocks in series, with gains A, B & C. Simple unity gain feedback. The input signal is Vin and the output is Vout. Now introduce 3 disturbance signals at the inputs of each gain block, call these Va, Vb & Vc.
(1 + ABC).Vout = ABC.Vin + ABC.Va + BC.Vb + CVc
The disturbances represent distortions. The rejection ratios of the disturbance signals are not the same. Va is reduced the least. You can imagine Va being the distortion introduced by the input subtractor. The closer the error signal is to the input signal, in terms of total forward gain, the less it will be reduced.
Does this make sense?
Brian
Imagine you have 3 linear, forward path gain blocks in series, with gains A, B & C. Simple unity gain feedback. The input signal is Vin and the output is Vout. Now introduce 3 disturbance signals at the inputs of each gain block, call these Va, Vb & Vc.
(1 + ABC).Vout = ABC.Vin + ABC.Va + BC.Vb + CVc
The disturbances represent distortions. The rejection ratios of the disturbance signals are not the same. Va is reduced the least. You can imagine Va being the distortion introduced by the input subtractor. The closer the error signal is to the input signal, in terms of total forward gain, the less it will be reduced.
Does this make sense?
Brian
traderbam said:
Yes, makes sense. And it also means, the closer the disturbance is to the output, the less impact it has on the output distortion.
In my example I had two amps, one with the disturbance at the input side, another with the disturbance at the output side. Assuming that the open loop distortion was equal in both cases, closing the loop would lead to identical amplifiers. So, from this pov, the location of the disturbance, at the front or at the rear, doesn't make a difference. And that was the statement I questioned.
BUT, in the case of the amp with the disturbance at the input side, that disturbance has to be smaller than the disturbance at the output side in the other amp if the two amps had to have identical output disturbances in OL, right? So I guess what we could say is that the further to the input a disturbance is generated, the more it impacts on the OL output distortion. But we already knew that, didn't we ?
Jan Didden
Yes, Jan. But I interpreted Hugh's point as arguing the case of the importance of the input subtractor because its error is not fixed by feedback. I interpreted your responses as suggesting that no source of error is more important than another just because of its location in the loop (which is not right):
Perhaps you and Hugh were talking about different things. I didn't want any readers to accidentally get the idea that the location of the error doesn't matter to the transmission of that error to the output. Eg: -30dB of distortion generated at the LTP stage is much more significant than -30dB generated at the output stage, in general.
and
Perhaps you and Hugh were talking about different things. I didn't want any readers to accidentally get the idea that the location of the error doesn't matter to the transmission of that error to the output. Eg: -30dB of distortion generated at the LTP stage is much more significant than -30dB generated at the output stage, in general.
Hi,
I tried a little sim on this, 3 stages with Aol=10 each (no poles), and injecting noise/distortion (a fraction of the squared input voltage, here) at one of this stages, in turn.
When the injection is made with the same levels at each stage, the reduction gets less the closer to the input I move (as expected), but with weighted injection (so that each stage has the same local THD which then gives the same THD at the output also) there is no change in resulting distortion. That means, if each stage adds the same amount of distortion with respect to its own output voltage it does not matter where this distortion takes place? Say, 10% THD in the IS, VAS or OS resp., one at time, doesn't change things?
Did I confuse things?
- Klaus
I tried a little sim on this, 3 stages with Aol=10 each (no poles), and injecting noise/distortion (a fraction of the squared input voltage, here) at one of this stages, in turn.
When the injection is made with the same levels at each stage, the reduction gets less the closer to the input I move (as expected), but with weighted injection (so that each stage has the same local THD which then gives the same THD at the output also) there is no change in resulting distortion. That means, if each stage adds the same amount of distortion with respect to its own output voltage it does not matter where this distortion takes place? Say, 10% THD in the IS, VAS or OS resp., one at time, doesn't change things?
Did I confuse things?
- Klaus
Attachments
KSTR said:
I think it makes sense. With weighting down the THD depending on how close you are to the input, you decrease the distortion levels if they have more gain to go through towards the output, if I understand you correctly. I guess this weighting down is with the ratio of the OL gain of each block?
Jan Didden
Hi Jan,
Yes, the weighting is corresponding to the OLG of each stage.
But, then we have some (semantical) contradiction, somehow:
It seems one needs to qualify that "smaller". Smaller only in absolute amounts (say, injected PSU ripple etc). From the distortion perspective, the errors should be referred to the actual signal levels at each stage and then we have a "same in, same out" black-box situation which is "blind" to what actually happens inside...
- Klaus
Yes, the weighting is corresponding to the OLG of each stage.
But, then we have some (semantical) contradiction, somehow:
janneman said:
It seems one needs to qualify that "smaller". Smaller only in absolute amounts (say, injected PSU ripple etc). From the distortion perspective, the errors should be referred to the actual signal levels at each stage and then we have a "same in, same out" black-box situation which is "blind" to what actually happens inside...
- Klaus
No you didn't confuse things, you are quite right. I confused things.
In a LINEAR system, it does not make any difference where an error is injected, downstream of the subtractor, if the error % is the same. That's right, because everything is simply multiplied together.
The absolute subtraction error will be added at the output (scaled by the CL gain).
Refer to Table 1, the third and fourth data set. He breaks out the effects of NFB separately from the combined feedback, which is what caught my attention at your talk. As I had guessed at the time, the feedback is relatively low (11dB without the regeneration). Yet every harmonic is decreased.
My point was not an apples to apples comparison- it's an observation that what you have in your hand is never actually an apple, it's a tangerine. The conclusions drawn from a simple square-law model are probably not generally applicable to real, multistage amplifiers, whether transistor or tube. Low feedback factors may indeed be beneficial.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.