How audible is distortion?

[Blumlein 88]

So the change at 50 seconds was from covering the DiMo membrane covered hole? BTW, what is the material of the membrane?

EDIT to add:

Okay found what the membrane is.

Whereas most simple flutes have only a blowing hole (known as chui kong (吹孔) in Chinese) and finger-holes, the dizi has a very different additional hole, called a mo kong (膜孔), between the embouchure and finger-holes. A special membrane called dimo (笛膜, lit. "di membrane"), made from an almost tissue-like shaving of reed (made from the inner skin of bamboo cells), is made taut and glued over this hole, traditionally with a substance called ejiao, an animal glue. Garlic juice may also be used to adhere the dimo, but it is not recommended as a permanent replacement. This application process, in which fine wrinkles are created in the centre of the dimo to create a penetrating buzzy timbre, is an art form in itself.

The dimo covered mokong has a distinctive resonating effect on the sound produced by the dizi, making it brighter and louder, and adding harmonics to give the final tone a buzzing, nasal quality. Dizi have a relatively large range, covering about two-and-a-quarter octaves.

 

[bennetng]

There are four demos in the video. The first three were using a real "di mo" but with different applying techniques, from loose to tight.

The last one was not using a "di mo" but a plastic tape.

Also, while the video is demonstrating a Chinese instrument, the subtitle is in Japanese and saying that the second demo is the ideal adjustment.

 

[Don Hills]

So it's basically a fancy kazoo...

 

[Cosmik]

at 50 seconds in western ears will appreciate it, for the first part more Asian ears will appreciate, its a cultural thing IMO

Therefore you highlight a major problem with 'scientific' listening tests. Especially the "preference" based ones.

And was the dizi used during the development of lossy codecs? And a dizi ensemble? And a dizi-guitar-drums combo? And a dizi-ocarina duet? etc. With Western listeners? And Asian? etc. etc. If listening tests really are key to the development of lossy codecs, then this is the minefield the developers are entering. If, on the other hand, these codecs are based on something more fundamental than listen-tweak-listen-tweak-, etc., then listening tests are just confirmation that they more-or-less work.

 

[bennetng]

That's why synthesized music are quite common in "killer samples" since a synthesizer sound designer can create something that no one ever heard before and have no limitation in terms of acoustic design.

Your comment is a good indication that listening tests play an important role on lossy codec development, even if you don't agree with this approach.

 

[Cosmik]

That's why synthesized music are quite common in "killer samples" since a synthesizer sound designer can create something that no one ever heard before and have no limitation in terms of acoustic design.

Absolutely right. I think you could go further than that. By looking at what the codec does, and with our knowledge of human hearing, it would be perfectly possible to synthesize "killer samples" deliberately to trip them up. (This goes counter to the common notion that scientific audio is all about observation of things that have almost come into being by some chance process - like nature. Rather, audio hardware and software can be designed deliberately with high levels of predictability).

Anyway, it's lucky that audiophiles are in the position to regard lossy codecs as curiosities of the early digital age. But they will always be useful in other applications.

 

[RayDunzl]

A distorting audio system cannot add harmonic distortion to anything except a recording of a single monophonic instrument. If two instruments are playing at the same time, it must add intermodulation distortion, which is not harmonic and not pleasing.

Since timbre'd instruments (that's all of them) are not generating single frequency sines, but are generating a harmonic series, there are IMD products generated during the playback of a single instrument, as the speaker tries to produce the fundamental and the overtones simultaneously.

Experiment:

Single sine at 500 Hz, 70dB - no noticeable (not measurably rising above noise floor, anyway) harmonic components

Two sines, 500 and 600 Hz, left speaker 500 and right 600 at 70dB - no noticeable harmonic components - so, IMD doesn't occur "in the air".

Two sines, 500 and 600 Hz, both playing from one (or both) speakers - modulation distortion products noted at 400 and 700 Hz about 40dB down, and not obviously audible.

---

Experiment:

Deliberately adding tones at 400 and 700 required an additional 10dB or above the IMD frequency level to become audible.

It did create a bunch of new low level IMD products higher up, though. Below, not sure, the noise floor is higher.

 

[Cosmik]

Since timbre'd instruments (that's all of them) are not generating single frequency sines, but are generating a harmonic series, there are IMD products generated during the playback of a single instrument, as the speaker tries to produce the fundamental and the overtones simultaneously.

Experiment:

Single sine at 500 Hz, 70dB - no noticeable (not measurably rising above noise floor, anyway) harmonic components

Two sines, 500 and 600 Hz, left speaker 500 and right 600 at 70dB - no noticeable harmonic components - so, IMD doesn't occur "in the air".

Two sines, 500 and 600 Hz, both playing from one (or both) speakers - modulation distortion products noted at 400 and 700 Hz about 40dB down, and not obviously audible.

---

Experiment:

Deliberately adding tones at 400 and 700 required an additional 10dB or above the IMD frequency level to become audible.

It did create a bunch of new low level IMD products higher up, though. Below, not sure, the noise floor is higher.

Ooer, I'm trying to get my head round it. In my mind's eye, if I feed a repeating i.e. harmonic(?) waveform into a 'bent' transfer function, I still get out a repeating (i.e. harmonic?) waveform.

Which has led me onto this fascinating little Wikipedia entry.

Edit: OK, I think I'm beginning to understand. Are you sure you get actual IMD within a single harmonic-rich tone? Surely if I have a 100 Hz fundamental plus some harmonics at 200 Hz, 300 Hz, 400 Hz and so on, the only sums and differences (i.e. IMD) I am going to get will be at multiples of 100 Hz i.e. all harmonically related to the fundamental. Even if the original waveform is 'missing the fundamental', IMD will merely add some fundamental back into the waveform - as the Wikipedia image shows, even if the fundamental is 'missing' the waveform is still periodic at the fundamental frequency.

 

[fas42]

Ooer, I'm trying to get my head round it. In my mind's eye, if I feed a repeating i.e. harmonic(?) waveform into a 'bent' transfer function, I still get out a repeating (i.e. harmonic?) waveform.

Which has led me onto this fascinating little Wikipedia entry.

Yes, the brain creating the fundamental works very nicely with items such as pipe organ recordings, on a system with "no bass" - with the setup in good tune, I never feel anything is missing when playing this sort of material on "tiny" speakers.

 

[Cosmik]

Yes, the brain creating the fundamental works very nicely with items such as pipe organ recordings, on a system with "no bass" - with the setup in good tune, I never feel anything is missing when playing this sort of material on "tiny" speakers.

I went to a talk by some of the people who do mobile phone audio development, and the synthesis of harmonics to indicate the missing fundamental is one of the tricks they do.

 

[RayDunzl]

I couldn't make pictures a while ago, but can now.

Two tones in the air. One tone per speaker. No IMD noted. A little 3nd harmonic of the fundamental seen at 1200Hz and the 2nd of 800 at 1600Hz (below 20dB on this display)

Combine tones into one (left) speaker, imitating a single 'instrument' which has its first overtone level approximately equal to the fundamental. IMD products appear.

Add the third harmonic of the fundamental at 1200Hz, all three tones from one speaker:

More IMD present. Interestingly, the 2kHz product is suppressed here.

 

[Cosmik]

I couldn't make pictures a while ago, but can now.

Two tones in the air. One tone per speaker. No IMD noted. A little 3nd harmonic of the fundamental seen at 1200Hz and the 2nd of 800 at 1600Hz (below 20dB on this display)

View attachment 9579

Combine tones into one (left) speaker, imitating a single 'instrument' which has its first overtone level approximately equal to the fundamental. IMD products appear.

View attachment 9580

Add the third harmonic of the fundamental at 1200Hz, all three tones from one speaker:

View attachment 9582

More IMD present. Interestingly, the 2kHz product is suppressed here.

Of course. Yes, it is IMD, but is only generating harmonics of the fundamental - I was fixated on the inharmonic aspect, but that isn't the 'definition' of IMD.

 

[bennetng]

Anyway, it's lucky that audiophiles are in the position to regard lossy codecs as curiosities of the early digital age. But they will always be useful in other applications.

Yeah. For example the opus audio codec used in the Dizi and ocarina Youtube videos, and music videos with billions of views like Descapito. So members who watched the Dizi and ocarina videos might be hearing the biased sound, and consumers who purchased the lossless version of Descapito because of Youtube could be disappointed because the Youtube version might sound better to them.

You can see how the opus developers deliberately put themselves into the minefield and having their professional knowledge of human hearing getting challenged by some casual users here:

https://hydrogenaud.io/index.php/board,71.0.html

BTW, I am an Asian. Sorry if my English is hard to understand.

 

[amirm]

BTW, I am an Asian. Sorry if my English is hard to understand.

Your English is perfect....

 

[bennetng]

Your English is perfect....

Thanks. I frequently use Google to check my spelling and grammar. However I almost misspelled Descapito since Google suggested "Despacito". I don't know if it is also a bias or not.

[edit] sigh. The correct spelling is "Despacito".

 

[Cosmik]

Yeah. For example the opus audio codec used in the Dizi and ocarina Youtube videos, and music videos with billions of views like Descapito. So members who watched the Dizi and ocarina videos might be hearing the biased sound, and consumers who purchased the lossless version of Descapito because of Youtube could be disappointed because the Youtube version might sound better to them.

You can see how the opus developers deliberately put themselves into the minefield and having their professional knowledge of human hearing getting challenged by some casual users here:

https://hydrogenaud.io/index.php/board,71.0.html

BTW, I am an Asian. Sorry if my English is hard to understand.

Not sure what you're saying. That lossy compression will always be relevant to audiophiles because a popular video streaming service uses it? That the cleverness of the codec's developers means that the audible quality must be as good as lossless - or better..?

All I was saying was that seekers after the best quality can listen to lossless (i.e. by definition the highest quality) versions of their music at almost no penalty these days. Whether or not Youtube's lossy codecs are close to transparent is not the issue - these I place in the "other applications" category that I mentioned as being "still useful".

 

[bennetng]

Not sure what you're saying. That lossy compression will always be relevant to audiophiles because a popular video streaming service uses it?

Audiophiles or not, it is normal for potential customers to listen to lossy platforms like free streaming services, radio and TV, before they decide to purchase something, so they at least have some ideas about the content of the song, like lyrics, arrangement and genre. People who buy music, lossy or not, can be a form of showing their support to the artists. The intention may not be totally or solely about quality.

That the cleverness of the codec's developers means that the audible quality must be as good as lossless - or better..?

Not again, the possibilities are subjectively better/worse, no opinion but audible, and no differernce. All of them are possible, since lossy codecs are not always transparent.

All I was saying was that seekers after the best quality can listen to lossless (i.e. by definition the highest quality) versions of their music at almost no penalty these days. Whether or not Youtube's lossy codecs are close to transparent is not the issue - these I place in the "other applications" category that I mentioned as being "still useful".

I understand and agree with your point. For example, we are still very far from having lossless video, even on offline storage devices due to the insane file size, but for audio, it is a non-issue.

 

[Wombat]

Yes, the brain creating the fundamental works very nicely with items such as pipe organ recordings, on a system with "no bass" - with the setup in good tune, I never feel anything is missing when playing this sort of material on "tiny" speakers.

I do.

I have Altec 416Bs in 13.75 Cu.ft.(net) enclosures(32Hz tuned) for comparison with my 'little guys'. However, maybe that effect is why I don't add a sub-woofer? No, lack of space for another large cab is why. I like large cabs. and high efficiency drivers.

 

[oivavoi]

Benchmark recently published an interesting write-up on distortion: https://benchmarkmedia.com/blogs/application_notes/interpreting-thd-measurements-think-db-not-percent?utm_source=Benchmark's+Application+Notes&utm_campaign=f9856e624b-EMAIL_app+note+11-20-17&utm_medium=email&utm_term=0_7c8c792ee5-f9856e624b-194245873

One of the points that were striking to me, was their point on reproducing the sound of pianos. It is often assumed that solo piano is one of the most difficult things to get "right" for a sound reproduction system. I tend to agree with that. According to Benchmark's John Siau, this might be the explanation:

Some instruments, such as the piano, produce overtones that are not perfectly harmonic. The overtones produced by a piano string are stretched by slightly more than octave ratios. This is why the tuning of a piano must be stretched to prevent beating between the overtones and the note one octave above. The stretched overtone spacing can also beat with the perfect octave spacing produced by harmonic distortion. The piano string's 1st overtone may beat against the 2nd harmonic distortion produced by the electronics (they are at slightly different frequencies). To my ears, this beating mimics the sound of a poorly tuned piano. I became sensitized to this after tuning my own piano. I believe this effect is one of the reasons why it is so difficult to accurately reproduce the sound of the piano. If you can reproduce a realistic piano sound, you have a very good audio system.

I would perhaps be inclined to think that distortion in the speakers is a more likely culprit than distortion in the electronics. But then again, who knows.

---------
Btw, a fellow over at Diyaudio has created a really useful blind test on non-linear distortion: http://www.diyaudio.com/forums/everything-else/314218-test-ears-abx-test-44.html#post5242441

 

[DonH56]

I am usually careful to use "overtones" and "timbre" rather than "harmonic series" to describe the sound of musical instruments. Many of them have non-harmonic components, including brass and most wind instruments. Strings can generate beats among the actual strings on the instrument as they are played, etc. Drums are impulse generators and yet do have fundamental'ish terms depending upon how the head is tuned. Piano is tough for many reasons: the percussives when hammers strike strings, there are 2-3 strings per note and those strings are tuned to sound "richer" so are not necessarily exactly the same note, different strings (different material/construction) have slightly different sounds, the sound comes from strings and sound board, the sound can vary along the string since there are nodes and antinodes, and of course that also affects how nearby strings interact, the sound is modified by interaction of the sound waves from various strings within the piano and sound board depending upon the position of the lid, octaves are generally (intentionally) "stretched" at the top and bottom, it is a very wide frequency range to cover, etc. etc. etc.

I made some files years ago with added distortion for people to try. IMD is much easier to hear than THD, but in music any distortion is harder to hear. The rule of tens is not quite right but decades ago I was curious and did some blind tests with friends and customers of the store and it was not too far off. That is, for a pure (single) tone in the midrange or vocal band folk could detect down to about 0.1% distortion with moderately high volume; with a few multiple tones more like 1% depending on how they were spaced, and for musical selections or many tones more like 5% - 10%. Music includes so many harmonic and non-harmonic frequencies anyway that telling what is distortion can be hard until it gets pretty high. And then add the whole loudness curve bit and any general comments about "x% distortion is audible" pretty much fly out the window.

IMO - Don