24/192 Music Downloads ...and why they make no sense

[wolster]

(2012-03-19 16:43)SaltyDog Wrote:  I don't get the format not dropping in price at the same rate as other technology. Nothing physical is being produced, shipped, or inventoried. Vinyl has higher material costs and are produced individually. CDs have to cost more to distribute too.

One of the best made points, IMO.
I have the same issue with eBooks -but that's another story

[Delboy]

(2012-03-19 21:20)hcl Wrote:  

(2012-03-18 23:38)Evo Wrote:  

hcl Wrote:I guess you did not quite get the point the first time then? It is not a matter of confidence when you reach a 50/50 result. It shows exactly just what it shows. No difference where detected in that specific experiement. It says nothing else. A statistically significant difference result shows on the other hand that a defference is possible to detect and that there allways is a possibility to do so (even if another experiement happen to show the opposite). Actually one experiment showing a statistically significant detectability have precedence over 1000 non conclusive tests. One can make ones own conclusions from it, but fact is. A 50/50 result is only indicating something. It is no proof.

If you perform a double blind test to determine if a subject can tell the difference between two different resolution music files and they get it wrong as often as they get it right i.e., in 100 trials they get it wrong 50 times, and right 50 times, then it is reasonable to conclude that they cannot tell the difference between the files. If you then get more people to take the test and they too get it right as often as they get it wrong, then you can conclude that they too cannot tell the difference. Assuming the study is sufficiently well designed to optimise the chance of telling a difference if there were one, then the more people who cannot tell the difference, the more confidence you can have in the outcome, which is that even if there is a difference, it is one that cannot be heard by the participants in the trial, and, if the sample size is big enough, a significant proportion of the population at large. If you cannot follow this then I will ask one of my statisticians to suggest some references you can read about the design of equivalence and non-inferiority tests. I suspect the confusion lies in the use of the term 50/50 and the association of this with chance.

Suppose we change amp in that set-up and the result come up with a 100/0 that the format difference is detectable. What does that say about the extensive tests and the (at that point obvious) pre-mature conclusions made, about the format differences, based on the first tests?

Good point, but I did say "Assuming the study is sufficiently well designed to optimise the chance of telling a difference if there were one", and that would include using the best quality equipment to ensure it was not a variable. Also the only conclusion made based on this study is that it did not show a difference, and, given it appears to have been quite well designed, this may well apply to a wider population than the participants in the study. This does not mean that another study may alter this conclusion, that is how science moves forwards. I would be delighted to see others repeat this study and have an open mind with respect to the outcome. However, at present, given no-one has provided contrary evidence from another equally well designed study, we are left with the possibility that most people cannot hear a format difference. And it is because this is only a possibility and not a definitive conclusion that we are involved in this most enjoyable debate.

[MnM]

(2012-03-20 00:17)Warren Wrote:  

(2012-03-19 22:01)MnM Wrote:  your workings and explanations are completely barmy!

...

I won't pretend I really knowing what's going on in the digital filter

These two comments seem to cancel each other out.

No mate. I go up to where I'm comfortable and then stop.

[maralunga]

For those of you who want to compare or randomly test some different files there is a link from GIMELL records and e-classical (go to bottom of the page). It only includes up to 24/96 though.

http://www.gimell.com/musicstore-downloa...files.aspx

http://www.eclassical.com/pages/24-bit-faq.html

Have fun.

EDIT: Forget about the Gimell; it only gives you a test if your DAC can manage the format. eclassical works.

[VicJayL]

(2012-03-19 15:20)simoncn Wrote:  

(2012-03-19 15:05)VicJayL Wrote:  

(2012-03-19 14:18)simoncn Wrote:  That isn't quite what I meant, though I'd be pleased to try ABX on one or two recordings if you could point me to a simple way of doing this.

Congratulations on a refreshingly open mind for this thread Simon.

I outlined the procedure I used in post number 42 and as I wasn't deluged with objections to it have assumed that it was as fair and objective as I could make it.

The two key points seem to be listening to a randomly generated playing order (that you are unaware of, obviously) and that you have no aural or visual communication with the person who made the list and operates the progress through it. I had my wife toss a coin 15 times to get the playing order in advance (in another room) and then operate the iPad at my signal after I had noted down "16bit" or "24bit". The "playlist" of the two tracks were, of course, one for "heads" and one for "tails". I made my choice and asked for the next track. She followed the previously generated order through the list to the end. I honestly can't see how this could be any more objective. Then I made another list out of the room and she sat in the hot seat. The result was: her - 8 correct; 7 wrong. Mine - 7 correct; 8 wrong. ie No better than guesswork.

Imagine how different this debate would be if a couple of dozen members did the same and posted their results!

Good luck with yours. I hope you will want to post your results too.

Vic.

Thanks for this. I'm trying to imagine my wife's reaction if I asked her to go through this procedure.

(pause for a moment)

OK, I think I've got that now. Maybe I'll spend an hour cooking up a software solution to randomly generate a .dpl file to play in Kinsky. Are there any issues with using this as an ABX test?

Can't get my head around this one, but...

I would think one way of doing it if you are on your own is to generate that random list (tossing a coin) then load a playlist in that order with the two 16bit and 24bit tracks (choose a short one!) The longer the list the more accurate the result. When you are ready to listen, tap one to start, averting your eyes so you are not sure which one begins playing. Record your opinion of which is which right to the end of the playlist then compare your result with the playlist, starting from the last track, moving backwards up the list till you reach the one you started with.

This ensures random play order as well as your not knowing which is which when they are playing them. Can anyone see anything wrong with this as a fair blind test?

It might take a while to set up and do, but if it produces the result it seems likely to (and you act upon it) you are nearly doubling the money you have to spend on great download music.

Vic.

[simoncn]

(2012-03-21 12:07)VicJayL Wrote:  Can't get my head around this one, but...

I would think one way of doing it if you are on your own is to generate that random list (tossing a coin) then load a playlist in that order with the two 16bit and 24bit tracks (choose a short one!) The longer the list the more accurate the result. When you are ready to listen, tap one to start, averting your eyes so you are not sure which one begins playing. Record your opinion of which is which right to the end of the playlist then compare your result with the playlist, starting from the last track, moving backwards up the list till you reach the one you started with.

This ensures random play order as well as your not knowing which is which when they are playing them. Can anyone see anything wrong with this as a fair blind test?

It might take a while to set up and do, but if it produces the result it seems likely to (and you act upon it) you are nearly doubling the money you have to spend on great download music.

Vic.

If I have two versions of the same track, it's simple to randomly generate a .dpl playlist with 15 entries where each entry is created by an "electronic coin toss" that has an equal chance of selecting the high-res or low-res version of the track. The names of the playlist items would be "Track 1" to "Track 15" so there's no indication of whether "Track 1" is high-res or low-res, etc. It would be necessary to set the DS display to not display the bitrate and also to use a control point that doesn't show the bitrate while the track is playing.

As I listen to each track, I would write down 1=low-res, 2=high-res, etc. At the end of the session I would look at the .dpl file to see which tracks were low-res and which were high-res, and mark my answers right or wrong.

[VicJayL]

(2012-03-21 15:02)simoncn Wrote:  If I have two versions of the same track, it's simple to randomly generate a .dpl playlist with 15 entries where each entry is created by an "electronic coin toss" that has an equal chance of selecting the high-res or low-res version of the track. The names of the playlist items would be "Track 1" to "Track 15" so there's no indication of whether "Track 1" is high-res or low-res, etc. It would be necessary to set the DS display to not display the bitrate and also to use a control point that doesn't show the bitrate while the track is playing.

As I listen to each track, I would write down 1=low-res, 2=high-res, etc. At the end of the session I would look at the .dpl file to see which tracks were low-res and which were high-res, and mark my answers right or wrong.

Neat! Go for it. And do tell us the result.

Well, that's two ways - one for the savvy and one for the dinosaurs like me! Looking forward to a seeing some results. Would anyone bet their mortgage on a result?

Vic.

[ukandrewf]

(2012-03-21 15:02)simoncn Wrote:  It would be necessary to set the DS display to not display the bitrate and also to use a control point that doesn't show the bitrate while the track is playing.

Am sure there is a low tech solution to this - stick some card over the info?

Also all of this will depend on the quality of the system used for testing.

Some systems may not be able to show the difference.

I was just playing a/b testing my media centre pc against the renew ds - the difference was very clear .. But if I only had the media centre I doubt I would really hear the benefit of 24 bit.

Someone once said to me that the difference between cd and lp is in the silences, they seem deeper and more characterful, more real. I suppose this should be true as compressed media will compress the highs and lows while still playing the middle where most info is. The same must also be true of 24 bit that it's the high and low info which add realism and the playback system must be able to fully realise those highs and lows to make it identifiable.

Just fwiw

Andrew

[fbee]

OTOH be sure to compare the redbook format with "real" high res recording files , not just upsampled files !
You might check the frequency content of the files with the free "Spek" program.
Even some downloads from HD Tracks showed limited spectra.
Frank

[Lunch]

(2012-03-23 05:53)fbee Wrote:  OTOH be sure to compare the redbook format with "real" high res recording files , not just upsampled files !
You might check the frequency content of the files with the free "Spek" program.
Even some downloads from HD Tracks showed limited spectra.
Frank

As someone mentioned earlier, you should start with a good 24bit version and make your own 16 bit from that. Then you know it's a decent 24 bit and that the mastering is the same.