Yesterday I tested with the following setup:
Source: NAD CDP541i
Preamp: Nil
Amplifiers: Rotel RB991, AKSA clone, Anthony Holton's N-Channel, same N-channel with cascode front end & Stochino Amp.
Speakers: Klipsch RF3
Given that Interconnects and Speaker Cables were not swapped, yes, each was a different performer with a different presentation of the music and sound stage; while the Stereophile recommended Class A Rotel was clearly the last in the race, the AKSA and Stochino were neck to neck in almost every aspect - ie., timbral accuracy, low level resolution, dynamics etc., even the width and depth of the soundstage were almost as good as the other; however, the AKSA clone had a very forward soundstage (and hence, sounded more robust and brawny) while the Stochino had a more laid back sound stage and hence, sounded finer and more presentable.
I remember Grey saying something about the "forwardness" of the sound stage with the Alephs at some point.
Any comments or observations?
Source: NAD CDP541i
Preamp: Nil
Amplifiers: Rotel RB991, AKSA clone, Anthony Holton's N-Channel, same N-channel with cascode front end & Stochino Amp.
Speakers: Klipsch RF3
Given that Interconnects and Speaker Cables were not swapped, yes, each was a different performer with a different presentation of the music and sound stage; while the Stereophile recommended Class A Rotel was clearly the last in the race, the AKSA and Stochino were neck to neck in almost every aspect - ie., timbral accuracy, low level resolution, dynamics etc., even the width and depth of the soundstage were almost as good as the other; however, the AKSA clone had a very forward soundstage (and hence, sounded more robust and brawny) while the Stochino had a more laid back sound stage and hence, sounded finer and more presentable.
I remember Grey saying something about the "forwardness" of the sound stage with the Alephs at some point.
Any comments or observations?
Would someone want to comment about soundstage if finally, with all of its ramifications, that is what Hi-End Audio is all about?we should give up stereo for the potentially wider and more stable sound stage of 5.1?
- not that 5.1 is really designed to improve this audiophile illusion, nor are quality speakers commonly used esp. center, and the mixing is usually for exagerated effects rather than improved imagng
- not that 5.1 is really designed to improve this audiophile illusion, nor are quality speakers commonly used esp. center, and the mixing is usually for exagerated effects rather than improved imagng
Samuel,
While your comments on the AKSA are generally flattering, you have either a clone - not sold by me, possibly copied from my intellectual property - or a very early pcb from ESP. With some reservations, I disown this clone because the component choice is necessarily different. Component choice (and layout) makes a marked difference to this design. If it's a copy, did you (or someone else?) copy the Aspen artwork; if so may I ask from whom, and are there other examples of my amplifier 'out there'?
I make my living from this amplifier, various other products, and associated service and my company has never released circuit diagrams, artwork or documentation into the wild. Only paying customers have this information, which is all copyrighted. I am deeply concerned when I hear of clones out there in the 'wild'. Despite its apparent simplicity it took a long time and considerable investment to develop this amplifier, and I would appreciate it if you did not identify it as an AKSA because it is not the genuine product and does not perform the same way.
Of course, it's possible you bought one of the early pcbs from Rod Elliott at ESP. If so, no problem at all; please ignore the above. Otherwise I would appreciate a PM on this if you have the time.......
Cheers,
Hugh
While your comments on the AKSA are generally flattering, you have either a clone - not sold by me, possibly copied from my intellectual property - or a very early pcb from ESP. With some reservations, I disown this clone because the component choice is necessarily different. Component choice (and layout) makes a marked difference to this design. If it's a copy, did you (or someone else?) copy the Aspen artwork; if so may I ask from whom, and are there other examples of my amplifier 'out there'?
I make my living from this amplifier, various other products, and associated service and my company has never released circuit diagrams, artwork or documentation into the wild. Only paying customers have this information, which is all copyrighted. I am deeply concerned when I hear of clones out there in the 'wild'. Despite its apparent simplicity it took a long time and considerable investment to develop this amplifier, and I would appreciate it if you did not identify it as an AKSA because it is not the genuine product and does not perform the same way.
Of course, it's possible you bought one of the early pcbs from Rod Elliott at ESP. If so, no problem at all; please ignore the above. Otherwise I would appreciate a PM on this if you have the time.......
Cheers,
Hugh
Well Hugh, when I think about Pass clones, I try to imagine
myself as Bill Gates, still worried that somebody's running
a pirate copy of Dos 3.0, and it always makes me feel better.
myself as Bill Gates, still worried that somebody's running
a pirate copy of Dos 3.0, and it always makes me feel better.
Nelson,
The situation is very different for you, Sir. You are established, published, big marketing setup, you even hold patents.
I see the humor in all this too, but there is a dark side I cannot ignore.
Cheers,
Hugh
The situation is very different for you, Sir. You are established, published, big marketing setup, you even hold patents.
I see the humor in all this too, but there is a dark side I cannot ignore.
Cheers,
Hugh
I, being all too human, have strong feelings about Bill Gates...and they aren't positive.
Okay, lemme see. Soundstage. The comments that Samuel Jayaraj is referring to were in comparison to my tube amps. The Aleph 2s that I built (note the DIY aspect, these are not to be confused with Nelson's commercial product) imaged roughly a foot further forward than my tube amps, perhaps two feet. Meaning that if you were to close your eyes and imagine throwing a wad of paper at a given musician, you would tend to aim a foot or two closer. I traded a few e-mails with Nelson over this at the time and my recollection is that his conclusion was 'it's a mystery.'
My working assumption is that image is rooted primarily in two things: detail retrieval and phase relationships. Since, to my knowledge, no one has ever come up with a way to measure image width or depth, we have to gauge such things subjectively. There are a few truisms that tend to hold, although there are enough counterexamples that you can't state them as absolute rules. One is that tube amps tend to image better than solid state. Why? Who knows? A rabid solid state fan will tell you that tubes image the way they do because of distortion products that create a false sense of spaciousness. My feeling is that such reasoning is spurious at best and may boil down to a simple case of sour grapes. It's curious that these so-called distortion products always sound more like the way real music sounds in a real hall. That said, I've heard tubes that didn't image very well, and solid state that did much better than expected.
The Aleph design sounds very tube-ish for a solid state design. It doesn't sound "like tubes" but it leans strongly in that direction. There are going to be at least two known things involved here. One is that it's a simple design, with only two stages. The other is that it tends towards second harmonic. Comparatively low feedback will probably figure in there as well, although the Alephs still use more feedback than most tube equipment; 20dB NFB (if I recall) vs. about 10-15dB for a lot of the better tube lines.
Parenthetically, note that tube equipment gets pretty good specs for using such low feeback. For some reason, the myth persists that tube stuff has several percent distortion. The truth is that most of the good stuff is well under 1%, and achieves that with very modest feedback. You can get really, really low distortion specs out of tube gear if you're willing to use feedback ratios similar to what solid state uses. Be prepared to build in another gain stage or three in order to build up the open loop gain.
So, does NFB have anything to do with image? It's not an open and shut case. Tube stuff is marvellously forgiving. You can do things with tubes that causes solid state people to cringe, like varying feedback ratios on the fly. Yes, the gain changes, although you can jigger things to offset that if it bothers you. At least between about 6dB and 20dB NFB, there is not an enormous, orders-of-magnitude difference in the image. It changes some, but not as much as you might expect. (It does get brighter tonally, but that's another matter entirely.)
Miscellaneous factoids:
Film caps can improve image for either solid state or tubes, but the basic circuit design has to be capable of decent imaging first.
Fewer gain stages tend to improve imaging.
Lower feedback yields a better image.
Low distortion and imaging are not directly related. This one gets complicated, in that distortion is inescapably related to negative feedback, which does influence imaging. Just don't fall into the trap of assuming that low disortion necessarily means that a purer signal is getting through and hence the image must improve. Ain't so. It's a balancing act. At some point you realize that this falls into the category of voicing an amp and you make compromises.
Grey
Okay, lemme see. Soundstage. The comments that Samuel Jayaraj is referring to were in comparison to my tube amps. The Aleph 2s that I built (note the DIY aspect, these are not to be confused with Nelson's commercial product) imaged roughly a foot further forward than my tube amps, perhaps two feet. Meaning that if you were to close your eyes and imagine throwing a wad of paper at a given musician, you would tend to aim a foot or two closer. I traded a few e-mails with Nelson over this at the time and my recollection is that his conclusion was 'it's a mystery.'
My working assumption is that image is rooted primarily in two things: detail retrieval and phase relationships. Since, to my knowledge, no one has ever come up with a way to measure image width or depth, we have to gauge such things subjectively. There are a few truisms that tend to hold, although there are enough counterexamples that you can't state them as absolute rules. One is that tube amps tend to image better than solid state. Why? Who knows? A rabid solid state fan will tell you that tubes image the way they do because of distortion products that create a false sense of spaciousness. My feeling is that such reasoning is spurious at best and may boil down to a simple case of sour grapes. It's curious that these so-called distortion products always sound more like the way real music sounds in a real hall. That said, I've heard tubes that didn't image very well, and solid state that did much better than expected.
The Aleph design sounds very tube-ish for a solid state design. It doesn't sound "like tubes" but it leans strongly in that direction. There are going to be at least two known things involved here. One is that it's a simple design, with only two stages. The other is that it tends towards second harmonic. Comparatively low feedback will probably figure in there as well, although the Alephs still use more feedback than most tube equipment; 20dB NFB (if I recall) vs. about 10-15dB for a lot of the better tube lines.
Parenthetically, note that tube equipment gets pretty good specs for using such low feeback. For some reason, the myth persists that tube stuff has several percent distortion. The truth is that most of the good stuff is well under 1%, and achieves that with very modest feedback. You can get really, really low distortion specs out of tube gear if you're willing to use feedback ratios similar to what solid state uses. Be prepared to build in another gain stage or three in order to build up the open loop gain.
So, does NFB have anything to do with image? It's not an open and shut case. Tube stuff is marvellously forgiving. You can do things with tubes that causes solid state people to cringe, like varying feedback ratios on the fly. Yes, the gain changes, although you can jigger things to offset that if it bothers you. At least between about 6dB and 20dB NFB, there is not an enormous, orders-of-magnitude difference in the image. It changes some, but not as much as you might expect. (It does get brighter tonally, but that's another matter entirely.)
Miscellaneous factoids:
Film caps can improve image for either solid state or tubes, but the basic circuit design has to be capable of decent imaging first.
Fewer gain stages tend to improve imaging.
Lower feedback yields a better image.
Low distortion and imaging are not directly related. This one gets complicated, in that distortion is inescapably related to negative feedback, which does influence imaging. Just don't fall into the trap of assuming that low disortion necessarily means that a purer signal is getting through and hence the image must improve. Ain't so. It's a balancing act. At some point you realize that this falls into the category of voicing an amp and you make compromises.
Grey
Hugh,
I see both sides, yours and Nelson's. I have not heard one of your amps. I will not build one even if I happen to come across a known "good" schematic. (I seem to recall that there are various incorrect or obsolete versions floating around on the web.)
Folks, if you want an AKSA, go to Hugh. Please.
Grey
I see both sides, yours and Nelson's. I have not heard one of your amps. I will not build one even if I happen to come across a known "good" schematic. (I seem to recall that there are various incorrect or obsolete versions floating around on the web.)
Folks, if you want an AKSA, go to Hugh. Please.
Grey
My previous post isn't quite as clear as I'd like.
Although I haven't heard one, I have heard enough people say that the AKSA is a good amplifier that I believe it to be the case. My saying that I wouldn't build one should be understood to mean that I would not build an unauthorized copy.
Hugh is one of the good guys. Please support him.
Grey
Although I haven't heard one, I have heard enough people say that the AKSA is a good amplifier that I believe it to be the case. My saying that I wouldn't build one should be understood to mean that I would not build an unauthorized copy.
Hugh is one of the good guys. Please support him.
Grey
Hugh, thanks for your comments and expressed concerns. My apoligies for referring to the "AKSA clone". This was only to let the forum members know, well, exactly what it is.
This is certainly one of your earlier versions of the AKSA as published by Rod Elliot before it was removed from his site. The thermal tracking is via a pair of diodes rather than the modern version that has a Vbe multiplier. If one were to take that erstwhile schematic and juxtapose it with the one published by Carlos (destroyerx) with your permission, a very close copy of the present AKSA would emerge; yes, only a copy but I am not interested to make gains out of that. As you rightly point out the track layout and component choice do make a difference and in my case, I have chosen both going by what I felt would be best, just to make a comparison.
Given the above, the current version of the AKSA must be fabulous sounding, since what I heard out of 'my copy' was itself the best in terms of both timbral accuracy and capturing the stage nuances. I was only trying to figure out the slight forwardness of the whole sound stage. Thanks again.
Nelson, thank you very much for your comments. You are indeed a happy man; you have a very high degree of security feelings and one of the great reasons for your happy state of affairs is yours blessing the diy community out of your abundance, as much as your commercial success is. You are highly regarded and appreciated.
Grey, thanks for your insightful comments on the relationship between NFB and soundstage. That does throw a lot of light and should provide designers with some guidance while taking into consideration other aspects of circuitry.
Would you say that there is any such thing as a "correctness" to the soundstage in as far as the forward and relatively laid back aspects are concerned?
This is certainly one of your earlier versions of the AKSA as published by Rod Elliot before it was removed from his site. The thermal tracking is via a pair of diodes rather than the modern version that has a Vbe multiplier. If one were to take that erstwhile schematic and juxtapose it with the one published by Carlos (destroyerx) with your permission, a very close copy of the present AKSA would emerge; yes, only a copy but I am not interested to make gains out of that. As you rightly point out the track layout and component choice do make a difference and in my case, I have chosen both going by what I felt would be best, just to make a comparison.
Given the above, the current version of the AKSA must be fabulous sounding, since what I heard out of 'my copy' was itself the best in terms of both timbral accuracy and capturing the stage nuances. I was only trying to figure out the slight forwardness of the whole sound stage. Thanks again.
Nelson, thank you very much for your comments. You are indeed a happy man; you have a very high degree of security feelings and one of the great reasons for your happy state of affairs is yours blessing the diy community out of your abundance, as much as your commercial success is. You are highly regarded and appreciated.
Grey, thanks for your insightful comments on the relationship between NFB and soundstage. That does throw a lot of light and should provide designers with some guidance while taking into consideration other aspects of circuitry.
Would you say that there is any such thing as a "correctness" to the soundstage in as far as the forward and relatively laid back aspects are concerned?
Samuel,
Many thanks for your balanced, friendly and thoughtful response. I retract all my earlier comments about clones; this circuit was indeed published early on by Rod until we agreed it should come down, and indeed there are a couple of details missing from that circuit. It has now evolved to the Nirvana Plus version with some very fancy components and performance is somewhat enhanced over the original.
In fact, George O., also a native of Bangalore as you may know, has an AKSA, though not a N+. I am talking with Sridhar Reddy at present (who is also from Bangalore), and we might well do something into the future when he returns to his country from the States. Recently at a Trade Fair here in Melbourne I met some PCB manufacturers from Bangalore, so I'm slowly getting familiar with your industrial environment and frankly, it is impressive. Of all the Asian countries I would prefer to deal with India because of the language issues and expertise I have found; so watch this space......
The issues of sound stage you mention are vexing, difficult problems. It is still difficult to identify what parts in the schematic are responsible for sound stage integrity, however, I've found lag compensation and phase lead to be the big ones, as imaging is related to the phase integrity of the signal as it passes through the amplifier. Since these directly impact on amplifier stability, you want the minimum compensation/lead to achieve proper operation into real world loads whilst ensuring that imaging is preserved. The phase margin of the amplifier is VERY important to this pursuit, and particularly tricky with a global feedback design. I have found imaging performance to be the toughest aspect of amplifier design.
Cheers,
Hugh
Many thanks for your balanced, friendly and thoughtful response. I retract all my earlier comments about clones; this circuit was indeed published early on by Rod until we agreed it should come down, and indeed there are a couple of details missing from that circuit. It has now evolved to the Nirvana Plus version with some very fancy components and performance is somewhat enhanced over the original.
In fact, George O., also a native of Bangalore as you may know, has an AKSA, though not a N+. I am talking with Sridhar Reddy at present (who is also from Bangalore), and we might well do something into the future when he returns to his country from the States. Recently at a Trade Fair here in Melbourne I met some PCB manufacturers from Bangalore, so I'm slowly getting familiar with your industrial environment and frankly, it is impressive. Of all the Asian countries I would prefer to deal with India because of the language issues and expertise I have found; so watch this space......
The issues of sound stage you mention are vexing, difficult problems. It is still difficult to identify what parts in the schematic are responsible for sound stage integrity, however, I've found lag compensation and phase lead to be the big ones, as imaging is related to the phase integrity of the signal as it passes through the amplifier. Since these directly impact on amplifier stability, you want the minimum compensation/lead to achieve proper operation into real world loads whilst ensuring that imaging is preserved. The phase margin of the amplifier is VERY important to this pursuit, and particularly tricky with a global feedback design. I have found imaging performance to be the toughest aspect of amplifier design.
Cheers,
Hugh
Hugh,
Thank you immensely for your kind words. Thanks also for sharing on your findings with regard to Soundstage related to phase margin in amplifiers.
Nelson, Grey since both of you might have the most experience with Alephs and Aleph-Xs, how do you see each of these presenting the soundstage since both have typically common characteristics except for the X with all of its ramifications in the latter. This might help us further know if there are other criteria that significantly alter soundstaging.
Thanks.
Thank you immensely for your kind words. Thanks also for sharing on your findings with regard to Soundstage related to phase margin in amplifiers.
Nelson, Grey since both of you might have the most experience with Alephs and Aleph-Xs, how do you see each of these presenting the soundstage since both have typically common characteristics except for the X with all of its ramifications in the latter. This might help us further know if there are other criteria that significantly alter soundstaging.
Thanks.
Today, I was glancing through the white papers on the XA released by Pass Labs and found that it is already documented that the XAs (in our context the Aleph-X) produces a soundstage far superior to both the Alephs and well as the X amps. So I guess that is the answer as far as this comparison goes.
Would Geoff, Graham Maynard or anyone else want to comment on the soundstage of the JLH or modified versions. This is also a Class A low feedback design with great resolution being reported. How about the soundstage?
Hugh, would you like to add your worthy comments with regard to your Glass Harmony and soundstage?
This discussion might give everyone insight into the real critical parameters in amplifier design and show that conventional measurements of THD, IMD, DF, Slew Rate among others don't convey everything there is to great audio.
Would Geoff, Graham Maynard or anyone else want to comment on the soundstage of the JLH or modified versions. This is also a Class A low feedback design with great resolution being reported. How about the soundstage?
Hugh, would you like to add your worthy comments with regard to your Glass Harmony and soundstage?
This discussion might give everyone insight into the real critical parameters in amplifier design and show that conventional measurements of THD, IMD, DF, Slew Rate among others don't convey everything there is to great audio.
Hi, Samuel Jayaraj,
This is the answer to your question
Too long? This is the summary :
I've got difficult times with friends who talks about this. They can hear (with eyes closed) that this amp has wide stage, this amp has "infront of face" stage, that amp has "deep" sound stage, etc. Until now I cannot make an amp that has the "desired" sound stage. Is there anything can be "methodically done" to make the stage is like 5m deep or 1m deep?
Once I comparing CD players. Nakamichi, Denon, etc. The one that has "most deep" sound stage is Denon. When I look at the block diagram, it has a "Denon Processor" in its digital cct.
It seems even company like Denon could not make this "deep" soundstage analogly, they have to lean on DSP.
If anyone has other opinion besides Grey's, it would be very nice. How to intentionally create depth of soundstage with analog power amp.
This is the answer to your question
Too long? This is the summary :
I've got difficult times with friends who talks about this. They can hear (with eyes closed) that this amp has wide stage, this amp has "infront of face" stage, that amp has "deep" sound stage, etc. Until now I cannot make an amp that has the "desired" sound stage. Is there anything can be "methodically done" to make the stage is like 5m deep or 1m deep?
Once I comparing CD players. Nakamichi, Denon, etc. The one that has "most deep" sound stage is Denon. When I look at the block diagram, it has a "Denon Processor" in its digital cct.
It seems even company like Denon could not make this "deep" soundstage analogly, they have to lean on DSP.
If anyone has other opinion besides Grey's, it would be very nice. How to intentionally create depth of soundstage with analog power amp.
Yes, I suspect that 'It's a mystery' is on the money......😕
However, Grey makes very sage comment, and I have difficulty disagreeing with any of it. Since this is such an abstract topic, I will to make it more accessible by using bullet points.
1. Imaging performance is related to phase shift as the signal passes through the amplifier. In light of the phase reversals in the (usually two with SS, three or four with tubes) common emitter (cathode) stages, and the well known propagation delays, effects on imaging are inevitable.
2. The problem seems to be that a constant group delay through the amplifier represents very different phase shifts across three decades of audio frequencies. 10 degree at 20KHz is 1 degree at 2KHz and 0.1 degrees at 20Hz. My anecdotal understanding is that the imaging information is contained in the 2-6 KHz region, though Blumlein was able to show in the thirties that all spatial information in a stereo recording was amplitude, not phase, related. The result is that different frequencies experience different phase shifts, meaning that the waveform interrelationships are not faithfully preserved. Just how this affects imaging, at least to me, remains a mystery.
3. In a practical Bailey/Lin amplifier, Bode/Nyquist constraints impose the need for single pole dominant lag compensation, normally a cap at the primary voltage amplifier between base and collector. This pulls back open loop gain below unity by the pole frequency, where phase reversal would otherwise turn negative into positive feedback. While this maintains stability and avoids amplifier self-immolation, the fact is the pole frequency is also dependent on the load - hence the need with many amps to match speakers - and there is no doubt that more compensation badly affects vitality, lifeforce, crispness, imaging, whatever the heck you want to call it. This fact is pivotal to sonics.
4. The vast majority of speakers are designed for use with voltage sources, and such electrodynamic speakers must have a low impedance source. NP has done a lot of work on this, and has articles on his website discussing source impedance at length, but to make an amp work well with the usual run of speakers we need a Zout of 300 milliohms or less. While this can sometimes be done with local feedback on the output stage, this is unusual, and most SS amps use global negative feedback to achieve this low source impedance. (For an example of a low Zout output stage buffer, see Pavel Macura's interesting 'Error Correction' regime, which in fact is a negative feedback system operating solely within the output stage). A low source impedance with strong current delivery yields sonic impact and slam, which are essential to the listening experience for rock music, and a few other genres as well, so this is very important at the consumer level. You will notice that most tube amps, particularly SETs, suffer a little in this department, and while their sonics are often outstanding, the rock head-bangers amongst us find the tube amps a little, ahem, unsatisfying......
5. Slam and impact requires, generally speaking, global feedback which necessarily encloses the voltage amp and the output stage. Usually this loop also includes the input stage, since the LTP is almost universal and a most elegant solution to the offset problem. The LTP also permits direct coupling with high reliability to the speaker. With at least three separate stages, and four if you use drivers for the output devices, a global feedback loop of this size is going to be tricky, because the signal path is long, and the load will have profound influence on the dominant pole, leading to possible instability. Yet more compensation will demand more and more miller cap charge current, which in extreme cases will immobilize the negative feedback loop and leave the output uncontrolled. This is a particularly malevolent compromise, damned if you do, damned if you don't.
6. There are only two CRO measurements which I have found useful in predicting the 'sonics' of an amplifier. They are square wave into a real world load at full power and 5KHz, and X-Y input/output plot of phase at 20Khz into a real world load at 1 watt and again at full power. This second is the more informative; you are looking for a straight line across the entire swept audio band, NOT an ellipse. Of course, I'm assuming the usual frequency response plots already achieve the required specification.
7. Any circuit manipulations which can reduce lag compensation usually result in better sound. I particularly like the bootstrap as it can be set up to roll off naturally at around 300KHz, pulling back VAS gain at HF without much affecting sonics.
8. CFPs (also known as Sziklai pairs, after George Sziklai, the Hungarian-American who at RCA in the fifties discovered it) look great, but suffer transient instability and are extremely tetchy to bias up. You can make them behave, but you need to slap them in capacitive manacles and this really damages the music.
9. The good ol' double emitter follower is my chosen output stage. But you must use stopper resistors as they too are inclined towards transient instability, particularly with reactive loads.
10. A combination of VAS and output stage voltage feedback can reduce load dependence and enhance imaging very considerably, but at the expense of slam and impact. You can't have it all - where would you put it??
11. My Glass Harmony design uses a tube front end for voltage amplification, and a single ended, trick unity gain output buffer for current amplification. There is no lag compensation - it's usually not needed for open loop tube designs - and this I believe is the reason for its exceptional imaging.
A great deal of energy is expended discussing clever topologies. I've tried quite a few of them, but in almost all cases simplicity is better. The real answer seems to me to lie in operating points and component selection. There is no easy answer. The fact that the results are not always explainable in terms to gladden the hearts of the math and measure brigade means audio design is probably as much art as science.
When you add to this potent mixture the fevervish marketing departments of yore you have the potential for cubic miles of !@#$ fed daily to unsuspecting audiophiles the world over........
Cheers,
Hugh
However, Grey makes very sage comment, and I have difficulty disagreeing with any of it. Since this is such an abstract topic, I will to make it more accessible by using bullet points.
1. Imaging performance is related to phase shift as the signal passes through the amplifier. In light of the phase reversals in the (usually two with SS, three or four with tubes) common emitter (cathode) stages, and the well known propagation delays, effects on imaging are inevitable.
2. The problem seems to be that a constant group delay through the amplifier represents very different phase shifts across three decades of audio frequencies. 10 degree at 20KHz is 1 degree at 2KHz and 0.1 degrees at 20Hz. My anecdotal understanding is that the imaging information is contained in the 2-6 KHz region, though Blumlein was able to show in the thirties that all spatial information in a stereo recording was amplitude, not phase, related. The result is that different frequencies experience different phase shifts, meaning that the waveform interrelationships are not faithfully preserved. Just how this affects imaging, at least to me, remains a mystery.
3. In a practical Bailey/Lin amplifier, Bode/Nyquist constraints impose the need for single pole dominant lag compensation, normally a cap at the primary voltage amplifier between base and collector. This pulls back open loop gain below unity by the pole frequency, where phase reversal would otherwise turn negative into positive feedback. While this maintains stability and avoids amplifier self-immolation, the fact is the pole frequency is also dependent on the load - hence the need with many amps to match speakers - and there is no doubt that more compensation badly affects vitality, lifeforce, crispness, imaging, whatever the heck you want to call it. This fact is pivotal to sonics.
4. The vast majority of speakers are designed for use with voltage sources, and such electrodynamic speakers must have a low impedance source. NP has done a lot of work on this, and has articles on his website discussing source impedance at length, but to make an amp work well with the usual run of speakers we need a Zout of 300 milliohms or less. While this can sometimes be done with local feedback on the output stage, this is unusual, and most SS amps use global negative feedback to achieve this low source impedance. (For an example of a low Zout output stage buffer, see Pavel Macura's interesting 'Error Correction' regime, which in fact is a negative feedback system operating solely within the output stage). A low source impedance with strong current delivery yields sonic impact and slam, which are essential to the listening experience for rock music, and a few other genres as well, so this is very important at the consumer level. You will notice that most tube amps, particularly SETs, suffer a little in this department, and while their sonics are often outstanding, the rock head-bangers amongst us find the tube amps a little, ahem, unsatisfying......
5. Slam and impact requires, generally speaking, global feedback which necessarily encloses the voltage amp and the output stage. Usually this loop also includes the input stage, since the LTP is almost universal and a most elegant solution to the offset problem. The LTP also permits direct coupling with high reliability to the speaker. With at least three separate stages, and four if you use drivers for the output devices, a global feedback loop of this size is going to be tricky, because the signal path is long, and the load will have profound influence on the dominant pole, leading to possible instability. Yet more compensation will demand more and more miller cap charge current, which in extreme cases will immobilize the negative feedback loop and leave the output uncontrolled. This is a particularly malevolent compromise, damned if you do, damned if you don't.
6. There are only two CRO measurements which I have found useful in predicting the 'sonics' of an amplifier. They are square wave into a real world load at full power and 5KHz, and X-Y input/output plot of phase at 20Khz into a real world load at 1 watt and again at full power. This second is the more informative; you are looking for a straight line across the entire swept audio band, NOT an ellipse. Of course, I'm assuming the usual frequency response plots already achieve the required specification.
7. Any circuit manipulations which can reduce lag compensation usually result in better sound. I particularly like the bootstrap as it can be set up to roll off naturally at around 300KHz, pulling back VAS gain at HF without much affecting sonics.
8. CFPs (also known as Sziklai pairs, after George Sziklai, the Hungarian-American who at RCA in the fifties discovered it) look great, but suffer transient instability and are extremely tetchy to bias up. You can make them behave, but you need to slap them in capacitive manacles and this really damages the music.
9. The good ol' double emitter follower is my chosen output stage. But you must use stopper resistors as they too are inclined towards transient instability, particularly with reactive loads.
10. A combination of VAS and output stage voltage feedback can reduce load dependence and enhance imaging very considerably, but at the expense of slam and impact. You can't have it all - where would you put it??
11. My Glass Harmony design uses a tube front end for voltage amplification, and a single ended, trick unity gain output buffer for current amplification. There is no lag compensation - it's usually not needed for open loop tube designs - and this I believe is the reason for its exceptional imaging.
A great deal of energy is expended discussing clever topologies. I've tried quite a few of them, but in almost all cases simplicity is better. The real answer seems to me to lie in operating points and component selection. There is no easy answer. The fact that the results are not always explainable in terms to gladden the hearts of the math and measure brigade means audio design is probably as much art as science.
When you add to this potent mixture the fevervish marketing departments of yore you have the potential for cubic miles of !@#$ fed daily to unsuspecting audiophiles the world over........Cheers,
Hugh
Hi, AKSA,
Am I reading correctly at your comment? Those informations are "secrets" to audio designers, yet you are willing to open them up 😀
Do you suggest more "inverting stages" in your point #1? Grey suggest minimal stages, I found this rather contradictive. Or are you saying that with more stages (any stages) with feedback is not good, but maybe we can use many "inverting stages" but without global feedback.
I've got another possibility, like you said in last part, it could be the factor of feedback. Less or no global feedback could result in better imaging. But I hear SS without feedback and tubes without feedback never sound the same. Why is this?
But I don't know how this fits in EC configuration. EC is a kind of local feedback, so will it be better or worse for imaging/staging? Sure better for the "Slam" or rock music, how about imaging/staging?
Could the bias current play a role here? I found that with less/minimal bias at all stages (but sufficient) the amp becomes more musical. More bias, like to differential or VAS, gives more fatique sound.
Am I reading correctly at your comment? Those informations are "secrets" to audio designers, yet you are willing to open them up 😀
Do you suggest more "inverting stages" in your point #1? Grey suggest minimal stages, I found this rather contradictive. Or are you saying that with more stages (any stages) with feedback is not good, but maybe we can use many "inverting stages" but without global feedback.
I've got another possibility, like you said in last part, it could be the factor of feedback. Less or no global feedback could result in better imaging. But I hear SS without feedback and tubes without feedback never sound the same. Why is this?
But I don't know how this fits in EC configuration. EC is a kind of local feedback, so will it be better or worse for imaging/staging? Sure better for the "Slam" or rock music, how about imaging/staging?
Could the bias current play a role here? I found that with less/minimal bias at all stages (but sufficient) the amp becomes more musical. More bias, like to differential or VAS, gives more fatique sound.
Hi Jacco,
In my view the Glass Harmony hybrid Class A is not a commercial proposition. It consumes 150W of energy for just 28W rms of output; this is unviable, particularly in the Class D world we are now entering.
However, I'm working on a SE bridged version of the GH, using forced air cooling and inductive loading. Just today I despatched the proto pcbs for manufacture. It will have about 45W into 8R for 180W of consumption; this might be more like it. Preliminary testing with one channel shows great promise, but we can't be completely sure until we have two channels with proper pcb construction. Paul and I will build it up in the next couple of months and have a good listen, who knows, it might be just the cat's meow..........
Lumenauw,
These are not trade secrets, they are audio philosophies. Not very useful really; it's the detail which is important - circuit schemats, dimensions, component grade and type. Besides, I'm not a math and measure man, I'm more intuitive, not an engineer (my academic qualifications are in Literature, Education and IT), and I'm not prepared to venture into the deep math to justify my conclusions. I doubt any PSpice device model correctly identifies the myriad parasitics in modern semiconductors anyway. Nonetheless, I do use a fair bit of maths, but nothing too complicated. Most calculations merely back up or double check empirical choices. I have always felt that THD measures the wrong thing - but always remember, nothing is more dangerous than one man's opinion.
Cheers,
Hugh
In my view the Glass Harmony hybrid Class A is not a commercial proposition. It consumes 150W of energy for just 28W rms of output; this is unviable, particularly in the Class D world we are now entering.
However, I'm working on a SE bridged version of the GH, using forced air cooling and inductive loading. Just today I despatched the proto pcbs for manufacture. It will have about 45W into 8R for 180W of consumption; this might be more like it. Preliminary testing with one channel shows great promise, but we can't be completely sure until we have two channels with proper pcb construction. Paul and I will build it up in the next couple of months and have a good listen, who knows, it might be just the cat's meow..........
Lumenauw,
These are not trade secrets, they are audio philosophies. Not very useful really; it's the detail which is important - circuit schemats, dimensions, component grade and type. Besides, I'm not a math and measure man, I'm more intuitive, not an engineer (my academic qualifications are in Literature, Education and IT), and I'm not prepared to venture into the deep math to justify my conclusions. I doubt any PSpice device model correctly identifies the myriad parasitics in modern semiconductors anyway. Nonetheless, I do use a fair bit of maths, but nothing too complicated. Most calculations merely back up or double check empirical choices. I have always felt that THD measures the wrong thing - but always remember, nothing is more dangerous than one man's opinion.
Cheers,
Hugh
AKSA said:
Hello Hugh
About;
"and X-Y input/output plot of phase at 20Khz into a real world load at 1 watt and again at full power"
Can you say more on how to do it practicaly ?
Thank
Bye
Gaetan
Hi Gaetan,
Hey, this thread is 3 years old!!
Dual channel CRO, one channel on input, one on output, switch to X-Y, set up sensitivity for a 45 degree straight line with 2Vpp input at 1KHz.
Set up a resistive 8R load (for safety for initial test).
Now sweep from 1KHz to 100KHz. The straight line angle should steepen, and the single line should open out into an ellipse.
The point at which this begins to take place is an indication of pole frequency. I find this a useful check, and when repeated with a reactive load, say 8R//0.47uF, it should show if there are stability problems. When phase shift is bad and non-linear, the ellipse becomes a circle, and if oscillation begins, looks like a rhombus.
Hugh
Hey, this thread is 3 years old!!
Dual channel CRO, one channel on input, one on output, switch to X-Y, set up sensitivity for a 45 degree straight line with 2Vpp input at 1KHz.
Set up a resistive 8R load (for safety for initial test).
Now sweep from 1KHz to 100KHz. The straight line angle should steepen, and the single line should open out into an ellipse.
The point at which this begins to take place is an indication of pole frequency. I find this a useful check, and when repeated with a reactive load, say 8R//0.47uF, it should show if there are stability problems. When phase shift is bad and non-linear, the ellipse becomes a circle, and if oscillation begins, looks like a rhombus.
Hugh
- Status
- Not open for further replies.