Hello everyone.
Arguments about value of sampling rate > 44.1 Khz and bit depth > 16 bits has filled thousands upon thousands of pages on various audio forums. There has been very little formal testing but what there is, such as Meyer and Moran point to no audible difference (yes, the test was flawed but this thread obsoletes what they did anyway).
Personally I must have wasted hundreds of hours talking about the need for transparency in our audio equipment and I don't know how to get there if it is limited to 16 bits/44.1K. I have provided the math, research, and tons of explanation to no avail. Folks demand "listening tests" when they run out of technical arguments to make.
As it turns out there was major advancement in this discussion in the last week or so. I was engaged on AVS Forum in yet another dog fight on this front with folks arguing that even 16-bit/44.1 Khz is "too much" let alone having something at higher spec. On this front, Arny Krueger claimed that he had "brickwall" filtered audio down to 16 Khz (32 K sampling) and no one could tell it apart from the original. And that he had run this test blind and double blind. Here is the original claim:
Turns out this is an experiment that I had also run years ago but to my surprise, I found clear audible difference. Yet, I could not hear the frequencies that I had taken out!
The discussion went on for a bit and then Arny post a recording of keys jingling that he had recorded and then sampled down to 44.1 Khz and 32 Khz. He then resampled them back out to the same 24/96 Khz of the original track. So all files had the same sample rate but the other two versions had their frequency response chopped down to 22 and 16 Khz respectively. Here is the link to the post and the files in dropbox:
Despite my success in hearing differences I expected this comparison to be a tough one. After all, I would think that Arny would first listen to these files and once he was assured there was no audible difference, he would post them. Such was not the case.
The challenges on AVS always demanded use of Foobar2000 ABX test and while I don't like that tool, I used it anyway. Here are my results. The format is N correct votes/Trial number M and (probability of chance). A score of say 10/11 says that 10 correct votes were made out of 11 tries (i.e. one mistake).
As you see, this is a perfect score when it comes to finding the difference between the original 96 Khz and down sampling to 44.1 and 32 Khz.
I found the 32 Khz/16 K max frequency comparison to be trivial. The difference was easily heard. Indeed another poster on AVS repeated the same perfect score I had. The second one with limiting the sampling to 44.1 Khz was much harder but I was still able to identify the right segment and as you see, achieve perfect score of 7/7. I could have kept going but hit the wrong button in foobar and it stopped there. The other poster however was not able to differentiate the files.
Please download the files and see how you do. My testing was all on my HP Zbook 14 laptop and etymotic ER4p headphone. The DAC was the internal sound of the laptop so nothing special there either. Foobar was in default mode using Windows Audio stack so additional dither was being applied. The sampling rate in Windows Sound control panel was set to 24/96 Khz to match the files.
In Part 2 I will provide the results of a second listening test that cements these findings.
Arguments about value of sampling rate > 44.1 Khz and bit depth > 16 bits has filled thousands upon thousands of pages on various audio forums. There has been very little formal testing but what there is, such as Meyer and Moran point to no audible difference (yes, the test was flawed but this thread obsoletes what they did anyway).
Personally I must have wasted hundreds of hours talking about the need for transparency in our audio equipment and I don't know how to get there if it is limited to 16 bits/44.1K. I have provided the math, research, and tons of explanation to no avail. Folks demand "listening tests" when they run out of technical arguments to make.
As it turns out there was major advancement in this discussion in the last week or so. I was engaged on AVS Forum in yet another dog fight on this front with folks arguing that even 16-bit/44.1 Khz is "too much" let alone having something at higher spec. On this front, Arny Krueger claimed that he had "brickwall" filtered audio down to 16 Khz (32 K sampling) and no one could tell it apart from the original. And that he had run this test blind and double blind. Here is the original claim:
Yes. Take the best audio system you can find. Take the best recordings you can find - recordings that sound great and also have significant content > 20 KHz, even > 35 KHz. Switch a 16 KHz brick wall filter in and out of the signal path. Nobody notices nuttin'
People say: "But I can hear pure tones at 21 KHz". Probably true. But that is without content at other frequencies masking it. Music is composed of many tones at many different frequencies. Masking in the upward direction frequency-wise is very strong.
Turns out this is an experiment that I had also run years ago but to my surprise, I found clear audible difference. Yet, I could not hear the frequencies that I had taken out!
The discussion went on for a bit and then Arny post a recording of keys jingling that he had recorded and then sampled down to 44.1 Khz and 32 Khz. He then resampled them back out to the same 24/96 Khz of the original track. So all files had the same sample rate but the other two versions had their frequency response chopped down to 22 and 16 Khz respectively. Here is the link to the post and the files in dropbox:
Thanks for the encouragement.
Here is the link for my idea of killer 24/96 .wav files files for using ABX to hear potential audible differences due to differences in bandpass and resolution:
https://www.dropbox.com/sh/l86f7oc7cdsb0xr/AAB4feU8z3Xww14Mrta1zs3Ca
Be careful with these files - the full bandpass version has so much ultrasonic energy that it can possibly take out your tweeters and you won't even think that it is playing very loud. They are also highly diagnostic for high frequency IM.
The source is my keychain ca. Y2K being jangled about a foot away from a DPA 4006 measurement mic, as precisely on axis as I could get it in a low reverberent room. This particular test has been well known among studio engineers for about half a century, and it has the advantage that for testing microphones and speakers, the only special test equipment has the right price!
Despite my success in hearing differences I expected this comparison to be a tough one. After all, I would think that Arny would first listen to these files and once he was assured there was no audible difference, he would post them. Such was not the case.
The challenges on AVS always demanded use of Foobar2000 ABX test and while I don't like that tool, I used it anyway. Here are my results. The format is N correct votes/Trial number M and (probability of chance). A score of say 10/11 says that 10 correct votes were made out of 11 tries (i.e. one mistake).
32 Khz versus 96 Khz
=================================
foo_abx 1.3.4 report
foobar2000 v1.3.2
2014/07/09 06:10:07
File A: C:\Users\Amir\Music\Arnys Filter Test\keys jangling band resolution limited 3216 2496.wav
File B: C:\Users\Amir\Music\Arnys Filter Test\keys jangling full band 2496.wav
06:10:07 : Test started.
06:10:38 : 01/01 50.0%
06:10:50 : 02/02 25.0%
06:11:07 : 03/03 12.5%
06:11:23 : 04/04 6.3%
06:11:36 : 05/05 3.1%
06:12:00 : 06/06 1.6%
06:12:14 : 07/07 0.8%
06:12:26 : 08/08 0.4%
06:12:38 : 09/09 0.2%
06:12:49 : 10/10 0.1%
06:13:00 : 11/11 0.0%
06:13:23 : 12/12 0.0%
06:13:42 : 13/13 0.0%
06:13:48 : Test finished.
----------
Total: 13/13 (0.0%)
44.1 versus 96 Khz
---------------------------------
foo_abx 1.3.4 report
foobar2000 v1.3.2
2014/07/09 06:32:02
File A: C:\Users\Amir\Music\Arnys Filter Test\keys jangling band resolution limited 4416 2496.wav
File B: C:\Users\Amir\Music\Arnys Filter Test\keys jangling full band 2496.wav
06:32:02 : Test started.
06:33:07 : 01/01 50.0%
06:33:17 : 02/02 25.0%
06:33:24 : 03/03 12.5%
06:33:36 : 04/04 6.3%
06:33:47 : 05/05 3.1%
06:33:58 : 06/06 1.6%
06:34:12 : 07/07 0.8%
06:34:15 : Test finished.
----------
Total: 7/7 (0.8%)
As you see, this is a perfect score when it comes to finding the difference between the original 96 Khz and down sampling to 44.1 and 32 Khz.
I found the 32 Khz/16 K max frequency comparison to be trivial. The difference was easily heard. Indeed another poster on AVS repeated the same perfect score I had. The second one with limiting the sampling to 44.1 Khz was much harder but I was still able to identify the right segment and as you see, achieve perfect score of 7/7. I could have kept going but hit the wrong button in foobar and it stopped there. The other poster however was not able to differentiate the files.
Please download the files and see how you do. My testing was all on my HP Zbook 14 laptop and etymotic ER4p headphone. The DAC was the internal sound of the laptop so nothing special there either. Foobar was in default mode using Windows Audio stack so additional dither was being applied. The sampling rate in Windows Sound control panel was set to 24/96 Khz to match the files.
In Part 2 I will provide the results of a second listening test that cements these findings.