Can you participate to a listening test?

[got tinnitus]

I'm a bit late to the party it seems as I just downloaded the files 20mins ago and listened to both only to see the test is closed!!

 

However I found the files not identical and preference was for A.

B in comparison seemed a tad squashed and round-off in the top end (or not as crunchy).

 

Pretty much what others have noted I believe(?).

[bibo01]

Users can still do the test. Also those who did not vote can post their impressions.

[manisandher]

bibo01, thanks for the detailed explanation - very interesting stuff. But what exactly do you mean by 'needles'? Can you perhaps give an example to help me understand?

 

Looking at the differences in what I heard (I did absolutely no analysis on the files whatsoever), it really is amazing that such a 'small' difference is audible. But for me, it clearly was. As I wrote, A (I called it 1) sounded more harmonically rich and 'analogue' and B more detailed and 'digital'. Is this what we would expect from triangular dither?

 

Mani.

[bibo01]

bibo01, thanks for the detailed explanation - very interesting stuff. But what exactly do you mean by 'needles'? Can you perhaps give an example to help me understand?

 

Looking at the differences in what I heard (I did absolutely no analysis on the files whatsoever), it really is amazing that such a 'small' difference is audible. But for me, it clearly was. As I wrote, A (I called it 1) sounded more harmonically rich and 'analogue' and B more detailed and 'digital'. Is this what we would expect from triangular dither?

 

Mani.

manisandher,

don't worry, it does take some time to read carefully all the detailed measurements.

 

If you look at the 3rd picture, the top graph is A-B overlay; the bottom graph is the z axis. The single samples that, following dither, did not sync up with sampling stick out and look like...needles. :-)

 

Dither is a rather complicated matter. It's normally used for truncation during downsampling to randomize quantization error because our ears find hard to discern random noise. I wouldn't be ready to generalize like that.

[Julf]

To be honest such test does not pretend to be a scientific test at all.

 

Unfortunately I have to agree. The problem is that you are asking people to subjectively evaluate if they hear a difference or not. You have no way of knowing if they *really* hear a difference. Instead of doing an A vs B test, you should have done an ABX test - with 3 files, 2 of which are identical, and asked people to try to hear *which* of the two are identical. Even then there are problems, but at least it is much better than a pure "A vs B".

 

So what you have shown is that some people think that they can hear a difference. I think I could have told you that in advance

[sandyk]

"Instead of doing an A vs B test, you should have done an ABX test - with 3 files, 2 of which are identical, and asked people to try to hear *which* of the two are identical. Even then there are problems, but at least it is much better than a pure "A vs B"."

 

Even better may have been to also include the original before it was stuffed around with, to give people some kind of a reference ?

Also it would have been far better to have started out with a decent quality 24/96 digital file than a digitised Vinyl rip with it's poor channel separation etc.

[bibo01]

I agree.

However, I explained how casually it has all started and the nature of the "test".

 

The track was chosen (not by me) because it was a sample which could be distributed freely.

I am sure though that the full version has the same quality.

[Julf]

However, I explained how casually it has all started and the nature of the "test".

 

Sure - appreciated. But just in case you redo it elsewhere - it would be just as easy to do an ABX instead of an AB, but the results would actually be at least somewhat meaningful.

[sandyk]

"The initial concept was to take a 24/96 file (file A), apply triangular dither (no noise shaping) and create a new file from it (file B). The two files are obviously different, although minutely, but could such differences be heard?! The listening test was born!

 

It took place in the open air in a small audio forum and then trasferred to a larger audience on another forum as well. So no trick was involved. Later on, it was proposed to ComputerAudiophile audience. The rest is history! :-) "

 

I have had a shower, and perhaps my brain is already half asleep,(embarassed smile) but does this mean that we are simply comparing the ORIGINAL File "A" with a NEW file "B" created by adding triangular Dither to the original File "A" ?

In other words, is file "A"untouched ?

IF so, then why do file "A" and file "B" both look like the .jpeg that I posted the link to in my previous post ?

I was using the .wav files DL from the link given in the post by Peter later on.

I have now DL the files from the previous RAR again, and get the same results in Sound Forge 9.

Alex

[bibo01]

...does this mean that we are simply comparing the ORIGINAL File "A" with a NEW file "B" created by adding triangular Dither to the original File "A" ?

...

Alex

Yes.

I do not know what happened in your SoundForge pic, but any null test, even by people on this forum who have DL the files from the links provided, shows very very small differences.

[bibo01]

Sure - appreciated. But just in case you redo it elsewhere - it would be just as easy to do an ABX instead of an AB, but the results would actually be at least somewhat meaningful.

Not by much though given we are on a forum.

[sandyk]

Then something is VERY wrong,because both A and B appear as in the .jpeg link that I provided earlier.

http://imageshack.us/photo/my-images/17/tracka.jpg/

I have double checked as stated in my last post, by freshly downloading from the RAR, and I get the same results for both the files from the rar and the .wav file DLs.

IF I am correct, then it's not surprising that the results appear to be random ! !

Alex

 

P.S.

I have now checked again, and all 4 files look similar in SF9 to the .jpeg that I linked to.

I will be happy to provide SF9 screen grabs of all 4 files via email on request.

The time and date of each indicates they were DL several minutes apart.

[jhwalker]

Then something is VERY wrong,because both A and B appear as in the .jpeg link that I provided earlier.

http://imageshack.us/photo/my-images/17/tracka.jpg/

I have double checked as stated in my last post, by freshly downloading from the RAR, and I get the same results for both the files from the rar and the .wav file DLs.

IF I am correct, then it's not surprising that the results appear to be random ! !

Alex

 

P.S.

I have now checked again, and all 4 files look similar in SF9 to the .jpeg that I linked to.

I will be happy to provide SF9 screen grabs of all 4 files via email on request.

The time and date of each indicates they were DL several minutes apart.

 

Not sure what you are getting at - the whole point is that the files should look almost identical, as the OP clarified. The differences are very subtle and there was even dispute over whether people would be able to hear them at all.

 

So I'm not sure what point you're trying to make - I'm either just missing it or you need to spell it out a bit

[goldsdad]

Then something is VERY wrong,because both A and B appear as in the .jpeg link that I provided earlier.

http://imageshack.us/photo/my-images/17/tracka.jpg/

I have double checked as stated in my last post, by freshly downloading from the RAR, and I get the same results for both the files from the rar and the .wav file DLs.

IF I am correct, then it's not surprising that the results appear to be random ! !

Alex

 

P.S.

I have now checked again, and all 4 files look similar in SF9 to the .jpeg that I linked to.

I will be happy to provide SF9 screen grabs of all 4 files via email on request.

The time and date of each indicates they were DL several minutes apart.

 

 

Of course the waveforms look the same at the scale of your image. It's the least significant of 24 bits that differs between those samples which have any difference. The samples have positive and negative values, therefore where there is a difference, it's equivalent to 2 to the power of minus 23, i.e. 0.00000012 units on a linear scale of -1 to 1.

[sandyk]

jhwalker.

 

The amplified low level "noise" (>-70dB ) at the beginning, as seen in the SF9 screen grab is not normal.

It should not appear in the untouched "A" file as well , unless perhaps it was a very poor digital conversion.

It appears in all 4 files that I downloaded.

I have never previously seen low level noise that looked like that in other .wav files..

It certainly wouldn't help with identifying differences between both files where low level Triangular dither was added?

Perhaps there is some other explanation ?

Alex

[goldsdad]

Examination in Sound Forge 9 reveals they are sourced from Vinyl with quite a high level of Wow and Flutter,and that although SUPPOSEDLY 24/ 96, the contents , judging by the digital looking waveforms suggest something perhaps more like MP3 ? ( I'm no expert !!).

 

They are not from vinyl. The digitally captured audio has been mixed in an analogue console and redigitized.

They do not contain wow and flutter. If you hear wow and flutter then I guess you are imagining it because you mistakenly think it is visible in your graph.

They are genuine 24/96.

The waveforms are no more "digital looking" than any other 24/96 audio.

A spectrogram reveals that they have never been anything like MP3.

 

 

 

[Julf]

Not by much though given we are on a forum.

 

Well, AB totally relies on subjective self-assessment, while ABX allows for independent, objective verification - forum or not.

[Alkan]

Ciao Gianluca,

Besides technical stuff is away my comprehension, it was very interesting to accompany all the feelings involved in your test. To commemorate it ends I will open a bottle of Brunello and listen Mahler's 9th with Gergiev

Dalton

[Julf]

Ciao Gianluca,

Besides technical stuff is away my comprehension, it's was very interesting to accompany all the feelings involved in your test. To commemorate it ends I will open a bottle of Brunello and listen Mahler's 9th with Gergiev

 

If it is the Society Of Sound version, I am sure the Brunello smoothens the sharp cut-off of high frequencies

[Alkan]

If it is the Society Of Sound version, I am sure the Brunello smoothens the sharp cut-off of high frequencies

 

It is a ripped CD with dBpoweramp. Has it the same cut-off?

[Julf]

It is a ripped CD with dBpoweramp. Has it the same cut-off?

 

Well, certainly a bit lower, as a RedBook CD won't contain anything above 22 kHz, but the Brunello should still taste OK

[Alkan]

Well, certainly a bit lower, as a RedBook CD won't contain anything above 22 kHz,...

 

Thanks. Anyway, the music is great.

 

...but the Brunello should still taste OK

 

Specially this 78's one

[bibo01]

Just to add something to the talks after the test...

 

Following the possibility that it is likely that at this level one is hearing differences in artifacts created by one's DAC, which may create greater differences in the analog output, the following is a NTD (Null Test Difference) on the DAC analog output by Tom Gefrusti (he has a very optimized system for this type of measurement). The graph is very zoommed in order to show the difference.

 

According to it, harmonic distribution on the DAC outputs changes continuously, even more than the change between A and B. In practice, there is a randomization of the harmonic content. So much so that the difference can be greater from A to A again.

 

 

In fact, contrary to ladder DACs, where the waveform output will not vary from play to play, except for noise, voltage and temperature changes, delta-sigma DACs introduce a new source of randomness - the state of the delta-sigma modulator, which is a pseudo-random number generator. One gets a completely different pattern of output noise when feeding them two constant input signals that differ by one in the least significant bit. The ability to hear these differences (especially in the DC case) must be taken as a fault in the DAC, not a "resolving" feature.

[Julf]

Bibo01,

 

I know it is the morning after Queen's Night here in Amsterdam, and I haven't had my second cup of coffee yet, so I might be especially thick, but I just can't manage to understand what you are trying to say here.

 

According to it, harmonic distribution on the DAC outputs changes continuously

 

Changes based on what?

 

In fact, contrary to ladder DACs, where the waveform output will not vary from play to play, except for noise, voltage and temperature changes, delta-sigma DACs introduce a new source of randomness - the state of the delta-sigma modulator, which is a pseudo-random number generator. One gets a completely different pattern of output noise when feeding them two constant input signals that differ by one in the least significant bit.

 

Yes. And that randomness of the delta-sigma noise manifests itself as just that - noise. Just like the "analog" noise from a ladder DAC. Both are random noise, and both are perceived as noise by the ear. Not sure what your point is.

[bibo01]

Changes based on what?

On/Off switch, time between tests, PS, how clean and stable current is...

 

This is an hypothesis of what can happen downstream.