Is there an official definition of Distortion? Should it be harnessed or avoided?

[caesar]

What exactly is distortion?

Some say vinyl and tubes are "distorted", but many love them none the less. Do the best engineers understand this reality and design products to harness the distortion? Or do the best engineers take out the distortion?

 

[Atmasphere]

^^ Some of this is true. Most digital engineers are still struggling with in-harmonic distortions, those that arise that are related to the scan frequency rather than the fundamental.

SET amplifier designers understand that the ear cares very little about lower ordered harmonics, in terms of objectionability. This is true but the ear does hear all harmonic distortion as tonality- this is why SETs sound 'rich'. Its also why they sound so dynamic for their low power.

Transistor amp designers have to be quite focused on getting low distortion, as transistors often tend to have poor linearity. So many of them use negative feedback to acheive linearity. There is a problem with this- the ear uses odd ordered harmonics to detirmine how loud a sound is. When you add negative feedback to an amplifer, which lowers THD, it can add to the odd orders. Those pesky 5th, 7th and 9th harmonics are the troublemakers. Since our ears use them to detirmine sound pressure, our ears are more sensitive to them than the best test equipment. This is why transistor amps can sound bright and harsh, even though they have flat frequency response. Its not because there is a frequency response abberation, its because of odd ordered harmonics.

This is why tubes failed to go obsolete 50 years ago- they sound more pleasent because they make less odd ordered harmonics. But if you want tubes to sound neutral, you have to get rid of the distortion they *do* make which is mostly the 2nd 3rd and 4th. The only way I have found to do this is to build the tube circuits fully differentially, which cancels out all the even orders throughout the entire circuit. And don't use feedback.

The problem with no feedback of course is that its very hard to get low 'output impedance', especially if you want the amp to behave like a voltage source. But behaving like a voltage source does not guarantee that it will sound like music, so we opted for a different approach, which is that if the speaker demands that of the amp, the chances are very high that it will thus never be able to sound like real music.

Overall audio engineering is not as well thought out, cut and dried as we are often lead to believe, simply because the human hearing rules that are emerging due to ongoing research have yet to be exploited. But if past history is any indication, it won't be happening anytime soon. We have known how distortion affects tonality and that the ear uses odd orders to detect volume levels since the 1960s, but as an industry have not done anything about that since then other than argue. So don't hold your breath!

 

[Steve Williams]

Great explanation

Welcome back Ralph. Good to see you in our neck of the woods

 

[opus111]

Transistor amp designers have to be quite focused on getting low distortion, as transistors often tend to have poor linearity. So many of them use negative feedback to acheive linearity. There is a problem with this- the ear uses odd ordered harmonics to detirmine how loud a sound is. When you add negative feedback to an amplifer, which lowers THD, it can add to the odd orders. Those pesky 5th, 7th and 9th harmonics are the troublemakers.

I've seen the math for this and its only for a limited range of amounts of feedback that the higher harmonics increase. I think there's a paper by Bruno Putzeys where he covers this, but he didn't originate the math. In my understanding most SS amps use more than 20dB of feedback and if I remember correctly that's about the amount where the higher harmonics start going down again.

Since our ears use them to detirmine sound pressure, our ears are more sensitive to them than the best test equipment. This is why transistor amps can sound bright and harsh, even though they have flat frequency response. Its not because there is a frequency response abberation, its because of odd ordered harmonics.

I haven't found this to be so in my experience. I start out with SS amps which sound bright and harsh, modify them for appropriate noise control and they then sound smooth and non-fatiguing. I doubt very much that their harmonic distortion spectrum has changed much, although I haven't measured. The harshness I associated with transistors (particularly low-noise bipolar opamps, the LM4562 is one of the worst) is I believe related to the input stage's sensitivity to low-level RF signals. Deft use of ferrite beads, common-mode chokes, attention to PSU cleanliness and star grounding removes it to my satisfaction. The LM4562 has practically immeasurable distortion according to its datasheet yet still sounds harsh and fatiguing without these kinds of countermeasures being applied.

<edit> I realized I'd like to add that in my estimation the reason that SS amps are harsh and fatiguing is that's the sound that customers want. We discussed this on the recent RMAF thread - people who want smooth sound buy valve amps, in general. I'm something of an exception, preferring the low maintenance, weight and better energy efficiency of transistors so I design my own stuff.

 

[treitz3]

Hello, Atmasphere. Welcome back and thank you for such a well educated and informed post. I do have a question for you, if you don't mind. I do personally avoid tubes in the amplification part [preference] but I do love tubes in other parts of the system. I.E. a tubed vinyl pre, a tubed CDP and tubes in line with other digital sources. How might this type of combination relate with what you mentioned?

I harness and avoid.

Tom

 

[opus111]

in digital audio, engineers seem to be pursuing eliminating distortions (that would be the integrated circuit engineers)).

Its true - that's why most DACs designed latterly suck in terms of SQ. Because they optimized low distortion and forgot about another parameter - noise modulation. That's because I think there's agreement about how to measure THD+N, no agreement about how to measure noise modulation. Its not that noise modulation isn't known about - Dolby I rather suspect know quite a lot about it and optimize their systems taking it into account. I also suspect that in part its their understanding and appreciation of noise modulation that allowed them to become dominant in the NR field. That's not to minimize Ray Dolby's phenomenal marketing abilities mind.

high end audio engineers (some) claim that only discrete circuits can be the most distortion free, yet after reducing said distortions fairly low, they shift bias point on a transisitor etc to generate more distortions, ie voiceing the amp or circuit.

I suspect that this is wide of the mark as high end sells on sound (reviewer's appreciation of it) not on distortion. I've not heard a high end designer say that discrete offers lower distortion, rather they tend to say that it sounds better. I have a handle I think on why, see my earlier post than this one.

the high end press almost to a person abhors low distortion, it always sounds cold, sterile, etc. because, I think, if they felt that was the answer, there could only really be one best low distortion component, and the rest losers, ans so you write about your one best unit for the year and all the rest are the losers.....

I rather get the impression here that you're confusing causation with correlation. In general engineers optimizing for distortion won't listen. Those who listen notice that the lowest measured distortion is no longer so necessary and focus on other aspects.

 

[Soundminded]

In an audio amplifier or preamplifier any change to the shape of the electrial waveform between the input and output is distortion. All those devices are supposed to do is make the signal larger, more powerful. Some distortions are audible, some aren't. It's important to sort them out. Addressing reducing distortions you can't hear is a waste of time, money, and effort. It makes for better advertising copy than engineering. There are thresholds below which distortions are not audible. One fallacy audiophiles fall for is not recognizing this. This makes them prey to ads claiming reduction in distortions that can't be heard. It's virtually useless. Wires are supposed to do the same thing but without any change in the signal size. One problem in testing distortion is performing the test under conditions the equipment will actually be used in. Another is testing for all forms of distortion. That's a real shortcoming of many of the products audiophiles buy. An amplifier's frequency response at one watt into a 4 or 8 ohm resistor won't tell you anything about what its frequency response will be feeding 50 watts into a speaker system whose impedance dips below one ohm and produces strong reverse emf. This is why equipment that tests identically on a test bench can sound very different in use.

There are other devices that are used to deliberately alter electrical signals in specific controllable ways. For example, the alteration to the waveform produced by an equalizer is intended to be the compliment or inverse of distortions that exist elsewhere. This is usually unavoidable, it's inherent in the limitations of the technology. For example, analog tape recordings and LPs could not exist without altering the signals fed to the recording equpment and altering them again on playback. Various forms of equalization, filtering (including crossover networks), digital signal processing, and dynamic compression and expansion (Dolby nose reduction for example) are deliberate changes to waveforms we do not customarily thing of as distortion.

Testing speakers for distortion is much harder. A speaker's job is to recreate a sound field at the listener's ears which is subjectively identical to an event the listener most likely never attended. Even memory does not serve well, you can't remember something you never experienced. And so people who have much experience listening to live music frequently have a better recollection of similar sounds. The analysis of sound fields is much more complex than the analysis of electrical signals. Unlike electrical signals they exist in three dimensions in space and interact in time and frequency with the environment. As of the current state of the art, the spatial, temporal, and spectral distortions compared to similar sound fields produced by live music is nothing short of horrendous for all speaker systems and arrays of speaker systems in a home listening environment. The key to better high fidelity will come from scientists and engineers who are qualified spending much more time studying and understanding sound than tinkering with electronics. That's the only way they can define design goals that will have meaning in terms of what we can hear.

 

[Raffles]

distortion is any change to the signal that is not directly linear in time and level.

Isn't this the nub of the question? From the definition above, I believe that systems which change phase and/or frequency response can still be counted as zero distortion i.e. a sine wave fed into them still comes out as a sine wave, and if the amplitude doubles at the input, it does at the output. But feed in a square wave, and it can come out unrecognisable (to the eye); the assumption is, however, that it still sounds like a square wave to the ear. Maybe. But does a transient, like the start of a piano note or a drum, still sound the same when processed by this system?

 

[Phelonious Ponk]

In an audio amplifier or preamplifier any change to the shape of the electrial waveform between the input and output is distortion. All those devices are supposed to do is make the signal larger, more powerful. Some distortions are audible, some aren't. It's important to sort them out. Addressing reducing distortions you can't hear is a waste of time, money, and effort. It makes for better advertising copy than engineering. There are thresholds below which distortions are not audible. One fallacy audiophiles fall for is not recognizing this. This makes them prey to ads claiming reduction in distortions that can't be heard. It's virtually useless. Wires are supposed to do the same thing but without any change in the signal size. One problem in testing distortion is performing the test under conditions the equipment will actually be used in. Another is testing for all forms of distortion. That's a real shortcoming of many of the products audiophiles buy. An amplifier's frequency response at one watt into a 4 or 8 ohm resistor won't tell you anything about what its frequency response will be feeding 50 watts into a speaker system whose impedance dips below one ohm and produces strong reverse emf. This is why equipment that tests identically on a test bench can sound very different in use.

There are other devices that are used to deliberately alter electrical signals in specific controllable ways. For example, the alteration to the waveform produced by an equalizer is intended to be the compliment or inverse of distortions that exist elsewhere. This is usually unavoidable, it's inherent in the limitations of the technology. For example, analog tape recordings and LPs could not exist without altering the signals fed to the recording equpment and altering them again on playback. Various forms of equalization, filtering (including crossover networks), digital signal processing, and dynamic compression and expansion (Dolby nose reduction for example) are deliberate changes to waveforms we do not customarily thing of as distortion.

Testing speakers for distortion is much harder. A speaker's job is to recreate a sound field at the listener's ears which is subjectively identical to an event the listener most likely never attended. Even memory does not serve well, you can't remember something you never experienced. And so people who have much experience listening to live music frequently have a better recollection of similar sounds. The analysis of sound fields is much more complex than the analysis of electrical signals. Unlike electrical signals they exist in three dimensions in space and interact in time and frequency with the environment. As of the current state of the art, the spatial, temporal, and spectral distortions compared to similar sound fields produced by live music is nothing short of horrendous for all speaker systems and arrays of speaker systems in a home listening environment. The key to better high fidelity will come from scientists and engineers who are qualified spending much more time studying and understanding sound than tinkering with electronics. That's the only way they can define design goals that will have meaning in terms of what we can hear.

This is a refreshingly agenda-free answer to the actual question. I'd only add that yes, it should be avoided and/or harnessed according to your tastes. It can't be avoided anyway, you may as well use it. Personally I don't hear the harshness everyone talks about in solid state. I have heard it, plenty of it, but i don't hear it in my current system, and don't hear it in most good quality SS systems unless they are being pushed too hard (though I might add that they are pushed too hard a lot). What I hear, instead, is clarity and definition. I do not want my audio "smoothed," but YMMV. There is an incredible variety of sound available, especially in speakers. Buy what you like and enjoy it. Or keep buying/selling/trading/"upgrading" if that's your bliss, and enjoy that.

Feeling rather mellow this morning....

Tim

 

[Soundminded]

Isn't this the nub of the question? From the definition above, I believe that systems which change phase and/or frequency response can still be counted as zero distortion i.e. a sine wave fed into them still comes out as a sine wave, and if the amplitude doubles at the input, it does at the output. But feed in a square wave, and it can come out unrecognisable (to the eye); the assumption is, however, that it still sounds like a square wave to the ear. Maybe. But does a transient, like the start of a piano note or a drum, still sound the same when processed by this system?

Transient response, square wave response, frequency response are all different ways of looking at exactly the same thing. It is an exercise for undergraduate electrical engineering students givenan example of one type to determine the others. Low frequency response falloff can be seen as a tilt from left to right in a square wave. The steeper the tilt for any given frequency, the greater the falloff. A rounded leading edge means a falloff in high frequency response. A high frequency peak will show up as "ringing" that is damped oscillation to the shape of a square wave. Typically square waves at 100 hz and 10 khz are indicative of bass and treble response. A dual trace storage (memory) oscilloscope is often useful for capturing both the input and output and comparing them. Other indicators are impulse response (a single spike input) will show the degree of ringing in a component or system. Another is the tone burst, an sudden input of several cycles at a one frequency suddenly cut off also shows a lot. These are all different ways of looking at the same thing.

This method of analysis is very useful not just fror electronic equipment or electrical systems but for all types of systems, both human engineered and occurring in nature. But it seems to me only electrical engineering students get hammered with it over and over again. And you'd be surprised how many of them forget it later on in their careers.

 

[Gregadd]

the high end press almost to a person abhors low distortion, it always sounds cold, sterile, etc. because, I think, if they felt that was the answer, there could only really be one best low distortion component, and the rest losers, ans so you write about your one best unit for the year and all the rest are the losers.....its politics, just like blind listening, anything that removes the mystery of golden eared reviewers, their fuses, oils, magnets, etc, these things keep food at their tables.

I assume you are exaggerating for effect. A device that is "cold, sterile, etc" is by definition not "low distortion." Perhaps they refer to it as sounding "cold and sterile" because it does.
Only one winner? No co-champions? Even assuming low distortion as the only relevant parameter Could there not be many devices with equally low distortion?

We talk frequently about Sean Olive and his listening test. Sean Olive believes in trained listeners. I do not have the link but he referred to them as his "golden ears." Experience counts Access to a wide variety of equipment counts. Knowing what to listen for counts. No mystery there.

Sean Olive in trained listeners:http://seanolive.blogspot.com/2009/05/harmans-how-to-listen-new-listener.html

 

[NorthStar]

In an audio amplifier or preamplifier any change to the shape of the electrial waveform between the input and output is distortion. All those devices are supposed to do is make the signal larger, more powerful. Some distortions are audible, some aren't. It's important to sort them out. Addressing reducing distortions you can't hear is a waste of time, money, and effort. It makes for better advertising copy than engineering. There are thresholds below which distortions are not audible. One fallacy audiophiles fall for is not recognizing this. This makes them prey to ads claiming reduction in distortions that can't be heard. It's virtually useless. Wires are supposed to do the same thing but without any change in the signal size. One problem in testing distortion is performing the test under conditions the equipment will actually be used in. Another is testing for all forms of distortion. That's a real shortcoming of many of the products audiophiles buy. An amplifier's frequency response at one watt into a 4 or 8 ohm resistor won't tell you anything about what its frequency response will be feeding 50 watts into a speaker system whose impedance dips below one ohm and produces strong reverse emf. This is why equipment that tests identically on a test bench can sound very different in use.

There are other devices that are used to deliberately alter electrical signals in specific controllable ways. For example, the alteration to the waveform produced by an equalizer is intended to be the compliment or inverse of distortions that exist elsewhere. This is usually unavoidable, it's inherent in the limitations of the technology. For example, analog tape recordings and LPs could not exist without altering the signals fed to the recording equpment and altering them again on playback. Various forms of equalization, filtering (including crossover networks), digital signal processing, and dynamic compression and expansion (Dolby nose reduction for example) are deliberate changes to waveforms we do not customarily thing of as distortion.

Testing speakers for distortion is much harder. A speaker's job is to recreate a sound field at the listener's ears which is subjectively identical to an event the listener most likely never attended. Even memory does not serve well, you can't remember something you never experienced. And so people who have much experience listening to live music frequently have a better recollection of similar sounds. The analysis of sound fields is much more complex than the analysis of electrical signals. Unlike electrical signals they exist in three dimensions in space and interact in time and frequency with the environment. As of the current state of the art, the spatial, temporal, and spectral distortions compared to similar sound fields produced by live music is nothing short of horrendous for all speaker systems and arrays of speaker systems in a home listening environment. The key to better high fidelity will come from scientists and engineers who are qualified spending much more time studying and understanding sound than tinkering with electronics. That's the only way they can define design goals that will have meaning in terms of what we can hear.

-----I fully agree with that; and that's where we all should put our best efforts first at improving.

 

[NorthStar]

I assume you are exaggerating for effect. A device that is "cold, sterile, etc" is by definition not "low distortion." Perhaps they refer to it as sounding "cold and sterile" because it does.
Only one winner? No co-champions? Even assuming low distortion as the only relevant parameter Could there not be many devices with equally low distortion?

We talk frequently about Sean Olive and his listening test. Sean Olive believes in trained listeners. I do not have the link but he referred to them as his "golden ears." Experience counts Access to a wide variety of equipment counts. Knowing what to listen for counts. No mystery there.

Sean Olive in trained listeners:http://seanolive.blogspot.com/2009/05/harmans-how-to-listen-new-listener.html

-----Greg, could you please indicate the name of the quoter in your quote, thanx. :b

* Now I know it is Tom, but when new readers are reading our most interesting and intelligent threads, they deserve only the best. :b

 

[Gregadd]

Greg, could you please indicate the name of the quoter in your quote, thanx.

* Now I know it is Tom, but when new readers are reading our most interesting and intelligent threads, they deserve only the best.

I did not think it was important. It came from post #2.

 

[NorthStar]

I did not think it was important. It came from post #2.

-----For me, and many others, it is important. :b ...And yes I know.

 

[Gregadd]

Hi Bob! yeah, Greg keeps an eye on my posts, and sometimes agrees with me, and I him!

That green school bus in your avatar grabs my attention.:b

 

[Atmasphere]

I do personally avoid tubes in the amplification part [preference] but I do love tubes in other parts of the system. I.E. a tubed vinyl pre, a tubed CDP and tubes in line with other digital sources. How might this type of combination relate with what you mentioned?

I harness and avoid.

Tom

Tubes in the phono preamp is amplification. The main thing we are interested in is linearity. Tubes in signal applications can be quite low distortion so I advocate their use, especially since they often lend themselves to simpler circuits.

I've seen the math for this and its only for a limited range of amounts of feedback that the higher harmonics increase. I think there's a paper by Bruno Putzeys where he covers this, but he didn't originate the math. In my understanding most SS amps use more than 20dB of feedback and if I remember correctly that's about the amount where the higher harmonics start going down again.

I haven't found this to be so in my experience. I start out with SS amps which sound bright and harsh, modify them for appropriate noise control and they then sound smooth and non-fatiguing. I doubt very much that their harmonic distortion spectrum has changed much, although I haven't measured. The harshness I associated with transistors (particularly low-noise bipolar opamps, the LM4562 is one of the worst) is I believe related to the input stage's sensitivity to low-level RF signals. Deft use of ferrite beads, common-mode chokes, attention to PSU cleanliness and star grounding removes it to my satisfaction. The LM4562 has practically immeasurable distortion according to its datasheet yet still sounds harsh and fatiguing without these kinds of countermeasures being applied.

<edit> I realized I'd like to add that in my estimation the reason that SS amps are harsh and fatiguing is that's the sound that customers want. We discussed this on the recent RMAF thread - people who want smooth sound buy valve amps, in general. I'm something of an exception, preferring the low maintenance, weight and better energy efficiency of transistors so I design my own stuff.

I have seen people that don't like smooth sound- but in every case it was someone who had no intention of spending more than a few hundred dollars on the entire system, and also seemed to lack a lot of listening discernment. I think we often deal with this at the entry level- a public so used to seeing cheap equipment described as state of the art that they believe that rather than their own ears.

However it is a fact that our ears use odd ordered harmonics to determine the sound pressure of a sound and in the 21st century this should not be a matter of debate (its been known since the 1960s at the very least). It is also easily proven if one has access to a sine/square generator, an amplifier, a speaker and a VU meter. I can give the entire procedure if interested.

It is also true that at about 20 db on many amps, the odd-ordered harmonic distortion does indeed start to 'go back down'. However this is a very different comment from 'suddenly is totally gone' if you get my point. Nelson Pass has a wonderful article on the subject on his website- what you see is that you keep adding feedback and the odd orders do decrease from some peak value, but never to a point less than the original signal. Long before that you run out of gain, and so have to add another gain stage, which will require more feedback- ultimately you wind up always behind the goal.

It is true that there are other things that will cause an amplifier to sound smoother as you mention. Keeping an amplifier immune to RF difficulties, whether tube or transistor, is always good practice. But compared side by side the smoothest of the best transistor amps I have heard still have an unnatural hardness that the better tubes always seem to lack. I should point out that while I work with tubes, I am not of the opinion that is *has* to be done with tubes, only that its a hellava lot easier to do a competent design with low odd-ordered harmonic generation using them.

Its not rocket science to put the two concepts together; if you follow the data points a picture emerges of what has to happen (build an amplifier with low odd-ordered harmonic generation = more musical, more relaxed). I often hear about how tubes are all colored because they have a 2nd harmonic distortion, which is often true. But that can be eliminated by the use of fully differential operation, the same way it often is in transistor designs. While I have yet to hear a transistor amp that lacks odd orders in the way that tubes amps often do, that is not the same as saying that I don't think it can happen.

 

[opus111]

However it is a fact that our ears use odd ordered harmonics to determine the sound pressure of a sound and in the 21st century this should not be a matter of debate (its been known since the 1960s at the very least).

Do you have a link?

It is also easily proven if one has access to a sine/square generator, an amplifier, a speaker and a VU meter. I can give the entire procedure if interested.

I don't consider a sine or squarewave to be an interesting sound worth spending time listening to so no, no thanks

It is also true that at about 20 db on many amps, the odd-ordered harmonic distortion does indeed start to 'go back down'. However this is a very different comment from 'suddenly is totally gone' if you get my point. Nelson Pass has a wonderful article on the subject on his website- what you see is that you keep adding feedback and the odd orders do decrease from some peak value, but never to a point less than the original signal. Long before that you run out of gain, and so have to add another gain stage, which will require more feedback- ultimately you wind up always behind the goal.

No, I don't really get your point no. No amp is going to be able to take out distortion in the original signal, nor would I want it to. But perhaps I'm misunderstanding what's meant here by 'the original signal'. In any case I'm far more interested in intermodulation distortion than harmonics as I've yet to hear amp differences which I've correlated with distortion orders. Nevertheless I don't consider feedback an adequate substitute for intrinsic linearity so that might endear me somewhat to tube designers.

It is true that there are other things that will cause an amplifier to sound smoother as you mention. Keeping an amplifier immune to RF difficulties, whether tube or transistor, is always good practice. But compared side by side the smoothest of the best transistor amps I have heard still have an unnatural hardness that the better tubes always seem to lack.

Out of the SS amps you've listened to, which are the smoothest?

I should point out that while I work with tubes, I am not of the opinion that is *has* to be done with tubes, only that its a hellava lot easier to do a competent design with low odd-ordered harmonic generation using them.

Noted. I like challenges myself so for now, I'll stick with transistors. They may yet defeat me

 

[Atmasphere]

Do you have a link?

General Electric did their studies that found this and other interesting information back in the mid-60s, and they have not seen fit to publish their study online.

I don't consider a sine or squarewave to be an interesting sound worth spending time listening to so no, no thanks

No worries. And I guess, no arguments then about the principle either. I don't know anyone that spends any time listening to test tones, but they can easily create raised consicousness about how our ears actually behave. Decades ago I noticed that distorted sounds sounded a lot louder that pure sine waves when I was working as an audio technician while putting myself through school (it was a different world in the 70s...). It was not until the late 90s that I stumbled across GE's study that explained it.

No, I don't really get your point no. No amp is going to be able to take out distortion in the original signal, nor would I want it to. But perhaps I'm misunderstanding what's meant here by 'the original signal'. In any case I'm far more interested in intermodulation distortion than harmonics as I've yet to hear amp differences which I've correlated with distortion orders. Nevertheless I don't consider feedback an adequate substitute for intrinsic linearity so that might endear me somewhat to tube designers.

It does sound as if you missed my point.

'Original signal' in this case merely refers to the signal before amplified by the amp; nor was I implying that an amplifier somehow process the signal to remove distortion. And you are right- feedback added to a non-linear amp does not result in a good linear amp. Its best to start with something with good open loop specs already if you really want to do it right. BTW, low IM is not a function of feedback, its a function of linearity at all points in the amplifier, and is governed by good power supply design, good handling of ground and power distrubution, with linear active circuits and good component quality throughout. IOW its possible to build an amplifier that is zero feedback and very low IM at the same time.

Out of the SS amps you've listened to, which are the smoothest?

Easily the best transistor amp I have heard by a wide margin, and better than many tube amps (this is coming from a tube amp designer/manufacturer, take that FWIW), is the Ridley Audio power amplifier, which makes about 100 watts for about $100,000 or so. It employs a patent in which the output devices are heated to a controlled temperature, and is otherwise a Circlotron output like our amps, with no loop feedback of any sort, just like our amps. Naturally Dr. Ridley only dabbles in audio so no idea if his concepts will make it mainstream anytime soon. But his amplifier is proof that transistors need not sound like hifi- they can sound like real music instead, which IMO/IME makes it unique.