John,
I used to manage custom IC development projects. The big pros are size, manufacturability (sometimes!) and cost. There is a price to pay for this which, in addition to the items you cited, includes having to make all the components out of silicon. Silicon is non-linear and its electrical properties are variable from part to part. In a discrete circuit we are able to make resistors out of metal film or carbon and capacitors out of metal and ceramic for example. And we can make inductors easily. It seems to me that the primary source of distortion in audio is silicon so the fewer parts made out of it the better.
Brian
I used to manage custom IC development projects. The big pros are size, manufacturability (sometimes!) and cost. There is a price to pay for this which, in addition to the items you cited, includes having to make all the components out of silicon. Silicon is non-linear and its electrical properties are variable from part to part. In a discrete circuit we are able to make resistors out of metal film or carbon and capacitors out of metal and ceramic for example. And we can make inductors easily. It seems to me that the primary source of distortion in audio is silicon so the fewer parts made out of it the better.
Brian
traderbam said:
But this is completely untrue....i have given references here (Hurst, Panissi, Nordholt, Roberge, Rosenstark, Middlebrook...etc) that anyone can obtain at any time...
E.G:
http://www.ardem.com/gft.asp
http://diyaudio.com/forums/showthread.php?postid=463951#post463951
http://diyaudio.com/forums/showthread.php?s=&threadid=41110
mikeks said:
Mike, I'm aware of that. But for the purposes of discussion of the wide open-loop bandwidth issue, I think the comparison of two amplifiers with equal gain-bandwidth products but very different open-loop gain and bandwidth illustrates the concerns very well. They're similar in one important way and different in another. Not to mention that most commercial voltage feedback op-amps have simple single pole compensation, maybe with a zero thrown in that only affects things at much higher frequencies than what we're worried about. Also, since many of these ideas originated with Otala, it makes sense to discuss them in the same circuit design context.
I'd like to convince John of some of these ideas if I can, and in order to do so, I'd reallly, really like to keep things completely non-confrontational and purely technical, like the title of the thread suggests. And if he convinces me of his ideas in the process, that's fine with me too.
john curl said:
I do agree that the AD797 has been much better opamp for audio than the NE5534, but not perfect. Still there remains traces of harsh and smear. I have been investigating a number of opamps again when developing my new error correction amp - here is the review of our colleague:
http://www.diyaudio.com/forums/showthread.php?postid=482839#post482839
I have compared many VFB opamps (OPA627, OPA134, AD797, LT1028, LT1122) with the AD844 CFB opamp, and the AD844 is the winner. It was the only one that outperformed a discrete design in a listening test.
Oh, all the time. Have you not built a circuit and compared open and closed loop performance? Have you compared the realism of the music at differing levels of feedback?
I have.
I'm sure many in this forum will atest to this counter-intuitive experience.
I agree that it is known how feedback works in general, after all it is an extremely simple concept. The thing that isn't widely known is exactly why loop feedback saps life out of music when applied in more than moderate quantities to amplifier designs.
That it does is easily demonstrable. It is supported by the proliferation of low or no loop feedback designs. Take Pass and Krell for instance.
Now this isn't saying feedback is bad as a concept, just that the generally accepted behaviour of feedback is not applicable in these systems and new maths is required.
It is difficult, at this time, for me to describe PIM, except that it is FM distortion of audio signals; rather than AM distortion, which is the normally measured distortion. I now have Barrie Gilbert's paper, but I must reread it carefully, before I can further comment on it. I will try to contact Walt Jung as well, since he has studied this paper and first recommended it to me.
In my practical experience, I use global negative feedback, most of the time, in order to make consistent, easy to make, products. In doing so, I always try to make a relatively simple, balanced circuit that is as linear and as fast as possible. I don't expect negative feedback to do much more than give me better techincal specs. than I need to meet THX requirements.
For me, this makes a 400W amplifier difficult to design, without global negative feedback. Smaller amplifiers, adjusted carefully, and using very high class A operation, can be acceptable, not using global feedback, so long as technical specs are not considered very important.
In my practical experience, I use global negative feedback, most of the time, in order to make consistent, easy to make, products. In doing so, I always try to make a relatively simple, balanced circuit that is as linear and as fast as possible. I don't expect negative feedback to do much more than give me better techincal specs. than I need to meet THX requirements.
For me, this makes a 400W amplifier difficult to design, without global negative feedback. Smaller amplifiers, adjusted carefully, and using very high class A operation, can be acceptable, not using global feedback, so long as technical specs are not considered very important.
Folks.....
....I have taken on board all the constructive advice from members to tone down my approach.....
...I apologise for any offence caused....
This however was inardvertent, as heated exchanges sometimes occur when the truth and nothing but is at stake....
OK...here goes....
Double pole compensation, used correctly, does one thing, and one thing only...
It allows the designer to maintain as high a loop gain across as much of the audio band as possible.......and this without significantly erroding loop gain stability margins...
Now, assuming you had enough raw gain at low frequencies in the foward path (i.e. from input to output with the major loop open-circuit), then.....
...... using this compensation method will reduce ALL forms of distortion,....relative to single-pole compensation...
......provided the designer takes advantage of the enhanced foward path gain by applying sufficient feedback via the major loop.
Now...If you had a 100KHz foward path bandwidth when you applied your double pole compensation, it can only mean your DC foward path gain was significantly less than 60dB to begin with.........
This in turn means you were wasting your time applying double pole compensation, as the circuit to which it was applied had insufficient intrinsic foward path gain to begin with.............
ie: The circuit was sub-optimal..............
........for there is no reason why at least 90dB (foward path at DC) cannot be obtained with two stages of amplification....
....I have taken on board all the constructive advice from members to tone down my approach.....
...I apologise for any offence caused....
This however was inardvertent, as heated exchanges sometimes occur when the truth and nothing but is at stake....
OK...here goes....
john curl said:
Double pole compensation, used correctly, does one thing, and one thing only...
It allows the designer to maintain as high a loop gain across as much of the audio band as possible.......and this without significantly erroding loop gain stability margins...
Now, assuming you had enough raw gain at low frequencies in the foward path (i.e. from input to output with the major loop open-circuit), then.....
...... using this compensation method will reduce ALL forms of distortion,....relative to single-pole compensation...
......provided the designer takes advantage of the enhanced foward path gain by applying sufficient feedback via the major loop.
Now...If you had a 100KHz foward path bandwidth when you applied your double pole compensation, it can only mean your DC foward path gain was significantly less than 60dB to begin with.........
This in turn means you were wasting your time applying double pole compensation, as the circuit to which it was applied had insufficient intrinsic foward path gain to begin with.............
ie: The circuit was sub-optimal..............
........for there is no reason why at least 90dB (foward path at DC) cannot be obtained with two stages of amplification....
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.