noise shaping v. plain dither comparison tracks

2009-09-09 08:01:55

[attachment=5385:ETsample_shaped.wav]
[attachment=5386:ETsample_TPDF.wav]
Here are samples from 28.846154-43.461538 seconds into the tracks I discussed in my "listening tests" post.  Good luck!  The difference between these tracks is very subtle.

noise shaping v. plain dither comparison tracks

Reply #1 – 2009-09-09 17:44:51

And here, by popular request, is approximately the same passage at 44.1 kHz, floating point resolution.

[attachment=5387:ETsample_float.wav]

noise shaping v. plain dither comparison tracks

Reply #2 – 2009-09-09 19:28:30

For anyone interested: as discussend in this thread, here's the upload of the 24-bit sample, reduced to 16 bit with TPDF dither and the "DF UEN" noise shaper proposed in this AES paper. The sample is time aligned to match the above 16-bit items.

Chris

[attachment=5388:ETsample...Helmrich.wav]

noise shaping v. plain dither comparison tracks

Reply #3 – 2009-09-11 08:28:17

And here the TPDF dithered, unshaped version again, this time made from the above float version and delay-aligned to exactly match my shaped version.

Chris

[attachment=5390:ETsample_TPDF2.wav]

noise shaping v. plain dither comparison tracks

Reply #4 – 2009-09-11 08:42:46

Here is the original float version from above, but resampled by 2463.596841 samples so that it aligns with the two 16-bit samples (which are perfectly aligned with each other).

noise shaping v. plain dither comparison tracks

Reply #5 – 2009-09-15 06:37:07

Here is the passage from 6:02.308313-6:15.563495 on track 1 of the CD layer of BIS-SACD-1516  .  Both the French and  the American Classics Today sites marked the CD down largely because this passage is so soft.  (The reviewers, Christophe Huss and David Hurwitz, frequently discuss recordings with each other, and then write separate reviews.  So they're semi-independent reviews.)

[attachment=5400:Eroica.wav]

noise shaping v. plain dither comparison tracks

Reply #6 – 2009-09-15 13:46:25

A clarification: the first files I posted were samples made from complete 16/44.1 files generated from the original source file

8-EtudesTableaux Little Red Ri.flac

obtained at

http://www.hdtracks.com/index.php?file=cat...=HD030911109622

noise shaping v. plain dither comparison tracks

Reply #7 – 2010-07-03 21:08:57

To go with my test of this music under a standard antialiasing lowpass filter for 44.1kHz sampling, here are samples of the same music in 96/24:

From the commercially available file,
[attachment=5993:littlered_sample.flac]

and the same with an effective 20kHz lowpass done by resampling twice:

[attachment=5994:littlere...s_sample.flac]

Sorry they aren't perfectly aligned!