I must say, Charles, you are 1 helluva nice guy. The apology to Pavel Dudek was very nice of you.
I read through Pavel Macura's comments on feedback, and, to my surprise, I found I agreed with everything he said! I've had very similar experiences.
John, agree emphatically on CFP comments. Less stable than the DEF. I find that any residual instability of the DEF can be cured completely with base stoppers; 100R on drivers, and 10R on outputs. The DEF bias set is easy; the CFP bias set has hair trigger, almost thermonuclear, sensitivity.
Cheers,
Hugh
I read through Pavel Macura's comments on feedback, and, to my surprise, I found I agreed with everything he said! I've had very similar experiences.
John, agree emphatically on CFP comments. Less stable than the DEF. I find that any residual instability of the DEF can be cured completely with base stoppers; 100R on drivers, and 10R on outputs. The DEF bias set is easy; the CFP bias set has hair trigger, almost thermonuclear, sensitivity.
Cheers,
Hugh
Hugh, I agree with you on this. Now, when it comes to distortion byproducts, I don't know how to compare the two. I am pretty sure that a typical CFP has a lower output impedance than a dual emitter follower. This was the reason that Otala found that this was the best performing output stage for IIM distortion, which requires a low 'open loop output impedance' for lowest distortion.
Personally, I am concerned with any kind of feedback, because it tends to multiply the intrinsic harmonic series, generating higher order harmonic and IM distortion products. I don't know if this happens with a follower, however.
I agree with your ' base stopper resistors'. Many people don't realize that dual followers can create so much phase shift at very high frequencies, that you can actually generate - R, or negative resistance. This means that oscillation is going to happen. + R in the input leads of the transistors usually cancels with any - R generated.
Personally, I am concerned with any kind of feedback, because it tends to multiply the intrinsic harmonic series, generating higher order harmonic and IM distortion products. I don't know if this happens with a follower, however.
I agree with your ' base stopper resistors'. Many people don't realize that dual followers can create so much phase shift at very high frequencies, that you can actually generate - R, or negative resistance. This means that oscillation is going to happen. + R in the input leads of the transistors usually cancels with any - R generated.
My new creation
Xcuse me for the big schematic
Inspired by this very thread I made the crossed distortion killer circuit without feedback loop around the output fets M1 & M3.
The difference between leftside input and left side output is added to the right side and vice versa.
Gives K2 -158 dB and K3 125 dB across load.
With feedback it is even much better.
What ya think ?
Xcuse me for the big schematic
Inspired by this very thread I made the crossed distortion killer circuit without feedback loop around the output fets M1 & M3.
The difference between leftside input and left side output is added to the right side and vice versa.
Gives K2 -158 dB and K3 125 dB across load.
With feedback it is even much better.
What ya think ?
john curl said:
If it does, that would certainly dispel the notion that a simple emitter-follower is an "open loop" circuit with no feedback, wouldn't it?
Andy, would you care to do a distortion spectra for a simple emitter follower versus a common-emitter with the same value of load resistor?
se
To Charles Hansen
Charles, you are gentleman and I am glad, that all between us is clear. In last hour I was reading some magazines, where are tests of your products ( before I had any information, only that exist fa Ayre - you know, newcoming " easteuropean " ). Please explain me, how is possible, that you don't hear distortion, which I saw there, but please don't talk about some tubelike nice sound - this words I heard manytimes and for me they are only synonymum for distorted sound. We are friends with Pavel Macura ( PMA ). His amp A - 1 ( see above ) have theoreticaly optimal " feedbackless " configuration ( three independent maximaly linearized stages ), what is searched on this thread several days. We both with him commonly test our " babyes " and I must say, that I agree with all his words. Parametres of A - 1 are excelent and for many amps here showed hardly accessible, but our listening experience exactly follow measured curve of distortion : in low and middle level, when measured distortion was below 0,01 %, was all OK. When by higher level distortion was rised, we both hear it quite clearly as " offocusiation and loss of space picture ". In this case was distortion still below 0,1 % ! Newest his amp A - 2, which have thanks to overall feedback distortion ten times lower, hold basic character of sound of this amp in every levels the same. Many people here and not only here, take distortion as necessary for " listening enjoy " and becouse any amps haven't the same rate of harmonics and the same dependability of this on output power, people evaluate some as "musical" and some as " less musical ". They don't know, that musical apparatus is only this one, which reproduce only music, not one , which reproduce music + " something". With amp which have ultralow distortion, you can hear " silence between notes " or " fart of mouse in studio ", but I'm sorry, without feedback is hard to get it .
Charles, you are gentleman and I am glad, that all between us is clear. In last hour I was reading some magazines, where are tests of your products ( before I had any information, only that exist fa Ayre - you know, newcoming " easteuropean " ).
Please explain me, how is possible, that you don't hear distortion, which I saw there, but please don't talk about some tubelike nice sound - this words I heard manytimes and for me they are only synonymum for distorted sound. We are friends with Pavel Macura ( PMA ). His amp A - 1 ( see above ) have theoreticaly optimal " feedbackless " configuration ( three independent maximaly linearized stages ), what is searched on this thread several days. We both with him commonly test our " babyes " and I must say, that I agree with all his words. Parametres of A - 1 are excelent and for many amps here showed hardly accessible, but our listening experience exactly follow measured curve of distortion : in low and middle level, when measured distortion was below 0,01 %, was all OK. When by higher level distortion was rised, we both hear it quite clearly as " offocusiation and loss of space picture ". In this case was distortion still below 0,1 % ! Newest his amp A - 2, which have thanks to overall feedback distortion ten times lower, hold basic character of sound of this amp in every levels the same. Many people here and not only here, take distortion as necessary for " listening enjoy " and becouse any amps haven't the same rate of harmonics and the same dependability of this on output power, people evaluate some as "musical" and some as " less musical ". They don't know, that musical apparatus is only this one, which reproduce only music, not one , which reproduce music + " something". With amp which have ultralow distortion, you can hear " silence between notes " or " fart of mouse in studio ", but I'm sorry, without feedback is hard to get it .
Re: To Charles Hansen
Hello -
This is a difficult question to answer, because we cannot directly compare our experiences. However, I would like to point out a few things:
1) Some people prefer the sound of a feedback circuit. Some of this may be due to personal taste, but perhaps some is also because that is what people are used to. Remember that 99.9999% of all amplifiers in use today use high feedback. There is an old saying in this country, "People know what they like, and they like what they know." So I think there may also be a period of adjustment (perhaps some weeks or even months) to fully understand what the zero feedback circuit does for music.
2) The test that was performed did not compare a feedback circuit with a no feedback circuit. Instead it compared an amplifier with 3 high feedback modules connected in series to an amplifier with 2 high feedback modules in series (the last two modules were joined with a feedback loop to form one module). So I really don't think this was a test of "feedback vs. no feedback", but instead "one type of feedback vs. a similar type of feedback".
In my personal experience I have found that it is possible to achieve all of the sonic performance of a feedback design from a zero feedback design. But I have not experienced the converse. That is, there is always something special about the way that a zero feedback design conveys the meaning of the music that cannot be duplicated with a feedback design.
It is difficult to experience this for yourself, as there is so little equipment that is designed with zero feedback. When you have the opportunity to experience a playback system where the source, the preamp, and the power amp are all zero feedback, I think you will understand what I am saying. In the meantime, it is useful to build some zero feedback designs for yourself and try some different ideas.
In this thread we have been focusing on power amplifiers. Earlier I pointed out some threads in the Digital Forum that were focused on zero feedback D/A converters. Please refer to those threads and see how people respond to the sound of zero feedback in that application:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=16323
http://www.diyaudio.com/forums/showthread.php?s=&threadid=22206
Best regards,
Charles Hansen
Upupa Epops said:We are friends with Pavel Macura ( PMA ). His amp A - 1 ( see above ) have theoreticaly optimal " feedbackless " configuration ( three independent maximaly linearized stages ), what is searched on this thread several days. With amp which have ultralow distortion, you can hear " silence between notes " or " fart of mouse in studio ", but I'm sorry, without feedback is hard to get it
Hello -
This is a difficult question to answer, because we cannot directly compare our experiences. However, I would like to point out a few things:
1) Some people prefer the sound of a feedback circuit. Some of this may be due to personal taste, but perhaps some is also because that is what people are used to. Remember that 99.9999% of all amplifiers in use today use high feedback. There is an old saying in this country, "People know what they like, and they like what they know." So I think there may also be a period of adjustment (perhaps some weeks or even months) to fully understand what the zero feedback circuit does for music.
2) The test that was performed did not compare a feedback circuit with a no feedback circuit. Instead it compared an amplifier with 3 high feedback modules connected in series to an amplifier with 2 high feedback modules in series (the last two modules were joined with a feedback loop to form one module). So I really don't think this was a test of "feedback vs. no feedback", but instead "one type of feedback vs. a similar type of feedback".
In my personal experience I have found that it is possible to achieve all of the sonic performance of a feedback design from a zero feedback design. But I have not experienced the converse. That is, there is always something special about the way that a zero feedback design conveys the meaning of the music that cannot be duplicated with a feedback design.
It is difficult to experience this for yourself, as there is so little equipment that is designed with zero feedback. When you have the opportunity to experience a playback system where the source, the preamp, and the power amp are all zero feedback, I think you will understand what I am saying. In the meantime, it is useful to build some zero feedback designs for yourself and try some different ideas.
In this thread we have been focusing on power amplifiers. Earlier I pointed out some threads in the Digital Forum that were focused on zero feedback D/A converters. Please refer to those threads and see how people respond to the sound of zero feedback in that application:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=16323
http://www.diyaudio.com/forums/showthread.php?s=&threadid=22206
Best regards,
Charles Hansen
SY said:
Hello -
Well, it's really not too different than the positive tempco that exists in bipolar output stages. So you need to mount the bias device (in this case a Vgs multiplier) on the heatsink.
The problem you run into is the variability of the Vgs threshold points for the different devices, from sample-to-sample. With a bipolar, the variations from part-to-part are negligible. But with MOSFETs, the variation can easily be +/-20% (or more). The Vgs multiplier easily accomodates the different bias voltages required for any particular output device to be set to its desired quiescent current.
But the tricky part is that the bias circuit doesn't just multiply the Vgs, it also multiplies the *tempco* of the Vgs. So in theory you need to watch for this. How to do so is, as they say, an exercise left for the reader.
Best regards,
Charles Hansen
Re: Re: To Charles Hansen
I suppose that something might be probably understood from here ("between the rows info"): http://www.stereophile.com/amplificationreviews/635/index2.html
We often (Pavel Dudek and me) perform listenning tests with people familiar and fond of live presentation of symphonical classical music. The results are always similar - they (and we to) prefer the amps with the lowest possible distortion, and overall NFB. The best results were obtained with Pavel Dudek's dpa386 like amp (mosfet output stage + error correction) or my class A with NFB. We can compare with well known amps and have access to speakers like Wilson Maxx or B+W Matrix 801.
Charles Hansen said:
I suppose that something might be probably understood from here ("between the rows info"): http://www.stereophile.com/amplificationreviews/635/index2.html
We often (Pavel Dudek and me) perform listenning tests with people familiar and fond of live presentation of symphonical classical music. The results are always similar - they (and we to) prefer the amps with the lowest possible distortion, and overall NFB. The best results were obtained with Pavel Dudek's dpa386 like amp (mosfet output stage + error correction) or my class A with NFB. We can compare with well known amps and have access to speakers like Wilson Maxx or B+W Matrix 801.
PSU
To Mike : Certainly, PSU is very important part of amp, because all is connected in series : PSU, amp and load. If you have good amp, but PSU is wrong, you can't get good results. Often I see DIY, but often profi product too, which have this part build wrong. What I mean ? Mostly exist this mistake : PSU is too far from amp, wires are too thin and they all makes big loop. Often are DC rails which are leaded on PCB really thin - good example is original PCB for JLH amp. Long wire have big inductance, thin wire big DC resistance : both is not good for amp, both have negative influence on parameters of amp and for low feedback types much more. In my opinion is best solution of PSU to get it on the same PCB as amp ( AC part of PS - trafo - may be far ). For really good rails with low DC resistance you can't use PCB trace - try to count, which is cross - section of PCB trace with wide for example 5 mm / 35 or 75 microns ! In my amps I'm using as DC rails gold platted copper prism ( 10*20 mm for TO 220 or 10*30 mm for TO 264 ). This solution have one another anvantage : very good ( and quick ) temperature connection of practicaly all power devices, 'cos they are here without any insulation and copper is better conductor then aluminium ). For this solution ( PSU and amp on common PCB ), when ground traces are lead apart as star, is not problem with normal transistors ( and standard rectifiers ) to get SNR over 120 dB ( refered to 1 V on input ). If you can, look at my amp on pages of PMA. Now I'm going away ( till sunday ), but in future we can talk about more.
To Mike : Certainly, PSU is very important part of amp, because all is connected in series : PSU, amp and load. If you have good amp, but PSU is wrong, you can't get good results. Often I see DIY, but often profi product too, which have this part build wrong. What I mean ? Mostly exist this mistake : PSU is too far from amp, wires are too thin and they all makes big loop. Often are DC rails which are leaded on PCB really thin - good example is original PCB for JLH amp. Long wire have big inductance, thin wire big DC resistance : both is not good for amp, both have negative influence on parameters of amp and for low feedback types much more. In my opinion is best solution of PSU to get it on the same PCB as amp ( AC part of PS - trafo - may be far ). For really good rails with low DC resistance you can't use PCB trace - try to count, which is cross - section of PCB trace with wide for example 5 mm / 35 or 75 microns ! In my amps I'm using as DC rails gold platted copper prism ( 10*20 mm for TO 220 or 10*30 mm for TO 264 ). This solution have one another anvantage : very good ( and quick ) temperature connection of practicaly all power devices, 'cos they are here without any insulation and copper is better conductor then aluminium ). For this solution ( PSU and amp on common PCB ), when ground traces are lead apart as star, is not problem with normal transistors ( and standard rectifiers ) to get SNR over 120 dB ( refered to 1 V on input ). If you can, look at my amp on pages of PMA. Now I'm going away ( till sunday ), but in future we can talk about more.
Hi traderbaum,
you seemed to take exception to my points posted earlier, I'm not quite sure why. were they really that unusual ? perhaps a bit unclear - hopefully this will be clearer
I thought it was generally accepted that a good phase margin is bebeficial for a good sound - do you disagree ? JHL made the point in one of his articles that extra phase margin at the expence of some more HF distortion was a benificial trade off.
However If the amp if very fast then you can good phase margins and good HF distortion figures. i.e. compensation is operating way above audio band. Best of both worldls - why settle for less ? Also a fast amp i.e. short propigation time will be least affected by any problems apart from stability that might arise from the delay between i/p & o/p when using GFB
Now noise. I have found that when-ever I have taken measures to reduced PSU noise there have been great sonic rewards. I personally really like the sound of choke regulation.
To my ears any class AB amp where the o/p transistors are switching off and creating noise, sounds unatural.
memory distortion - i.e. change of gain with temperature with time delays - in a feedback cct will create all manner of different effects depending on the type of distortion that the feedback is trying to deal with - in a word Chaos. Any amp using feedback that does not address memory distortion is at best ( low noise class A ) going to sound slightly vague or muddled. If some switching noise and Xover distortion is thrown into the equation. God only knows what is going on. For me this has to be reduced to an absolute minimum.
short signal & feed backpath. I don't understand this one but if the likes of hugh deans & peter daniels recommend it I am happy to give it a try, after all nothing to loose.
I hope this is now less wishy washy.
I don't know where you are within UK but perhaps when I have finnished my next amp I will bring it over and see if you think I have succeeded.
regards
mike
you seemed to take exception to my points posted earlier, I'm not quite sure why. were they really that unusual ? perhaps a bit unclear - hopefully this will be clearer
I thought it was generally accepted that a good phase margin is bebeficial for a good sound - do you disagree ? JHL made the point in one of his articles that extra phase margin at the expence of some more HF distortion was a benificial trade off.
However If the amp if very fast then you can good phase margins and good HF distortion figures. i.e. compensation is operating way above audio band. Best of both worldls - why settle for less ? Also a fast amp i.e. short propigation time will be least affected by any problems apart from stability that might arise from the delay between i/p & o/p when using GFB
Now noise. I have found that when-ever I have taken measures to reduced PSU noise there have been great sonic rewards. I personally really like the sound of choke regulation.
To my ears any class AB amp where the o/p transistors are switching off and creating noise, sounds unatural.
memory distortion - i.e. change of gain with temperature with time delays - in a feedback cct will create all manner of different effects depending on the type of distortion that the feedback is trying to deal with - in a word Chaos. Any amp using feedback that does not address memory distortion is at best ( low noise class A ) going to sound slightly vague or muddled. If some switching noise and Xover distortion is thrown into the equation. God only knows what is going on. For me this has to be reduced to an absolute minimum.
short signal & feed backpath. I don't understand this one but if the likes of hugh deans & peter daniels recommend it I am happy to give it a try, after all nothing to loose.
I hope this is now less wishy washy.
I don't know where you are within UK but perhaps when I have finnished my next amp I will bring it over and see if you think I have succeeded.
regards
mike
Re: Re: Re: To Charles Hansen
Hello -
This review was brought up in another thread. Please refer to my reply there:
http://www.diyaudio.com/forums/showthread.php?s=&postid=336175#post336175
Best regards,
Charles Hansen
PMA said:
Hello -
This review was brought up in another thread. Please refer to my reply there:
http://www.diyaudio.com/forums/showthread.php?s=&postid=336175#post336175
Best regards,
Charles Hansen
PMA, Thanks for the schematics. You have chosen to either place the output stage inside or outside the loop of the second op-amp. The way you've implemented the roll-off is particularly good - a byproduct of using an op-amp. I assume the op-amp OL response is a typical differentiator - what is the OL bandwidth of the op-amp? I can believe the feedback improves the sound. Of course this rather depends upon how you've selected the output devices.
Mike,
I'm taking exception in order to encourage you to think for yourself and not fall into the pervasive trap of trusting the publicists. In my experience if you are ambitious about designing something really good you must question every assertion until you get to some common-sense reasoning to support it. If you don't then you can't trust the assertion, regardless of how many articles the person saying it has written.
I'm taking exception in order to encourage you to think for yourself and not fall into the pervasive trap of trusting the publicists. In my experience if you are ambitious about designing something really good you must question every assertion until you get to some common-sense reasoning to support it. If you don't then you can't trust the assertion, regardless of how many articles the person saying it has written.
I do disagree. Why should phase margin matter, unless it causes gross frequency response changes within the audio band.
I don't agree. The faster the circuit is the higher the energy of frequencies beyond audibility. What benefit is this? One benefit of high bandwidth is to allow greater feedback to correct distortion. More often than not the higher bandwidth leads to more distortion.
I agree with your liking for choke regulation - it produces less HF noise. Injecting HF noise into a wide-band non-linear amp can produce audible artifacts.
You are assuming it is the "switching" that causes noise and that this noise makes an AB amp sound unatural. Are you sure this is why an AB amp sounds unatural?
All amps contain hysterisis effects, I agree. Charles' transconductance instability issue is an example. Often these affects are at a relatively low frequency where feedback can dispatch them.
Bristol. Please do.
traderbam said:
Hi
I was thinking of the temperature induced non linearities actually in the input / feedback transistor (s).
If these had been dealt with then I agree, the feedback may well do a pretty good job of sorting out the problems elsewhere, but the feedback cct will not be able to "see" it's own internal non- linearities so I believe that these have to be reduced as much as is possible to make this a viable approach.
On the AB switching thing...
There may be some class AB amps ( that switch ) out there that I would like, but to date, I have not heard one that sounds as natural as a good class A design. Based on this my current view is that switching transistors don't sound good. I would be happy to be prooved wrong ( mostly for the sake of the environment and my electric bill.....
)thanks for your input and the invite
mike
Ayre V5
To Charles Hansen:
There has been a "High End 2004" Exhibition in Prague now and I have had a chance to listen to AYRE V5 + Avalon speakers for quite a long time. Yes, this is a remarkable amp. Especially for some kind of music, like jazz bass, acoustic guitar etc. I have also tried Hector Berlioz's "Symphonie Phantastique", Telarc, Baltimore Orchestra directed by David Zinman. For this kind of music, I have preferred the sound of Classé CAM350 + Wilson Sophia. Especially during steep dynamic changes like in "March to the Scaffold" part, Ayre looses integrity and purity.
Best regards,
Pavel Macura
To Charles Hansen:
There has been a "High End 2004" Exhibition in Prague now and I have had a chance to listen to AYRE V5 + Avalon speakers for quite a long time. Yes, this is a remarkable amp. Especially for some kind of music, like jazz bass, acoustic guitar etc. I have also tried Hector Berlioz's "Symphonie Phantastique", Telarc, Baltimore Orchestra directed by David Zinman. For this kind of music, I have preferred the sound of Classé CAM350 + Wilson Sophia. Especially during steep dynamic changes like in "March to the Scaffold" part, Ayre looses integrity and purity.
Best regards,
Pavel Macura
Hello Pavel,
I have no reason to doubt your observations. Based on these observations, I can think of at least three different possibilities that would explain them:
1) A zero-feedback amplifier will never sound as "dynamic" as a conventional amplifier with feedback.
2) A zero-feedback amplifier is capable of sounding as "dynamic" as a conventional feedback amplifier, but in this specific instance it did not. (Keep in mind that you were comparing monoblock feedback amps with over twice the rated power, nearly twice the price, and roughly three times the weight as the Ayre V-5x.)
3) Wilson Sophias are more "dynamic" than an unspecified Avalon model.
It seems to me that it would be difficult to draw a conclusion from this particular listening experience.
Best regards,
Charles Hansen
I have no reason to doubt your observations. Based on these observations, I can think of at least three different possibilities that would explain them:
1) A zero-feedback amplifier will never sound as "dynamic" as a conventional amplifier with feedback.
2) A zero-feedback amplifier is capable of sounding as "dynamic" as a conventional feedback amplifier, but in this specific instance it did not. (Keep in mind that you were comparing monoblock feedback amps with over twice the rated power, nearly twice the price, and roughly three times the weight as the Ayre V-5x.)
3) Wilson Sophias are more "dynamic" than an unspecified Avalon model.
It seems to me that it would be difficult to draw a conclusion from this particular listening experience.
Best regards,
Charles Hansen
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.