DSPRelated.com
Forums

G726 ITU test vector strangeness

Started by Jeff Brower December 21, 2009
All-

Has anyone ever noticed the G726 ITU u-Law input test vector (a file called nrm.m on
ITU pages) has what appears to be an inversion or wrap-around? Not saying this is a
problem, bit-exact testing using the vector works Ok. But it looks odd, and the
inversion doesn't occur around 128. Here is a pic that shows an example in the 504
Hz tone section of the test vector:

http://www.signalogic.com/images/G726_test_vector_strangeness.jpg

Another odd thing is that if a capture of linear output after back-to-back
encode/decode (for example at 16 kbps), the inversion is "fixed" and the tones look
"normal".

If anyone knows how this came to be part of the ITU standard, or whether there are
other ITU G726 input test vectors that ITU has since come up with that do not exhibit
this, please let me know. Thanks.

-Jeff
Hi All

I did lot of google search but not able to find how to differentiate between the files encoded with GSM Coders like HR, FR and EFR.

What is file header content for that?

Thanks in advance.

Regards
Sridhar Artham
Sridhar-

> I did lot of google search but not able to find how to differentiate
> between the files encoded with GSM Coders like
> HR, FR and EFR.
>
> What is file header content for that?

Not sure but you might check SoX related web pages and docs. A quick check of the feature list shows "GSM files",
".WAV GSM files", and "AMR WB" and "AMR NB" file formats.

Sox is open source and cross-platform, possibly the closest thing to a "de facto standard" there is for handling
various audio/waveform file formats.

-Jeff