How to increase Harmonic distortion for clean sounds

[Alan0354]

Try expanding the first 3 terms will make your head spin already, That's the reason people have to use simulation program, but still it's not intuitive at all looking at the graph in the paper beyond the first few peaks.

[Chuck H]

Regardless of the link provided by Ken (which is actually about a driver and power tube and not cascade preamp stages) I still don't see how series stages cancels 2nd order harmonics. Yes each stage is out of phase. But if that is the only criteria for cancellation there should be equal detriment to the fundamental!!! Ergo... Whatever goes into one end of a voltage amplifier comes out the other side bigger. Including 2nd order harmonics. Unless you distort them and then they become higher order harmonics. Just as the fundamental begins to lose ground to harmonic orders higher than itself! The idea that a specific harmonic can be "cancelled" in a typical series preamp equates to me like thinking you can reduce the miles on your car by going in reverse. No, the harmonic cannot be cancelled. It can be tuned into other harmonics, but it doesn't go away. Avoid distorting it and you still achieve amplification of 2nd order harmonics. Ok. I'm done.

The results of the VOX amp showing predominantly odd order harmonics is interesting. Would that be odd numbered harmonics or actually odd order harmonics? I ask because all of a sudden I'm not even sure what even order and odd order mean. I always assumed that odd order harmonics were dissonant rather than harmonious. Now I'm not at all sure. Because loud AC30's sound plenty musical to me. I don't think it was specified whether this particular AC30 was a TB being used on the TB channel or not. Obviously this makes a difference since the cathode follower is said to add even order harmonics!?!

[Alan0354]

It does, and it's obvious. You don't have to go through the power series, like what I explained in #53 in a very simple way.

1)The triode stretch the top half of the input sine wave and compress the bottom part.

2) The output of the first stage being inverted, you'll see stretch on the bottom part of the "sine" wave and compressed the top part.

3) The second triode stage stretch the top half of the sine wave from the first stage ( which is the compressed part from the first stage) and compress the bottom part ( which is the stretched part from the first stage).

So, the first order approximation is the second harmonics is cancelled out.

I know those IMD stuff, how the (x+y) and (x-y) frequencies from the IMD. But this is a very simple and intuitive way of seeing how the second harmonics being cancelled.

As I said, AC30 has even number of preamp stages, that's the reason the second harmonics is lower. Look at Fig. 5 of Jonston's paper and it shows. I absolutely agree with the paper and the plot of Fig. 5. I just don't have the software to do that.

[Alan0354]

RWhatever goes into one end of a voltage amplifier comes out the other side bigger. Including 2nd order harmonics. Unless you distort them and then they become higher order harmonics. Just as the fundamental begins to lose ground to harmonic orders higher than itself! The idea that a specific harmonic can be "cancelled" in a typical series preamp equates to me like thinking you can reduce the miles on your car by going in reverse. No, the harmonic cannot be cancelled. It can be tuned into other harmonics, but it doesn't go away. Avoid distorting it and you still achieve amplification of 2nd order harmonics. Ok. I'm done.

Think of it, if the first stage compress the bottom of the input wave by 30% and stretch the top by 30%. then inverted the signal. So the bottom of the output of first stage is being stretched 30% and the top compressed 30%.

The second stage compressed the bottom of the wave 30% which is stretched 30% by the first stage. The second stage stretch the top of the wave 30% which is compressed 30% by the first stage. You net out with no stretch or compression of the original input wave!!!

Notice I use % of stretch and compression( as it is how it works) ? Because this make it independent to the gain of the stages. Even if the signal at the output of the first stage is much smaller than the second stage, but it does not matter in the first approximation. But you are right about the signal has to go through more gain. Of cause in real life, it's more complicated as the gain of the second harmonics is amplitude dependent also. The signal level the second stage is much larger than the first stage. That's the reason you see in Fig 5 that the second harmonics is only down 25dB, not totally cancelled.

Of cause, this is a very simplistic way of looking at it that does not take into account of the IMD that creates tons of other harmonics. But that explain the cancellation of the second harmonic.

[Enzo]

I don't see it as compressing and extending 30% or some other percent. You feed it a sine wave, and it clips one side of the waveform, you now have a distorted waveform. You run that through a second stage, it now wants to clip the unclipped side, but it doesn't expand the previous clipped side to its earlier shape.

[Alan0354]

I don't see it as compressing and extending 30% or some other percent. You feed it a sine wave, and it clips one side of the waveform, you now have a distorted waveform. You run that through a second stage, it now wants to clip the unclipped side, but it doesn't expand the previous clipped side to its earlier shape.

No, it's not clipping. We are still 100% in the linear ( relative linear) region. The sound is absolute clean, no clipping distortion. We consider this is clean sound with added harmonics. This is all clean sound we are talking about, that how to add some harmonics to give it the magic.

Once the signal clips, it is a totally different ball game. I am using 30% as an example that make it more obvious. You can think of just compress or extend 3% or even less. result is the same.

[Enzo]

But it is not clean. The output waveform is not the same as the input. By definition, that is distortion. Your whole quest is for pleasing distortion of the waveform.


Is it hard clipping like a diode across the path? No, it is a soft edged clip, a tube in its glory, but nonetheless, the wavepeak is truncated, and that is what a clip is.

[Alan0354]

But it is not clean. The output waveform is not the same as the input. By definition, that is distortion. Your whole quest is for pleasing distortion of the waveform.


Is it hard clipping like a diode across the path? No, it is a soft edged clip, a tube in its glory, but nonetheless, the wavepeak is truncated, and that is what a clip is.

Yes, this is what we talk about, and this is what most people call clean sound. There is no tube amp that has no distortion!!! It's the nature of the tube amp to have distortion.

No we are not talking about truncate at all here. There is no clipping, no truncate.

[kg]

the paper which appeared in sound practices was addressing a "fad" which was in favor around that time, namely using a power tube as a voltage amp for the same power tube. the reasoning was that the two gm transfer functions (ie, figure 9.1 here http://ken-gilbert.com/techstuff/spa...--page_203.jpg) would be so similar that the even order harmonics would cancel each other out.

indeed, as ivan demonstrated, this is the case. however, the tradeoff is that the odd harmonics are increased in relative value, and the order of all harmonic distortions is pushed up in frequency.

sounds a lot like global NFB loops, come to think of it.

anyhow, ime the quest for the holy grail of a completely even harmonic distortion amp is a fairly fruitless one. at the end of the day, multiple cascaded stages with "just enough" attenuation between stages has been used for over half a century for a reason: it works.

my advice is to spend a bit less time focusing on the distortion/clipping side and more on the tone shaping side (bypass caps, coupling caps, etc). of course they are two sides of the same coin and changing one will affect the other, but personally i have found it to have a much more profound effect on the feel of the amp than say shifting the operating point of a stage here or there.

i also find the even/odd #of stage thing to be a red herring, fwiw.

ken

[Enzo]

OK Alan, have it your way, the output waveform is not altered at all, it is the exact shape as the input waveform. One side of the wave is then NOT lopped off short of where it would have been.

There is clean and there is clean. it is a fairly slippery definition when you can have it both ways. If you want to use "clean" to mean a pleasing distortion, then we cannot also use the term to mean undistorted. If we are discussing distortion, we should reserve the term clean to mean distortion-free.

[Enzo]

OK Alan, have it your way, the output waveform is not altered at all, it is the exact shape as the input waveform. One side of the wave is then NOT lopped off short of where it would have been.

There is clean and there is clean. it is a fairly slippery definition when you can have it both ways. If you want to use "clean" to mean a pleasing distortion, then we cannot also use the term to mean undistorted. If we are discussing distortion, we should reserve the term clean to mean distortion-free.

[Chuck H]

Alan, your definition of clean isn't lost on me. And I think your terms are valid. Clean WRT guitar amp tone generally means "not clipping to the point of causing sharp edges on the waveform". We all know that to be true and you'll get no argument here. So...

Your example assumes a non linear operating point for the tube. This DID get past me. In your example the 2nd harmonic isn't cancelled, it's transformed via further distortion into higher order harmonics. I think Kens article explains that. If you stay in the linear area of a tubes operation you have more latitude. For example... If you skew bias the first triode to achieve a larger amount of second harmonic via a slightly truncated wave form and then center bias the following stage but do not distort the wave form on the truncated end you WILL achieve further amplification of the second harmonic. Since guitar amps come in all manner of circuitry (Marshall and Fender circuits that don't skew the bias on series stages come to mind) it's erroneous to say that amps with an even number of amplification stages cancel preamp generated second order harmonics. You really need to get your head wrapped around the reality that there is no relevant phase cancellation of anything in series amplifiers that do not employ any feedback. Research it. I'm right on this.

[Alan0354]

indeed, as ivan demonstrated, this is the case. however, the tradeoff is that the odd harmonics are increased in relative value, and the order of all harmonic distortions is pushed up in frequency.

sounds a lot like global NFB loops, come to think of it.

anyhow, ime the quest for the holy grail of a completely even harmonic distortion amp is a fairly fruitless one. at the end of the day, multiple cascaded stages with "just enough" attenuation between stages has been used for over half a century for a reason: it works.

my advice is to spend a bit less time focusing on the distortion/clipping side and more on the tone shaping side (bypass caps, coupling caps, etc). of course they are two sides of the same coin and changing one will affect the other, but personally i have found it to have a much more profound effect on the feel of the amp than say shifting the operating point of a stage here or there.

i also find the even/odd #of stage thing to be a red herring, fwiw.

ken

I personally agree with you, in fact, I already tried introducing even harmonic by design and I don't exactly like the sound. But I do believe even and odd stages do make a difference, but not better or worse. I personally like two stages over three stages of the BF/SF Fender amps.

But as I said before, this is a thread of how to introduce distortion ( not clipping), not how to make it sound better!!!. Sounding better is subjective, introducing distortion is theoretical, whether sounds better is a totally different matter. This is a theory section, I look at it as a brain exercise.

[Alan0354]

Alan, your definition of clean isn't lost on me. And I think your terms are valid. Clean WRT guitar amp tone generally means "not clipping to the point of causing sharp edges on the waveform". We all know that to be true and you'll get no argument here. So...

Your example assumes a non linear operating point for the tube. This DID get past me. In your example the 2nd harmonic isn't cancelled, it's transformed via further distortion into higher order harmonics. I think Kens article explains that. If you stay in the linear area of a tubes operation you have more latitude. For example... If you skew bias the first triode to achieve a larger amount of second harmonic via a slightly truncated wave form and then center bias the following stage but do not distort the wave form on the truncated end you WILL achieve further amplification of the second harmonic. Since guitar amps come in all manner of circuitry (Marshall and Fender circuits that don't skew the bias on series stages come to mind) it's erroneous to say that amps with an even number of amplification stages cancel preamp generated second order harmonics. You really need to get your head wrapped around the reality that there is no relevant phase cancellation of anything in series amplifiers that do not employ any feedback. Research it. I'm right on this.

Maybe I should be careful in using the word "cancel", I should use "reduce" instead. If you read my post, you'll see I did said "on the first pass" which means just on the first order approximation.....just roughly. I did not even try to use power series as in the paper, I just use the nature of the tube that stretch one side and compress the other side. I assume we all know the IMD theory as this is very important in any power amp design, that any distortion that generates harmonics will have the (x+y) and (x-y) effect and generate lots of side band. That is a given. I did not say anything about the other harmonics.

Whether the two stages are identical or not, the nature of the tube is such that it tends to reduce second harmonics in even stages unless you intentionally operating one stage in very linear region. Maybe, I should say reduce, not cancel..........which I did say about the paper that the result of even the identical stage ONLY reduce the second harmonics by 25dB only, not cancelling. There is no total cancellation.

[Alan0354]

OK Alan, have it your way, the output waveform is not altered at all, it is the exact shape as the input waveform. One side of the wave is then NOT lopped off short of where it would have been.

There is clean and there is clean. it is a fairly slippery definition when you can have it both ways. If you want to use "clean" to mean a pleasing distortion, then we cannot also use the term to mean undistorted. If we are discussing distortion, we should reserve the term clean to mean distortion-free.

I never said the output wave form is not alter!!!! Maybe I should not use "cancel" and use "reduce" instead. I think you are assuming what I said. All I said is the the stretching one side and compressing the other side, that the second stage have the opposite effect and the result is the output is more symmetrical. AND I did say it is only the first pass approximation. Of cause, we all design amps and I assume it's common knowledge about power series approximation and the IMD as result. This is a given that we all know that!!! That we create plenty of (x+y) and (x-y) harmonics.

I think people in guitar amp refers clean as not clipped only, tube amp by definition, introduce amplitude distortion ( not clipping). This is not exactly my way. I think we are all talking about clean sound as none clipping sound here.

Think of it this way, all the shifting of operating point is is about changing the gain of the stage at different voltage level along the input sine wave. Another way to look at it is the gain change with the input voltage, that the gain is different along the lower portion of the sine wave and the upper portion and vary continuously along the sine wave.