X is the signal voltage, not the level. Making the output depend partly on X^2 is how second order distortion is added. If it was X^3 it would be third order etc.Jay said:
To understand distortion you have to understand algebra and trigonometry.
Yes, I meant voltage level, the X. The question was what happened to db level and dynamics. I'm not yet clear whether it has to be affected or not. I don't know the mechanism to add the distortion, technically... Just to see if possible error could creep in.
I want to make sure the above is perfect and well controlled (as SY like to say) so I can draw an accurate and meaningful conclusion from what I have perceived in the listening test.
I expect PMA ensured equal level - he usually knows what he is doing. To achieve this, he would need to set b equal to 1.Jay said:
Dynamics are not affected by second order distortion. You would need third order for that - and it can be compressive or expansive, depending on the sign of the X^3 coefficient.
You do - he told us. He applied a polynomial to the signal.
The test appears to show a slight preference for the more distorted version. Not too surprising.
Pavel, very nice to see the results, thank you for putting your time and effort into this.
Since this test is about preference, I have to admit I may have lied. I made a choice for the clean signal not because it sounded better, but because it sounded cleaner (or less loaded, difficult to put in words).
The interesting variable to test is: till what point is added H2 preferred by most, before it becomes detrimental by popular vote. WAVAC may have hit the sweet spot.
Since this test is about preference, I have to admit I may have lied. I made a choice for the clean signal not because it sounded better, but because it sounded cleaner (or less loaded, difficult to put in words).
The interesting variable to test is: till what point is added H2 preferred by most, before it becomes detrimental by popular vote. WAVAC may have hit the sweet spot.
-
Thanks, tho I still cannot see the why/how... yet. And the db level? The process change the level, then independent manual adjustment is performed?
I mean the technicality, the process, like what tool (device or software) to use...
I still doubt it. There are many reasons why we picked 04b. Even tho the instruction was to choose what we prefer but we can choose simply based on what we think is undistorted, for example.
I believe that drawing conclusion is a sophisticated process.
Thanks, tho I still cannot see the why/how... yet. And the db level? The process change the level, then independent manual adjustment is performed?
I mean the technicality, the process, like what tool (device or software) to use...
I still doubt it. There are many reasons why we picked 04b. Even tho the instruction was to choose what we prefer but we can choose simply based on what we think is undistorted, for example.
I believe that drawing conclusion is a sophisticated process.
This is what I meant. We cannot quickly conclude that preference was for H2.
But I'm not "giving up". There must be explanation to why 04b sounded cleaner and more relax (even tho boring). It's hard to believe if there is a certain way to add H2 to make the sound more softer... But if so, don't we have something extra-ordinarily good "inventtion" here??? 😕😀
As I said, if he did what he said he did (simply applied a polynomial) then all he needed to do was ensure that the linear term had a coefficient of 1. No need for any adjustment, manual or automatic.Jay said:
Provided that it did what it was supposed to do, no more and no less, then it doesn't matter at all what did it. Of course, this assume sensible behaviour when the polynomial causes overflow - if that occured anywhere.
Of course, if people choose on some other basis than preference then they were not doing the test properly. However, for whatever reasons, when asked to choose on the basis of preference there was a small bias in favour of the distorted version. It could be that they really did prefer it. They might have thought it was less distorted, even though they didn't prefer it. Some might have chosen randomly.
It is unsurprising that some people picked the distorted version. It is also unsurprising that some of those who did this will now wish to offer excuses. It has been known for about 60 years that some people (I forget whether it is a majority or not) prefer some low order distortion with their music.
Well, I thought 04a sounded cleaner (not 04b), but not better per se. But again: a different piano might have the same timbre as the piano of 04b, and the same goes for the violin. So, this test is not about naturalness, but rather about whether you like a violin and piano with a slightly more complicated timbre. Just wait for the flute to come in.
Got it (No need adjustment), thanks.
Overflow. So it's a software... My point in asking is because, like I said in my first post, it didn't make sense... even now (whichever is the distorted one, it still doesn't make sense). You know, I trust my ears. I tried to relate theory/Physics with what I heard, and something just didn't "click" till now.
GRANTED, I preferred 04b when I was at work listening using Windows Media Player with (I think) a Logitech desktop speaker. But at home using my new speaker, I didn't think about preference, but most probably I prefer 04a because I enjoyed it. I had only 2 or 3 minutes listening to 04b in the morning, and I could perceive that it was calmer, tho boring.
If you think that I wish to offer excuses, you are WRONG. I stand by what I said and heard. If I got sufficient confirmation that everything is in order, then the conclusion was the same or close with what I have stated many times in the past...
With average system, H2 is preferred, but once you got sophisticated system (I'm more a speaker guy), H2 is not so different with other distortion.
But I'm here seeking explanation about something that is unexpected to me of course...
So you perceived that 04a (the undistorted one) to sound cleaner... But you choose 04b (?) tho you knew that 04a is undistorted... Okay, my mistake.
I don't like the music so I didn't pay attention at all to naturalness etc (I don't know how a flute should sound like). I pay attention to DISTORTION only, or whether the sound is DISTURBING or not... And I thought that 04a is the distorted one but I was wrong... But I still cannot understand, that's why I'm seeking confirmation before making conclusion or even trying to listen again (whether 04b is really calmer than 04a).
Or may be my wife turned down the volume such that when I listened to 04b in the morning I felt as if 04b is calmer 😀 But, no, it's not an excuse, because I thought my wife wouldn't have done that so I didn't ask.
There is no other way to modify a data file!Jay said:
If the music is represented as a voltage, then you can apply distortion in other ways - some deliberate, some accidental. The advantage of a software-added distortion is that you can know exactly what has happened, while hardware added distortion always has higher order terms plus added noise and a change in frequency response. However, people often prefer hardware-added distortion too - so the source of the distortion is not too important.
I am still unclear what it is that puzzles you. PMA took a file of music, applied a simple second-order polynomial, and then asked people to choose which they preferred. A small majority (of a small sample) chose the distorted version.
If you don't know what a flute should sound like, how can you hope to detect distortion? Maybe a flute is supposed to sound disturbing? (It isn't, but some instruments do!)
Last edited:
Jay, it's always going to be combinations - the style and quality of the source material, plus the quality of the playback chain. Depending upon everything, that which nominally is more distorted may sound better in the particular circumstances - I just tried listening to 4a and 4b again, and 4b still comes across better in key areas for me; 4a is just too stark, threadbare for my laptop configuration at the moment.
Wow! I understand now. So a software (What software has this functionality BTW?) I need confirmation because if some error happened during the process, then case closed, I have nothing left unanswered.
Read my post before Pavel revealed the answer:
How can I answer that?
Several days ago I posted in Blowtorch thread how I did DBT or designing speaker using ears. To know which sound is better, A or B, I can compare them with C, which has been pretty well stored inside my memory. This is regarding CORRECT TONALITY. In this listening test I don't have the C. But I don't need to listen to tonality to detect distortion. It's different.
No, I compared my feeling when I listen to A and my feeling when I listen to B. Distortion is always fatiguing. In this test 04a was not fatiguing (as I said) that's why I was confused... (and it is not the only thing that did not make sense)
Exactly. And the file is dithered (TPD), as everyone can see from posted distortion plots, that are results of one pure sine and twin tone through the same function and exhibit no digital artifacts, just correct 2nd harmonic distortion components as they should be.
Guys will complain, that was expected.
This is also well expected 🙂 Thanks anyway for this test. Now I know how a H2 sound (with no other harmonics present). Very useful because I often am confused when I have to assess distortion performance (I don't know whether I was listening to H2, H3 or THD...). Now I can keep these files and do some "research" with them to understand more.
Is it? If you expect distortion (even low order) to always be fatiguing then I can understand your puzzlement.Jay said:
What experience? You were simply asked to express a preference. Experience would only be relevant if you were asked to say which you thought sounded more like the real thing, or which had more distortion. What is unnatural about added distortion to sound - lots of people choose to use 'tube buffers' or SET amps, are they unnatural too?
Indeed.
Why should I make a preference between ANY two files that is far from sounding musical, from a desktop computer??? Of course I had my own agenda 😀
In doing the analysis, I need to apply my knowledge, regarding sound. What knowledge? Empirical experience! I have never heard a H2 (alone without other harmonics) for example...
"Softening" the sound was not what I expect from so called distortion.
And that's only a few...
Now after this test... I think I need to revise my project list. I need to put Vifa XT25 on the queue. Crossing above 3KHz, it's high order distortion performance (H3 and above) outperform so many top class tweeters...
It's the player (speaker, amp, soundcard, etc.)
And I don't know whether the file is good or not for you. All my other music, thru the same system, sound much better than these sample files.
Your preference is only a function of your system. In my office desktop speaker, 04a was so disturbing. But not thru my low distortion speaker (even using horrible laptop sound card). So what have you expected from this test??? Get 5 audience and then draw a meaningful conclusion based on the result??? You don't know Pavel 😀
- Status
- Not open for further replies.