Bob Cordell Interview: Negative Feedback

PMA said:
Once again, my non-global NFB, real-world amplifier, 2dB under limitation. I believe it is a question of implementation.


Very nice. I always agree it is a matter of implementation.

Remind me again, what's in the output stage and what are the conditions here, such as idle current and load resistance. Is it just a straight AB output stage, or do you have error correction or feedback in there?

Thanks!
Bob
 
Re: BJT distortion vs emitter degeneration

Bob Cordell said:
So if this Baxendall phenomena is making NFB bad, then the so-called no-NFB designs are suffering from it just as well!

Cheers,
Bob

Very interesting observation for this thread to circle around to.

So now that it's been shown (simulated) that, quite possibly, distortion characturistics are not the predominent sonic driver between different approaches to amp design, where does one look for the source of the obvious sonic differences between differing approaches to amplifier design?

I hope everyone is enjoying the Holidays, Mike.
 
Re: Re: BJT distortion vs emitter degeneration

MikeBettinger said:
So now that it's been shown (simulated) that, quite possibly, distortion characturistics are not the predominent sonic driver between different approaches to amp design, where does one look for the source of the obvious sonic differences between differing approaches to amplifier design?
Mike,
the question is here if we agree that this very simplified steady-state distortion test/simulation is covering a real amp's situation or not. Details of actual implementation, especially with high current class-A/B designs and their non-signal-like supply currents might easily dominate conceptual differences, then there are thermal issues, choice of components, etc. Maybe one could even say that different actual implementations of a certain concept might sonically differ as much or even more than different concepts with like implementations (say, one single designer does the circuits and the layout/wiring along his typical preferences)...

- Klaus
 
Re: Re: BJT distortion vs emitter degeneration

MikeBettinger said:


Very interesting observation for this thread to circle around to.

So now that it's been shown (simulated) that, quite possibly, distortion characturistics are not the predominent sonic driver between different approaches to amp design, where does one look for the source of the obvious sonic differences between differing approaches to amplifier design?

I hope everyone is enjoying the Holidays, Mike.


Hi Mike,

You raise a very good question, and you put it in the right terms, using the word different rather than better or worse.

The data that I showed above does not answer this question, although it may shed some light on a piece of the puzzle. That data only suggests that if feedback amplifiers sound different from no-feedback amplifiers, it is not due to NFB creating or exacerbating high-order distortions. The simulations don’t show that distortion is not the predominant driver of sonic differences.

I believe that there are many sources of sonic differences among amplifiers, some of which may only be coincidentally associated with an amplifier design approach. Moreover, the origin of many of these differences may not be well understood, or isolated. I don’t think that the reasons for sonic differences among interconnects, speaker cables and capacitors are well-understood in a measurable sonic cause-and-effect way either.

I do believe that the amplifier interface to the loudspeaker is often a contributor to sonic differences, both in terms of frequency response colorations and amplifier misbehavior. I believe that the frequency response effects due to output impedance interactions, which are readily measurable, are often under-estimated, while at the same time some are claiming hearing acuity that can detect differences that are not measurable.

Have you ever seen a reviewer measure the in-situ frequency response of the review amplifier driving the review loudspeakers through the review cables, as seen at the speaker terminals? Not a difficult thing to do, but I have never seen it.

Apart from distortion, output impedance is a very big typical difference between NFB and non-NFB amplifiers. Even the Ayre MX-R has an output impedance of 0.23 ohms and shows a +/- 0.2 dB response variation into Stereophile’s loudspeaker simulation load, whose impedance dips down to only about 4 ohms. Most well-designed NFB amplifiers are quite flat when driving such a load (but not all!).

Many high-end speakers are designed without any regard for the load which they present to the amplifier, and many dip to 2 ohms or less. Now you might be talking about a potential 0.8 dB peak-to-peak variation for that amplifier. The Ayre MX-R is one of the better no-feedback amplifiers out there.

The Ayre V3 shows a 0.8 dB p-p variation between 2 kHz and 5 kHz on the simulated loudspeaker load. It might show 1.6 dB p-p into a difficult high-end load. It has a high 1.1 ohm output impedance at 20 kHz, and is down fully 3 dB at 20 kHz. And these are some of the more responsibly-designed no-NFB amplifiers out there. We must not neglect the obvious sources of sonic differences.

If the two speakers of the stereo pair do not have identically nasty impedance curves (production variations and component tolerances), then relative channel-channel frequency-dependent phase differences can be introduced, which can alter imaging.

There is another interesting issue out there having partly to do with frequency response. Have you ever wondered what amplifier the speaker designer used when he voiced his speakers? That is the reference, at least according to his ears. If he voiced the speakers with a soft-sounding tube amp with a high output impedance, and then the reviewer plays the speakers with an absolutely neutral straight-wire-with-gain solid state amp with virtually zero output impedance, what will it sound like? Oooohhh, that solid-state glare!

Why LP’s sound different is a whole ‘nother interesting topic of discussion and speculation.

Cheers,
Bob
 
Re: Re: Re: BJT distortion vs emitter degeneration

KSTR said:
Mike,
the question is here if we agree that this very simplified steady-state distortion test/simulation is covering a real amp's situation or not.

Details of actual implementation might easily dominate conceptual differences.

Maybe one could even say that different implementations of a concept might sonically differ as much or even more than different concepts with like implementations (say, one single designer does the circuits and the layout/wiring along his typical preferences)...

- Klaus

Klaus,

As I've been following this there has been a heavy reliance on simulations as sort of a benchmark for theoretical validation of the merit of the various approaches. Bob's simulations are a solid piece of this part of the picture. The details of the implementations have not really entered into the picture... Hence my observation.

I agree 100% with your observations as to the details of the implementation being a significant portion of the end result (that is if I did not mis-interpert your comments), although my experience supports this.

I guess my real point is that the physical world has not entered into the picture and that where does one start, if the long standing hypothosis as to the generation and sonic effect of the higher order distortions has been nixed by the Myth Busters (Bob), to identify the the root cause of the obvious differences between amplifiers.

Regards, Mike.
 
Re: Re: Re: BJT distortion vs emitter degeneration

Bob Cordell said:
Hi Mike,

You raise a very good question, and you put it in the right terms, using the word different rather than better or worse.

The data that I showed above does not answer this question, although it may shed some light on a piece of the puzzle. Cheers,
Bob


Hi Bob,

I'm too tired to respond tonight, although I'd like to. We'll see what tomorrow brings.

Mike.
 
Re: Re: Re: BJT distortion vs emitter degeneration

Bob Cordell said:
Hi Mike,

You raise a very good question, and you put it in the right terms, using the word different rather than better or worse.

The data that I showed above does not answer this question, although it may shed some light on a piece of the puzzle. That data only suggests that if feedback amplifiers sound different from no-feedback amplifiers, it is not due to NFB creating or exacerbating high-order distortions. The simulations don’t show that distortion is not the predominant driver of sonic differences.
Cheers,
Bob

Hi Bob,

I finally got a bit of free time and a chance to follow up on your thoughtful response.
I really did not expect a response to my post; I considered my comments simply a left field observation in a room filled with feedback experts - merely food for thought from the cheap seats.

I do think the work you put into this data is very valuable and it follows my tendency to distill issues down to a practical means of testing a premise and keeping the experiments in an apples-to-apples comparison. The posted results are indeed food for thought


You’re right in that the data does not answer my question, but the data does pose a question as to the audibility of the higher orders of distortion. My interpretation of the results (within the frame work of any of today’s well-executed designs) are: if the distortion were a major player in the sound of an amplifier, a high feedback design would have more in common with the low feedback design. Your data points to (as Edmund points out) feedback is feedback, It appears that the spectral distribution is very similar in both approaches.

Yes, I know it’s been proven that the human ear can more readily perceive the higher orders, but I would question whether they are as easily perceived at normal listening levels within the full body of the music. Grey questioned, in an earlier post, our ability to clearly perceive effects below -60 to -80db (I forget the exact db numbers he discussed, but they were not -110db plus measurements normally associated with current designs). I tend to agree with this thought based on my experiences over the years.

My question was as much about what other aspects of the design might be more influential as it was to question the validity of chasing lower and lower levels of distortion through increased circuit complexity and the resulting demands on the electrical environment that the circuits operate in.

In your response, you list system interactions, the amplifier/speaker interface (output impedance interactions), and the designer’s own reference and experiences all as having a marked effect on the character of the design.

I would add that the power supply interface to the circuit, the physical implementation of the layout, the ground interface to the outside world (the other components in the system), component choices, wiring quality (to a lesser degree than most), as a simple list, all having as much an effect (in my opinion, more of one) on the sound of a design as the feedback levels and distribution.

To bring this full circle, the impetus for my comments was based more on the observation that the topic of feedback can be discussed in infinite detail based on its theoretical body of documentation to bring to the discussion. When the discussion veers into the sonic differences and other possible approaches to circuit design and implementation, as was attempted by Grey in the Feedback Question/Clarification thread, the results are a slow pounding with a textbook.

There should be room for both.

Regards, Mike.
 
IMHO, it is neither the very low THD of current sophisticated designs, nor the frequency response small deviations that matters.

I believe that the problem is a different kind of mess we can see at lower levels. That means PSU and mains interference residuals, voltage drops on signal cable shield and HF induced to cables and systems. All these issues are measurable.

To support my opinion, I would mention cosiderable sonic influence of isolation transformers and mains conditioners, as well as signal cables.
 
PMA said:
IMHO, it is neither the very low THD of current sophisticated designs, nor the frequency response small deviations that matters.

I believe that the problem is a different kind of mess we can see at lower levels. That means PSU and mains interference residuals, voltage drops on signal cable shield and HF induced to cables and systems. All these issues are measurable.

To support my opinion, I would mention cosiderable sonic influence of isolation transformers and mains conditioners, as well as signal cables.

Did you experience with balanced signal cables mainly with very good audio transformers. If yes did you find a sonic difference.

Jean-Pierre
 
Re: Re: Re: Re: BJT distortion vs emitter degeneration

MikeBettinger said:


Hi Bob,

I finally got a bit of free time and a chance to follow up on your thoughtful response.
I really did not expect a response to my post; I considered my comments simply a left field observation in a room filled with feedback experts - merely food for thought from the cheap seats.

I do think the work you put into this data is very valuable and it follows my tendency to distill issues down to a practical means of testing a premise and keeping the experiments in an apples-to-apples comparison. The posted results are indeed food for thought


You’re right in that the data does not answer my question, but the data does pose a question as to the audibility of the higher orders of distortion. My interpretation of the results (within the frame work of any of today’s well-executed designs) are: if the distortion were a major player in the sound of an amplifier, a high feedback design would have more in common with the low feedback design. Your data points to (as Edmund points out) feedback is feedback, It appears that the spectral distribution is very similar in both approaches.

Yes, I know it’s been proven that the human ear can more readily perceive the higher orders, but I would question whether they are as easily perceived at normal listening levels within the full body of the music. Grey questioned, in an earlier post, our ability to clearly perceive effects below -60 to -80db (I forget the exact db numbers he discussed, but they were not -110db plus measurements normally associated with current designs). I tend to agree with this thought based on my experiences over the years.

My question was as much about what other aspects of the design might be more influential as it was to question the validity of chasing lower and lower levels of distortion through increased circuit complexity and the resulting demands on the electrical environment that the circuits operate in.

In your response, you list system interactions, the amplifier/speaker interface (output impedance interactions), and the designer’s own reference and experiences all as having a marked effect on the character of the design.

I would add that the power supply interface to the circuit, the physical implementation of the layout, the ground interface to the outside world (the other components in the system), component choices, wiring quality (to a lesser degree than most), as a simple list, all having as much an effect (in my opinion, more of one) on the sound of a design as the feedback levels and distribution.

To bring this full circle, the impetus for my comments was based more on the observation that the topic of feedback can be discussed in infinite detail based on its theoretical body of documentation to bring to the discussion. When the discussion veers into the sonic differences and other possible approaches to circuit design and implementation, as was attempted by Grey in the Feedback Question/Clarification thread, the results are a slow pounding with a textbook.

There should be room for both.

Regards, Mike.


I agree with your comments. Indeed, we can only address one myth at a time. I have really done nothing more than show that, in the conditions and amplifiers I explored, the myth is not true that the introduction of negative feedback in practice causes larger high-order distortions than were already there. That is one of the favorite pseudo-technical arguments that the no-feedback camp likes to cite, so they will now have to look elsewhere for reasons why NFB is bad.

There always needs to be room for things we just don't understand. If NFB is bad, there is no technical or pseudo-technical reason I know of, that has been properly tested to be true. I always encourage hypotheses about why something is perceived to sound the way it does, but often those hypotheses can and should be checked out. Where we can, we always want to plug in the numbers and do the measurements to test a technical hypothesis. We often learn a lot along the way.

I also think there is a lot of room for discussion about the audibility of different kinds of distortions.

I am a very big believer in fundamental linearity, and detail and design understanding that must be brought to bear on an amplifier to get its distortions way down. I believe that such attention for detail leaves less room for other effects that we may not understand as well.

One also might ask about the popularity of Class-A. Why do so many people like it, in spite of the heat? Is not Class-A's reason d'etre to reduce output stage distortion? Many questions like these remain very intriguing.

Cheers,
Bob
 
PMA said:
IMHO, it is neither the very low THD of current sophisticated designs, nor the frequency response small deviations that matters.

I believe that the problem is a different kind of mess we can see at lower levels. That means PSU and mains interference residuals, voltage drops on signal cable shield and HF induced to cables and systems. All these issues are measurable.

To support my opinion, I would mention cosiderable sonic influence of isolation transformers and mains conditioners, as well as signal cables.


These are good points, although I do think you under-estimate the audible differences due to loudspeaker interface frequency response colorations.

If we are talking about the effect of RF ingress, we should be able to ask and answer some difficult questions.

RF ingress through the input, the mains line, or via the speaker cable is an example of a hypothesis that we ought to be able to test in some way, not just waive our hands about it.

Indeed, one of the pseudo-technical reasons why Charles Hansen doesn't like NFB is his theory that the speaker cable acts as an antenna and that the picked-up signals are fed back to the input of the amplifier where they can intermodulate within the amplifier. Sounds reasonable. Has he or anyone ever measured the effect?

Just because there is a plausible causal effect does not mean that it rises to the level of real-world discernability.

I know that when I turn on my cellphone it causes an interference noise in my cheap computer speakers. That would clearly seem to be a case of RF ingress. But this raises another question. If RF ingress is affecting the sound of the music being played, how is it that we do not hear the effect of the RF ingress as some sort of noise in the loudspeaker in the absence of the music? Usually, in equiipment suffering from some sort of rectification of RF garbage, a noise is readily heard?

Let's hear some suggestions on how we can test for these RF ingress effects in an objective way?

Should we do some wideband RF measurements at the output of our amplifier when it is driving a loudspeaker through 10 feet of cheap speaker cable?

Should we inject some RF twin tones into the output of the amplifier at some current level to see if some intermodulation results.

Other ideas?

Cheers,
Bob
 
PMA said:
I believe that the problem is a different kind of mess we can see at lower levels. That means PSU and mains interference residuals, voltage drops on signal cable shield and HF induced to cables and systems. All these issues are measurable.
I second that. At least it is the only way my engineering alter ego can handle the fact that there are sometimes differences where they aren't supposed to be...

I agree with Bob that it would be good to have some measurement standards, and we might eben find a case where an amplifier gives consistent results both from listening and from measurents when subjected with HF.

The mains issue is also very interesting. In Germany there is/was a very hefty debate about a specific product, a mains regenerator. The problem here is that there is enough positive user feedback (from the pro audio scene also) that it is unlikely that it's all faked, OTOH as an engineer this raises the question "are all those PSU designers idiots", at least in the top class, where cost and effort should not be much of an issue in the first place? Not very likely, either....

Klaus.
 
Re: Re: Re: Re: Re: BJT distortion vs emitter degeneration

Bob Cordell said:
One also might ask about the popularity of Class-A. Why do so many people like it, in spite of the heat? Is not Class-A's reason d'etre to reduce output stage distortion? Many questions like these remain very intriguing.
Besides "absence" of low level signal output stage distortion the most prominent characteristic of class-A is that there not much uncorrelated signal flying around, that is, no rectified load currents in the PSU lines. And in a constant current class-A bridge the PSU completely is out of the picture, WRT load currents. That makes class-A more tolerant to layout/wiring compromises, it seems...

Klaus
 
IMHO, it is neither the very low THD of current sophisticated designs, nor the frequency response small deviations that matters.

I believe that the problem is a different kind of mess we can see at lower levels. That means PSU and mains interference residuals, voltage drops on signal cable shield and HF induced to cables and systems. All these issues are measurable.

To support my opinion, I would mention cosiderable sonic influence of isolation transformers and mains conditioners, as well as signal cables.

Hi, PMA,

A few days ago, me and my friend just experimented with some "hocus-pocus" stuff (that many members here don't believe in). The stuff are various accessories from Harmonix (www.combak.net). My friend buys everything from their catalogue for this test.

I have to admit some of their items have positive obvious impact on sonic performance. Some are confusing (whether make difference of not). But one thing, they all don't make any sense from electrical engineering POV (20hz-20khz).

I make one conclusion about this. Their products indeed focussed on high frequency garbage outside the electronic equipment. Whether it is vibration, electricity mess, or accoustical vibration.

They seems focussing in material properties (as we know all material has spring+damper properties that sets their fo). Like a transformer, they choose material which has less hf junk than ordinary steel plate used in average transformer cores.
They use alot of sandwiched materials (wood, carbon, something), I think it is for tuning the spring+damper properties of the final product.

Like HF bypass cap-Enacom (I measure the value is about 3.3nF), it work when placed in mains, speaker out. We tested it with fire spark from a relay, it disappears completely with this 3.3nF. So, we can safely say that this product is intended to filter HF in electric power/signal.

Something makes very obvious difference yet I cannot make any reason why it is happened. You just have to experience it yourself
. They make a 6 room tuning buttons. It must be placed by 6/set.

Looks like it made from metal covered with somekind of felt outside. We try putting those in and out so many times (because we cannot find any reason why it should make any difference logically), that in the end we have to say that indeed those buttons make a difference in perceived sonics. Yet we cannot say why.
 
KSTR said:

The mains issue is also very interesting. In Germany there is/was a very hefty debate about a specific product, a mains regenerator. The problem here is that there is enough positive user feedback (from the pro audio scene also) that it is unlikely that it's all faked, OTOH as an engineer this raises the question "are all those PSU designers idiots", at least in the top class, where cost and effort should not be much of an issue in the first place? Not very likely, either....

Klaus.

Good point. The mains issue is at once intriguing and a dilemma. Once again, garbage on the rails should be easily measurable. And if the power amplifier has poor PSRR to it, the garbage should be measurable at the output.

Are the guys out there not designing good PSRR into their amplifiers? The answer is probably yes for those designers of the simpler-is-better school. On the other hand, it is not difficult or expensive to design good PSRR into an amplifier. It should also not be that difficult to keep HF mains garbage from getting to the supply rails.

Go figure.

Cheers,
Bob
 
lumanauw said:


Hi, PMA,

A few days ago, me and my friend just experimented with some "hocus-pocus" stuff (that many members here don't believe in). The stuff are various accessories from Harmonix (www.combak.net). My friend buys everything from their catalogue for this test.

I have to admit some of their items have positive obvious impact on sonic performance. Some are confusing (whether make difference of not). But one thing, they all don't make any sense from electrical engineering POV (20hz-20khz).

I make one conclusion about this. Their products indeed focussed on high frequency garbage outside the electronic equipment. Whether it is vibration, electricity mess, or accoustical vibration.

They seems focussing in material properties (as we know all material has spring+damper properties that sets their fo). Like a transformer, they choose material which has less hf junk than ordinary steel plate used in average transformer cores.
They use alot of sandwiched materials (wood, carbon, something), I think it is for tuning the spring+damper properties of the final product.

Like HF bypass cap-Enacom (I measure the value is about 3.3nF), it work when placed in mains, speaker out. We tested it with fire spark from a relay, it disappears completely with this 3.3nF. So, we can safely say that this product is intended to filter HF in electric power/signal.

Something makes very obvious difference yet I cannot make any reason why it is happened. You just have to experience it yourself
. They make a 6 room tuning buttons. It must be placed by 6/set.

Looks like it made from metal covered with somekind of felt outside. We try putting those in and out so many times (because we cannot find any reason why it should make any difference logically), that in the end we have to say that indeed those buttons make a difference in perceived sonics. Yet we cannot say why.

VEEERRRY interesting! Would that we could somehow isolate the phenomenon.

Cheers,
Bob
 
Yes, Bob, it's very interesting 😀
Some products are easily indentified why they perform better. Like the line-conditioner, when we tested them and compared them with other brands, their products makes the cleanest sinewave (seen with osciloscope after distortion analyzer).
We lack of measuring equipment to know why some products make a difference. Like there is a tuning sticker for CD plates, we don't know why it makes a difference. The sticker is made from plastic, cut like a "propeller" shape here and there.
With ordinary sticker (paper) it makes no difference in sound, but with this sticker we can hear a difference. Still don't know why, though.

Some years ago, we tested the cryogenic-ed cables. This cable makes a different perceived sonics, but in the wrong way. Adding HF.

I think PMA is in the right direction. Audio cct's are already matured these days. What will make a difference is when a device is connected with real outside world (with all their junk) from cables, power lines, radiation, etc. This area still have little exploration, IMHO.
 
PMA said:
IMHO, it is neither the very low THD of current sophisticated designs, nor the frequency response small deviations that matters.

I believe that the problem is a different kind of mess we can see at lower levels. That means PSU and mains interference residuals, voltage drops on signal cable shield and HF induced to cables and systems. All these issues are measurable.

To support my opinion, I would mention cosiderable sonic influence of isolation transformers and mains conditioners, as well as signal cables.

My first real system was an HK Citation 12 amp and a DIY copy of a Citation 11 preamp, I was working in a stereo repair shop that serviced the HK gear and was able to purchase the original transformer, lineamp and phono cards from HK. A pretty straight forward aluminum box with a couple inputs and an output. Anything about it that was correct from a layout standpoint was purely an accident. I met a guy who owned an original HK version, actually his system was a pretty good system for the mid 70's, and both of us were curious as to how the DIY version would stack up.

Long story short, the DIY version sounded noticably better. Surprized me, pissed him off, alot. I wasn't expecting it but it started a lifelong interest in the execution details.

My point is, one can design a circuit with vanishing low levels of distortion and still have cloudly/congested mess if the various stages have unintended noise or interference from the other stages circulating through them.

As Pavel has pointed out, this is the low level mess that is also part of the equation. The noise pickup by the cabling, to which I'll add noise voltages created between the various system power grounds that are also circulating through the interconnects. These effects are very difficult to separate from the signal because they appear to be part of the signal. Either being impressed on the signal before it enters the gain stages or injected as part of the ground reference or supply lines to the various stages.

Feedback will have no affect on these but maybe the different gain structures of the low feedback amp relative to the high has something to do with a designs susceptability to these effects.

Just a thought or two.

Regards, Mike.