24/96 Files?

[Brooko]

Hey thanks for doing it  A lot of people won't.
 
 
Although the 7/10 sounds good - it's actually a statistical fail with an overall probability you were guessing.  To have a statistically relevant pass, you need to be under 5% probability of guessing.  Obviously the more iterations the better (so 15 would be a good number for a really good statistical sample).  Did you actually try it with the 'hide results' box checked?  Makes it even harder ..... and more revealing.
 
If you don't want to go any further - thanks for coming this far anyway.  It is very enlightening isn't it?
 
Would be nice if you could share the two files somehow as well - so other people could try same files for themselves.  Have a great Xmas.
 
 
Cheers - Paul

 

[nelamvr6]

Quote:

Hey thanks for doing it  A lot of people won't.
 
 
Although the 7/10 sounds good - it's actually a statistical fail with an overall probability you were guessing.  To have a statistically relevant pass, you need to be under 5% probability of guessing.  Obviously the more iterations the better (so 15 would be a good number for a really good statistical sample).  Did you actually try it with the 'hide results' box checked?  Makes it even harder ..... and more revealing.
 
If you don't want to go any further - thanks for coming this far anyway.  It is very enlightening isn't it?
 
Would be nice if you could share the two files somehow as well - so other people could try same files for themselves.  Have a great Xmas.
 
 
Cheers - Paul

 
 
Well you can count it as a fail, I'm counting it as a win! 

 
I wouldn't mind sharing the files, but I wouldn't want to get Head-Fi in trouble for harboring file sharers.
 
If you know of a way that I could share them without getting the site into trouble I'm game.
 
It was interesting.  The SQ was very close, and I never denied that it would be.  But you know how it goes, that final 1% is the most difficult.  
 
The differences I heard were VERY subtle.  If it was music I wasn't familiar with, I probably would have done a lot worse.
 
The way I look at it, I'm going to a lot of trouble making needle drops, and I'm spending a lot of money on audio equipment.  Hard drive space and flash memory is dirt cheap right now, why skimp on the resolution?

 

[stv014]

Quote:

Well you can count it as a fail, I'm counting it as a win! 

 
As the log says, there is still a 17.2% chance that you were only guessing. Doing more trials and combining the scores (for example, if you try it again 10 times and guess right 6 times out of that, then the overall score becomes 13/20), or simply increasing the number of attempts from 10 to 20 would give a more clear result.
 
Quote:

I wouldn't mind sharing the files, but I wouldn't want to get Head-Fi in trouble for harboring file sharers.

 
You can share a sample that is shorter than 30 seconds. This also has the useful side effect of keeping the file sizes reasonably small. Extract the part of the track where you think the difference is most audible.

 

[stv014]

Quote:

I think that  after, say, 30 iterations our hearing system  (ears/brain) is exhausted and therefore we can't always tell the difference.

 
It is not necessary to do all the iterations at once. You can do 3*10 trials over 3 days, for example, and then sum the results to get an overall score.

 

[nelamvr6]

Yeah, I'm a little busy listening to music and sipping bourbon.
 
I'll look into cutting the files down to 29 seconds and posting them here.

 

[JakeJack_2008]

Quote:

Yeah, I'm a little busy listening to music and sipping bourbon.
 
I'll look into cutting the files down to 29 seconds and posting them here.

 
David Gilmore, please do it. I'll be getting the ODAC and O2 and  HDTracks and also will do my comparisons.
What I know, I can hear   the sound improvement on DVD-audio and SACD over the sound on  regular CDs.

 

[nelamvr6]

OK, attachment size limitations prevent me from attaching the files here, so I've uploaded them to SendSpace.
 
Here are the download links:
 
http://www.sendspace.com/file/92q1c4
 
http://www.sendspace.com/file/f69d2e
 
http://www.sendspace.com/file/216ckg
 
http://www.sendspace.com/file/mp6u57
 
 
There are 4 files in all.  I was able to listen to the entire track while I was doing the ABX test, so I didn't think it was fair to limit anyone else taking the test to just 30 seconds.  So there are two 30 second snippits of the track, one each at 96/24 and at redbook.
 
The needle drop was made using a VPI Scoutmaster turntable with a Dynavector DV-20X2 low output moving coil cart.  The phono stage is a Trigon Vanguard II.  The DAC is an E-Mu 0404 USB.
 
The original was recorded using 96kHz sample rate.  Software used was Adobe Audition 3 for the recording and editing, ClickRepair, and the file was converted to FLAC using dbPoweramp.  Foobar was used to add per track ReplayGain tags to these 4 snippits.

 

[Puranti]

But the samples are not the same, this is not possible to do a valid abx
But the 1st hi rez sample I can hear a peak at a high frequency.

 

[nelamvr6]

The only thing that was done to those files was the low res (redbook) files were downsampled from 96/24.  Nothing else.

 

[Puranti]

Yep but what I mean is that it's not the same part of the music, hence the samples are not the same, to do a proper abx I need the same tracks.

 

[nelamvr6]

If you look at the file names you'll see that there are files with Snip 1 in the name and also files with snip 2 in the name.
 
Also, the snip 1 files are tagged as track 01, the snip 2 files are tagged as track 02.

 

[stv014]

Ideally, the low resolution file should be converted back to 96/24 for the comparison, so that potential playback differences of the two formats are eliminated; this might not be necessary, though, and it does increase the download size. By the way, the high resolution file has a steep cut-off at about 20 kHz, maybe the track was originally recorded at CD quality ?

 

[nelamvr6]

Quote:

Ideally, the low resolution file should be converted back to 96/24 for the comparison, so that potential playback differences of the two formats are eliminated; this might not be necessary, though, and it does increase the download size. By the way, the high resolution file has a steep cut-off at about 20 kHz, maybe the track was originally recorded at CD quality ?

 
 
Are you talking about attempting to eliminate any differences in the playback equipment or software?  I can't say that methodology makes a lot of sense to me.  I understand what you're trying to accomplish, but I'm not sure that's the best way to do it, or if there even is a good way to do it.
 
I'll have to think about that.
 
At any rate, you have the files, if you want to upsample them feel free, just be sure to re-label them if you distribute them...
 
The original file , as I have mentioned, was recorded from vinyl.  You're not likely to get a lot of energy above 20 kHz on vinyl...

 

[stv014]

Quote:

Are you talking about attempting to eliminate any differences in the playback equipment or software?  I can't say that methodology makes a lot of sense to me.  I understand what you're trying to accomplish, but I'm not sure that's the best way to do it, or if there even is a good way to do it.

 
I meant potential differences in DAC performance (filtering etc.), even though they should not be audible with a good DAC, it is best to minimize the number of variables in the test, especially if the samples are meant for public testing on a variety of unknown hardware and software configurations. Other, partly software related issues are different latency at the two sample rates, or artifacts (clicking etc.) when switching from one sample rate to another; problems like these could provide cues to find out which file is playing. By converting back to the original format, and making sure that the files are accurately level matched and synchronized, it is only possible to hear differences that actually result from the conversions. One may object to the idea of additional processing, but it obviously cannot restore information that was lost in the downsampling and quantization, so it cannot make the test unfairly harder.
 
Quote:

The original file , as I have mentioned, was recorded from vinyl.  You're not likely to get a lot of energy above 20 kHz on vinyl...

 
What I meant is that it drops off abruptly above 20 kHz, that is not what one would expect from a purely analog recording:

 

[nelamvr6]

Quote:

 
What I meant is that it drops off abruptly above 20 kHz, that is not what one would expect from a purely analog recording:

 
 
Well, I definitely recorded it from vinyl.  I can't say anything about the master tape used to master the vinyl is it possible that is the source of that drop off?
 
here is a view of the spectral frequency display from Audition 3:
 

 
 
I'm not sure what that artifact is around 15k either...