Randy R wrote:
> I am working with several SIP projects that use g722, or are trying to
> do so, with pjsip library.
>
> According to pjsip team's interpretation of g722, it works with 14bits
> PCM for input/output, so pjsip basically 'converts' the audio
sample
> from 16 bits to 14 when encoding and vice-versa. Some implementations
> don't do 16<->14 bits conversion, so when pjmedia talks to one of
> those the over-driven audio problems appear.
>
> What we need to know is what's the most used implementation:
14<->16
> bits conversion or not.
>
> Any pointers to help clear this up? We'd really like to see more
> g722-capable SIP clients for our own conference on ZipDX.
As far as I am aware, for ITU-T compliance the codec only cares about 14
significant bits, but the reference source code needs those 14 bits in
the *top* 14 bits of each 16-bit word that it supplies/produces. The
Asterisk implementation does not do any bit-shifting or masking at all,
and seems to interoperate with quite a few endpoints just fine, so
presumably that means it's the correct implementation :-)
--
Kevin P. Fleming
Digium, Inc. | Director of Software Technologies
445 Jan Davis Drive NW - Huntsville, AL 35806 - USA
skype: kpfleming | jabber: kpfleming at digium.com
Check us out at www.digium.com & www.asterisk.org