Distortion is usually avoided for music playback, but it is extensively used artistically, and the tube amp and NOS DAC nuts seem to enjoy their euphonic distortion quite a bit.

I see a general trend that many forms of distortion lead to a more enjoyable sound for most people. Witness the extensive use of tube amps for guitars, and less so for home systems. The continued use of analog tape (and tape-like distortion) in music production. And of dynamic range compression, when extensively applied, can also be considered a form of nonlinear distortion that producers, artists and many consumers seem to enjoy (all comments on the loudness war notwithstanding).

Subjectively, I get the impression that people think that music is occasionally too "thin" and bright unless some form is distortion is used to juice up the sound.

Is there a psychoacoustic reason behind liking distortion in a sound? Can euphonic distortion be objectively defined? Or is it an ad-hoc explanation to rationalize peoples' decisions?

With tube amps, it's a matter of squashing the sound rather than clipping. I've seen it suggested by many that tubes produce even harmonics (2x, 4x, 6x, ...) rather than odd harmonics (1x, 3x, 5x, ...) which is what happens when there's clipping.

Regarding the over-exuberance by many to make the loudness race synonymous with clipping, the compression algorithms these days usually don't produce any clipping to speak of.

I don't think anybody would argue that clipping is desirable under any rational context, ie it's only explainable as an artistic effect (and a somewhat irregularly used one at that, as you've pointed out). I'm talking more about dynamic range reduction and intermodulation, which tend to be quite commonly used.

Nobody's ever explained why even harmonics sound better. As far as compression is concerned, I see that it allows low-level details to peek above the noise floor, which is nice, but it doesn't explain the use of compression to "fatten" up the sound of individual instruments, or complete tracks. People talk about mastering making a mix sound "professional" and the exact mechanism how that happens is completely opaque to me.

I do understand what you're asking and apologize for focusing on just a small part of what people consider distortion.

I take it you understand why people like to apply compression to voices that aren't very well trained, correct?

To me the rest of it is like me figuring out why blue is the favorite color of most people that I ask. I usually don't come up with terribly useful things to say in the Scientific/R&D forum.

As far as the difference between tube distortion vs. clipping, I think it may have as much to do with the magnitude and frequency of the harmonics generated as with even vs. odd. The sharp edges of clipping produce very strong harmonics at frequencies well separated from the original. whereas rounding the top of the peaks produces harmonics that are much weaker and closer to the original. Why the latter would produce a coloring of the sound that some people find pleasing I can only guess.

Distortion in the context of electric guitar is a wholly different matter. In that case there is a lot of history involved - early electric guitar amps were very low wattage tube affairs that easily broke up when playing with drums and horns, and players liked the sound. As the instrument matured, players continued to exploit that sound and found new ways to implement it.

The distortion in a good guitar amp provides level compression and sustain, which makes the instrument respond more like a horn or bowed string instrument. The additional harmonics add information to an otherwise dry and clinical signal coming from the pickups - the magnetic pickups report string vibration but do not provide very much information about body resonances and other attributes that make an acoustic instrument interesting. Artificial harmonics provide color that can be musically useful.

I have been a professional guitarist for 30 years. It has been my experience that while modern solid state amps can work very well, it is usually easiest to get an interesting tone with a tube amp, even at lower volume levels where the distortion is not readily evident. A typical "old school" tube guitar amp runs at about 5% THD even played "clean", and this difference is clearly audible as additional harmonic color in the sound and subtle intermodulation effects.

The fact that the sound is pleasing is really an accident of history. Oh, and amplifying an acoustic instrument is a different kettle of fish entirely!

Analog distortion (generally) has the nice side effect of adding harmonics and if done cleanly very little nonharmonic noise, so it's musically pleasing and gives you more to work with when shaping the sound. But talking about music is like dancing about architecture

Complicated question.

Systems that distort more at high levels (nearly all analog systems that don't center clip) create an interesting effect, in that the loudness goes up while the system starts to compress (soft clip) for most signals.

Why? Because the distortion spreads the spectrum, and if you recall, loudness compresses inside a critical band, but adds between critical bands. So if you spread any meaningful amount of energy to some place it wasn't in the original signal, the loudness goes up disproportionately.

Try this: Take a 150 Hz sine wave. Without changing the gain, try clipping it first a little and then a lot, and observe that first the loudness goes UP, even though the signal energy is substantially decreasing.

Eventually of course the loudness will go down regardless, because you've reduced the signal energy so much.

Clipping creates very high harmonics, including, digitally, harmonics that alias back down and do ugly, ugly things by being utterly inharmonic after aliasing back in. This tends to sound spectacularly ugly.

Consider the spectrum of zero followed by the top 10% of a sine wave, followed by zeros (with the sine wave shifted down so that the peak starts from zero). This is an example of an error signal from clipping. It's UGLY, man.

Soft (analog) clipping, on the other hand, grows in order as the energy rises, giving rise to the kind of performance mentioned above.

There are other distortion effects that people like.

All of this can be tested in your basic octave or matlab, so try it out.

It is a matter of the harmonic series and the interval relationships between the different harmonics of the fundamental. I'll try to simplify this. 2nd harmonic is double of the fundamental, the second most consonant interval (after the unison). 4th is the double of the double of the fundamental, and 6th is the double of the third harmonic and makes with the 4th a „perfect fifth" (6/4, that is 3/2, again traditional consonant interval—a musical perfection for Pythagoras). And so on. That's why even harmonics (if they are only present) sound more pleasant to the human ear.

Understanding this may require understanding what this distortion overally does to the sounds "meaning". Look around your surroundings and you will - besides of electronic stuff - nowhere find "clean" and highly ordered sounds. Neither are those sounds an exact copy of something else. In nature, everything is an original and it always contains some degree of "chaos". It is always an organic mix of structure and chaos.

When the music world started to "go digital for the sake of going digital", they completely forgot that aspect in the sound-creation phase. When they realized that i.e. their digital synthesizers were sounding "lifeless and synthetic", they didnt notice the root of the problem and instead invested tens of years of development in trying to "emulate" the behaviour, which i.e. analogue synths expose natively.

So, i would say that part of the reason why you see "distortion-stages" in nowadays music creation, is because with the move towards digital sound creation, something went missing which they are now trying to "add back" afterwards.

- Lyx

So lemme try to distill this.

• Purely harmonic distortion tends to, subjectively, yield pleasing sounds. The best objective explanation for this effect is that it allows the perceptive loudness to increase while keeping the output power about the same, combined with an appeal to the ear's love of louder music. Beyond that, we need to appeal to some emotional response to harmony.
• Purely inharmonic distortion tends to, subjectively, yield dissonant sounds. This can mostly be explained on emotional and not objective grounds.
• Nonlinear distortion, of all types, yields both harmonic and inharmonic distortion in varying degrees.
• High order distortion tends to yield more inharmonic distortion than low order distortion, because the resulting frequencies are more likely to be strongly inharmonic.
• Because the distortion order of soft clipping (typically analog) is usually lower than that of hard clipping (typically digital), it tends to yield relatively less inharmonic distortion than hard clipping, and so is more pleasing to the ear.
• If the inharmonic distortion of soft clipping is low enough, it may sound better than if no distortion were present, due to loudness boosting from distributing the signal power among different critical bands, and due to the ear favoring louder noises in general.
• Dynamic range compression can make the signal seem more pleasing to the ear because it allows more detail to be heard; more critical bands are excited for the same maximum input power. Multiband compression is a somewhat extreme example of this principle.

Now the questions.

• Is this a reasonably objective/scientific explanation? Can the masking of critical bands, combined with assuming the ear likes louder and more harmonic sounds, provide a complete explanation for the love of distortion?
• Is a fairly complete (read: much more complicated than ITU 1770) loudness model necessary to estimate the loudness issues described here? ie, critical band compression? Or will a 50ms-windowed RMS measurement be sufficient?
• Can the inharmonic quality of hard clipping be accurately estimated with a spectral analysis of harmonic vs inharmonic sound?

(EDIT: partially answered my own question about loudness modelling)

People are different. I can’t stand even the smell of coffee but I’m fairly sure the world is not full of gastronomical masochist. For audio I am not so certain. I really like acoustic guitar but a great deal of that deliberate distortion on electric is at best annoying and often unpleasant.

The simplest explanation is that even the best sound reproduction loses so much of the original event that the pale imitation we end up with sounds better (more convincing? or just more fun?) when we try and put something back in.

I think there are (also) auditory streaming phenomena at work. Where the distortion is perceived as part of the original sound, "nice" distortion will add to that sound and make it more "full" (it being, basically, a pale imitation otherwise). However, where the distortion is perceived as being separate from the original sound, it is objectionable. The auditory system is able to stream the distortion as a separate entity more eaisly both when it is "nasty" distortion, and when the playback equipment is so good that the difference between the original sound and the added distortion is more obvious, and less hidden by the playback equipment itself.

Hope this makes sense. It's just a theory.

Cheers,
David.

Are you also referring to artist using vinyl crackle etc. intentionally?

No.. more like why many CDs are bounced to tape during mastering. Or for that matter, why electric guitar amps still use tubes and speaker boxes and mics.

(I'm still waiting for any comments on my last post.)

Yes! This was the answer I was going to give.

The hugest losses in the recording chain are at recording and playback...where the electromechanical transducers come into play. Perhaps some 'euphonic' distortion happen to have a compensatory perceptual effect, for some people.

That explanation is completely ad hoc.

That's oversimplified, but in fact we record a very few places in a soundfield and try to make them reproduce through one set of speakers with one specific and very inflexible ratio of direct to rebererant sound.

So there is a large loss in information in the reproduced soundfield, in both the perceptual and analytic sense.

But distortion in general doesn't improve this, what it does is create a sense of more dynamic range.

Some kinds of specific distortion (M/S distortion differences in LP's for instance) create what many people regard as a desirable set of effects on the soundstage.

Low-level out of phase rumble, if reproduced, mimics the lf behavior in a very, very wide hall, and makes things sound less "closed-in". Ditto out of phase bass, which of course we almost never see except as rumble on an LP, and which we rarely see on a CD because everyone has been trained to mono their bass.

And, of course, such signals get eaten completely by systems that use 1 (completely insufficient in number) subwoofer. One can prove this to themselves, by listening in headphones to in-phase vs. out of phase 50Hz sine waves out of octave or matlab, or whatever your program of choice might be.

OK, this has been something I thought for years now. Let’s see if I got this right by telling something I seen many years ago. I had a test vinyl record with a low frequency sine wave in phase on both channels. I feed the signal to an oscilloscope with the one channel going to the vertical input and the other channel going to the horizontal input. Now with both the vertical and horizontal adjusted to deflect the same one should get a straight diagonal line that looks something like this “ / “. But no, one gets a line that is very squiggly, but doing the same thing with a CD one does get a straight diagonal line. So there is the so-called depth that my friend hears on vinyl records, and the funny thing is that he hears the same thing on a CD that is a recording of a vinyl record along with all the other distortions that he loves so well.

Paul

Normal people like distortion as "special effects" ONLY.

I don't like added distortion by my equipment, I want as little as possible.

Paul

Whatever, Mr. Classic Rock Fans.

By special effects I mean like a guitar player playing a distorted guitar during recording session and not the listener "distorting" the whole recording when listening to it.

It's not immediately clear to me if such a distinction can be made - between "special effects" distortion, playback distortion, and natural sound. The "natural" sound of many instruments often has a timbre rich in harmonic content, deliberately chosen by the manufacturer (and artist). Any line made between that and a distortion effect must be fairly arbitrary.

Take pianos. Earlier pianos are far more percussive than modern pianos, with less sustain (read: reverb). Pianos kept evolving until they had an energy output capable of amply filling a concert hall. This required designing so much reverberation into the things that many pieces from the baroque, classical, and romantic periods cannot be played as originally noted because they will sound like mush. (Beethoven's Moonlight Sonata, for instance.) That is, you can't say that a grand piano is an inherently "natural" sound - there's nothing natural about 10+ tons of tension on a piano frame!

Actually, you could even draw a parallel between pianos and the modern loudness war. Pianos have a large amount of dynamic range compression built into them, due to this resonance, and one could argue that this compromises the artistic effect of older works, for the sake of playing the newer works for a wider consumer audience - which is a strong parallel to the modern loudness war. I guess that "war" only truly ended with the period performance movement, and everybody just got along with the new sound.




Well, that's evidence of rumble (probably a tonearm resonance), and that's the same thing Woodinville's talking about. Whether or not that completely explains "depth" is another issue entirely.

Wait, I'm confused. When would I ever see out of phase bass in an unamplified environment? The wavelength of a 20hz pressure wave is 17.2 meters. Assuming a 10cm ear-to-ear distance, the max phase offset between ears is something like 2 degrees. That seems like a pretty strong argument to sum to mono to me.

It's also tentative as written, rather than assertion of fact...as long as there's a 'perhaps' in there, I'm OK with it.

Very interesting. I want to sort out in my mind though -- are you saying that while distortion in general doesn't improve the large losses incurred during record and playback stages, the distortions you specify here do?

If so, that would seem to support the idea mentioned above, that some 'euphonic' distortions
happen to be perceived as 'compensatory' to at least some of the kinds of losses incurred during the lossiest stages of the chain.

If not, then how does the information lost at record&playback differ from the dynamic range and soundsage effects mimiced by LP distortion?

In a word "HI-FI", to reproduce the original sound without adding and/or subtracting from the sound, the concept is simple and easy to understand.

Now, how close we come to that is another matter.

Paul

I went to a rock-concert last night, everything and everybody was using a microphone, so what is so real about a lot of rock-concerts to begin with.

Paul

For some people. Some of the time. We're talking about preference, there is no universal, really, but some things (out of phase rumble, different distortion in M and S for instance) seem to be more commonly considered "sounding nice".


Analytically they differ enormously. Perceptually they differ, and even the best system (with or without euphonic distortions) is limited by the radiation pattern of the loudspeaker vs. the recording venue (assuming there is one such thing).

Hm. I have no idea where the slide deck for the Johnston, J. D., “Perceptual soundfield reconstruction,” Audio Engineering Society Oct. 2000 Fall Convention Program, Los Angeles, CA, October 2000.
talk is, I think it was still (gasp) foils, not a .ppt. There was a bit more work, Johnston, J. D., “Automatic loudspeaker directivity control for soundfield reconstruction,” 19th AES International Conference, Bavaria, Germany, 21-24 June 2001. (that's missing an author, I think??) but again, I don't know where any presentiation would reside.

Finally, http://www.aes.org/sections/pnw/ppt/other/2005aud.ppt discusses some of the issues.

Beranek discusses the properties of concert halls in enormous detail, and that would help as well. There are a lot of issues in the question you just asked, to say the least.