Long post, please read it if you want to 'test' your own hearing capabilities (not to me or anyone else but to yourself).
When NOT interested in testing your own hearing SKIP this entire post !
The main discussion about the ability of hires vs lesser formats is divided into 2 separate issues.
Frequency extension (sample rate) and resolution (bit depth).
For CD both are 'compromised' when made from a hiresolution master.
Some claim the extended frequencies is what they are hearing (higher sample rate) and others say it is the bit depth (24 bit vs 16 bit).
Most do not want to test blind as it involves other peoples time.
One has to be able to do this in their own time and comfort and have both the reference file and 'compromised' one to play with to find the differences.
If a test would be made that allows this and only address these separate issues it would be eaier to determine if the 24 or 192kHz part (or both) are the degrading ones.
A problem rises when 2 different bitrates are handled by one DAC or one choose the wrong resampler which could result in different sound, not because of the format but because of the DAC's properties.
The same problem is there when a 'blind' test is included with 2 different files as they would give themselves away in file size and shown bitrates/depth.
So I cooked up 3 tests that comply to these rules.
I used a real 192/24 recording and analysed it to see if it has enough ultrasonics.
The 192/24 recording is dropping off steeply around 40kHz which makes me suspect this recording was made on 96/24 or 88.2/24 and was upsampled to be sold as 192/24.
Good thing is that this is a DR16 live recording (thus a little more noise than studio recordings can have) and has all kind of 'audiophile' instruments and a well recorded female voice.
It may not be everyones cup of tea though.
I took an excerpt from the file (as 192/24 files are huge) and converted that file to 48/24 removing the ultrasonics while maintaining bit-depth.
Also I used the same original file and converted that to 192/16 so ONLY the bit depth has been altered.
The file is truncated to achieve a maximal 'difference'.
Truncting means REMOVING the lowest 8 bits without adding dither.
As a third test I used the same file and ran it through LAMEXP set to CBR/320kbs/Highest Quality.
Now the biggest issue is the used resampler.
As can be seen
HERE there are substantial differences between resamplers and some of which can cause VERY audible artefacts.
These added sounds that have no relation to the original signal may lead to wrong conclusions as about resampling.
This means that the differences in SQ one hears, going back and forth between files from the same master, may NOT be due to the format but are INTRODUCED by the used resampler.
For this reason I used a good performing resampler to make the files used in this test.
Awave Studio was chosen to do the resampling bit (check on the linked site for its qualities)
The files were left untouched on all other aspects so no trickery or me meddeling/rigging original files as some people have accused me of in the past.
I am quite serious with tests and will NEVER rig a test merely to 'prove' my point.
To prevent the DAC (or playback software) to alter the sound with their built-in digital filtering it is important to convert the files back to 192/24 so all files are played back under the same circumstances.
It also makes it possible to include a small blind test to test YOURSELF.
This blind test can NOT be used as 'evidence' of above average hearing capabilities, as with some knowledge one can easily find out which file is what and 'present' a perfect score without even using the ears.
It is meant to see if YOU can also detect the differences between the files if you do not KNOW which file is the 'lesser' one.
You can compare test files to known files and 'doctored' files and can take all the time you need and stop when you like.
So all 'changed' files were converted back (Awave resampler) so the files all play back
under the same circumstances and cannot be told apart on first glance by file size, extension, bit rate, bit depth etc. while they ARE substantially and very measurably different.
For those that may fear the resampling to 192/24 nullifies the previous operations than I can assure you this is NOT the case.
The bandwidth of the 48/24 has been limited (simply all ultrasonics above 21kHz are NO longer present and will not be 're-invented' by upsampling again, it is lost (and thus absent) forever which is the ONLY difference between those files.
The 16 bit file has NOT been resampled ONLY the bottom 8 bits (that contain the finest nuances level wise) have been REMOVED and adding 8 bits WITHOUT upsampling to higher bitrates (as done here) will simply add 8 empty bits (they simply are all '0')
The MP3 file should be obvious... it is resampled by LAMEXP to 48/16 and converted to 320kbs/CBR and can assure you things are very measurably LOST forever during this process.
Resampling to 192/24 cannot add what has been lost.
The tests consist of the original file and the 'doctored' file (that has been converted to a 'lesser' quality format and converted back)
It also includes 5 files named 1 to 5.
I determined the sequence of the 2 different files by flipping a coin.
Once you have heard the differences (if you can) then you can blind test yourself and see if you can determine which of the files is 'original' and which had been 'modified' specifically for the mentioned test/parameters.
So 1-5 can either be original or 'modified' acc to the test foldername.
The files are packed so have to be unzipped, but they are packed WITHOUT any compression so they form a 'folder' or 'package' with all the relevant files.
The files SHOULD be played back on a 192/24 capable DAC but as the original file does not contain info it can be played back on 96/24 DACs as well.
This is problem free as the files are all resampled by the DAC (or software) in the SAME way.
The test folders are almost 1.2GB each.
The files are available till March 15th so IF you would really like to know IF you can hear differences between the 'formats' and not between different masters or 'badly' resampled masters files I can recommend to at least DOWNLOAD those files before March 15th and listen later if needed.
After a few weeks I will post what the 'blind' test order is and how to check for yourself the test was not rigged and the files you have downloaded INDEED differ substantially in contents.
You can compare this with your notes and see if your hearing is equally good when comparing files sighted (the 2 reference files with names) and when done blind (the numbered files)
I don't need to know your score or which files one thinks is what, as said with simple tricks it is easy to find out which files are original and modified.
IF the download time has expired (after 7 days) I can upload them again if someone is interested and did not download the files (takes me a day to upload!) but with another 1-5 sequence of course.
download link has expired.