[PaulFolbrecht]

Hello again,

To start, I know that theory isn't everything, and furthermore that class D amps simply change a lot of the rules and our biases must change with them.

That said, one discovery I had made in years of listening to amps is that negative feedback is simply always detrimental to sound quality.  I don't care what it does to distortion specs, it destroys correct imaging and spaciousness (compared to no neg feedback, that is).  It is used as a crutch to circumnavigate design faults.  I'm talking class A and AB amps here, of course.

I know all class D amps do implement negative feedback and what I have been wondering is if there is some reason why the rules are different here: why negative feedback doesn't necessarily result in the same problems as with the traditional topologies, because it does seem that it does not.

Perhaps someone (Dusty?) could point me to link or offer a brief explanation here.  I think it perhaps might be interesting to a lot of folks.

[CIAudio]

To start, I know that theory isn't everything, and furthermore that class D amps simply change a lot of the rules and our biases must change with them.

That said, one discovery I had made in years of listening to amps is that negative feedback is simply always detrimental to sound quality.  I don't care what it does to distortion specs, it destroys correct imaging and spaciousness (compared to no neg feedback, that is).  It is used as a crutch to circumnavigate design faults.  I'm talking class A and AB amps here, of course.

I know all class D amps do implement negative feedback and what I have been wondering is if there is some reason why the rules are different here: why negative feedback doesn't necessarily result in the same problems as with the traditional topologies, because it does seem that it does not.

Perhaps someone (Dusty?) could point me to link or offer a brief explanation here.  I think it perhaps might be interesting to a lot of folks.
 

Class D amplifiers do use feedback in the same ways as Class A/B types. Saying that negative feedback is simply always detrimental to sound quality is a crazy thing to say...some the world's best amplifiers use negative feedback. It really depends on the complete circuit design and implementation. UcD uses equal feedback at all frequencies, which is why distortion does not rise with frequency. The fact that the models you didn't like had negative feedback doesn't necessarily mean that was the reason.

[JoshK]

I think there is a difference, as you say, in using gNFB as a design crutch and using it carefully. I am not designer but my understanding is a lot of poorer designs use NFB to make a non-linear circuit linear as opposed to taking a design that can stand on its own two feet without it and adding that last touch which lowers distortion even further.  I think, unfortunately, that the commercial market is flooded with poorer designs even if good designs are out there. 

I think that class D is a bit different from traditional designs, maybe cause it is faster?  Don't know.

[PaulFolbrecht]

Dusty,

You may be ENTIRELY right and I should have prefaced my question with "I'm NOT a designer and this is merely an impression I've got from experience and reading".  (You must admit that some designers/brands (Cary, for one) tout 0 NF as a big selling point.  And - that made sense to me - as I found their products to sound completely wonderful, and the theoretical reasons for disliking NF made sense too.  But - there's a theory to support anything.)

So, anyway, sure did not intend it as a dig of any sorts - I know you have some of the best-sounding amps around, at any price, from what so many say.  I have much to learn, I am sure.

Paul

[jon_010101]

JoshK makes some great points.  There are a lot of bad feedback amps, where it is used casually without proper tuning to get good specs out of a mediocre design.  It is easier and cheaper to design a good-sounding 0-feedback amp, especially with tubes: use good simple parts and a triode output stage (PP or SE), which already has a nice low output impedance and reasonable distortion figures. 

But, there are some seriously amazing tube amps with tons of negative feedback (NFB).  The circuit on the HK Citation II is insane... symmetric local feedback between the push-pull halves of the amplifier, and a tiny bit of global, but ~30dB NFB overall! ... one of these will be mine some day .  And, when properly implemented, a triode Williamson amp will sound and measure very darn good, even with ~20dB global NFB.  Circuits like these require careful tuning and perfect output transformer specifications; making a *good* feedback amp will certainly cost more than making a good zero-feedback amp.  Perhaps that's why good, affordable tube feedback amps are increasingly rare?   

Sorry though, regarding your real question, I don't know a thing about Class D amps   Just had to share my tube opinions.

[richidoo]

Cary SLI-80 is a great performing tube design, It handily beats some famous tube amps in head to head comparo, and it uses a fair amount of NFB. I was surprised to learn this after reading all the "no-NFB" hype out of the other side of Cary's marketing mouth. Can't argue with the results, though, somehow Dennis got away without it in the 6 pacs, although SLI-80 sells a hell of a lot more units than 6packs. I have never heard the six pacs or other 0 NFB Cary deisgns (are there any others?)
Rich.

[JeffB]

I am not an audio engineer, so perhaps I shouldn't even comment.
However, I thought class D designs analyzed their output coming out of the chip on the way to the speaker.
And I thought negative feedback was related to measuring what comes back after going through the speaker.
I have read that with negative feedback, by the time the signal has gone through a crossover there is too much time and phase shift in the signal to make any meaningful use of it.
Perhaps someone with more knowledge can correct my feable understanding.

[welwynnick]

Interesting thread.  As a longstanding electronics engineer, audiophile and videophile, I’ve been very curious about the relationship between design, measurement and performance.  It appears to me that once you reach a reasonable standard of fidelity, there is very little correlation between specs and quality.

However, one thing that consistently stands out is that the more processing you apply to a signal, the worse it gets.  By processing, I mean recording, replay, transcoding, AD & DA conversion, amplification, filtering, attenuation, switching, connections.  Everything. 

In this context, I think the important process is amplification.  While NFB does have benefits in improving the linearity of the frequency and amplitude response, reducing output impedance etc, I think you can have too much of a good thing.  Some manufacturers seem to apply so much NFB that THD is reduced to very low levels, yet sound quality gets worse.  However, to achieve the same closed loop gain, higher levels of NFB require correspondingly more open loop gain – more gain stages within the loop, if you like. 

In my book that is simply more processing, and the inevitable slight degradation in transparency may not be offset by the benefits of improved linearity.  Indeed, I don’t think the latter has to be very good at all before you start trading off the former – witness the specs and measurements of valve amps. 

How this applies to class D amps is not so clear, but I understand that many designs use NFB to control linearity.  However, I don’t think this is so necessary as with transistor amps, at least.  Digital power amps are not so intrinsically non-linear as transistors; it’s a question of engineering rather than inheritance, if you like.  It seems that a class D stage can be adequately linear if it is well developed, and to the best of my understanding, NOT all class D amps use NFB.  I’ve used a few digital amps in my system, and I understand that my current Sony DA9000 does not use NFB at all (and sounds all the better for it to my ears, despite mediocre THD figures).  I believe the TACT amplifiers didn’t use any NFB either.

I don’t regard low distortion as a goal to be achieved at any cost, and if a digital amp can achieve reasonably good open loop linearity, then I suspect it may be better off without any NFB.

Regards,   Nick