Earl Geddes has done distortion testing similar to this and published with AES.
He contends that distortion measurements as a quality measurement for predicting preference are useless.
One test involved auralization of compression driver distortion with 15seconds of music. Music was recorded at varying levels through driver, results normalized and used for testing.
A second test generated a series of distortion profiles of multiple types and applied them mathematically to source signal for generating test tracks.
He freely acknowledges that certain types of distortion are audible at extremely low levels, and points out that other types of distortion at extremely high levels, leads to no preference in blind testing.
Pavel, does this distortion when applied to two tones produce IMD?
He contends that distortion measurements as a quality measurement for predicting preference are useless.
One test involved auralization of compression driver distortion with 15seconds of music. Music was recorded at varying levels through driver, results normalized and used for testing.
A second test generated a series of distortion profiles of multiple types and applied them mathematically to source signal for generating test tracks.
He freely acknowledges that certain types of distortion are audible at extremely low levels, and points out that other types of distortion at extremely high levels, leads to no preference in blind testing.
Pavel, does this distortion when applied to two tones produce IMD?
IMD is only different view of non-linearity, once we have harmonic distortion we MUST have intermodulation distortion. It is impossible to have one without the second. Anyway, I will try to post IMD plots generated by these two distortion profiles.
Here is the distortion of 250Hz + 7kHz signal distorted by pure 3rd harmonic 1% FS. Of course harmonic distortion creates intermodulation distortion as well, they both are only different visualization of the SAME non-linearity. The IMD here is 0.97%.
3rd harmonic at 1% FS and higher is definitely audible.
Attachments
Last edited:
With same purely 3H distortion, we can see what happens with two originally 250Hz + 800Hz twintone. New, non-harmonic components can be seen quite distanced from original tones.
It is a general belief that 2nd and 3rd harmonic are audibly innocuous. I am sure, that above certain level it is not true. They create harmonic tones ("tuned") only in case of a single tone. When a simple single instrument like flute is reproduced, it probably only changes a tonality. But in case of more complex music + voices even 2nd and 3rd harmonic, above certain level, create audible mess and worsen the resulting sound.
It is a general belief that 2nd and 3rd harmonic are audibly innocuous. I am sure, that above certain level it is not true. They create harmonic tones ("tuned") only in case of a single tone. When a simple single instrument like flute is reproduced, it probably only changes a tonality. But in case of more complex music + voices even 2nd and 3rd harmonic, above certain level, create audible mess and worsen the resulting sound.
Attachments
OK, no more votes, so I am posting the right answers I received in a PM:
You have received a new private message at diyAudio from fas42, entitled "Listening Test - Added Harmonic".
This is the message that was sent:
***************
Very fast off the mark getting that up, I'm impressed!
Okay,
music2 - original
music3 - less distorted
music1 - more distorted
Didn't want more egg on my face, 😀, so quite a few rounds of shifting from one to the other, the sound kept subtlely changing as the DAC settled down. At first I had music1 and music3 swapped around, but as the DAC got into its rhythm the final ranking was established.
What's interesting is that the mathematically fiddled distortion has quite a different character from that going through conversion chains - the playback has to be 'sharper' to poperly assess them.
-------------------
You have received a new private message at diyAudio from wintermute, entitled "added distortion test results.".
This is the message that was sent:
***************
Hi Pavel, I'm going to go with my first impressions on this one. I've only listened with my cheap headphones, but I'm going to say.
track 2 is the original. The vocals seem the smoothest.
Picking which distortion is added to which of the other two tracks is more tricky, the more I listened the less I could tell the difference so my first impression was:
Track1 has the most distortion and track three is the one with only one spike added.
It's funny though, if you had given me any one of those tracks on its own, I'd have said it was excellent quality!
I think I can hear a difference between the three tracks on the first bit (I assume it is a person hitting the open end of bottles??) but mostly I'm picking it on the vocals. As I said track two sounds smoothest, track 1 sounds the most husky and track three a little bit husky. But all three sound natural!
When I had only listened to track one I thought "This must be the original!"
------------
My question on "more distorted" and "less distorted" was a bit tricky. Anyway, I agree with fas42 and wintermute subjective impression hereabove. My last question remains, do you consider the 3rd harmonic distortion that I have used here as "more" or "less" distorted than the the comb of odd harmonics distortion? Please thoroughly check the distortion profiles I have shown at the beginning of the thread. Numerical THD of the 3rd order distortion is 1% and for comb distortion it is 0.36%.
You have received a new private message at diyAudio from fas42, entitled "Listening Test - Added Harmonic".
This is the message that was sent:
***************
Very fast off the mark getting that up, I'm impressed!
Okay,
music2 - original
music3 - less distorted
music1 - more distorted
Didn't want more egg on my face, 😀, so quite a few rounds of shifting from one to the other, the sound kept subtlely changing as the DAC settled down. At first I had music1 and music3 swapped around, but as the DAC got into its rhythm the final ranking was established.
What's interesting is that the mathematically fiddled distortion has quite a different character from that going through conversion chains - the playback has to be 'sharper' to poperly assess them.
-------------------
You have received a new private message at diyAudio from wintermute, entitled "added distortion test results.".
This is the message that was sent:
***************
Hi Pavel, I'm going to go with my first impressions on this one. I've only listened with my cheap headphones, but I'm going to say.
track 2 is the original. The vocals seem the smoothest.
Picking which distortion is added to which of the other two tracks is more tricky, the more I listened the less I could tell the difference so my first impression was:
Track1 has the most distortion and track three is the one with only one spike added.
It's funny though, if you had given me any one of those tracks on its own, I'd have said it was excellent quality!
I think I can hear a difference between the three tracks on the first bit (I assume it is a person hitting the open end of bottles??) but mostly I'm picking it on the vocals. As I said track two sounds smoothest, track 1 sounds the most husky and track three a little bit husky. But all three sound natural!
When I had only listened to track one I thought "This must be the original!"
------------
My question on "more distorted" and "less distorted" was a bit tricky. Anyway, I agree with fas42 and wintermute subjective impression hereabove. My last question remains, do you consider the 3rd harmonic distortion that I have used here as "more" or "less" distorted than the the comb of odd harmonics distortion? Please thoroughly check the distortion profiles I have shown at the beginning of the thread. Numerical THD of the 3rd order distortion is 1% and for comb distortion it is 0.36%.
Last edited:
Thanks Pavel! 🙂
On the 3rd vs spray of harmonics I thought (quite possibly incorrectly) that the one that sounded more distorted would naturally be the one with all the extra added harmonics. I did not look closely at the relative levels of distortion between the two plots earlier, and I'm not really experienced in analysing distortion mechanisms and how they actually manifest.
Looking more carefully at the spray of harmonics, and plugging them into a db to percentage calculator makes me realize that the contribution that those extra added harmonics would be having is very low ( I don't think I fully understood that when you questioned me via pm). The 5th harmonic is only about 0.1% and the 7th about 0.03%
Tony.
On the 3rd vs spray of harmonics I thought (quite possibly incorrectly) that the one that sounded more distorted would naturally be the one with all the extra added harmonics. I did not look closely at the relative levels of distortion between the two plots earlier, and I'm not really experienced in analysing distortion mechanisms and how they actually manifest.
Looking more carefully at the spray of harmonics, and plugging them into a db to percentage calculator makes me realize that the contribution that those extra added harmonics would be having is very low ( I don't think I fully understood that when you questioned me via pm). The 5th harmonic is only about 0.1% and the 7th about 0.03%
Tony.
Well done you two... so no hope for me then. 'wanna buy a Micromega or Sony, one careful owner 😀
Pavel. Which is which then ? I liked track 3 the best as you know, although I felt it the most distorted. What was track 3 ?
Pavel. Which is which then ? I liked track 3 the best as you know, although I felt it the most distorted. What was track 3 ?
Not really quite sure what you're asking, Pavel, I already mentioned in the 2nd PM that the first track was subjectively further away in tone from the original, compared to the third - and this is my interpretation of "more" distortion. What the numbers are, what the precise spectral nature of the harmonic distortion is not relevant - it's how my ears heard it that counts ...
Is that what you're after?
Is that what you're after?
music2 ... no distortion added. Both Frank and Tony got it right.
music1 ... added pure 3rd harmonic distortion of 1% (-40dB) at full scale amplitude. This makes the biggest audible difference when subtracted (after very careful alignment) from the music file. The distortion at high amplitudes becomes quite audible and creates audible intermodulations. Please check the difference file I posted yesterday, it is the music2 - music1. I would call this music2 file as "more distorted".
music3 ... the "comb" distortion with odd harmonics. Added content:
3H -50dB
5H -60dB
7H -70dB
9H -80dB
11H -90dB
13H -100dB
15H -110dB
17H -120dB
19H -100dB
Remember this is all referred to full scale. I will show that at -20dB of signal amplitude only one component remains.
This distortion creates less audible artifacts that the pure 3H -40dB distortion, so I call it "less distorted".
I have realized to evaluate only your subjective impressions, expressed as "more" or "less" distorted.
In this post, you will see my "3H distortion" and "comb distortion" how it applies to -20dB signal. You can see, in both cases, there is only the 3rd harmonic component. Everything else is below 16-bit (dithered) resolution. The "comb" distortion 0.36% profile creates much lower 3rd component, than the 3H 1%. Subjectively, the "comb distortion 0.36% FS" is less audible than "pure 3H distortion 1% FS".
Attachments
Thanks Pavel. I listened to the difference file yesterday, it was very "small" or low amplitude level. Had to use headphones and turn the volume up to listen to it.
So I liked the one with added comb distortion 😀 I find this interesting tbh. Is this why I don't like the sound of so called low distortion amplifiers (I'm thinking of the Pioneer A80 and Blameless amp) and seem to prefer designs with more obvious "colour".
Thanks... I'll study these a little more.
So I liked the one with added comb distortion 😀 I find this interesting tbh. Is this why I don't like the sound of so called low distortion amplifiers (I'm thinking of the Pioneer A80 and Blameless amp) and seem to prefer designs with more obvious "colour".
Thanks... I'll study these a little more.
Karl, I hope you do not mind if I copy and paste from your reply:
Hi Pavel,
Had a very good listen to these... OK
music 1 sounds the cleanest to me.
You liked the music1, which is the file with pure 3rd harmonic 1% FS (-40dB)added. It is not the "comb distortion" file. The comb distortion file has all odd harmonics starting from 3rd, added.
comb distortion (music3)
3H -50dB
5H -60dB
7H -70dB
9H -80dB
11H -90dB
13H -100dB
15H -110dB
17H -120dB
19H -100dB
P.S.: please check the next link, especially y=x3, y=x5 😉
http://mpec.sc.mahidol.ac.th/radok/physmath/MAT1/mat131.HTM
Last edited:
Not at all, here is the full pm 🙂 I found 3 had an appealing quality to it. I tend to associate odd harmonics with "cross over distortion" type sound but it didn't sound at all like that to me. I seemed to prefer it over track 1 and 2. It had more colour to it... more "twangy" on the bass notes.
Hi Pavel,
Had a very good listen to these... OK
music 1 sounds the cleanest to me.
music 2 is pretty similar but "less clean".
music 3 sounds the most "distorted" and yet has an appealing quality to it. I think I can detect a "noise envelope" around the bass notes at the start of the track on this one. At least more pronounced on music 3 than the others.
Hope that helps
Hi Pavel,
Had a very good listen to these... OK
music 1 sounds the cleanest to me.
music 2 is pretty similar but "less clean".
music 3 sounds the most "distorted" and yet has an appealing quality to it. I think I can detect a "noise envelope" around the bass notes at the start of the track on this one. At least more pronounced on music 3 than the others.
Hope that helps
Very interesting results, Pavel. My guess was, as stated in my 2nd PM, that the comb distortion would have sounded "worse" - but, not so ...
The fact that 'mulched' sound often appeals more matches my own experiences over the years - nominally 'super clean' reproduction is often the hardest barrier to overcome, because it's "squeaky clean", with the emphasis on squeaky - the remanent artifacts can be so annoying at this point, that it would be preferable for many to have some 'soothing' distortion added ... 🙂
The fact that 'mulched' sound often appeals more matches my own experiences over the years - nominally 'super clean' reproduction is often the hardest barrier to overcome, because it's "squeaky clean", with the emphasis on squeaky - the remanent artifacts can be so annoying at this point, that it would be preferable for many to have some 'soothing' distortion added ... 🙂
Last edited:
No, because at lower signal level the "comb" distortion profile creates only 3rd harmonic (with very very low amplitude), as you can see from my plots at -20dBFS signal posted earlier.
Well I didn't really say in all fairness to yourself... open to interpretation 🙂
I did think 1 the cleanest but if 1 and 3 were say amplifiers I would probably have bought 3 on sound quality alone.
I did think 1 the cleanest but if 1 and 3 were say amplifiers I would probably have bought 3 on sound quality alone.
I would completely agree if the distortion amplitudes resulted in both files in 1% FS
As the 3H had -40dB at 3H distortion file,
and 3H had -50dB at comb file and higher were decreasing with rate -10dB
The comb distortion was less audible.
Your subjective comments in this test are very valuable and I agree with them, based on my experience with amplifier design and evaluation.
- Status
- Not open for further replies.