View RSS Feed


Listening test - bits and kiloherz

Rate this Entry
As some posters have doubted the benefit of resolutions and sample rates higher than 16 bit and 44.1 kHz, I set up a listening test.

Robert von Bahr of BIS has very kindly arranged for the permission to use a track from BIS-SACD-1949, "endBeginning" by New York Polyphony (available from eClassical).

I made available 8 versions of the track "Lamentationes Jeremiae - IV. Mem.". In addition to the original 24/96 track, there are versions that have been first downsampled and/or truncated to various permutations of 96, 48 and 44.1 kHz and 24 and 16 bits, and then upsampled back to original resolution, thus discarding varying amounts of information.

I asked forum members to listen to the different tracks, and for each, write down a score between 1 and 10 for the sound quality, as well as a verbal assessment of each track, and send me the results in a private message (or by email).

Along with the tracks the download page also contained a link to a document containing a list of what track is what resolution. I encrypted the file with gpg in order to provide the digital equivalent of a "closed envelope", and I will post the password to decrypt the file as soon as the test period is over.

There is no "trick" here - this is a serious test, and there are differences between all the files in terms of resolution or processing. In all cases but one, the differences are measurable and ought to be audible.

I also made it clear that this is a *listening* test, not a measuring test. Yes, it would be interesting to see if it is possible to determine the resolutions of the tracks by measurements, but that can be done after the listening test is completed.

After user "goldsdad" sent me results that he got by measuring that showed that it was relatively easy to cheat by using analysis software, I decided to stop collecting "official" results, but still wait with revealing the content of the files, so that people who wanted to do the test for themselves could still do that - thus I will publish the file contents, and the results I had already collected, after end of day, Monday, March 19.

Many thanks to Robert, The New York Polyphony, BIS and eClassical for making this possible!
Personal Blogs


  1. goldsdad's Avatar
    After some misunderstanding in the forum thread regarding this test, I want to make clear that I contacted Julf with a warning and demonstration that cheating would be easy. There was no initial pretence that my identification of the processing performed on the files was a result of listening. My opening words were that I had analysed the files with software instead of listening.
  2. Julf's Avatar
    Sorry for not being more explicit about that - yes, goldsdad was making it absolutely clear that he was not submitting his results as part of the test, but to show how somebody could cheat if he/she wanted to.