complementary input and VAS aren't the only tools in the box, nor is slew symmetry established as required for low distortion - including PIM once you exceed some minimum slew rate threshold for audio signals
even Self can figure out how to fix up slew rate symmetry in his Blameless, some of us are aware of other topologies as well
remember that any amp with low IMD has low PIM - because PIM is Intermodulation Distortion and does show up in the IMD products
even Self can figure out how to fix up slew rate symmetry in his Blameless, some of us are aware of other topologies as well
remember that any amp with low IMD has low PIM - because PIM is Intermodulation Distortion and does show up in the IMD products
Last edited:
We may have different opinion. Even Self does not achieve >100V/us in a power amplifier, which I get quite easily. It is questionable if high slew rate is not necessary, non-linear distortion is not the reason for me. When I compare complementary-differential to simple differential of similar design, the complementary-differential sounds better to me, and this counts. Yes, both designs have almost same non-linear distortion.
have you built, compared a Cordell Mosfet Amp? - symmetry is maintained to a high level with a differential VAS, achieves very high slew rate without a complementry input http://www.cordellaudio.com/papers/MOSFET_Power_Amp.pdf
JCX, Pavel was reporting his experiences. He knows his way around a circuit. I trust him to have kept the variables to a minimum, and he found that he preferred the sound of the complementary input. (He also did some excellent sims on it a year or two ago). There's nothing wrong with that.
There is not much point to compare it to a completely different design altogether (see edit). And I don't think even Bob would want to make a fully complementary version of his amp -- it's already complex enough!
As for me, I go completely overboard -- differential, complementary, and bridged outputs (ie, diff from input to output). If I could figure out a way to do more, I would. For a while I tried to think of ways to cross-couple signals and so forth, but the only thing I ever built like that was a differential current-to-voltage converter. It was a bear to build because everything had to be matched perfectly -- but it sure sounded good...
EDIT: The only reason to compare it to a completely different design would be if Pavel's non-complementary design failed some specific slew-rate criteria, and that one felt that this single criteria outweighed every single other difference between his design and Bob's design.
There is not much point to compare it to a completely different design altogether (see edit). And I don't think even Bob would want to make a fully complementary version of his amp -- it's already complex enough!
As for me, I go completely overboard -- differential, complementary, and bridged outputs (ie, diff from input to output). If I could figure out a way to do more, I would. For a while I tried to think of ways to cross-couple signals and so forth, but the only thing I ever built like that was a differential current-to-voltage converter. It was a bear to build because everything had to be matched perfectly -- but it sure sounded good...
EDIT: The only reason to compare it to a completely different design would be if Pavel's non-complementary design failed some specific slew-rate criteria, and that one felt that this single criteria outweighed every single other difference between his design and Bob's design.
Last edited:
I really do not know what the 'debate' is about when making power amp circuits. Many early designers, found improvements, all else being equal, back more than 40 years ago, in going to complementary differential input stages, after trying (everything else) first.
Different design approaches are necessary with IC design, since it appears that complementary symmetry jfet input is almost impossible, and usually single differential jfet designs are about the best that can be done. Then Bob Cordell's approach makes the most sense, along with Scott Wurcer's approach, but using jfets in the input, instead of bipolars. Personally, I used a version of Bob Cordell's basic topology from the very beginning, preferring it for lower noise input needs, until Toshiba came out with quiet Pchannel jfets, back around 1980.
Different design approaches are necessary with IC design, since it appears that complementary symmetry jfet input is almost impossible, and usually single differential jfet designs are about the best that can be done. Then Bob Cordell's approach makes the most sense, along with Scott Wurcer's approach, but using jfets in the input, instead of bipolars. Personally, I used a version of Bob Cordell's basic topology from the very beginning, preferring it for lower noise input needs, until Toshiba came out with quiet Pchannel jfets, back around 1980.
the point of the comments is simple, claims about circuit properties, distortion cancellation, slew symmetry are made - I point out that complementary symmetry doesn't better conventional diff pair in this
then slew rate symmetry is asserted as important - I point out that it can be trimmed or matched well in other topologies too
then you play your "trump" that complementary symmetry front ends just sound better
but people following the argument have to ask why you believe it has to be because of those circuit properties that the other circuits can match?
I am skeptical of the sound superiority claim - but even granting it - you aren't succeeding in explaining the circuit performance "why"
then slew rate symmetry is asserted as important - I point out that it can be trimmed or matched well in other topologies too
then you play your "trump" that complementary symmetry front ends just sound better
but people following the argument have to ask why you believe it has to be because of those circuit properties that the other circuits can match?
I am skeptical of the sound superiority claim - but even granting it - you aren't succeeding in explaining the circuit performance "why"
This is why I use resitively loaded LTP's. I also like fully balanced designs - but this is because I have built a few, I am comfortable with them and they overlocme some of the problems wrt unsymmetrical slew rates you see in the Lin. That said, you get 6dB of additional gain as well, and the 2nd harmonics cancel.
If you do some simms, you will see that fully balanced designs are quite tolerant of LTP pair to LTP balance (a few % gives only very low ppm increase in distortion). But, like a Lin topology, they show up distortion very quickly if the individual LTP's are unbalanced.
Re your comments about Bob's amp jcx, - in that design you go from a single ended LTP IP stage to a differential VAS, so the symmetrical slew issue is solved. Its a nice topology - I tried it about 10 years ago, but abandoned it in favour of the fully balanced topology.
I think this toplogy issue is a case of you go with what you like - no doubt there are fine amps using very different approaches
I can't speak for Pavel, but I have no way to explain, "Why?"
If everything could be reduced to some set of fully known, fully understandable requirements and parameters, then the perfect power amplifier would have already been built many years ago.
The only task left would be to figure out how to build it less expensively -- as a DIY'er, to save money for your speaker project, and for a manufacturer to be more competitive.
But most of the time I do listening tests, I get results that are inexplicable, at least to me or anyone I know. Maybe Peter Belt is right. Maybe photographs in the freezer is the key to the whole thing. But at this point I can only try to connect dots and come up with a statement like, "To me amplifiers that are fully complementary sound better than ones that are not."
Some would say I am deluded and that they sound the same. Others would say that I am deluded because the non-complementary one sounds better. Still others may agree with me.
My personal opinion is that what Nelson said is correct. Get your 10,000 hours of listening (not just to music, but to carefully controlled listening tests) and things will become much clearer.
Also comes pretty standard with a CFB design as Mark Alexander showed a couple of decades ago.
It must be understood that some designers achieved 100V/us slew rate in power amps more than 35 years ago. Yes, even before Bob Cordell came on the scene.
There is nothing wrong with optimizing or preferring a specific topology, but why try to push it forward on this thread? If you look at the original JC-2 phono stage, you will see the vestigial topology that Bob uses, e.g. a differential jfet input, bipolar differential second stage with current mirror load, and complementary fets on the output. This comes essentially from analog op amp designs, for example Dick Burwen's design for ADI, back in 1966. I have used it many times for tape recorders, microphone preamps, studio boards, and finally as an ultra low distortion line driver for test equipment (about .001% distortion at 100KHz). It works, but it is not my favorite, as it has a more complex series path, and a slight imbalance due to the current mirror. The complementary symmetry input is ideally more balanced, in my experience.
There is nothing wrong with optimizing or preferring a specific topology, but why try to push it forward on this thread? If you look at the original JC-2 phono stage, you will see the vestigial topology that Bob uses, e.g. a differential jfet input, bipolar differential second stage with current mirror load, and complementary fets on the output. This comes essentially from analog op amp designs, for example Dick Burwen's design for ADI, back in 1966. I have used it many times for tape recorders, microphone preamps, studio boards, and finally as an ultra low distortion line driver for test equipment (about .001% distortion at 100KHz). It works, but it is not my favorite, as it has a more complex series path, and a slight imbalance due to the current mirror. The complementary symmetry input is ideally more balanced, in my experience.
I certainly know this - everyone who contributes here knows this IMO. But I do not use the output coil and do not want to use it. Secondly, the complementary differential sounds better to me.
The gm's simply sum and seconds are cancelled for both individually or together, I would look to the second stage.
I meant the resistor and resistor/transistor combination on top (for npn/n-channel) input, I might be wrong but I don't think this arrangement sources and sinks current totally symmetrically. Only speaking of the classical JC-3/Hafler or whatever you call it input.
- Status
- Not open for further replies.