SET amp owners thread

[Atmasphere]

Class D bass sounds so different that I don't think it is a good blend with SET in the upper ranges...just IME...

If you want an amp to sound like an SET it has to have the same distortion signature as an SET. That means a prominent 2nd harmonic with an exponential reduction in amplitude as the harmonic order is increased. Essentially this is a quadratic non-linearity.

SETs actually have more high ordered harmonic content than any other amplifier. But you don't hear that as harshness and brightness because the 2nd harmonic is so profound- so it masks the higher orders.

You can get the same thing if the amp expressed a cubic non-linearity. In this case the 3rd will be most prominent and it is treated the same by the ear as the 2nd (innocuous, adds richness) on account of being so close to the fundamental. But in an amplifier like this, the succeeding harmonics fall off at a faster rate owing to overall less distortion passed from stage to stage in the amplifier. Also in this case the distortion falls off on an exponential curve which is desirable for masking, but falls off at a faster rate.

This type of amplifier will blend with an SET quite easily. There are tube amps and class D amps that have this type of signature FWIW.

 

[213Cobra]

Custom-built PX-25 SET + Zu Definition 4. Sometimes 16-ohms connected for PX-25 SET + Zu Druid 6. -Phil

 

[morricab]

If you want an amp to sound like an SET it has to have the same distortion signature as an SET. That means a prominent 2nd harmonic with an exponential reduction in amplitude as the harmonic order is increased. Essentially this is a quadratic non-linearity.

SETs actually have more high ordered harmonic content than any other amplifier. But you don't hear that as harshness and brightness because the 2nd harmonic is so profound- so it masks the higher orders.

You can get the same thing if the amp expressed a cubic non-linearity. In this case the 3rd will be most prominent and it is treated the same by the ear as the 2nd (innocuous, adds richness) on account of being so close to the fundamental. But in an amplifier like this, the succeeding harmonics fall off at a faster rate owing to overall less distortion passed from stage to stage in the amplifier. Also in this case the distortion falls off on an exponential curve which is desirable for masking, but falls off at a faster rate.

This type of amplifier will blend with an SET quite easily. There are tube amps and class D amps that have this type of signature FWIW.

No, you don't get the same thing...your ears should be able to tell you this easily...

 

[Atmasphere]

No, you don't get the same thing...your ears should be able to tell you this easily...

And they do. I've had SET amps on hand for over 20 years.

There is a persistent myth that we can hear things that are not measurable. What's really happening is the measurements you need to see never get shown and often never get made. Here's what to look for:
1) distortion VS frequency. It should be consistent across the band not rising with frequency
2) the distortion spectra; if distortion is not well below -100dB it will be important to have a prominent 2nd or 3rd harmonic to mask higher orders- see my prior post.

It is important to also understand the significance of the measurements as they relate to human hearing perceptual rules, for example how we sense sound pressure. These rules are invariant from human to human and have nothing to do with taste.

In any amplifier, once FR is out of the equation, the so-called 'sonic signature' is really the distortion signature. Put another way, in any amplifier there is the gain of the amplifier and there is a distortion block, both of which the signal passes through on its way to the speaker.

 

[morricab]

And they do. I've had SET amps on hand for over 20 years.

There is a persistent myth that we can hear things that are not measurable. What's really happening is the measurements you need to see never get shown and often never get made. Here's what to look for:
1) distortion VS frequency. It should be consistent across the band not rising with frequency
2) the distortion spectra; if distortion is not well below -100dB it will be important to have a prominent 2nd or 3rd harmonic to mask higher orders- see my prior post.

It is important to also understand the significance of the measurements as they relate to human hearing perceptual rules, for example how we sense sound pressure. These rules are invariant from human to human and have nothing to do with taste.

In any amplifier, once FR is out of the equation, the so-called 'sonic signature' is really the distortion signature. Put another way, in any amplifier there is the gain of the amplifier and there is a distortion block, both of which the signal passes through on its way to the speaker.

When you remove the 2nd/even orders you damage the sound...the pattern matters to brain expectations. All or mostly odd orders doesn’t really exist in nature.

 

[Atmasphere]

When you remove the 2nd/even orders you damage the sound...the pattern matters to brain expectations. All or mostly odd orders doesn’t really exist in nature.

To the latter: agreed! To the former: where does that happen??

Comment aside for a moment, its important to understand that the ear treats the 3rd the same as the second since it is so close to the fundamental tone. In a circuit that expresses a 3rd as the primary distortion product, an interesting thing happens: succeeding harmonics drop off at a faster rate, allowing the 3rd to mask them much more easily than seen in a circuit that expresses the 2nd.

This happens because distortion is not compounded as much from stage to stage as occurs in a single-ended circuit. The brain expectations to which you refer were mentioned by Olsen back in the 1950s or 60s- he stipulated that succeeding harmonics must fall off exponentially. This requirement is satisfied if the amplifier cancels even ordered harmonics. Where tube amps get into trouble in this regard is when single-ended input circuits are combined with PP outputs- due to algebraic summing the 5th harmonic tends to show up in larger amounts. Norman Crowhurst wrote of this 60 years ago.

Since your comment is written the way it is I feel compelled to point out that no amplifier made is able to 'remove the 2nd/even orders'; they are only ever able to add less of them.

 

[morricab]

To the latter: agreed! To the former: where does that happen??

Comment aside for a moment, its important to understand that the ear treats the 3rd the same as the second since it is so close to the fundamental tone. In a circuit that expresses a 3rd as the primary distortion product, an interesting thing happens: succeeding harmonics drop off at a faster rate, allowing the 3rd to mask them much more easily than seen in a circuit that expresses the 2nd.

This happens because distortion is not compounded as much from stage to stage as occurs in a single-ended circuit. The brain expectations to which you refer were mentioned by Olsen back in the 1950s or 60s- he stipulated that succeeding harmonics must fall off exponentially. This requirement is satisfied if the amplifier cancels even ordered harmonics. Where tube amps get into trouble in this regard is when single-ended input circuits are combined with PP outputs- due to algebraic summing the 5th harmonic tends to show up in larger amounts. Norman Crowhurst wrote of this 60 years ago.

Since your comment is written the way it is I feel compelled to point out that no amplifier made is able to 'remove the 2nd/even orders'; they are only ever able to add less of them.

Well balanced push pull by definition cancels out even orders...

 

[Atmasphere]

Well balanced push pull by definition cancels out even orders...

It does, but does not do what you say here:

remove the 2nd/even orders

I don't know of an amplifier that does that. So I'm assuming that you meant the first quote in this post when you wrote the 2nd quote in this post.

In which case, you have to be careful about whether the amp is fully differential or merely has a PP output; those are two very different animals! Also the issue of feedback and what kind of tubes has to be considered so the playing field is level. You can't compare a circuit like a Dynaco ST-70 to an SET and get anything meaningful out of it; if you used a PP amp that had DHTs of the same type as the SET in question that would be a good start. And the PP amp would have to be class A and built of the same part quality as used in the SET, with output transformers of similar quality but designed for PP operation. That's a lot of variables; I've not read any articles where anyone has done that sort of thing. I've done a bit of that on my own to satisfy my curiosity, but other than anecdotal I didn't document it.

At any rate if the PP amp is also fully differential, class A and zero feedback, it will have an exponential decay of higher ordered harmonics, but the exponent in the exponential function will cause the higher orders to fall off at a faster rate. So such an amplifier would actually be just as smooth and musical as an SET, while having considerably less distortion since it would be capable of nearly 4 times the power and still only 1/10th to 1/3rd the distortion at full power. The reduced distortion manifests as greater detail.

 

[morricab]

It does, but does not do what you say here:

I don't know of an amplifier that does that. So I'm assuming that you meant the first quote in this post when you wrote the 2nd quote in this post.

In which case, you have to be careful about whether the amp is fully differential or merely has a PP output; those are two very different animals! Also the issue of feedback and what kind of tubes has to be considered so the playing field is level. You can't compare a circuit like a Dynaco ST-70 to an SET and get anything meaningful out of it; if you used a PP amp that had DHTs of the same type as the SET in question that would be a good start. And the PP amp would have to be class A and built of the same part quality as used in the SET, with output transformers of similar quality but designed for PP operation. That's a lot of variables; I've not read any articles where anyone has done that sort of thing. I've done a bit of that on my own to satisfy my curiosity, but other than anecdotal I didn't document it.

At any rate if the PP amp is also fully differential, class A and zero feedback, it will have an exponential decay of higher ordered harmonics, but the exponent in the exponential function will cause the higher orders to fall off at a faster rate. So such an amplifier would actually be just as smooth and musical as an SET, while having considerably less distortion since it would be capable of nearly 4 times the power and still only 1/10th to 1/3rd the distortion at full power. The reduced distortion manifests as greater detail.

You are mincing words, Ralph. Cancellation of even order harmonics is what push/pull does. Cancellation = removal of those harmonics. If the two halves of the push/pull are not perfectly matched then you will get some even order creeping in.

Krell Full Power Balanced 350mc monoblock amplifier Measurements part 2

Fig.4 reveals that at low levels, this distortion is primarily third-harmonic. (The amplitude of the lower trace in this graph is very much exaggerated for clarity; its absolute level was below 0.005%.) At very high levels (fig.5), the third harmonic is joined by a regularly decreasing series of...

www.stereophile.com

This Krell demonstrates this nicely...please note that the 50Hz FFT shows very little even harmonic (it is never completely zero). This is fundamental to a complementary push/pull design (here executed just about perfectly). You will also note that while the odd harmonics reduce as the order increases, it is not an exponential decay.

Constellation Inspiration Stereo 1.0 power amplifier Measurements

Sidebar 3: Measurements

www.stereophile.com

Another example...again, some even orders but dominated by odd orders...this is fundamental to push/pull.

The third harmonic here will not significantly mask the substantial higher order odd harmonics.

 

[LL21]

Hi Morricab,

I am 'trying' to follow this, but if I may ask, I would love to know if there is a layperson's explanation of what Robert Koda (and Constellation Hercules series...where you reference Inspiration Series above) are doing...are they specifically addressing the push-pull issue that you are describing? What is the implication of what Constellation and Robert Koda are doing below?

1. Description of Constellation Hercules from Alan Sircom: https://hifiplus.com/articles/constellation-audio-hercules-ii-stereo-power-amplifier/

"...This is an amplifier that uses all of the strengths of a balanced amplifier (such as a low noise floor, greater dynamic range, and an ability to use long cable without significantly changing the sound), and combines them with the refined and sophisticated sound of a single-ended design. Instead of the usual arrangement of P-type (Positive-Negative-Positive) transistors handling the positive rail of the balanced line, and N-type (Negative-Positive-Negative) transistors handling the negative rail, the Hercules II modules are two single-ended amplifier designs, ending in N-type transistors in the output stage. This design has significant advantages over other circuits, because N-type and P-type transistors behave very differently, and that difference undermines the sound quality.

Constellation’s amplifier design has an additional advantage to the circuit: it makes it inherently stable into all kinds of loudspeaker loads...."

2. I note that Robert Koda also seems to be a design (which similar to above) that enables its power supply to always see itself as driving an easy, static load (ie. not 'seeing the speaker load or music signal') which then is devoid of dependence on music content or loudspeaker interaction.

 

[morricab]

Hi Morricab,

I am 'trying' to follow this, but if I may ask, I would love to know if there is a layperson's explanation of what Robert Koda (and Constellation Hercules series...where you reference Inspiration Series above) are doing...are they specifically addressing the push-pull issue that you are describing? What is the implication of what Constellation and Robert Koda are doing below?

1. Description of Constellation Hercules from Alan Sircom: https://hifiplus.com/articles/constellation-audio-hercules-ii-stereo-power-amplifier/

"...This is an amplifier that uses all of the strengths of a balanced amplifier (such as a low noise floor, greater dynamic range, and an ability to use long cable without significantly changing the sound), and combines them with the refined and sophisticated sound of a single-ended design. Instead of the usual arrangement of P-type (Positive-Negative-Positive) transistors handling the positive rail of the balanced line, and N-type (Negative-Positive-Negative) transistors handling the negative rail, the Hercules II modules are two single-ended amplifier designs, ending in N-type transistors in the output stage. This design has significant advantages over other circuits, because N-type and P-type transistors behave very differently, and that difference undermines the sound quality.

Constellation’s amplifier design has an additional advantage to the circuit: it makes it inherently stable into all kinds of loudspeaker loads...."

2. I note that Robert Koda also seems to be a design (which similar to above) that enables its power supply to always see itself as driving an easy, static load (ie. not 'seeing the speaker load or music signal') which then is devoid of dependence on music content or loudspeaker interaction.
View attachment 89659

BAT was one of the first to use all N-type transistors. All of their VK SS amps use only N-type transistors. Apparently, due to differences in the behavior of the two types of transistors (particularly their turn off time) there is difficulty in getting a match at the zero crossing. Complementary means using P and N types as they "complement" each other but they are not truly equivalent. This problem doesn't afflict tube amps because tubes don't have P and N types. The designers consider N-types to be better behaved and so anytime a push-pull design is done with one type of transistor they always choose N types. So, there is nothing totally new here but the argument can be made that you get less distortion at zero crossing for an amp that is not truly Class A (for Class A I am not sure that it matters as the transistors never totally shut off).

HOWEVER, this does not mean these amps are single ended...they are not. There is still a handoff between two (or more) transistors for the positive and negative halves of the waveform. The same harmonic distortion patterns are created as this has a lot (but not all) to do with the topology of the circuit. A symmetric circuit cancels out even order harmonics and an asymmetric circuit will have all harmonics but in a descending level with increasing order. The better matched the two halves of a push-pull circuit, the more cancellation of even orders will occur.

It MIGHT give a somewhat more tube like sound as the design concept is more tube amp like in design (because, as I said there are no P and N type tubes...a tube is a tube) and the zero crossing improvement might reduce need for negative feedback (or possibly eliminate it?) or other error correction. I think you would probably agree that your RK amps don't really sound like tubes but they might also not sound quite like a standard PP SS amp. I had a BAT VK200 back in the day that had an all N-type design and it was rich and warm sounding but also a bit dark overall tonality...it was clear there was a desire to get a somewhat tube like sound. However, to my ears it was a tube like sound that I would not really want (more of a stereotype of tube sound and not the tube sound of a really excellent tube amp).

The best sounding amp I have ever had that contained transistors was a single ended one. It was actually a hybrid where it was a tube input, tube driver and a single large MOSFET on the output stage. It was huge, it was hot and it sounded amazing after playing music for about 2 hours straight. Not sure why it took so long to come on song but that is why I sold it ultimately. That said...it still did not sound like an all tube SET...it was something different again...not surprising given the output device was a MOSFET with a totally different distortion character from a triode.

 

[cal3713]

Hi Morricab,

I am 'trying' to follow this, but if I may ask, I would love to know if there is a layperson's explanation of what Robert Koda (and Constellation Hercules series...where you reference Inspiration Series above) are doing...are they specifically addressing the push-pull issue that you are describing? What is the implication of what Constellation and Robert Koda are doing below?

1. Description of Constellation Hercules from Alan Sircom: https://hifiplus.com/articles/constellation-audio-hercules-ii-stereo-power-amplifier/

"...This is an amplifier that uses all of the strengths of a balanced amplifier (such as a low noise floor, greater dynamic range, and an ability to use long cable without significantly changing the sound), and combines them with the refined and sophisticated sound of a single-ended design. Instead of the usual arrangement of P-type (Positive-Negative-Positive) transistors handling the positive rail of the balanced line, and N-type (Negative-Positive-Negative) transistors handling the negative rail, the Hercules II modules are two single-ended amplifier designs, ending in N-type transistors in the output stage. This design has significant advantages over other circuits, because N-type and P-type transistors behave very differently, and that difference undermines the sound quality.

Constellation’s amplifier design has an additional advantage to the circuit: it makes it inherently stable into all kinds of loudspeaker loads...."

2. I note that Robert Koda also seems to be a design (which similar to above) that enables its power supply to always see itself as driving an easy, static load (ie. not 'seeing the speaker load or music signal') which then is devoid of dependence on music content or loudspeaker interaction.
View attachment 89659

I think in both cases you've got a little truth combined with a bunch of hocus pocus quantum marketing speak. @morricab does a good job with #1, and as for #2... I'm sure someone with more technical knowledge can find the truth in the statement, but the amplifier will always be affected by loudspeaker load.

 

[LL21]

Thank you both! Very interesting reading...am re-reading as a non-techie. As for sound, the sound shares some traits of the CJ GAT 2/Gryphon Mephisto combination with some notable differences
- it carries on the CJ GAT 2s approach to a tonally very even band and evens it out notably further all the way through the bass
- much quieter, much more filigreed, but also notably more powerful (but not massively more) presentation where the Gryphon is happy to show off its dominant control (which I think is fantastic and sets a great foundation for music)...
- I think I have distilled it down to the observation that the Koda pairing is more unassuming. Much more unassuming.

- You dont really know it is there...until here is a particularly complex passage or something like a massive Kodo drum hammer strike, where the Koda surprises you with a more thunderous and more electric charge on bass impact and reverb. Even then you still dont actually notice anything because the RK is so unassuming. The only reason you do notice something is because it is the realization of how nonchalant a passage seems which came across as stressfully complex or dynamic on other amplification.

 

[morricab]

I think in both cases you've got a little truth combined with a bunch of hocus pocus quantum marketing speak. @morricab does a good job with #1, and as for #2... I'm sure someone with more technical knowledge can find the truth in the statement, but the amplifier will always be affected by loudspeaker load.

An amp with no negative feedback around the output stage will be relatively immune to the reactivity of the speaker load but of course this won't mean it doesn't need more or less current depending on the load at a given frequency.

There was interesting research into the interactivity between the amp and the speaker load by Matti Otala (not sure if I spelled that right) looking at how a reactive speaker load can have a significant impact on the distortion an amp makes when it has a global negative feedback loop with a high dB of feedback. The back EMF from the speaker injects signal back into the output stage of the amplifier and if there is a feedback loop then some of that return signal is put right back into the amplification pathway with the new incoming signal. Some speakers, like electrostats, send almost all of the energy put into them back into the output stage of the amp but it is garbled and distorted by the filters and/or panel capacitance. Same with complex crossovers...the signal that gets pumped back into the amp is nothing like what went out and a small part of that gets reamplified.

It was speculated that this could be why some amps will sound great on one type of speaker but horrible on another and a lot of the so-called 'synergy' is related to this phenomenon. An interesting idea but I am not sure they were able to prove this in a significant way.

 

[Atmasphere]

You are mincing words, Ralph. Cancellation of even order harmonics is what push/pull does. Cancellation = removal of those harmonics. If the two halves of the push/pull are not perfectly matched then you will get some even order creeping in.

Krell Full Power Balanced 350mc monoblock amplifier Measurements part 2

Fig.4 reveals that at low levels, this distortion is primarily third-harmonic. (The amplitude of the lower trace in this graph is very much exaggerated for clarity; its absolute level was below 0.005%.) At very high levels (fig.5), the third harmonic is joined by a regularly decreasing series of...

www.stereophile.com

This Krell demonstrates this nicely...please note that the 50Hz FFT shows very little even harmonic (it is never completely zero). This is fundamental to a complementary push/pull design (here executed just about perfectly). You will also note that while the odd harmonics reduce as the order increases, it is not an exponential decay.

Constellation Inspiration Stereo 1.0 power amplifier Measurements

Sidebar 3: Measurements

www.stereophile.com

Another example...again, some even orders but dominated by odd orders...this is fundamental to push/pull.

The third harmonic here will not significantly mask the substantial higher order odd harmonics.

I'm not mincing words. Your statement 'Cancellation = removal of those harmonics' is problematic. No amp 'removes' harmonics from the signal (I concede its possible that this interpretation isn't what you meant). If you want to state this more clearly, its correct to say that 'in a PP circuit, even ordered harmonic distortion is cancelled in the load'.

A cautionary point must be raised! I've pointed out that in order for the PP circuit to 2

Your example of the Krell is apples to oranges. The Krell lacks the Gain Bandwidth Product to support its feedback and so at lower frequencies it suppresses distortion fairly well. But as frequency goes down the feedback goes down with it. So higher orders are not suppressed so well so the 3rd is unable to mask them. 80dB down isn't enough, and also because the feedback is adding higher orders of its own we don't see the exponential decay which Olsen described as so essential (regarding SETs in particular). As a result this amp sounds nothing like an SET (no duh) nor meets the criteria I outlined.

Clearly the Constellation has a similar problem.

An amp with no negative feedback around the output stage will be relatively immune to the reactivity of the speaker load

This statement is false. Feedback is known to reduce output impedance via a servo effect (control theory, for those in the know following along; its weird how control theory is so well-known outside of audio but in the audio world it is apparently mysterious). If there is no feedback the output section will react more to the load, not less. Of course there is a question of what is meant by 'more'. 99% of the time 'more' means that the output voltage will be more variable.

But if the speaker doesn't subscribe to the Voltage Paradigm then an amplifier that behaves as a voltage source (IOW the output voltage does not change as the load does; this is an amp that either doubles power as the load impedance is halved or in the case of tube amps halves its output power as the load impedance is doubled) won't sound right on that speaker. ESLs are a classic example, since their impedance curve isn't based on a driver in a box with its attendant resonance. So it was serendipitously a poor choice on the part of Matti Otala to use ESLs since they are not a Voltage Paradigm device.

It was speculated that this could be why some amps will sound great on one type of speaker but horrible on another and a lot of the so-called 'synergy' is related to this phenomenon. An interesting idea but I am not sure they were able to prove this in a significant way.

I've dropped this link plenty of times; the speculation can end. This is why most amps really won't play an ESL properly and also why a certain number of speakers that SET users like to play won't work so well with higher power amps with feedback as well:
http://www.atma-sphere.com/en/resources-paradigms-in-amplifier-design.html
The Sound Lab is the only ESL I know that properly addresses the issue of using a voltage source to drive their speakers. The speaker is supplied with adjustments for the bass output and also has a brilliance control. These are provided to allow the speaker to be adjusted to the voltage response of the amplifier which will be quite different depending on feedback!

 

[DaveC]

Class D bass sounds so different that I don't think it is a good blend with SET in the upper ranges...just IME...

If you want an amp to sound like an SET it has to have the same distortion signature as an SET. That means a prominent 2nd harmonic with an exponential reduction in amplitude as the harmonic order is increased. Essentially this is a quadratic non-linearity.

SETs actually have more high ordered harmonic content than any other amplifier. But you don't hear that as harshness and brightness because the 2nd harmonic is so profound- so it masks the higher orders.

You can get the same thing if the amp expressed a cubic non-linearity. In this case the 3rd will be most prominent and it is treated the same by the ear as the 2nd (innocuous, adds richness) on account of being so close to the fundamental. But in an amplifier like this, the succeeding harmonics fall off at a faster rate owing to overall less distortion passed from stage to stage in the amplifier. Also in this case the distortion falls off on an exponential curve which is desirable for masking, but falls off at a faster rate.

This type of amplifier will blend with an SET quite easily. There are tube amps and class D amps that have this type of signature FWIW.

Brad, you keep saying this, but it makes it clear you haven't tried many class D amps. They don't all sound the same, not even close. Some are indeed a poor match with my SET amp but others sound very similar. Ralph's explanation here is very good, we may not agree on everything but on this topic he's correct imo.

 

[cal3713]

I'm not mincing words. Your statement 'Cancellation = removal of those harmonics' is problematic. No amp 'removes' harmonics from the signal (I concede its possible that this interpretation isn't what you meant). If you want to state this more clearly, its correct to say that 'in a PP circuit, even ordered harmonic distortion is cancelled in the load'.

Ralph. Are you suggesting that the intention was to argue that the original signal was modified to remove harmonic information?

I cannot imagine that this was Brad's intention as that would require some bizarre computer chip based algorithm akin to the 4k video resampling techniques that eliminate visual distortion from vintage footage.

 

[morricab]

Brad, you keep saying this, but it makes it clear you haven't tried many class D amps. They don't all sound the same, not even close. Some are indeed a poor match with my SET amp but others sound very similar. Ralph's explanation here is very good, we may not agree on everything but on this topic he's correct imo.

It's interesting how arrogance is manifested in phrases like "It's clear you haven't tried many class D amps"... as if you have any freaking clue how many ClassD amps I have heard or owned. I have owned or had at home at length 4 different ones and have friends and colleagues with them or had them as well. Not to mention dozens of demos at shows and shops over the years. My opinions are based on this experience and if it is too limited, well whose experience is universal?? I have, though heard implementations of all the major class D players, if not all examples of all tech. If you have found ones that give a good match with SET then great, simply state that and which one so that others can consider it. Trying to slap me with the " I lack experience" card is not constructive and frankly false.

 

[morricab]

Ralph. Are you suggesting that the intention was to argue that the original signal was modified to remove harmonic information?

I cannot imagine that this was Brad's intention as that would require some bizarre computer chip based algorithm akin to the 4k video resampling techniques that eliminate visual distortion from vintage footage.

This was clearly not what I meant.

I'm not mincing words. Your statement 'Cancellation = removal of those harmonics' is problematic. No amp 'removes' harmonics from the signal (I concede its possible that this interpretation isn't what you meant). If you want to state this more clearly, its correct to say that 'in a PP circuit, even ordered harmonic distortion is cancelled in the load'.

A cautionary point must be raised! I've pointed out that in order for the PP circuit to 2

Your example of the Krell is apples to oranges. The Krell lacks the Gain Bandwidth Product to support its feedback and so at lower frequencies it suppresses distortion fairly well. But as frequency goes down the feedback goes down with it. So higher orders are not suppressed so well so the 3rd is unable to mask them. 80dB down isn't enough, and also because the feedback is adding higher orders of its own we don't see the exponential decay which Olsen described as so essential (regarding SETs in particular). As a result this amp sounds nothing like an SET (no duh) nor meets the criteria I outlined.

Clearly the Constellation has a similar problem.

This statement is false. Feedback is known to reduce output impedance via a servo effect (control theory, for those in the know following along; its weird how control theory is so well-known outside of audio but in the audio world it is apparently mysterious). If there is no feedback the output section will react more to the load, not less. Of course there is a question of what is meant by 'more'. 99% of the time 'more' means that the output voltage will be more variable.

But if the speaker doesn't subscribe to the Voltage Paradigm then an amplifier that behaves as a voltage source (IOW the output voltage does not change as the load does; this is an amp that either doubles power as the load impedance is halved or in the case of tube amps halves its output power as the load impedance is doubled) won't sound right on that speaker. ESLs are a classic example, since their impedance curve isn't based on a driver in a box with its attendant resonance. So it was serendipitously a poor choice on the part of Matti Otala to use ESLs since they are not a Voltage Paradigm device.

I've dropped this link plenty of times; the speculation can end. This is why most amps really won't play an ESL properly and also why a certain number of speakers that SET users like to play won't work so well with higher power amps with feedback as well:
http://www.atma-sphere.com/en/resources-paradigms-in-amplifier-design.html
The Sound Lab is the only ESL I know that properly addresses the issue of using a voltage source to drive their speakers. The speaker is supplied with adjustments for the bass output and also has a brilliance control. These are provided to allow the speaker to be adjusted to the voltage response of the amplifier which will be quite different depending on feedback!

Push–pull output - Wikipedia

en.wikipedia.org

"Symmetrical construction of the two sides of the amplifier means that even-order harmonics are cancelled, which can reduce distortion.[1]"

"Symmetrical push–pull
Each half of the output pair "mirror" the other, in that an NPN (or N-Channel FET) device in one half will be matched by a PNP (or P-Channel FET) in the other. This type of arrangement tends to give lower distortion than quasi-symmetric stages because even harmonics are cancelled more effectively with greater symmetry."

What is clear from these descriptions (and others in the Wiki page) is that even order distortion is created in each half of the push pull stage due to the inherent behavior of the devices (Mosfets and bipolars will both make all harmonics as single ended devices...at least when not biased Class A...see paper by Boyk and Sussmann). This even order distortion due to the devices is cancelled by the symmetry of the circuit design ...therefore what I said about the distortion being removed by the circuit is essentially correct. It is being generated by the inherent qualities of the devices and cancelled (i.e removed) by the symmetry of the circuit and having those out of phase. The better the symmetry, the greater the cancellation.

Yes the feedback tends to accentuate the higher orders but it won't accentuate orders that have been cancelled by the push/pull design. So, odd high order harmonics (the worst kind) will be added and enhanced.

With regard to the Krell: https://www.stereophile.com/content/krell-full-power-balanced-350mc-monoblock-amplifier-measurements

As you can see it is only 1 db down at 100Khz and the distortion vs. frequency is quite flat...so this doesn't fit with your argument that the Krell lacks the Gain Bandwidth Product to support its feedback. There are many feedback amps that have a rising distortion with increasing frequency (some alarmingly so) and those I would agree with you meet your argument and that probably leads to problems in the highs and overall tonality. The Krell though, measurably doesn't have this issue. I have actually heard this amp many years ago with a pair of Wilson CUB speakers and although I don't like the overall tonality and "feel" I got from the amp I have to admit it's highs were very clean and the soundstage was very deep and wide with a superbly low noise floor. You can see this also with the high frequency IMD measurement which shows virtually nothing.

My point about feedback and interaction is different than what you are talking about. I am aware of the reduced output impedance, which means less variability for a voltage source and a speaker that likes that. I am talking about what comes back from the speaker, the so-called back EMF, and it's interaction with the amplifier circuit if global feedback is used. Have you read the papers form Otala? It is not about frequency response of the speaker it is about distortion products from the amplifier.

 

[morricab]

Ralph. Are you suggesting that the intention was to argue that the original signal was modified to remove harmonic information?

I cannot imagine that this was Brad's intention as that would require some bizarre computer chip based algorithm akin to the 4k video resampling techniques that eliminate visual distortion from vintage footage.

I was definitely not talking about removal of information from the signal...I am talking about distortion that is amp generated and the cancelllation of that distortion from the symmetry of the circuitry. All devices make both even and odd harmonics (theoretically the FETs would only make even harmonics single ended in Class A...but in reality they are not perfect and make both even and odd) as single devices. It is only through the push/pull in opposite phases that you get the cancellation of asymmetric (even) distortions. So, I am talking about cancellation or removal of SELF-generated distortions not signal related harmonics...although I am not sure if anyone has proven that some signal is also not lost and/or modified by this process.