I think if the endings are bjt the vbe multiplier requires bjt in TO126 container or similar. I think if the finals are MOSFETs the VBE multiplier requires MOSFETs like BS170 or similar. I think if the finals are hexfet the vbe multiplier requires hexfet type IRF840 or similar. More or less.
Last edited:
Anatech: I think you misunderstood my comment about Boulder.
They do use a lot of global feedback, but they have devided the amplifier in two separted amplifiers with their own feedback loops.
First one with around approx. 20dB of gain with output stage in class A avoiding crossover distortion.
The second amplifier containing the real output stage have only 6dB of GAIN, not feedback. Feedback probably quite high since Boulder usually have very low output impedance. They also adopt sliding bias to further avoid crossover distortion and keep the output transistors from turning off.
I think their application is clever. Avoiding crossover over distortion is essential and by deviding amplification in two serial amplifiers where 20dB of gain are outside the output stage feedback loop they get a very clean and good performance.
So even if using a lot of feedback Boulder have found a way to minimize the problem of crossover distortion.
They do use a lot of global feedback, but they have devided the amplifier in two separted amplifiers with their own feedback loops.
First one with around approx. 20dB of gain with output stage in class A avoiding crossover distortion.
The second amplifier containing the real output stage have only 6dB of GAIN, not feedback. Feedback probably quite high since Boulder usually have very low output impedance. They also adopt sliding bias to further avoid crossover distortion and keep the output transistors from turning off.
I think their application is clever. Avoiding crossover over distortion is essential and by deviding amplification in two serial amplifiers where 20dB of gain are outside the output stage feedback loop they get a very clean and good performance.
So even if using a lot of feedback Boulder have found a way to minimize the problem of crossover distortion.
Hi flex2,
I am not offended at all. What I am is really tired of hearing / seeing the same misunderstandings over and over again.
I agree, Counterpoint product had a very nice sound quality. If you investigate, the SA-3000 and SA-5000 preamplifiers have very low distortion. Most other products do not come close, although a couple do.
One constant weakness in Counterpoint designs is power supply, and anything to do with solid state. Michael considered himself an engineer (he wasn't) and designed by ear. He also paid far too much attention to audio fads, and his following "upgrades" were centred on that. Those were pure rip-offs to the consumer. The company failed due to reliability problems and the costs of trying to produce the SA-100 and SA-220 (the original Mosfets were discontinued on top, others suffered from pronounced gate charge issues). So this design was an extremely poor decision for starters, it was not manufacturable. Once the original Mosfet types were discontinued, repair and further production of this type of design became impossible. So, this is the base of your amplifier design. I am fully aware of all the poor design choices and have corrected them. The end result will be new PCBs completely and a design that is repairable (and very reliable). It sounds the same but cleaner.
Distortion and output impedance is reduced by the amount of overall feedback applied. Therefore if you don't have a mechanism that causes excessive phase shift, the more feedback you can apply - the better . This is simply true. Limiting factors will be a poor design (feedback can't fix that, and it will misbehave) and something like an output transformer. Poor PCB or wiring layout can also introduce severe problems.
About distortion. In the 80's we could easily see maybe -80 dB down. Today things have changed dramatically. I can easily see about -135 dB from 2.83 V without taking special care. I have a test and measurement background as well. Most of my equipment is HP and Keysight. Also, you are quoting numbers as we would get from a THD meter. I used them, but hung a spectrum analyzer on the monitor output way back in time. I have HP 3580A, 35665A, 3585A and 4195A, plus an RTX 6001 audio analyzer. My THD meters are HP 339A and ShibaSoku AS725C. My DVMs and scope are current (Keysight MSOX3104T and 34465A and 34461A) plus other gear. I do have the instrumentation required to support each and every claim I have made, and the experience to back it up also. The cost of admission is pretty high. Anyway, I interpret the entire audio spectrum and beyond, not simple numbers. I've also found this does in fact correlate to subjective opinion (not my own, I never assess my own work).
If you have complete information, then you will see what I have said is in fact correct.
Now, any signal stage does not differentiate between audio signals or anything other kind of signal. As long as the signals it sees is within the operational bandwidth, it will process that signal the way it was designed to. So crossover distortion is no different than anything else except it shows up in the inverting input side instead of the forward signal path. Since it is not the same as the input signal, it is inverted (made complimentary) and cancelled in the following stages that caused the distortion in the first place. Individual components function at far higher frequencies than the circuit overall, so forget fancy ideas of time displacement. Same for "a period of time when feedback lags the signal". That simply does not happen.
Most audio myths follow the line of "it could happen, we don't know everything". Bunk. We know enough about these signals and how circuits work. Most of this garbage was born in the 1960's and into the 1980's when it was true, we couldn't measure signals to the levels and frequencies we can today. We can measure so far below the level of the human body to even perceive - or sense in any way, claims that we can hear things that can't be measured are pure ignorance and fancy. I do it every single day. I can often predict how something will sound (and I'm correct).
A linear input stage is critical, so yes. Also, so it high transconductance. Under those conditions, and a reasonable design without severe defects, higher applied feedback that remains stable will always improve subjective and measured performance. You shouldn't be surprised by this. I also fully agree with you. Avoid creating distortion to begin with, I never once said anything to the contrary. That is precisely why I selected the output stage I went with. It has lower distortion than a mosfet stage does. More even loading as well so I don't create distortion in the circuit driving the output stage.
Try not to fall prey to fanciful stories and half truths. My world revolves around physics and facts. This agrees with subjective opinion, and my best subjects are children and women who do not consider the audio press. People I use are presented with items that may or may not have been modified. I never tell them, so they have learned to be honest. Your brain can be your biggest enemy, or greatest allie. Expectation bias is your greatest enemy, and so are half-true stories and "white papers".
Now the people you just quoted use audio myth to gain market share. They follow a design path that is romantic and marketable. There area number of designers that make better sounding products based in reality and physics (imagine that). These people do not suffer from ego issues at all either. The fact that an "audio designer" is well known has zero to do with what is correct. Most excellent audio designers are not well known, tend to be approachable and quiet about what they do. Also, teams of engineers do a better job than one. You also really do need the test equipment and knowledge to support your line of work. I didn't spend what I did on test equipment to brag (that would be pretty stupid). It was purchased in order to get the repeatable answers I need. Otherwise I'd be working in the dark and mumbling nonsense.
I am not offended at all. What I am is really tired of hearing / seeing the same misunderstandings over and over again.
I agree, Counterpoint product had a very nice sound quality. If you investigate, the SA-3000 and SA-5000 preamplifiers have very low distortion. Most other products do not come close, although a couple do.
One constant weakness in Counterpoint designs is power supply, and anything to do with solid state. Michael considered himself an engineer (he wasn't) and designed by ear. He also paid far too much attention to audio fads, and his following "upgrades" were centred on that. Those were pure rip-offs to the consumer. The company failed due to reliability problems and the costs of trying to produce the SA-100 and SA-220 (the original Mosfets were discontinued on top, others suffered from pronounced gate charge issues). So this design was an extremely poor decision for starters, it was not manufacturable. Once the original Mosfet types were discontinued, repair and further production of this type of design became impossible. So, this is the base of your amplifier design. I am fully aware of all the poor design choices and have corrected them. The end result will be new PCBs completely and a design that is repairable (and very reliable). It sounds the same but cleaner.
Distortion and output impedance is reduced by the amount of overall feedback applied. Therefore if you don't have a mechanism that causes excessive phase shift, the more feedback you can apply - the better . This is simply true. Limiting factors will be a poor design (feedback can't fix that, and it will misbehave) and something like an output transformer. Poor PCB or wiring layout can also introduce severe problems.
About distortion. In the 80's we could easily see maybe -80 dB down. Today things have changed dramatically. I can easily see about -135 dB from 2.83 V without taking special care. I have a test and measurement background as well. Most of my equipment is HP and Keysight. Also, you are quoting numbers as we would get from a THD meter. I used them, but hung a spectrum analyzer on the monitor output way back in time. I have HP 3580A, 35665A, 3585A and 4195A, plus an RTX 6001 audio analyzer. My THD meters are HP 339A and ShibaSoku AS725C. My DVMs and scope are current (Keysight MSOX3104T and 34465A and 34461A) plus other gear. I do have the instrumentation required to support each and every claim I have made, and the experience to back it up also. The cost of admission is pretty high. Anyway, I interpret the entire audio spectrum and beyond, not simple numbers. I've also found this does in fact correlate to subjective opinion (not my own, I never assess my own work).
If you have complete information, then you will see what I have said is in fact correct.
Now, any signal stage does not differentiate between audio signals or anything other kind of signal. As long as the signals it sees is within the operational bandwidth, it will process that signal the way it was designed to. So crossover distortion is no different than anything else except it shows up in the inverting input side instead of the forward signal path. Since it is not the same as the input signal, it is inverted (made complimentary) and cancelled in the following stages that caused the distortion in the first place. Individual components function at far higher frequencies than the circuit overall, so forget fancy ideas of time displacement. Same for "a period of time when feedback lags the signal". That simply does not happen.
Most audio myths follow the line of "it could happen, we don't know everything". Bunk. We know enough about these signals and how circuits work. Most of this garbage was born in the 1960's and into the 1980's when it was true, we couldn't measure signals to the levels and frequencies we can today. We can measure so far below the level of the human body to even perceive - or sense in any way, claims that we can hear things that can't be measured are pure ignorance and fancy. I do it every single day. I can often predict how something will sound (and I'm correct).
A linear input stage is critical, so yes. Also, so it high transconductance. Under those conditions, and a reasonable design without severe defects, higher applied feedback that remains stable will always improve subjective and measured performance. You shouldn't be surprised by this. I also fully agree with you. Avoid creating distortion to begin with, I never once said anything to the contrary. That is precisely why I selected the output stage I went with. It has lower distortion than a mosfet stage does. More even loading as well so I don't create distortion in the circuit driving the output stage.
Try not to fall prey to fanciful stories and half truths. My world revolves around physics and facts. This agrees with subjective opinion, and my best subjects are children and women who do not consider the audio press. People I use are presented with items that may or may not have been modified. I never tell them, so they have learned to be honest. Your brain can be your biggest enemy, or greatest allie. Expectation bias is your greatest enemy, and so are half-true stories and "white papers".
Now the people you just quoted use audio myth to gain market share. They follow a design path that is romantic and marketable. There area number of designers that make better sounding products based in reality and physics (imagine that). These people do not suffer from ego issues at all either. The fact that an "audio designer" is well known has zero to do with what is correct. Most excellent audio designers are not well known, tend to be approachable and quiet about what they do. Also, teams of engineers do a better job than one. You also really do need the test equipment and knowledge to support your line of work. I didn't spend what I did on test equipment to brag (that would be pretty stupid). It was purchased in order to get the repeatable answers I need. Otherwise I'd be working in the dark and mumbling nonsense.
In the case of sliding bias, that is a good technique. I think you might want to ask Nelson Pass about that. Even Nikko used methods to keep outputs from turning off. Check out the Alpha 220 (I did warranty for Nikko back then as well). However a standard class "AB" stage with proper bias does not create high amounts of crossover distortion.
Try this. Take an oscilloscope and hang it on the monitor output of a good THD meter. Connect and run a BJT amplifier at low bias (I assume this would be worst case from your perspective). I was doing this in the early 1980's, and I still do run this check. Increase the bias current slowly. Now, a good circuit with matched parts will eliminate crossover by about 5 mA bias or lower. A poor design will not show a minimum and crossover will never be great, but it will continue to be reduced as you crank up the standing current. Gee? That should show you something important. This is where the class "A" myth came from. Got a crappy design? Crank the bias, spin a story and you have a heavy, hot bestseller. This kind of class "A" has the same issues as "AB" does by the way. This should also not surprise you if you design electronics.
Now the Nikko Alpha 220 isn't a great sounding amplifier, but it sounds better than many. It has other issues too. I'm simply trying to illustrate that keeping an output stage in class "A" doesn't mean it will sound good. Everything is important for good performance. Component type (not brand!), circuit layout and topology. You can focus on a design concept, but execution matters more in some cases. There are so many factors that have to come together to make a circuit that performs well, audio or otherwise.
Anyway, what Boulder did is decades old as a design concept. Big deal.
Besides, we are talking about the Counterpoint SA-220. It has serious flaws, so okay. They can be fixed and you do end up with a much better product that still sounds like the original design, but much better with lower distortion and higher damping factor. It runs cooler too (failure rate doubles every 10 rise in temperature), and this you do know as a successful designer - right?
-Chris
Try this. Take an oscilloscope and hang it on the monitor output of a good THD meter. Connect and run a BJT amplifier at low bias (I assume this would be worst case from your perspective). I was doing this in the early 1980's, and I still do run this check. Increase the bias current slowly. Now, a good circuit with matched parts will eliminate crossover by about 5 mA bias or lower. A poor design will not show a minimum and crossover will never be great, but it will continue to be reduced as you crank up the standing current. Gee? That should show you something important. This is where the class "A" myth came from. Got a crappy design? Crank the bias, spin a story and you have a heavy, hot bestseller. This kind of class "A" has the same issues as "AB" does by the way. This should also not surprise you if you design electronics.
Now the Nikko Alpha 220 isn't a great sounding amplifier, but it sounds better than many. It has other issues too. I'm simply trying to illustrate that keeping an output stage in class "A" doesn't mean it will sound good. Everything is important for good performance. Component type (not brand!), circuit layout and topology. You can focus on a design concept, but execution matters more in some cases. There are so many factors that have to come together to make a circuit that performs well, audio or otherwise.
Anyway, what Boulder did is decades old as a design concept. Big deal.
Besides, we are talking about the Counterpoint SA-220. It has serious flaws, so okay. They can be fixed and you do end up with a much better product that still sounds like the original design, but much better with lower distortion and higher damping factor. It runs cooler too (failure rate doubles every 10 rise in temperature), and this you do know as a successful designer - right?
-Chris
Thanks for a long and nice answer. Appreciated.
But what about input stage bandwidth?
Don't you need to limit bandwidth to implement global feedback without getting into oscillation?
When limited too much into what we can hear isn't the input unable to amplify essential and recorded harmonics in the signal?
Feedback can't fix that problem...
In my book of good practice I usually will not accept a input stage bandwidth less than 40 Khz before entering eventual global feedback.
Regarding Counterpoint original design and reliability we are the same page. An unreliable design that sounds good when working...
I think we could have a very interesting discussion about amplifier technology and harmonics in distortion, but this thread is about SA-220 so it have to end here.
If you for some reason want me to simulate something regarding SA-12, SA-100, SA-20 or SA-220 please just ask.
Since I have the models built it probably would be quickly done.
Funny that they had different circuits for Europe compared to the rest of the world...
Do we listen differently in Europe???
Personally I don't see the need for sliding bias since I only build audio eqipment for personal use and can select class of operation.
I had some interesting discussion with Mr. Pass but sliding bias was not one of them.
I suspect he wouldn't support using them and say dead end...
I sent him a research paper I made regarding at what level we can detect different kind of harmonic distortion.
Have no idea if he used it for something. Just noted that he thought it was a bit interesting.
Pass is one of the nicest persons in the business and I follow his work closely.
Still have a F6 that needs to be built...
But what about input stage bandwidth?
Don't you need to limit bandwidth to implement global feedback without getting into oscillation?
When limited too much into what we can hear isn't the input unable to amplify essential and recorded harmonics in the signal?
Feedback can't fix that problem...
In my book of good practice I usually will not accept a input stage bandwidth less than 40 Khz before entering eventual global feedback.
Regarding Counterpoint original design and reliability we are the same page. An unreliable design that sounds good when working...
I think we could have a very interesting discussion about amplifier technology and harmonics in distortion, but this thread is about SA-220 so it have to end here.
If you for some reason want me to simulate something regarding SA-12, SA-100, SA-20 or SA-220 please just ask.
Since I have the models built it probably would be quickly done.
Funny that they had different circuits for Europe compared to the rest of the world...
Do we listen differently in Europe???
Personally I don't see the need for sliding bias since I only build audio eqipment for personal use and can select class of operation.
I had some interesting discussion with Mr. Pass but sliding bias was not one of them.
I suspect he wouldn't support using them and say dead end...
I sent him a research paper I made regarding at what level we can detect different kind of harmonic distortion.
Have no idea if he used it for something. Just noted that he thought it was a bit interesting.
Pass is one of the nicest persons in the business and I follow his work closely.
Still have a F6 that needs to be built...
Last edited:
lol!
Hi flex2,
Yes, you should have a wider circuit bandwidth, then reduce the input signal bandwidth. This is very common practice. My tube amp designs have much lower bandwidth than my solid state designs (no shocker there). Lower voltage, higher standing currents. Having said that, the output transformers I use were designed by Hammond especially for me. No one else can buy them and they have great high frequency extension and stupid good low frequency response. Thin laminations and very high quality steel for the stack. That designer left the company during COVID.
Your bandwidth is sufficient if you roll off 100 KHz, maybe a little higher. Several may complain about that, but signals at 20 KHz are lower in amplitude and don't require high slew rate. Full power? Sure! Then you are shopping for tweeters! Overall it isn't normally a problem to get higher slew rates in amplifier design these days. You are further ahead to roll it off if the circuit might distort.
We honestly don't sense much past 20 KHz as young folks. Most tweeters don't get much past 22 KHz, and if they do they are distorted. It's a good thing levels are low!
I could, but it would be the new design. There is little point in running a sim on the original design. I don't need a simulation - but many thanks.
European markets have different requirements from certification standpoints.
Hi flex2,
Yes, you should have a wider circuit bandwidth, then reduce the input signal bandwidth. This is very common practice. My tube amp designs have much lower bandwidth than my solid state designs (no shocker there). Lower voltage, higher standing currents. Having said that, the output transformers I use were designed by Hammond especially for me. No one else can buy them and they have great high frequency extension and stupid good low frequency response. Thin laminations and very high quality steel for the stack. That designer left the company during COVID.
Your bandwidth is sufficient if you roll off 100 KHz, maybe a little higher. Several may complain about that, but signals at 20 KHz are lower in amplitude and don't require high slew rate. Full power? Sure! Then you are shopping for tweeters! Overall it isn't normally a problem to get higher slew rates in amplifier design these days. You are further ahead to roll it off if the circuit might distort.
We honestly don't sense much past 20 KHz as young folks. Most tweeters don't get much past 22 KHz, and if they do they are distorted. It's a good thing levels are low!
I could, but it would be the new design. There is little point in running a sim on the original design. I don't need a simulation - but many thanks.
European markets have different requirements from certification standpoints.
Yes, I have met Nelson as well. He is one gentleman for sure.
He is very candid. He'll tell you he follows his interests as a design path. He has found a niche that pays (very) well. As long as you understand he doesn't design the way he feels is the best, you'll understand him better. He is one of the few designers that takes parts to the bench and plays with them. I have a great deal of respect for that man.
Douglas Self is another, as is Bob Cordell. In general, people who really do know what they are doing aren't saddled with huge egos and are actually very helpful people. They don't have to agree, but they do like to pass their knowledge and experience on. I've met many others that are too focused on how great they are to learn anything. Maybe a brutal thing to say, but I believe it is true.
He is very candid. He'll tell you he follows his interests as a design path. He has found a niche that pays (very) well. As long as you understand he doesn't design the way he feels is the best, you'll understand him better. He is one of the few designers that takes parts to the bench and plays with them. I have a great deal of respect for that man.
Douglas Self is another, as is Bob Cordell. In general, people who really do know what they are doing aren't saddled with huge egos and are actually very helpful people. They don't have to agree, but they do like to pass their knowledge and experience on. I've met many others that are too focused on how great they are to learn anything. Maybe a brutal thing to say, but I believe it is true.
Pass do us all a favour with his exploring mindset and new way of designing circuits. Many learn a lot from his First Watt designs.
Great man...
Sorry, but I have not liked amplifiers I built by Self or Cordell. Here is where we differs. I find them sterile, edgy and monotonic. Not my sonic preference.
I do not mind a bit of distortion as long as it is the right relationship between mainly 2nd and 3rd harmonic at listening level.
Design goal is less than 0,05% 2nd and 3rd 6-9dB lower than that. I detect 2nd at 0,6% and 3rd at 0,25% and keeping that relationship at listening level is essential for what I find to sound good. The ~10 fold marginal is necessary when pushing the amplifier with complex material.
I'm a strong believer that distortion creates distortion into a perfect storm of distortion the way Pass describe it in his paper about distortion.
So some marginal is essential.
Great paper about distortion...
Nice talking to you despite somewhat different view on distortion and principles of using feedback...
Great man...
Sorry, but I have not liked amplifiers I built by Self or Cordell. Here is where we differs. I find them sterile, edgy and monotonic. Not my sonic preference.
I do not mind a bit of distortion as long as it is the right relationship between mainly 2nd and 3rd harmonic at listening level.
Design goal is less than 0,05% 2nd and 3rd 6-9dB lower than that. I detect 2nd at 0,6% and 3rd at 0,25% and keeping that relationship at listening level is essential for what I find to sound good. The ~10 fold marginal is necessary when pushing the amplifier with complex material.
I'm a strong believer that distortion creates distortion into a perfect storm of distortion the way Pass describe it in his paper about distortion.
So some marginal is essential.
Great paper about distortion...
Nice talking to you despite somewhat different view on distortion and principles of using feedback...
Well, I think we probably don't differ if we have all the information.
No distortion is always best. However if other parts of the system aren't up to scratch, some kinds of distortion can mask it. Clean amplifiers are brutal if the speakers have a problem for example. Also, the perfect schematic on the perfect layout can sound "not great", as is the case of a stock Marantz 300DC. I changed a couple things and it became the amp it should be. Those things totally changed the sound.
I am lucky in that I can measure performance down to tiny details. Most people can't. So I know what the performance really is and not what is reported. The measured differences aren't large between the stock and current 300DC, but they are significantly different in the areas that matter. Same transistors, same PCB layout - no trace alterations. Two completely different amplifiers. So, what does that tell you? These amplifies do not sound sterile in case you're wondering.
I have not read Nelson's paper, but if he advances the idea that some kinds of distortion are desired, I'll strongly disagree with him. An amplifier that adds any deviation to the signal is an effects unit and colours all signals. You do not want this. If another component has an issue - fix that , but do not compensate somewhere else. It is popular to debate whether someone likes 2nd or 3rd order distortion and how much. They are both bad, so as little as possible. A lack of distortion does not sound sterile. Romantic ideas. However a clean amplifier design with some issues may sound sterile - due to distortion. It doesn't take much of the wrong kind. Again, you must see the entire spectrum at low levels to see where problems are.
Another issue is that you become accustomed to the "sound" you have right now. You will prefer that sound quality until you become accustomed to something better. It takes some folks longer than others. It's exactly like changing speakers. I've seen it time and time again.
The amplifiers you built may have issues you don't like, but they are probably correctable. Things like tightly matching input transistor pairs make a big difference. I designed a jig for that and gave it to this community. Getting complimentary driver and other transistors close in beta matters too. Even the feedback resistors and capacitors matter. So it's easy to say "I built xyz amplifier and it didn't sound good to my ears". But, did you really? Maybe not, and maybe not even close. Assumptions can lead you to incorrect conclusions.
No distortion is always best. However if other parts of the system aren't up to scratch, some kinds of distortion can mask it. Clean amplifiers are brutal if the speakers have a problem for example. Also, the perfect schematic on the perfect layout can sound "not great", as is the case of a stock Marantz 300DC. I changed a couple things and it became the amp it should be. Those things totally changed the sound.
I am lucky in that I can measure performance down to tiny details. Most people can't. So I know what the performance really is and not what is reported. The measured differences aren't large between the stock and current 300DC, but they are significantly different in the areas that matter. Same transistors, same PCB layout - no trace alterations. Two completely different amplifiers. So, what does that tell you? These amplifies do not sound sterile in case you're wondering.
I have not read Nelson's paper, but if he advances the idea that some kinds of distortion are desired, I'll strongly disagree with him. An amplifier that adds any deviation to the signal is an effects unit and colours all signals. You do not want this. If another component has an issue - fix that , but do not compensate somewhere else. It is popular to debate whether someone likes 2nd or 3rd order distortion and how much. They are both bad, so as little as possible. A lack of distortion does not sound sterile. Romantic ideas. However a clean amplifier design with some issues may sound sterile - due to distortion. It doesn't take much of the wrong kind. Again, you must see the entire spectrum at low levels to see where problems are.
Another issue is that you become accustomed to the "sound" you have right now. You will prefer that sound quality until you become accustomed to something better. It takes some folks longer than others. It's exactly like changing speakers. I've seen it time and time again.
The amplifiers you built may have issues you don't like, but they are probably correctable. Things like tightly matching input transistor pairs make a big difference. I designed a jig for that and gave it to this community. Getting complimentary driver and other transistors close in beta matters too. Even the feedback resistors and capacitors matter. So it's easy to say "I built xyz amplifier and it didn't sound good to my ears". But, did you really? Maybe not, and maybe not even close. Assumptions can lead you to incorrect conclusions.
I think you would understand my reasoning about distortion better if you read his paper.
No, Pass do not advocate that a certain level of distortion is desired.
He just shows that distortion is a much more complex matter than most of us know.
Distortion is bad, there we are on common ground, but I strongly believe that what ever distortion there is needs to be configured in a certain way to be "good" sounding.
Too much even harmonics will sound soft and euphonic, aka "tube like".
Too much odd harmonics will sound clinical, sharp and edgy, aka "solid state like".
The sound I like the most and find to be most "musical" is when even and odd harmonic distortion levels have a specific relationship at my prefered listening level, and as I wrote earlier I have learnt that I prefer amplifiers that have 6-9dB higher 2nd harmonic distortion than 3rd at prefered listening level.
That happends to be equal to the relationship I can detect the specific harmonic distortions.
A assume you think that keeping those harmonic distortion levels below a certain level will make them undetectable.
Not necessarily so...
The reason is that we measure distortion with a single sinus tone as reference signal.
In Mr. Pass paper he shows that adding multiple sinus tones each with it's own distortion will be added to each other creating much higher level of distortion in total. In his "perfect storm of distortion" up to 15% of distortion. Different distortions doesn't add, they multiply.
As music is very complex and have unlimited number of "sinus" tones music signal can create much higher distortion than we expect by measuring with a single tone reference. The more complex music the more detectable distortion.
So even if the distortion levels are very low they still can penetrate the detection floor when playing music and that when I believe we hear the sonic "character" of the amplifier, and that's when I have found that the relationship between even and odd distortion are so important to be in a certain way.
Feedback can of course lower distortion at the output stage but only at the cost of rising distortion at the feedback point (normally the input stage).
I have found that it's is OK to do so until the distortion of the input stage are at the same level of the output stage.
More feedback than that will have an influence on the sound and make it monotonic. Specially when crossover distortion is added into the feedback loop.
So that's why I have found that feedback can't be used without limits.
None of the different amplifier stages should be allowed to be higher than the output stage in my designs.
So in my book of "truth" even the amplifier with the least distortion at the output can have massive amount of distortion at the input stage and sound very bad.
If we look at Counterpoint in that perspective they built amplifiers with a distortion pattern that comply quite well with my understanding of how to handle distortion in a "good" sounding amplifier.
But as you say, they didn't know how to design their amplifiers to be reliable.
I hope you get some time over to read Nelson Pass paper about distortion.
It opened up my eyes about distortion and how they can multiply to very high levels that are very much hearable.
No, Pass do not advocate that a certain level of distortion is desired.
He just shows that distortion is a much more complex matter than most of us know.
Distortion is bad, there we are on common ground, but I strongly believe that what ever distortion there is needs to be configured in a certain way to be "good" sounding.
Too much even harmonics will sound soft and euphonic, aka "tube like".
Too much odd harmonics will sound clinical, sharp and edgy, aka "solid state like".
The sound I like the most and find to be most "musical" is when even and odd harmonic distortion levels have a specific relationship at my prefered listening level, and as I wrote earlier I have learnt that I prefer amplifiers that have 6-9dB higher 2nd harmonic distortion than 3rd at prefered listening level.
That happends to be equal to the relationship I can detect the specific harmonic distortions.
A assume you think that keeping those harmonic distortion levels below a certain level will make them undetectable.
Not necessarily so...
The reason is that we measure distortion with a single sinus tone as reference signal.
In Mr. Pass paper he shows that adding multiple sinus tones each with it's own distortion will be added to each other creating much higher level of distortion in total. In his "perfect storm of distortion" up to 15% of distortion. Different distortions doesn't add, they multiply.
As music is very complex and have unlimited number of "sinus" tones music signal can create much higher distortion than we expect by measuring with a single tone reference. The more complex music the more detectable distortion.
So even if the distortion levels are very low they still can penetrate the detection floor when playing music and that when I believe we hear the sonic "character" of the amplifier, and that's when I have found that the relationship between even and odd distortion are so important to be in a certain way.
Feedback can of course lower distortion at the output stage but only at the cost of rising distortion at the feedback point (normally the input stage).
I have found that it's is OK to do so until the distortion of the input stage are at the same level of the output stage.
More feedback than that will have an influence on the sound and make it monotonic. Specially when crossover distortion is added into the feedback loop.
So that's why I have found that feedback can't be used without limits.
None of the different amplifier stages should be allowed to be higher than the output stage in my designs.
So in my book of "truth" even the amplifier with the least distortion at the output can have massive amount of distortion at the input stage and sound very bad.
If we look at Counterpoint in that perspective they built amplifiers with a distortion pattern that comply quite well with my understanding of how to handle distortion in a "good" sounding amplifier.
But as you say, they didn't know how to design their amplifiers to be reliable.
I hope you get some time over to read Nelson Pass paper about distortion.
It opened up my eyes about distortion and how they can multiply to very high levels that are very much hearable.
Last edited:
Hi flex2,
Well then, do you have a link?
I do disagree with distortion being okay at any level, you just don't worry about it too much below certain levels.
We measure distortion two ways, both critical. A single tone sine of course - but we look at the entire spectrum. Then we hit the amp with 19 KHz and 20 KHz at equal levels and look at the IM component - plus the entire spectrum. THD and IMD will combine to create a distortion spectrum and I think you are referring to that. THD will magnify the IM components and make things much worse. These two measurements can fully characterize a system or amplifier. With amplifiers we measure at 1 watt into 8 ohms, non-inductive (I have three sets of industry standard dummy loads, the are not cheap). Then you can run the tests at higher power levels to see what happens when you get some current flowing.
Now as for making the input stage distort, not true if things are working properly. The input stage must always remain linear. As soon as the input stage becomes nonlinear you end up with massive distortion and that is going to sound horrible. It is a design error.
My read? Stop going over papers and ignore what people say for a period of time. Ignore the simulator for now, they can lie big time. Now, get some very good test equipment and put some amplifiers and signal stages to the test, then listen to them. Just drop expectations and listen while looking at the audio spectrum. You'll need to see below -100 dBV, most decent stuff will show -140 dBV. Look at the spectrum to at least 48 KHz, I can see 96 KHz and normally leave it wide open.
Thought experiments will lead you down the wrong path if you aren't very careful. Look at what is, not what might be. If there is a concept - prove it!
Well then, do you have a link?
I do disagree with distortion being okay at any level, you just don't worry about it too much below certain levels.
We measure distortion two ways, both critical. A single tone sine of course - but we look at the entire spectrum. Then we hit the amp with 19 KHz and 20 KHz at equal levels and look at the IM component - plus the entire spectrum. THD and IMD will combine to create a distortion spectrum and I think you are referring to that. THD will magnify the IM components and make things much worse. These two measurements can fully characterize a system or amplifier. With amplifiers we measure at 1 watt into 8 ohms, non-inductive (I have three sets of industry standard dummy loads, the are not cheap). Then you can run the tests at higher power levels to see what happens when you get some current flowing.
Now as for making the input stage distort, not true if things are working properly. The input stage must always remain linear. As soon as the input stage becomes nonlinear you end up with massive distortion and that is going to sound horrible. It is a design error.
My read? Stop going over papers and ignore what people say for a period of time. Ignore the simulator for now, they can lie big time. Now, get some very good test equipment and put some amplifiers and signal stages to the test, then listen to them. Just drop expectations and listen while looking at the audio spectrum. You'll need to see below -100 dBV, most decent stuff will show -140 dBV. Look at the spectrum to at least 48 KHz, I can see 96 KHz and normally leave it wide open.
Thought experiments will lead you down the wrong path if you aren't very careful. Look at what is, not what might be. If there is a concept - prove it!
https://www.firstwatt.com/articles.html
First post 2008...
I have some test equipment too, just haven't mentioned it.
Not on the same scale as you, but I can verify both my thoughts and to a high degree the simulations I often use when exploring both my own and others designs when I decide to build a project.
Notes about simulation is made simply because I have some Counterpoint models built in Multisim and offered to make some simulations if there where some interest for it.
No, I'm not refering to IMD and THD in the way you think.
I think you understand what I refer to when reading Mr. Pass paper.
First post 2008...
I have some test equipment too, just haven't mentioned it.
Not on the same scale as you, but I can verify both my thoughts and to a high degree the simulations I often use when exploring both my own and others designs when I decide to build a project.
Notes about simulation is made simply because I have some Counterpoint models built in Multisim and offered to make some simulations if there where some interest for it.
No, I'm not refering to IMD and THD in the way you think.
I think you understand what I refer to when reading Mr. Pass paper.
Last edited:
Hi flex2
Yes, probably. Thank you for the link. Leaving class A?
As far as test equipment is concerned, great! You have the ability to see something at least.
Now, can you see an audio spectrum? What is your noise floor? Can you give me a clue as to what you have and how you have looked at an audio signal?
Yes, probably. Thank you for the link. Leaving class A?
As far as test equipment is concerned, great! You have the ability to see something at least.
Now, can you see an audio spectrum? What is your noise floor? Can you give me a clue as to what you have and how you have looked at an audio signal?
Hi flex2,
That paper describes his design decisions.
Allow me to comment on what I understand to be the direction Nelson has taken. I respect his design choices, but as I said earlier he follows his interest. There is more than one way to "skin a cat" as the saying goes. Nelson would probably agree.
Nelson is correct in the fact that you want to reduce the distortion in the output stage as much as reasonably possible. His paper runs the output stage "open" without feedback to illustrate this. You would normally wrap the entire amplifier in the feedback loop. Nelson's designs are low open loop gain designs, or low transconductance. So he doesn't have as much feedback to use to correct problems as conventional designs do. That's a valid design choice and he has followed a very responsible design path with those considerations. He also uses mosfets, bipolar output sections can match better and have higher transconductance, but they are not square law devices. Also, earlier BJT transistors didn't have great beta vs current curves. The ones we had in the 1970s were not very good in that respect. Bias in mosfet output stages are always higher - and they have to be. This guides his designs.
Problem #1. Failure rate doubles every 10 ° rise in temperature. That includes everything inside the case. I've measured 50 °C and that is decidedly not good for the rest of the amplifier. So we are building in high maintenance, that's your choice.
Problem #2. You have to get rid of the heat. In Canada and your country, in the winter this isn't really too much of a concern. I have gas heating and that is less expensive than electric heat (hint). In the summer we need air conditioning - and that gets really expensive. I love music, so my systems are on when I am up.
I also went with an intrinsically low distortion output stage for the same reasons Nelson did. The stock Counterpoint has a low transconductance voltage amp stage, I increased this a lot. without adding anything. The power diamond buffer is lower distortion than the mosfet stage, and now I have feedback correction and a much more linear voltage amp stage. So I hit this amplifier in more than two areas where it was deficient. It also runs at much lower idle current, lowering the internal case temperature and extending the life of everything inside.
Now, if you have an output stage that offers less distortion, the voltage amp has less correction to do. If you can have very high transconductance the performance is even better. We are talking about a linear voltage amplifier of course. This does in fact sound a lot better. This is generally my design direction. The reasons others follow this ranges between the fact it does sound better down to they haven't designed the amp well and they are hoping feedback with cure their ills. Doing the latter will give you an amplifier that measures "okay" sort of, but sounds bad. If your distortion measuring equipment isn't at a high level you might say those amplifiers measure as well as others that sound fantastic. It isn't measurements that are misleading you, it is a question of resolution or maybe being able to interpret your results. Remember, I am looking at the entire spectrum, not the numbers.
That paper describes his design decisions.
Allow me to comment on what I understand to be the direction Nelson has taken. I respect his design choices, but as I said earlier he follows his interest. There is more than one way to "skin a cat" as the saying goes. Nelson would probably agree.
Nelson is correct in the fact that you want to reduce the distortion in the output stage as much as reasonably possible. His paper runs the output stage "open" without feedback to illustrate this. You would normally wrap the entire amplifier in the feedback loop. Nelson's designs are low open loop gain designs, or low transconductance. So he doesn't have as much feedback to use to correct problems as conventional designs do. That's a valid design choice and he has followed a very responsible design path with those considerations. He also uses mosfets, bipolar output sections can match better and have higher transconductance, but they are not square law devices. Also, earlier BJT transistors didn't have great beta vs current curves. The ones we had in the 1970s were not very good in that respect. Bias in mosfet output stages are always higher - and they have to be. This guides his designs.
Problem #1. Failure rate doubles every 10 ° rise in temperature. That includes everything inside the case. I've measured 50 °C and that is decidedly not good for the rest of the amplifier. So we are building in high maintenance, that's your choice.
Problem #2. You have to get rid of the heat. In Canada and your country, in the winter this isn't really too much of a concern. I have gas heating and that is less expensive than electric heat (hint). In the summer we need air conditioning - and that gets really expensive. I love music, so my systems are on when I am up.
I also went with an intrinsically low distortion output stage for the same reasons Nelson did. The stock Counterpoint has a low transconductance voltage amp stage, I increased this a lot. without adding anything. The power diamond buffer is lower distortion than the mosfet stage, and now I have feedback correction and a much more linear voltage amp stage. So I hit this amplifier in more than two areas where it was deficient. It also runs at much lower idle current, lowering the internal case temperature and extending the life of everything inside.
Now, if you have an output stage that offers less distortion, the voltage amp has less correction to do. If you can have very high transconductance the performance is even better. We are talking about a linear voltage amplifier of course. This does in fact sound a lot better. This is generally my design direction. The reasons others follow this ranges between the fact it does sound better down to they haven't designed the amp well and they are hoping feedback with cure their ills. Doing the latter will give you an amplifier that measures "okay" sort of, but sounds bad. If your distortion measuring equipment isn't at a high level you might say those amplifiers measure as well as others that sound fantastic. It isn't measurements that are misleading you, it is a question of resolution or maybe being able to interpret your results. Remember, I am looking at the entire spectrum, not the numbers.
All right. I have studied it for a long time, and I found that there is a problem with my previous bias. Exicon's mosfet Nmos vgs is 0.5v Pmos vgs is 1v. Well, so I have to pull vr2 to adjust the DC offset. After adjusting, the center point comes to 326mv in the simulation software. Well, if you maintain the R17 of 22.1k after adjusting the offset, even if you short R15, you can't let the voltage be lower than 1.2v. Obviously, this can't be used... Well, I simulated three of them, two of which are to remove R17. The only difference is that R15 is changed to 100k or 47.5k. I think 47 .5k is better because vgs can lower the modulation, and the bias at the largest point of the vr adjustment channel will only be about 1.48v and will not exceed 1.5v. The other is to change 15 to 47.5k R17 to 5.1k, which can also get the required bias 0.5v and 1v. Of course, when adjusted to the maximum, it will exceed 1.5v and reach about 1.75v. Well, should I choose R1547.5k, remove R17, or change R17 to 5.1k?
Well, I have seen a lot of mosfet exicon showing that he is a negative stability system and will reduce vgs by himself when the temperature rises. Really? So why do you need temperature feedback... Well, in short, the official said that it will drop about 0.1v-0.2v vgs. According to the temperature curve of 2n2222A, it will be about 0.4vbe 75 degrees and about 0.5vbe at 25 degrees. Well, it seems that the amount of feedback will not be too large.
Well, I also noticed that exicon wrote that ferrite can reduce oscillation. Should I change it? When replacing the burning gate-level resistor, the 16 ferrite bodies installed in the original factory were removed.
Finally, well, I have to make the selected version of mosfet yellow, so 135ma*8*70 will have nearly 75 standby power consumption. Will it be very warm?
In the simulation, in order to pair the mosfet vgs voltage and let the +-channel run at 135ma, is the 326mv offset voltage too high?
Well, do you think it is more reasonable to decide how to change R17 under this situation?
Thank you very much Anatech.
Thank you very much
Well, I have seen a lot of mosfet exicon showing that he is a negative stability system and will reduce vgs by himself when the temperature rises. Really? So why do you need temperature feedback... Well, in short, the official said that it will drop about 0.1v-0.2v vgs. According to the temperature curve of 2n2222A, it will be about 0.4vbe 75 degrees and about 0.5vbe at 25 degrees. Well, it seems that the amount of feedback will not be too large.
Well, I also noticed that exicon wrote that ferrite can reduce oscillation. Should I change it? When replacing the burning gate-level resistor, the 16 ferrite bodies installed in the original factory were removed.
Finally, well, I have to make the selected version of mosfet yellow, so 135ma*8*70 will have nearly 75 standby power consumption. Will it be very warm?
In the simulation, in order to pair the mosfet vgs voltage and let the +-channel run at 135ma, is the 326mv offset voltage too high?
Well, do you think it is more reasonable to decide how to change R17 under this situation?
Thank you very much Anatech.
Thank you very much
Attachments
Last edited:
Hi Willam,
At this point you are designing a new output stage. That's cool, really good on you actually.
Damping resistors normally work very well. Lossy ferrites are a good choice too. As for the famous resistor R17, make a space for it that you can short. Then play with the circuit for real - forget the simulator for your final decisions.
I could design this for you - except that I am a working technician and simply don't have time. I'm just restarting my own design that I don't have time for.
Temperature rise has everything to do with dissipated power and heatsink fin area * efficiency. If you look at some heat sink vendor web sites, they may have calculators for you. Measure your heat sink, pick a similar one from their stock to get the efficiency number and use their calculator to figure it out. Airflow is key, so is ambient temperature and that will determine inside chassis temperatures. The SA-100 was too high, measured at 50 °C at idle, ambient running about 21 °C (Canada in the winter with the furnace running, or air conditioning in the summer). So you must pay attention to the internal chassis temperatures as well. Counterpoint lost power transformers due to internal case temperatures, they have a thermal fuse. If you place that amplifier in an enclosed space, expect the temperatures to get really high, even open at the front and back.
Your design goal should be to reduce running temperatures. Run at the lowest bias current to give you good sound and low distortion if possible. A great sounding amp that has a blown power transformer doesn't make a lot of noise. But, it makes a great paper weight!
At this point you are designing a new output stage. That's cool, really good on you actually.
Damping resistors normally work very well. Lossy ferrites are a good choice too. As for the famous resistor R17, make a space for it that you can short. Then play with the circuit for real - forget the simulator for your final decisions.
I could design this for you - except that I am a working technician and simply don't have time. I'm just restarting my own design that I don't have time for.
Temperature rise has everything to do with dissipated power and heatsink fin area * efficiency. If you look at some heat sink vendor web sites, they may have calculators for you. Measure your heat sink, pick a similar one from their stock to get the efficiency number and use their calculator to figure it out. Airflow is key, so is ambient temperature and that will determine inside chassis temperatures. The SA-100 was too high, measured at 50 °C at idle, ambient running about 21 °C (Canada in the winter with the furnace running, or air conditioning in the summer). So you must pay attention to the internal chassis temperatures as well. Counterpoint lost power transformers due to internal case temperatures, they have a thermal fuse. If you place that amplifier in an enclosed space, expect the temperatures to get really high, even open at the front and back.
Your design goal should be to reduce running temperatures. Run at the lowest bias current to give you good sound and low distortion if possible. A great sounding amp that has a blown power transformer doesn't make a lot of noise. But, it makes a great paper weight!