Contrarians
I love that trailer, it summarizes what my Catholic lay-minister father thought of me...
Howie
I love that trailer, it summarizes what my Catholic lay-minister father thought of me...
Howie
For the same GBW product, no I don't. It's not bickering it seems like a technical discussion to me on points that folks disagree on, and I think people need to hear both sides.
Maybe you missed the original discussion about "improving" things by simply resistively loading the VAS, etc. The point was about constant GBW product and simply moving the low frequency pole around.
Scott,
Well, here is a little comment or two on this GBWP...
First, most of the distortion sources in "classic" feedback solid state circuitry rise with frequency and often with a quite steep slope (I think I came up with over 4th order final distortion rise towards high frequencies based on the stuff in Selfs WW Articles).
Second, a feedback amplifier that has low open loop linearity (because degeneration and thus wideband linearisation, reduced sensitivity to TIM/PIM et al. has been traded for increased loop gain and lower noise) and will not have any more loop gain available for a given high frequency (as we in the end need a stable amplifier), which is arguably where we need amplifier linearity most, but it will invariably have worse distortion at high frequencies.
So all else being equal, the same number of transistors etc. expended on a given problem will probably produce an amplifier with greater low frequency distortion but lesser high frequency distortion if the open loop gain is kept low by linearising the circuit using degeneration (not by throwing massive amounts of loop gain away by loading the VAS I might want to add) than an amplifier that maximises open loop gain and in my experience the subjective results of such a design are preferable to one that maximises DC open loop gain.
Of course, taking this logic to it's conclusions we end up with Amplifiers that have 0.004% THD @ 10W/10KHz and < 0.1% THD at rated power, (but compound transistors) and have easily > 1MHz open loop bandwidth with an open loop gain of 26dB (that would >20MHz GBWP)...
I would further add that in most common solid state feedback amplifiers the greatest source of distortion is the VAS non-linearity and the non-linear load onto the VAS by the output stage, especially the latter is highly complex. Other "interstage" loading is also not negligible (of course Self & Cordell cover this in some length, so this nothing new).
So if we want improvements overall in distortion we may be better off dealing with the non-linear interstage loads than we are cranking up negative loop feedback.
Of course, one may argue that such measures quite obviously also benefit Amplifiers employing a lot of looped feedback.
Yet once they are applied to open loop circuitry to a sufficient degree we need to ask if the performance gains we can make from applying more feedback are really relevant or equivalent to the vice of Master Bater from the UK Childrens TV Cartoon Captain Pugwash...
Ciao T
Well, here is a little comment or two on this GBWP...
First, most of the distortion sources in "classic" feedback solid state circuitry rise with frequency and often with a quite steep slope (I think I came up with over 4th order final distortion rise towards high frequencies based on the stuff in Selfs WW Articles).
Second, a feedback amplifier that has low open loop linearity (because degeneration and thus wideband linearisation, reduced sensitivity to TIM/PIM et al. has been traded for increased loop gain and lower noise) and will not have any more loop gain available for a given high frequency (as we in the end need a stable amplifier), which is arguably where we need amplifier linearity most, but it will invariably have worse distortion at high frequencies.
So all else being equal, the same number of transistors etc. expended on a given problem will probably produce an amplifier with greater low frequency distortion but lesser high frequency distortion if the open loop gain is kept low by linearising the circuit using degeneration (not by throwing massive amounts of loop gain away by loading the VAS I might want to add) than an amplifier that maximises open loop gain and in my experience the subjective results of such a design are preferable to one that maximises DC open loop gain.
Of course, taking this logic to it's conclusions we end up with Amplifiers that have 0.004% THD @ 10W/10KHz and < 0.1% THD at rated power, (but compound transistors) and have easily > 1MHz open loop bandwidth with an open loop gain of 26dB (that would >20MHz GBWP)...
I would further add that in most common solid state feedback amplifiers the greatest source of distortion is the VAS non-linearity and the non-linear load onto the VAS by the output stage, especially the latter is highly complex. Other "interstage" loading is also not negligible (of course Self & Cordell cover this in some length, so this nothing new).
So if we want improvements overall in distortion we may be better off dealing with the non-linear interstage loads than we are cranking up negative loop feedback.
Of course, one may argue that such measures quite obviously also benefit Amplifiers employing a lot of looped feedback.
Yet once they are applied to open loop circuitry to a sufficient degree we need to ask if the performance gains we can make from applying more feedback are really relevant or equivalent to the vice of Master Bater from the UK Childrens TV Cartoon Captain Pugwash...
Ciao T
Well, Charles might tell us a bit more, but if you read this, you know that the volume control in this gear is made of "loading the VAS". If you do it intelligently, the load ist nicely resistive, and if you isolate the VAS transistors from the load, you won't see an increase in distortion at lower levels (= higher loaded VAS).
And this example is open loop.
Ayre KX-R line preamplifier | Stereophile.com
I think it was Charles who said it here first. Putting a resistor to ground on the comp pin of the AD844 and moving the Aol corner out beyond the audio BW was in and of itself a sonic improvement. Considering I saw nothing to back that up beyond the usual listening tests, I continue to disagree with the reasoning. This then got rolled into the PIM discussion somehow and since counter opinion seems unwelcome I will leave you to it.
Hi Scott,
Not sure on that. Charles (and others, myself included) have used the AD844 Open-Loop by setting the transconductance with a resistor on the - In and loading the Comp Pin.
There are limits as the input must be degenerated quite a bit and overall transfer linearity in the AD844 could be better (load effect distortion?).
This is not what you describe, but I must admit, for whatever reason, I like the sound better that way. I have not tried to pad the AD844 gain down, as it stands it's CFB and has a ridish OLBW anyway.
Ciao T
Not sure on that. Charles (and others, myself included) have used the AD844 Open-Loop by setting the transconductance with a resistor on the - In and loading the Comp Pin.
There are limits as the input must be degenerated quite a bit and overall transfer linearity in the AD844 could be better (load effect distortion?).
This is not what you describe, but I must admit, for whatever reason, I like the sound better that way. I have not tried to pad the AD844 gain down, as it stands it's CFB and has a ridish OLBW anyway.
Ciao T
Hi,
Yes, doing that makes sense. But we are dealing with circuits designed ground up for open loop operation and you would have to move the volume control into the feedback loop if you where using one...
In addition, while I cannot be sure, I suspect Charles also uses the resistance between two differential halves source/cathodes.
I have done this using a Tube diff amp where either device sat on it's own CCS (C4S) and a resistor between the two halves cathodes was used for part of the gain setting (you cannot go too high, otherwise noise is way up) and the rest was done by loading between the anodes of the now extremely heavily degenerated diff pair.
I experimented with Photo Resistors but they have soo much distortion, it made my tubes look harmless, plus it did not sound good. Binary weighted relay matrixes worked rather well...
Ciao T
Yes, doing that makes sense. But we are dealing with circuits designed ground up for open loop operation and you would have to move the volume control into the feedback loop if you where using one...
In addition, while I cannot be sure, I suspect Charles also uses the resistance between two differential halves source/cathodes.
I have done this using a Tube diff amp where either device sat on it's own CCS (C4S) and a resistor between the two halves cathodes was used for part of the gain setting (you cannot go too high, otherwise noise is way up) and the rest was done by loading between the anodes of the now extremely heavily degenerated diff pair.
I experimented with Photo Resistors but they have soo much distortion, it made my tubes look harmless, plus it did not sound good. Binary weighted relay matrixes worked rather well...
Ciao T
I was informed that Charles Hansen has been very ill, above and beyond his normal condition, and this has precluded him from defending his position here.
I consider Charles Hansen my greatest competitor, with more 'gumption' than almost anyone else in the industry. Looking at his designs, and the times he has edged me out, convinces me he is on the right track.
Yes, Charles does not like global negative feedback. This comes from listening tests, more than any measurements. I must say that every time that I have followed Charles' lead and tried something 'exotic' myself, I have made a better product. I attribute most of the Blowtorch electronics to Charles doing it first. I did NOT, at first, think that it would beat a simple dual 10 turn wire wound volume control, that I used between my Vendetta and my power amp, for years, after putting the JC-80 in the closet. But it seemed that the open loop operation made the circuitry virtually invisible to notice, and I now have the advantage of a lower output Z and circuit gain, if necessary. That is where I sit today. I will most probably never make a better preamp than the Blowtorch, as I have only so many years left, in any case. If others want to follow in my footsteps, then listen up. If not, then you can always find a decent IC op amp that will measure well, and work pretty good. I might warn you not to take your IC design against the 'big boys' of the audio world, but if you are happy, that is enough.
Charles and I will NEVER be able to satisfy many here that we know what we are doing. It is sad, but that's life.
I consider Charles Hansen my greatest competitor, with more 'gumption' than almost anyone else in the industry. Looking at his designs, and the times he has edged me out, convinces me he is on the right track.
Yes, Charles does not like global negative feedback. This comes from listening tests, more than any measurements. I must say that every time that I have followed Charles' lead and tried something 'exotic' myself, I have made a better product. I attribute most of the Blowtorch electronics to Charles doing it first. I did NOT, at first, think that it would beat a simple dual 10 turn wire wound volume control, that I used between my Vendetta and my power amp, for years, after putting the JC-80 in the closet. But it seemed that the open loop operation made the circuitry virtually invisible to notice, and I now have the advantage of a lower output Z and circuit gain, if necessary. That is where I sit today. I will most probably never make a better preamp than the Blowtorch, as I have only so many years left, in any case. If others want to follow in my footsteps, then listen up. If not, then you can always find a decent IC op amp that will measure well, and work pretty good. I might warn you not to take your IC design against the 'big boys' of the audio world, but if you are happy, that is enough.
Charles and I will NEVER be able to satisfy many here that we know what we are doing. It is sad, but that's life.
John,
I am very sorry to hear that.
If you happen to talk to Charles or his family, please convey my personal best wishes for him to get better again soon, though I hope I can speak for all here to also wish all the best on behalf of all here at DIYA. I will also e-mail him of course.
Ciao T
I am very sorry to hear that.
If you happen to talk to Charles or his family, please convey my personal best wishes for him to get better again soon, though I hope I can speak for all here to also wish all the best on behalf of all here at DIYA. I will also e-mail him of course.
Ciao T
Folks,
A while back I mentioned in addition to Geddes and Geddes/Lee papers (which should be common knowledge) "that paper by the students at some skandinavian uni"...
Well, I happen-chanced on it again.
http://projekter.aau.dk/projekter/files/9852082/07gr1061_Thesis.pdf
It makes good reading both with regards to experimental design for blind audibility testing as well as statistical analysis and other such minutiae. The conclusions are hardly world-shaking, but the "references" sections also covers much excellent material.
Highly recommended.
Ciao T
A while back I mentioned in addition to Geddes and Geddes/Lee papers (which should be common knowledge) "that paper by the students at some skandinavian uni"...
Well, I happen-chanced on it again.
http://projekter.aau.dk/projekter/files/9852082/07gr1061_Thesis.pdf
It makes good reading both with regards to experimental design for blind audibility testing as well as statistical analysis and other such minutiae. The conclusions are hardly world-shaking, but the "references" sections also covers much excellent material.
Highly recommended.
Ciao T
It was here. The problem is that you can quite easily keep all the harmonics below -100dB (referred to 1st). The studies investigate very high distortion levels. Nothing has ever been published about audibility of -100dB distortions .... for the reason that results would be non transferable and not based on distortion as a reason of sonic differences. This paper does not bring any new light.
Well, I am a believer in incremental progress. It is not always a major breakthrough that happens every week, BUT I do not attempt to impede input, but showing a negative reaction to someone's real effort in making hi fidelity better. Just today Dick Sequerra told me the 'secret' to his latest success. A few days ago, Ron Quan told me of his progress in PIM measurement.
We do not sleep, just today I cracked a REAL audio headache. A power amp that I designed and built about 30 years ago. $200+ in technician time, and me working for free, fixed the accumulated problems. Some were very surprising. Now the amp works like it used to. It happens to be all fet, fully complementary, fully cascoded, fully power supply regulated. One of a kind. AND with more than many here would ever consider putting into a power amp.
We do not sleep, just today I cracked a REAL audio headache. A power amp that I designed and built about 30 years ago. $200+ in technician time, and me working for free, fixed the accumulated problems. Some were very surprising. Now the amp works like it used to. It happens to be all fet, fully complementary, fully cascoded, fully power supply regulated. One of a kind. AND with more than many here would ever consider putting into a power amp.
"Well, Charles might tell us a bit more, but if you read this, you know that the volume control in this gear is made of "loading the VAS". If you do it intelligently, the load ist nicely resistive, and if you isolate the VAS transistors from the load, you won't see an increase in distortion at lower levels (= higher loaded VAS). And this example is open loop."
If the VAS (transconductance stage) is a current mirror design (as is typically the case in a CFA), then this is an easy way to change the volume in an open loop design. Obviously CH's circuit is not available, but I would imagine this is how he is doing it.
You can use the same technique to reduce the loop gain in a CFA that is being used conventionally with feedback. By loading the current mirror node (I am talking here about a complementary structure), the OLG can be varied from very low - around 2x to 3x - all the way up to 80 or 100dB. I did some simulations a few years back and got about 300ppb with no mirror load (very high loop gain) and a CLG of 5, and then loaded down such that LG was about 10dB, around 50ppm 20K THD (also with closed loop gain of 5). For anyone wanting to experiment with changing loop gain easily this is quite an interesting technique.
Output level 2V into 600 Ohms.
If the VAS (transconductance stage) is a current mirror design (as is typically the case in a CFA), then this is an easy way to change the volume in an open loop design. Obviously CH's circuit is not available, but I would imagine this is how he is doing it.
You can use the same technique to reduce the loop gain in a CFA that is being used conventionally with feedback. By loading the current mirror node (I am talking here about a complementary structure), the OLG can be varied from very low - around 2x to 3x - all the way up to 80 or 100dB. I did some simulations a few years back and got about 300ppb with no mirror load (very high loop gain) and a CLG of 5, and then loaded down such that LG was about 10dB, around 50ppm 20K THD (also with closed loop gain of 5). For anyone wanting to experiment with changing loop gain easily this is quite an interesting technique.
Output level 2V into 600 Ohms.
Last edited:
We get the point, questioning the technical merit of something is an attack on "progress". You win.
Well, but it also shows that "loading the VAS", when done intelligently, doesn't rise the distortion in a way which would harm the sound and could therefore be applied (in this very form, but with a fixed value) also to loop feedback designs. It will not increase noise, compared to very high degeneration at the input for example.
And Scott, please be aware that the KX-R is NOT using AD844's.
Ayre has other products which uses this, and I believe that with the AD 844 it might be difficult to implement such a volume control ("load down the VAS") in such a drastic way.
Yes, if you want to change the gain, you can replace the resistor between the two balanced halves in the input differential.
There is a German article, Aus dem Messlabor: Ayre-Arbeitsprinzip und die neue Messung - stereoplay - Magnus.de
You could that do with two AD844 too, the two +Inputs are the diff input, and you connect the two -Inputs together with a resistor which suits your degeneration requirements. Conceptual schematic here
I wish Charles all the best.
- Status
- Not open for further replies.