[porthillsbomber]

Here are the files
 
http://www.mediafire.com/?t8b6pocc40yxb
 
I've done something to the files that will shed light on to something that is mentioned a lot but I have never seen any evidence on. I'm not going to tell you what the differences are but it's not compression related or any special effects etc. If you do the test I would like you to PM me which ones you think sound the best, in order too if you can. If you can't tell the difference say that too. Once I get 20 results or so I'll tell you what I've done.
 
The song is the first 25s of Chan Chan by the Buena Vista Social Club but if someone can suggest a good reference song for this test I can apply the same thing to that song too.
 
Thanks for your help.
 
Edit: Added first 35s of Evil Ways by Santana
 
Edit: Added the songs in zip files
Edit: Added In the water by Annie and the Beekeepers

 

[Yoga Flame]

I passed!

 

[porthillsbomber]

We've got three results so far. Keep 'em coming.

 

[Br777]

just did your test.. pm soon coming
 
btw, props for continuing to upkeep your thread through a freakin earth quake!
 
also.. if nothing else you have introduced me to buena vista social club ( after years of hearing about them but never hearing them) whose albums i will be buying.. thanks!
 

 

[porthillsbomber]

Thanks we're up to 5 now. Thankfully we're all fine at my place except the loss of a chimney and having to use a hole on our lawn as a toilet! Yeah they're great I was actually introduced to them from reading a beta22 review.

 

[porthillsbomber]

I've added another song, the first 35s of Evil Ways by Santana. Got 7 results so far and they're fairly consistent so I might reveal what I've done when we get to 10.

 

[Lenni]

question: is this something you did applies to the whole of a file, or to just some parts of it?
 
also: if you'd put all the files into one zipped folder you'd have more results, I think.

 

[porthillsbomber]

Lenni, I did it to the whole file. One of the files it the original by the way. Thanks I'll do that.

 

[porthillsbomber]

Well I've got about 10 results now and they're fairly consistent so I'll tell you what I did. I have added 5% harmonic distortion in the different harmonics from 2nd to 5th plus there is the original. The general consensus so far in best to worst sounding is
 
original -> 3rd -> 2nd -> 5th -> 4th
 
about 25% can't distinguish any difference (I suspect it may be more because I've had a lot more downloads than results) 50% can distinguish only the 4th and 5th and 25% can just distinguish between the rest also.
 
This is how I did it, please tell me if it was wrong.
 
Took the original flac I had and upsampled to 24 bit wav.
 
Imported to matlab and added harmonic distortion by using the Chebyshev n polynomial of the original waveform to create nth order harmonic distortion.
 
multiplied the by 0.05 and added it to the original wave form.
 
converted to 24bit flac so I could use replay gain
 
downsampled to 16 bit wav applying replay gain and dithering. 
 
 
So far the results seem to indicate that the odd harmonics are favoured to even ones isn't that the opposite of what we're told?
 

 

[nikongod]

 
Quote:

Well I've got about 10 results now and they're fairly consistent so I'll tell you what I did. I have added 5% harmonic distortion in the different harmonics from 2nd to 5th plus there is the original. The general consensus so far in best to worst sounding is
 
original -> 3rd -> 2nd -> 5th -> 4th
 
(snip)
 
 
So far the results seem to indicate that the odd harmonics are favoured to even ones isn't that the opposite of what we're told?
 

 
To answer your comment/question on the odd/vs even harmonics: Kind of sort of yes no
 
The general guideline is that LOWER order harmonics are preferable to higher order. Your test agrees with this, with BOTH 3rd & 2nd better than 5&4. There was a "weighted" distortion scale proposed in the (I want to say late 1950's, but I may be wrong) that reflected the way humans percieve distortion in the published distortion number VS the simple measurement technique that EVERYTHING can pass. Obviously the weighted scales got rejected... Its funny to think about this weighted scale when looking at the distortion spectra of a power amp that NEVER drops into the noise floor.
 
Anyways, after that even harmonics are generally perceived as being on the softer side, where odd harmonics can actually be perceived as being more detailed! at 5% THD added there is no doubt that they are audible... 
 
I wonder how the test would go at 0.5 to 1%THD, with "pure" harmonics added like this. 1% is still audible, and 0.5% may be (it certainly is in tones, Im not as sure about music). In any case, VERY few people listening to headphones run an amp hard enough to even get to 0.5%. Not even "open loop" tube guys, as much as the scope jockeys love to rag on them.
 
FWIW, the past few times I have tried (and others have tried too) the threshold of hearing THD in an ABx test on tones is 0.5% for a VAST majority of people. We did this at a NJ head-fi meet a few years back, it was quite fun.

 

[xnor]

Quote:

This is how I did it, please tell me if it was wrong.
 
Took the original flac I had and upsampled to 24 bit wav.
 
Imported to matlab and added harmonic distortion by using the Chebyshev n polynomial of the original waveform to create nth order harmonic distortion.

Ahhhm the chebyshev polynomials will output a defined set of harmonics only if the input signal is sinusoidal. More complex signals (like those tracks you posted) will get you plenty of intermod products.
 
I guess you'd need to split the input signal into several bands using BP filters and process each band individually to reduce IM. 

 
 
I'm sorry, but because of that I think it's best to declare the conclusions invalid.

 

[porthillsbomber]

Yes I thought about that but as far as I can tell you can't have something that produces harmonic but not IM. I've thought about this experiment from the amp point of view. If and amp is linear you get no harmonic or IM distortion if its non-linear you get harmonic and IM distortion.

 

[xnor]

Hmm could you run RMAA please, then in the top right configure 16 bit / 44.1 kHz, generate the test signal (floppy disc button), process the test file the same way you processed the tracks above and then analyze the resulting wav files (button next to floppy)?
 
Would be interesting to see what the THD and IM graphs look like. 
 
Or share your script and I will analyze it. 

 

[porthillsbomber]

Ok I'll do that but I'll post up my code soon once I've annotated it, it's not very complicated.

 

[porthillsbomber]

Here it is
 
 

  function hd(filename) l = 35; %length used in seconds [or,fs,N] = wavread(filename,[1 44.1e3*l]); %import first l seconds from wav hd2nd5p = or + 0.05*(2*or.^2 - 1); %adds 5% 2nd harmonic distortion hd3rd5p = or + 0.05*(4*or.^3 - 3*or); %adds 5% 3nd harmonic distortion hd4th5p = or + 0.05*(8*or.^4 - 8*or.^2 + 1); %adds 5% 4nd harmonic distortion hd5th5p = or + 0.05*(16*or.^5 - 20*or.^3 - 5*or); %adds 5% 5nd harmonic distortion n=length(filename); wavwrite(hd2nd5p,fs,N,[filename(1:n-4) 'hd2nd5p.wav']) %exports to wav with the same filename with hd2nd5p on the end wavwrite(hd3rd5p,fs,N,[filename(1:n-4) 'hd3rd5p.wav']) wavwrite(hd4th5p,fs,N,[filename(1:n-4) 'hd4th5p.wav']) wavwrite(hd5th5p,fs,N,[filename(1:n-4) 'hd5th5p.wav']) wavwrite(or,fs,N,[filename(1:n-4) 'or.wav']) %exports the original (first l seconds)