AVS/AIX High-Res Audio Test: Proposed Methodology - AVS Forum
Forum Jump: 
Reply
 
Thread Tools
post #1 of 146 Old 05-13-2014, 10:31 AM - Thread Starter
AVS Special Member
 
Scott Wilkinson's Avatar
 
Join Date: Sep 2001
Location: Burbank, CA
Posts: 1,135
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 195 Post(s)
Liked: 855

The thread entitled "Is High-Resolution Audio Irrelevant?" has inspired many interesting comments, as I thought it would. Among the participants in the discussion is fmaxwell, who is offering a downloadable zip file with three FLAC-encoded (losslessly compressed) audio files of the same recording—one at the original 24/96, one downsampled to 16/44.1 and then upsampled back to 24/96, and one at the downsampled 16/44.1 rate. All sample-rate conversions were performed using Sound Forge Pro 1.0 on a Mac with the highest-quality settings in both directions.

 

The original file was a clip from a recording made by Norwegian label 2L of Mozart's Violin Concerto in D Major - Allegro, performed by Marianne Thorsen and TrondheimSolistene. The original recording was made in DSD and edited in DXD, which is PCM at 352.8 kHz/24 bits. The recording is available in various formats, including 24/96, which is what fmaxwell used. I don't know if the original recording has any content beyond what CDs can represent.

 

This got me thinking about setting up a somewhat more formal listening test for AVS readers to see if they can hear the difference between a musical clip recorded at 24/96 and the same clip downsampled to 16/44.1. (Of course, this depends greatly on whether or not the listener's playback system is capable of reproducing frequencies above 20 kHz and dynamic range beyond 96 dB, but I'll save that issue for later.) I contacted Dr. Mark Waldrep, founder and chief engineer of AIX Records, to see if I could use some of his 24/96 recordings, which I know to have impeccable provenance, and some of which contain information beyond what CD is capable of representing.

 

The AIX Records Blu-ray release of trumpeter Wallace Roney's Stand includes Roney playing with a Harmon mute, which produces lots of ultrasonic partials that are faithfully captured at 24/96.

 

The first step is to establish a methodology that will satisfy even the harshest critics—impossible, perhaps, but that doesn't mean we shouldn't try. For example, how should the 24/96 original be downsampled? There are many ways to do this. Next, how should the two files be presented so there is little or no "dead air" between them? Auditory memory is very short, so listeners should be able to switch instantaneously between them.

 

Here's what we've come up with so far. First, downsample the 24/96 original using a well-known, well-regarded sample-rate converter; the specific algorithm TBD. Then, put both files in a single 24/96 "container" file that switches back and forth between them every 15 or 20 seconds, perhaps randomly in an ABX manner, making sure that listeners do not know which one is 24/96 and which is 16/44.1. (Of course, anyone with Audacity or a similar program can look at the file and determine which is which. When I actually post the test, I will appeal to everyone's sense of fairness and implore them not to do this and listen to the file blindly, but of course, that's no guarantee.)

 

The next question is, how to get the 16/44.1 clip into a 24/96 file? The 16/44.1 data could be upsampled, as fmaxwell did, but that might raise some eyebrows, depending on how it's done. Or both files could be converted to analog using a very high-quality DAC (digital-to-analog converter) and then immediately redigitized to 24/96 using a very high-quality ADC (analog-to-digital converter), though there are those who would object to that approach. However it's done, both files would then be edited into a single 24/96 file that would be made available from an ftp site.

 

Anyone who wishes to participate will be able to download the file and play it to see if they can hear the difference. They will also be encouraged to send their determination of which clip is which to a yet-to-be-named person, who will tabulate the correct and incorrect responses anonymously. Respondents will also be asked to include the specifics of their systems (make, model, and frequency-response and dynamic-range specs of the source device, DAC, preamp, power amp, and speakers) so we can get an idea of how many systems are actually capable of reproducing the content in the high-res clip. Each participant's determination and system will be held in confidence, so there is no risk of embarrassment or ridicule to anyone.

 

Admittedly, this is not a scientific test—there are way too many uncontrolled variables for that. But I still think it will be interesting to see what we come up with. And this is only the first in what I hope will be a series that could eventually include more rigorously scientific tests.

 

As I mentioned earlier, the playback system will make or break the test for listeners—if your system can't reproduce frequencies above 20 kHz and a dynamic range beyond 96 dB, there will be no acoustic difference between the clips. And headphones won't cut it, either. Then there's the noise floor in your room—if it's not sufficiently low, any extended dynamic range will be lost.

 

I'll address these issues more in a later post, but for now, I'd love to hear your thoughts on how to create the test file so that it provides an honest, unimpeachable way to determine if anyone can truly hear the difference between 24/96 and CD-quality digital audio.

 

Note: Rather than debate the whole high-res-versus-CD-audio thing here, head over to amirm's thread to discuss that. I'd like to keep this thread about the best way to create files and test whether or not high-res audio can be distinguished from CD-quality audio more reliably than random chance.

 

Like AVS Forum on Facebook

Follow AVS Forum on Twitter

+1 AVS Forum on Google+


Scott Wilkinson
AVS Editor
Scott Wilkinson is offline  
Sponsored Links
Advertisement
 
post #2 of 146 Old 05-13-2014, 10:41 AM
FMW
AVS Special Member
 
FMW's Avatar
 
Join Date: Jun 2007
Posts: 4,773
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 244 Post(s)
Liked: 700
Good for you, Scott. Glad to see AVS promoting a scientific listening test. I look forward to reading the results.
FMW is online now  
post #3 of 146 Old 05-13-2014, 10:56 AM
AVS Special Member
 
CharlesJ's Avatar
 
Join Date: Jan 2006
Posts: 3,232
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 83 Post(s)
Liked: 94
Looks interesting. I would suggest that if endeavor gets under way someone would double check that source for ultrasonic content, to what level that is and its DR. Nothing like double checking yourself as this would also eliminate any questions, just in case.
CharlesJ is offline  
post #4 of 146 Old 05-13-2014, 11:09 AM
AVS Special Member
 
Frank Derks's Avatar
 
Join Date: Jul 2002
Location: Region A,B,C
Posts: 1,891
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 98 Post(s)
Liked: 89
Just use the same re sampler tool to get 16/44 back to 24/96. This negates the up-sampling in the player that otherwise would occur anyway.
AVAdam likes this.
Frank Derks is offline  
post #5 of 146 Old 05-13-2014, 11:12 AM
AVS Addicted Member
 
amirm's Avatar
 
Join Date: Jan 2002
Location: Washington State
Posts: 18,022
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 695 Post(s)
Liked: 395
Hi Scott. Great to see the initiative. Alas, such tests do not work on the Internet these days. Reason being that machine analysis allows one to find out which file is which. Anyone with a copy of Audacity which is free can do such investigation and figure the different files. There are ways to get around this but require a lot of work to make sure it doesn't impact the test.

Please say hi for me to Mark. Hopefully he remembers me from my Microsoft days. I think I last saw him on home theater cruise.

Amir
Founder, Madrona Digital
"Insist on Quality Engineering"

amirm is offline  
post #6 of 146 Old 05-13-2014, 11:13 AM
Advanced Member
 
andyc56's Avatar
 
Join Date: Feb 2013
Posts: 576
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 17 Post(s)
Liked: 182
Sounds like a good plan, Scott. Regarding sample rate conversion, there is a web site that shows graphical performance comparisons of many of them here. I have found SoX, the freeware command-line utility, to have an extremely good resampling algorithm The evaluation was done via the very "round trip" approach you advocate, with the additional step of taking the original and round-trip-resampled 24/96 files as input to the free Audio DiffMaker program to form a file consisting of the difference between the two. In my experiments, the resulting difference file consisted of only a tiny amount of hiss, with nothing resembling music at all, even when turning the volume of my system all the way up (don't try this at home!).

The SoX command line I used was as follows:

sox input_file.wav -b 16 "01 - Uncle John's Band.wav" rate -v 44100 dither

The first argument is the original 24/96 WAV file name.
The "-b 16" says to convert to 16 bits.
The next argument is the output file name (which must be in quotes if it contains spaces).
Finally, the "rate -v 44100 dither" says to use the highest quality algorithm, 44.1k sample rate, and to use triangular PDF dither.

This was done on a file ripped from the Grateful Dead Workingman's Dead DVD-A.
A similar command line was used to resample back up.

Amir is right about the spectrum analysis thing. One experiment that's currently in progress only does bit depth conversion, which makes the machine analysis harder.
CruelInventions likes this.

Last edited by andyc56; 07-02-2014 at 05:59 AM.
andyc56 is offline  
post #7 of 146 Old 05-13-2014, 11:20 AM
Senior Member
 
Noman74656's Avatar
 
Join Date: May 2011
Location: Minneapolis, Minnesota
Posts: 259
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 25 Post(s)
Liked: 41
I don't have much advice for this, but I do fully support it and look forward to purchasing it! Fantastic idea.

"In nature there are neither rewards nor punishments; there are consequences."Robert Green Ingersoll

Noman74656's Martycube Builds
My Theater
Noman74656 is offline  
post #8 of 146 Old 05-13-2014, 11:51 AM - Thread Starter
AVS Special Member
 
Scott Wilkinson's Avatar
 
Join Date: Sep 2001
Location: Burbank, CA
Posts: 1,135
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 195 Post(s)
Liked: 855
Quote:
Originally Posted by amirm View Post

Hi Scott. Great to see the initiative. Alas, such tests do not work on the Internet these days. Reason being that machine analysis allows one to find out which file is which. Anyone with a copy of Audacity which is free can do such investigation and figure the different files. There are ways to get around this but require a lot of work to make sure it doesn't impact the test.


Yeah, I know that anyone with Audacity or similar program can look at the file and determine which is which. That's one reason why this is not a scientific test. When I actually post the test, I will appeal to people's sense of fairness and implore them not to do this and listen to the file blindly, but of course, that's no guarantee. What are the ways to get around this to which you refer?


Scott Wilkinson
AVS Editor
Scott Wilkinson is offline  
post #9 of 146 Old 05-13-2014, 11:54 AM - Thread Starter
AVS Special Member
 
Scott Wilkinson's Avatar
 
Join Date: Sep 2001
Location: Burbank, CA
Posts: 1,135
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 195 Post(s)
Liked: 855
Quote:
Originally Posted by andyc56 View Post

Sounds like a good plan, Scott. Regarding sample rate conversion, there is a web site that shows graphical performance comparisons of many of them here. I have found SoX, the freeware command-line utility, to have an extremely good resampling algorithm The evaluation was done via the very "round trip" approach you advocate, with the additional step of taking the original and round-trip-resampled 24/96 files as input to the free Audio DiffMaker program to form a file consisting of the difference between the two. In my experiments, the resulting difference file consisted of only a tiny amount of hiss, with nothing resembling music at all, even when turning the volume of my system all the way up (don't try this at home!).

The SoX command line I used was as follows:

sox track-01-01[0]-01-[L-R]-24-96000.wav -b 16 "01 - Uncle John's Band.wav" rate -v 44100 dither

The first argument is the original 24/96 WAV file name.
The "-b 16" says to convert to 16 bits.
The next argument is the output file name (which must be in quotes if it contains spaces).
Finally, the "rate -v 44100 dither" says to use the highest quality algorithm, 44.1k sample rate, and to use triangular PDF dither.

This was done on a file ripped from the Grateful Dead Workingman's Dead DVD-A.
A similar command line was used to resample back up.

Amir is right about the spectrum analysis thing. One experiment that's currently in progress only does bit depth conversion, which makes the machine analysis harder.


Thanks for the info! This addressed one of the questions I have: which type of downsampling to use. In this case, you have specified triangular PDF dither. There are many other options. do you think triangular PDF dither is best, and if so, why?

 

Regarding doing it only with bit depth, I believe that would be much harder to discern, since there's actually very little recorded content with a dynamic range beyond 96 dB, and even with such a recording, the noise floor in most people's equipment and room would preclude them from experiencing it.


Scott Wilkinson
AVS Editor
Scott Wilkinson is offline  
post #10 of 146 Old 05-13-2014, 11:58 AM - Thread Starter
AVS Special Member
 
Scott Wilkinson's Avatar
 
Join Date: Sep 2001
Location: Burbank, CA
Posts: 1,135
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 195 Post(s)
Liked: 855
Quote:
Originally Posted by Noman74656 View Post

I don't have much advice for this, but I do fully support it and look forward to purchasing it! Fantastic idea.


No purchase required; it will be free to download. Thanks!


Scott Wilkinson
AVS Editor
Scott Wilkinson is offline  
post #11 of 146 Old 05-13-2014, 12:09 PM - Thread Starter
AVS Special Member
 
Scott Wilkinson's Avatar
 
Join Date: Sep 2001
Location: Burbank, CA
Posts: 1,135
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 195 Post(s)
Liked: 855
Quote:
Originally Posted by CharlesJ View Post

Looks interesting. I would suggest that if endeavor gets under way someone would double check that source for ultrasonic content, to what level that is and its DR. Nothing like double checking yourself as this would also eliminate any questions, just in case.


I totally agree. We will make absolutely sure there is ultrasonic content and (hopefully) an expanded dynamic range in the source file.


Scott Wilkinson
AVS Editor
Scott Wilkinson is offline  
post #12 of 146 Old 05-13-2014, 12:28 PM
Senior Member
 
Dr. AIX's Avatar
 
Join Date: May 2004
Location: Los Angeles, CA
Posts: 430
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 14
I will get the spectrograph out for the tracks that will be used. The Wallace tracks have incredible high end and dynamic range too.
Dr. AIX is offline  
post #13 of 146 Old 05-13-2014, 12:29 PM
Senior Member
 
Dr. AIX's Avatar
 
Join Date: May 2004
Location: Los Angeles, CA
Posts: 430
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 14
Amir, it's been a very long time. Of course, I remember you.
Dr. AIX is offline  
post #14 of 146 Old 05-13-2014, 12:31 PM
Advanced Member
 
HockeyoAJB's Avatar
 
Join Date: Dec 2013
Posts: 622
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 139 Post(s)
Liked: 206
Quote:
Originally Posted by Scott Wilkinson View Post

...

Here's what we've come up with so far. First, downsample the 24/96 original using a well-known, well-regarded sample-rate converter; the specific algorithm TBD. Then, put both files in a single 24/96 "container" file that switches back and forth between them every 15 or 20 seconds, perhaps randomly in an ABX manner, making sure that listeners do not know which one is 24/96 and which is 16/44.1.

....

 

If we use a single file which contains both versions and they are played back at random then how does the listener know when the switch is made?  If switching back and forth multiple times, how does the listener keep track of which sample he is voting for?  Are you planning to add an audible cue to indicate "Now playing Sample A" and "Now playing Sample B"?  This way the listener can identify which sample he thought was hi res and which one he thought was not, without knowing for certain.

 

If using more than one clip, I would suggest switching up the order.  In the first comparison, sample A could be the hi res version and sample B the low res version or vice versa.  Then, for the second comparison, it might be the same or it might be the opposite.  Repeat for the 3rd and 4th comparisons.  This way a person is not inclined to continue voting for sample A, based solely off of the first comparison or two.

 

I look forward to participating.  My plan is to save the files to USB and play them directly from that since my AVR can play back FLAC and wav files at both 16 & 24 bit with sampling rates as high as 192 khz from USB.  Unfortunately, the speakers I will be using are only spec'd up to 30khz.  It's possible that there will be frequencies in the hi res version that can't be reproduced by them, though the majority of the higher frequency data (above CD capabilities) should be in the 22 khz to 30 khz range, so they might be good enough.

 

What volume setting do you recommend using for playback (assuming a 20 db noise floor) in order to determine if we can detect a greater dynamic range in one sample vs. the other?

HockeyoAJB is offline  
post #15 of 146 Old 05-13-2014, 12:38 PM
Advanced Member
 
andyc56's Avatar
 
Join Date: Feb 2013
Posts: 576
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 17 Post(s)
Liked: 182
Quote:
Originally Posted by Scott Wilkinson View Post

Thanks for the info! This addressed one of the questions I have: which type of downsampling to use. In this case, you have specified triangular PDF dither. There are many other options. do you think triangular PDF dither is best, and if so, why?

The dither is part of the re-quantization process rather than the sample rate conversion. It should be used even if no sample rate conversion is done. In Wannamaker's Ph.D thesis, The Theory of Dithered Quantization (PDF), he mathematically derived triangular PDF dither as the simplest PDF which sets the first and second moment of the quantization error to be independent of the input. IOW, if the moments of the quantization error are a function of the input instead of the error consisting simply of additive noise at a fixed level, that's a form of distortion. He says on page 198 of the thesis:
Quote:
In a [triangular] PDF dithered system, the first two moments of the total error are input independent and given by.. [formulas follow]
Quote:
Originally Posted by Scott Wilkinson View Post

Regarding doing it only with bit depth, I believe that would be much harder to discern, since there's actually very little recorded content with a dynamic range beyond 96 dB, and even with such a recording, the noise floor in most people's equipment and room would preclude them from experiencing it.

I agree, it would be much harder to discern.

Edit: PDF = "probability density function". Also, the specific form of triangular PDF dither derived by Wannamaker puts the span of the triangular PDF at +/- 1 LSB.
andyc56 is offline  
post #16 of 146 Old 05-13-2014, 12:51 PM - Thread Starter
AVS Special Member
 
Scott Wilkinson's Avatar
 
Join Date: Sep 2001
Location: Burbank, CA
Posts: 1,135
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 195 Post(s)
Liked: 855
Quote:
Originally Posted by HockeyoAJB View Post
 
Quote:
Originally Posted by Scott Wilkinson View Post

...

Here's what we've come up with so far. First, downsample the 24/96 original using a well-known, well-regarded sample-rate converter; the specific algorithm TBD. Then, put both files in a single 24/96 "container" file that switches back and forth between them every 15 or 20 seconds, perhaps randomly in an ABX manner, making sure that listeners do not know which one is 24/96 and which is 16/44.1.

....

 

If we use a single file which contains both versions and they are played back at random then how does the listener know when the switch is made?  If switching back and forth multiple times, how does the listener keep track of which sample he is voting for?  Are you planning to add an audible cue to indicate "Now playing Sample A" and "Now playing Sample B"?  This way the listener can identify which sample he thought was hi res and which one he thought was not, without knowing for certain.

 

If using more than one clip, I would suggest switching up the order.  In the first comparison, sample A could be the hi res version and sample B the low res version or vice versa.  Then, for the second comparison, it might be the same or it might be the opposite.  Repeat for the 3rd and 4th comparisons.  This way a person is not inclined to continue voting for sample A, based solely off of the first comparison or two.

 

I look forward to participating.  My plan is to save the files to USB and play them directly from that since my AVR can play back FLAC files at both 16 & 24 bit with sampling rates as high as 192 khz.  Unfortunately, the speakers I will be using are only spec'd up to 30khz.  It's possible that there will be frequencies in the hi res version that can't be reproduced by them, though the majority of the higher frequency data (above CD capabilities) should be in the 22 khz to 30 khz range, so they might be good enough.

 

What volume setting do you recommend using for playback (assuming a 20 db noise floor) in order to determine if we can detect a greater dynamic range in one sample vs. the other?


Actually, I was thinking of doing, say, 10 runs through the clip, randomly playing the 24/96 or 16/44 version each time. Each run would be numbered 1-10, and the listener would try to identify each run as either 24/96 or 16/44. This would be akin to an ABX approach. Or we could identify sample A and B, which would not be ABX, but perhaps it would be okay to do it that way. Either way, an audible cue is a good idea, since some folks, such as yourself, will want to listen without a computer.

 

We will make volume-setting recommendations when we post the test.


Scott Wilkinson
AVS Editor
Scott Wilkinson is offline  
post #17 of 146 Old 05-13-2014, 12:58 PM - Thread Starter
AVS Special Member
 
Scott Wilkinson's Avatar
 
Join Date: Sep 2001
Location: Burbank, CA
Posts: 1,135
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 195 Post(s)
Liked: 855
Quote:
Originally Posted by andyc56 View Post
 
Quote:
Originally Posted by Scott Wilkinson View Post

Thanks for the info! This addressed one of the questions I have: which type of downsampling to use. In this case, you have specified triangular PDF dither. There are many other options. do you think triangular PDF dither is best, and if so, why?


The dither is part of the re-quantization process rather than the sample rate conversion. It should be used even if no sample rate conversion is done.

 

You're absolutely correct. In this case, we will be doing sample-rate conversion, which will include dithering.


Scott Wilkinson
AVS Editor
Scott Wilkinson is offline  
post #18 of 146 Old 05-13-2014, 01:00 PM
AVS Special Member
 
Frank Derks's Avatar
 
Join Date: Jul 2002
Location: Region A,B,C
Posts: 1,891
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 98 Post(s)
Liked: 89
I see flac mentioned but I think it is best to keep the samples in wav. The flac compression ratio might give a clue due to a possible difference in file size. The upsampled 16//44 bit content might be more effectively compressed.
Frank Derks is offline  
post #19 of 146 Old 05-13-2014, 01:04 PM
Advanced Member
 
mightyhuhn's Avatar
 
Join Date: Nov 2013
Posts: 566
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 228 Post(s)
Liked: 81

how about Foobar2000 ABX plugin:

i used the Mozard samples above. i think these files are mastered in a way people will hear the difference. i did only 3 runs it's just an example...

my audio devices:

soundcard: asus essence stx DAC: http://www.ti.com/product/pcm1792a&lpos=See_Also_Container&lid=Alternative_Devices

Headset: beyerdynamics mmx300

mightyhuhn is online now  
post #20 of 146 Old 05-13-2014, 01:10 PM
Advanced Member
 
HockeyoAJB's Avatar
 
Join Date: Dec 2013
Posts: 622
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 139 Post(s)
Liked: 206
Quote:
Originally Posted by Frank Derks View Post

I see flac mentioned but I think it is best to keep the samples in wav. The flac compression ratio might give a clue due to a possible difference in file size. The upsampled 16//44 bit content might be more effectively compressed.

 

There might be a difference in file size even with wav files.  However, I agree.  It's probably best to keep compression out of this, if possible, as someone could claim that there was loss of hi res information due to compression.  I think it was me who mentioned FLAC, but I can also do wav, so no issues either way.

HockeyoAJB is offline  
post #21 of 146 Old 05-13-2014, 01:13 PM
AVS Special Member
 
RobertR's Avatar
 
Join Date: Dec 2000
Location: California
Posts: 6,116
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 142 Post(s)
Liked: 386
Thanks for this, Scott!
RobertR is offline  
post #22 of 146 Old 05-13-2014, 03:17 PM
AVS Special Member
 
cuzed2's Avatar
 
Join Date: Nov 2006
Location: Chicago Area
Posts: 4,406
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 6 Post(s)
Liked: 64
Sat in on Mark W's discussion at AXPONA a year ago,
Very much looking forward to this comparison

Cuzed2
Check out a video of my theater here
CuznEddy Cinema
Officially Hanesamatized on 8/1/09

cuzed2 is offline  
post #23 of 146 Old 05-13-2014, 03:35 PM - Thread Starter
AVS Special Member
 
Scott Wilkinson's Avatar
 
Join Date: Sep 2001
Location: Burbank, CA
Posts: 1,135
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 195 Post(s)
Liked: 855
Quote:
Originally Posted by Frank Derks View Post

I see flac mentioned but I think it is best to keep the samples in wav. The flac compression ratio might give a clue due to a possible difference in file size. The upsampled 16//44 bit content might be more effectively compressed.

I don't think differences in file size will be an issue, since there will be only one file with both 24/96 and upsampled 16/44 content.


Scott Wilkinson
AVS Editor
Scott Wilkinson is offline  
post #24 of 146 Old 05-13-2014, 03:38 PM - Thread Starter
AVS Special Member
 
Scott Wilkinson's Avatar
 
Join Date: Sep 2001
Location: Burbank, CA
Posts: 1,135
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 195 Post(s)
Liked: 855
Quote:
Originally Posted by HockeyoAJB View Post
 
Quote:
Originally Posted by Frank Derks View Post

I see flac mentioned but I think it is best to keep the samples in wav. The flac compression ratio might give a clue due to a possible difference in file size. The upsampled 16//44 bit content might be more effectively compressed.

 

There might be a difference in file size even with wav files.  However, I agree.  It's probably best to keep compression out of this, if possible, as someone could claim that there was loss of hi res information due to compression.  I think it was me who mentioned FLAC, but I can also do wav, so no issues either way.

I agree that using no compression is probably the best way to go, even though FLAC is lossless and should not lose any information whatsoever. But you're right, someone might complain about that, so better to avoid it altogether.


Scott Wilkinson
AVS Editor
Scott Wilkinson is offline  
post #25 of 146 Old 05-13-2014, 07:03 PM
Member
 
rogergraham's Avatar
 
Join Date: Feb 2006
Posts: 38
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 12
Why not just provide separate 24/96 and 16/44.1 files, and get people to use a free ABX testing program?

As you say, this test is completely non-scientific, but using an ABX program would at least preclude people from analysing the files to know which samples they were listening to.
rogergraham is offline  
post #26 of 146 Old 05-13-2014, 08:03 PM - Thread Starter
AVS Special Member
 
Scott Wilkinson's Avatar
 
Join Date: Sep 2001
Location: Burbank, CA
Posts: 1,135
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 195 Post(s)
Liked: 855
Quote:
Originally Posted by rogergraham View Post

Why not just provide separate 24/96 and 16/44.1 files, and get people to use a free ABX testing program?

As you say, this test is completely non-scientific - but using an ABX -style program would at least preclude people from analysing the files to know which samples they were listening to.

 

Several reasons:

 

- It would be even easier for people to determine which was which by file size. And simply selecting a file on a Mac will reveal the sample rate and bit depth in the Finder; I don't know if this is true in Windows, but I wouldn't be surprised if it is.

 

- Providing separate files as you suggest would certainly not preclude people from analyzing them to know which one they are listening to, it would simply make that task unnecessary.

 

- It might take a few seconds for an ABX program to adjust to different sample rates and bit depths, and we want the transition from one to the next to be as seamless as possible so auditory memory remains clear.

 

- I would rather not require anything of listeners other than to simply play a file; no need to acquire an ABX program or any other software.


Scott Wilkinson
AVS Editor
Scott Wilkinson is offline  
post #27 of 146 Old 05-13-2014, 08:05 PM - Thread Starter
AVS Special Member
 
Scott Wilkinson's Avatar
 
Join Date: Sep 2001
Location: Burbank, CA
Posts: 1,135
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 195 Post(s)
Liked: 855
Quote:
Originally Posted by FMW View Post

Good for you, Scott. Glad to see AVS promoting a scientific listening test. I look forward to reading the results.


Thanks; I hope you look forward to participating as well!


Scott Wilkinson
AVS Editor
Scott Wilkinson is offline  
post #28 of 146 Old 05-13-2014, 08:05 PM
Member
 
rogergraham's Avatar
 
Join Date: Feb 2006
Posts: 38
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 12
Quote:
Originally Posted by Scott Wilkinson View Post

I agree that using no compression is probably the best way to go, even though FLAC is lossless and should not lose any information whatsoever. But you're right, someone might complain about that, so better to avoid it altogether.

"...should not"? Does not! smile.gif

The comments of anyone who complains about file-size compression (e.g. FLAC files) are not to be taken seriously.
rogergraham is offline  
post #29 of 146 Old 05-13-2014, 08:09 PM
Member
 
rogergraham's Avatar
 
Join Date: Feb 2006
Posts: 38
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 12
Quote:
Originally Posted by Scott Wilkinson View Post

Several reasons:

- It would be even easier for people to determine which was which by file size. And simply selecting a file on a Mac will reveal the sample rate and bit depth in the Finder; I don't know if this is true in Windows, but I wouldn't be surprised if it is.

- Providing separate files as you suggest would certainly not preclude people from analyzing them to know which one they are listening to, it would simply make that task unnecessary.

- It might take a few seconds for an ABX program to adjust to different sample rates and bit depths, and we want the transition from one to the next to be as seamless as possible so auditory memory remains clear.

- I would rather not require anything of listeners other than to simply play a file; no need to acquire an ABX program or any other software.

I'm not sure I understand you. Of course people can look at the individual files and see which is which, but when using the ABX program they won't know which one they are listening to; that's the whole point?

I guess the real point here, any tests not done in a controlled environment are not useful.
rogergraham is offline  
post #30 of 146 Old 05-13-2014, 08:09 PM - Thread Starter
AVS Special Member
 
Scott Wilkinson's Avatar
 
Join Date: Sep 2001
Location: Burbank, CA
Posts: 1,135
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 195 Post(s)
Liked: 855
Quote:
Originally Posted by rogergraham View Post
 
Quote:
Originally Posted by Scott Wilkinson View Post

I agree that using no compression is probably the best way to go, even though FLAC is lossless and should not lose any information whatsoever. But you're right, someone might complain about that, so better to avoid it altogether.

"...should not"? Does not! smile.gif

The comments of anyone who complains about file-size compression (e.g. FLAC files) are not to be taken seriously.


Of course, you are right, but I can't rule out that there could be some who believe differently (despite facts to the contrary), so I think it's just better to avoid the issue entirely, especially since there's no real penalty for doing so.


Scott Wilkinson
AVS Editor
Scott Wilkinson is offline  
Reply Audio theory, Setup and Chat

User Tag List

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off