u-law, a-law verses G.723, G.729, GSM etc

Lennart Sorensen lsorense-1wCw9BSqJbv44Nm34jS7GywD8/FfD2ys at public.gmane.org
Tue Dec 16 22:07:01 UTC 2008

On Tue, Dec 16, 2008 at 02:40:21PM -0500, William Muriithi wrote:
> I am not sure if what I have in mind is correct and nothing picked up
> by google seriously satisfied me and that is why I am posting this.
> Its on VOIP. There is these compounding algorithms - according to
> wikipedia, u-law and a-law. Then there is the voice compression
> algorithms or codecs? I mean G.723, G.729, GSM to name a few.
> Now I kind of know how they work but a little mixed up as to whether
> they are inclusive, exclusive or just one thing with different names.
> For example, G.711 is in some on line article also called u-law. This
> is not a single incident and the whole business has made me feel
> confused.
> My take of it is this, when a call is received, its taken through
> either u-law or a-law depending on geographical location. The
> resulting audio signal is then taken again through the later group -
> G.723 or GSM. The result is then packed as an IP packet and delivered
> to destination where the reverse happen. Now, is this the correct or I
> could as well be smoking pot?
> audio <-> a-law <-> G.723 <-----------------> G.723 <-> a-law <->

I think it would be:

audio <-> a-law <-> G.723 <-> a-law (on regular phone system) <-> a-law
(on regular phone system) <-> G.723 <-> a-law <-> audio

After all a regular phone system is involved when GSM calls are made, so
the GSM compression only applies between the phone and the cell phone
provider, after which it is back to normal *-law again for transfer on
the normal 64kbps circuits.

> Ah, now I even noticed something else. Someone could call from Canada
> - read a-law to UK where u-law is used. Now that would result to using
> one algorithm for compressing and another for expanding and I do not
> believe the result would be intelligible. Someone please educate me
> here.

Normally when you do PCM encoding, you just use a linear map of voltages
to bits.  u-law and a-law use a logorithmic rather than linear mapping
of voltages to bits to get better coverage of both quiet and load sounds
with decent representations at both ends of the scale, while using less
bits than linear would require for the same thing.  u-law and a-law use
different values for each bit, but they are as far as I know fairly
close together.  Conversion does take place when moving between regions
and does cause a bit of loss of quality.  u-law is north america and
japan, a-law is everywhere else.

Len Sorensen
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists

More information about the Legacy mailing list