Andy comments thus in post 221:
I would say there is a problem anyway in a Class AB output stage during the dead zone crossover period. At this point loading on the VAS is up and down like a yo-yo. I confine my comments to AB, something I know reasonably well. Class A is a bit different.
I think this is true. By focussing the VAS load into a fixed resistor, abruptly changing impedances looking into the output stage would be swamped out, and while the VAS would then suffer drastic reduction in its OLG, the lower feedback ratio may not be a bad thing.
Usually, of course, we do reduce OLG on the VAS by use of a lag compensation capacitor, but the brute force regime suggested would certainly swamp the crossover impedance peturbation. This might be no bad thing, and might reduce creation of higher order harmonics.
Moving ahead, Charles in post 222 makes this comment:
This same experience was my epiphany too, but I came to it early on, so wasn't so jaded. It certainly made me think about audio design, and as time has gone on I've come to the conclusion that almost anything can be made to work well. Taking a slavish line, SET/Zero feedback/Class A is probably not too appropriate, because there are examples from all these genres which sound good, and some which sound bad. They can all be done badly, that is certain.
The problem is that audio designers are operating in an area where measurement is helpful, but not definitive. SETs produce horrific distortion, yet sound marvellous. Clearly something else is at work here, and we've not fully figured out what it is.
Andy's math analyses are stupendous, hats off to him. Pavel's topology insights are deep, very clear, Self's analysis is an excellent start point. But when it comes to producing a good amp, this (from Charles) applies:
I suspect the whole process is best done iteratively, experience is king, and listening is priceless. Just how you describe this process, and what precisely you hear, is the challenge.
Audio design is an art, because it cannot subscribe to the engineer's dictum of 'If you can't measure it, it isn't there'.
Cheers,
Hugh
I would say there is a problem anyway in a Class AB output stage during the dead zone crossover period. At this point loading on the VAS is up and down like a yo-yo. I confine my comments to AB, something I know reasonably well. Class A is a bit different.
I think this is true. By focussing the VAS load into a fixed resistor, abruptly changing impedances looking into the output stage would be swamped out, and while the VAS would then suffer drastic reduction in its OLG, the lower feedback ratio may not be a bad thing.
Usually, of course, we do reduce OLG on the VAS by use of a lag compensation capacitor, but the brute force regime suggested would certainly swamp the crossover impedance peturbation. This might be no bad thing, and might reduce creation of higher order harmonics.
Moving ahead, Charles in post 222 makes this comment:
This same experience was my epiphany too, but I came to it early on, so wasn't so jaded. It certainly made me think about audio design, and as time has gone on I've come to the conclusion that almost anything can be made to work well. Taking a slavish line, SET/Zero feedback/Class A is probably not too appropriate, because there are examples from all these genres which sound good, and some which sound bad. They can all be done badly, that is certain.
The problem is that audio designers are operating in an area where measurement is helpful, but not definitive. SETs produce horrific distortion, yet sound marvellous. Clearly something else is at work here, and we've not fully figured out what it is.
Andy's math analyses are stupendous, hats off to him. Pavel's topology insights are deep, very clear, Self's analysis is an excellent start point. But when it comes to producing a good amp, this (from Charles) applies:
I suspect the whole process is best done iteratively, experience is king, and listening is priceless. Just how you describe this process, and what precisely you hear, is the challenge.
Audio design is an art, because it cannot subscribe to the engineer's dictum of 'If you can't measure it, it isn't there'.
Cheers,
Hugh
andy_c said:
On the contrary, i fail to see how a shunt resistor could possibly 'improve' the non-linear loading of a class-B/AB output stage on the TIS......
....in fact, the shunt resistor would merely reduce the overall input impedance of the output stage as 'seen' by the TIS....
This would merely excerbate the problem...methinks...
fab said:
It's all a matter of personal preferences I guess. If you're the type that likes to modify amplifers based on listening tests, then why not? I wouldn't want to impose a design approach on anyone. But if an incorrect technical argument is made, such as in this link that started the discussion about collector resistors here: www.elecdesign.com/Articles/ArticleID/7207/7207.html then I might speak up. In that case, the author was making the argument that wide open-loop bandwidth was necessary for low TIM distortion (false), and that the fix is to add collector resistors. The collector resistors do make the open-loop bandwidth wider but don't improve the TIM distortion. Anyway, the collector resistors came up in a context different from your point regarding nonlinear VAS loading.
It's certainly true that the collector resistors make the DC gain of the VAS more well defined. But is this necessary? The objectivist argument is "no". Because the gain-bandwidth product is not affected by these resistors (GBP in rad/sec is gm/Ccomp), they don't affect the closed-loop bandwidth even though they affect the open-loop bandwidth. So even though the DC gain and open-loop bandwidth are hard to predict, their product is easy to predict, and it's this product, combined with the closed-loop gain, that sets the closed loop bandwidth of the amplifier. But I certainly wouldn't want to tell you whether to include them or not. I just disagreed with the technical arguments that were made in the linked article.
In some cases, such as Nelson's, the designer may have liked the sound better with them than without them, despite the degraded distortion performance. In other cases, it could just be that the designer was following the myth about wide open-loop bandwidth and TIM distortion that has persisted for about 30 years. Or it could be for some other reason that I don't know about. If it's a non-feedback amplifier, you must have them or the amplifier gain is uncontrolled. If one considers only objective criteria and restricts consideration to feedback amplifiers, it's best to leave them out except in possible cases of inadequate output stage current gain. But I don't wish to impose my design criteria on you.
mikeks said:
I was thinking of a Taylor series expansion of V vs. I with a parallel R and a nonlinear element. I would expect the parallel R to straighten out the curve a bit. But I could be wrong.
jam said:
Here's an example of a blatantly unoriginal design. Note the stolen output stage 🙂 http://www.diyaudio.com/forums/showthread.php?postid=198587#post198587
A lot of this stuff is vaporware jam, I admit it! 🙂 I'm struggling through board design now, as I've always had technicians do that at work. But it will eventually happen.
andy_c said:
I haven't had time to respond over the weekend, but thanks
for correcting me. I wasn't sure if it was Barrie Gilbert or
Gilbert Barrie and willingly admit I was to lazy to check it.
Thanks, that seems to confirm my understanding of what
Leach wrote, and it seems to make sense, to me at least.
john curl said:
Only you an Barrie Gilbert know what you discussed, of course,
so I don't mean to argue with you, just speculate a bit. Could
it be that you both were right but speaking different
languages, so to say? Perhaps the effects of slewing limitations
were known to Gilbert and certain other analog designers but
they never thought of it in terms of distorsion? Similarly, Otala
seems not to have though of TIM in terms of slewing limitations
for a start at least, but seems (if I understand it right) to have
thought of it as a somewhat mysterious new form of distorsion.
Perhaps it could have been more of a misunderstanding
between you because of different terminology and different
ways of looking at the same thing? It wouldn't be the first
time in history such things happen. It is certainly a common
reason for many heated debates amonst people on this
forum who in the end turn out to mean the same thing.
Charles Hansen said:
Really Charles, is that the reason? That he can hear "spectral consistency"? Not to talk down Nelson's achievements, far from it, but I think you agree there's other, more significant reasons.
Jan Didden
Re: Theory versus Practice
# 1 is verifiable, reproducable.
# 2 is just one man's opinion.
You are really comparing apples with carburettors.
Jan Didden
[snip]Charles Hansen said:
# 1 is verifiable, reproducable.
# 2 is just one man's opinion.
You are really comparing apples with carburettors.
Jan Didden
AKSA said:
what did you call the device to mix fuel and air on an engine in the days before fuel injection?
AKSA said:
Just consulted my Oxford just to check that the word indeed
exists also in english, so I could just as well quote the
definition from Oxford rather than use my own words:
"that part of an internal combustion engine in which petrol and
air ar mixed to make an explosive mixture".
Edit: Seems Ouroboros posted an explanation in parallell
with this post.
Christer said:
/OT on
What do you know! I invented a perfect English word! I guess I should leave audio like that other guy mentioned earlier!😉
Jan Didden
PS I like the "explosive mixture" part...
janneman said:
You have probably heard the word before. Maybe you even
have it in dutch? We have the word "karburator" in swedish,
but it is very old-fashioned and not very much used for the
past 50 years or so.
Christer, I was in the discussion with Barrie Gilbert, thirty years ago, in 1974. He lives in Oregon, USA. I was born in California, USA. We essentially speak the same language.
Also, I had dinner with Walt Jung for the first time that very evening, and WE discussed both TIM and Matti Otala. Later, in 1976 Matti and I wrote a paper on TIM together. Still later, about 1970, Matti, Marshall Leach, Walt Jung and I got together to write a paper on TIM. This was to head off Bob Cordell, who was chipping at our efforts, in 'Audio' magazine Still later, yet, Dr. Cherry started talking about TIM and wanted to call it 'slope distortion'. And so it goes! It is always easy to find minor faults in the original paths to new understanding. Still, TIM stands today as an important contributor to audio distortion.
Also, I had dinner with Walt Jung for the first time that very evening, and WE discussed both TIM and Matti Otala. Later, in 1976 Matti and I wrote a paper on TIM together. Still later, about 1970, Matti, Marshall Leach, Walt Jung and I got together to write a paper on TIM. This was to head off Bob Cordell, who was chipping at our efforts, in 'Audio' magazine Still later, yet, Dr. Cherry started talking about TIM and wanted to call it 'slope distortion'. And so it goes! It is always easy to find minor faults in the original paths to new understanding. Still, TIM stands today as an important contributor to audio distortion.
john curl said:
This is verifiably untrue......Slew induced distortion (before the onset of slew limiting) only occurs with any significance, in single-pole Miller compensated amplifiers..
This is due to the increased loading on the input stage by the capacitor as frequency increases......
To suggest therefore, that SID is an important contributor to audio distortion is an over-generalization......
......as many amplifiers, (which do not use this inferior frequency compensation technique), including ALL of Nelson Pass' designs are completely immune to this non-problem in their pass-band.....
Re: revised phase distortion calculation
Hi Don....
Sorry i missed your reply to my question....
I am afraid this explanation is deeply flawed....
In both implementations subtraction occurs in the input stage, with the foward path of the amplifier in both instances remaining blissfully unaware that anything at all has changed.....
smoking-amp said:
Hi Don....
Sorry i missed your reply to my question....
I am afraid this explanation is deeply flawed....
In both implementations subtraction occurs in the input stage, with the foward path of the amplifier in both instances remaining blissfully unaware that anything at all has changed.....
- Status
- Not open for further replies.