Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!apple!bionet!hayes.fai.alaska.edu!accuvax.nwu.edu!nucsrl!telecom-request
From: optilink!elliott@uunet.uu.net (Paul Elliott x225)
Newsgroups: comp.dcom.telecom
Subject: Re: An Introduction to ISDN From the CERFnet News
Message-ID: <13198@accuvax.nwu.edu>
Date: 8 Oct 90 23:56:30 GMT
Sender: news@accuvax.nwu.edu
Organization: Optilink Corporation, Petaluma, CA
Lines: 60
Approved: Telecom@eecs.nwu.edu
X-Submissions-To: telecom@eecs.nwu.edu
X-Administrivia-To: telecom-request@eecs.nwu.edu
X-Telecom-Digest: Volume 10, Issue 724, Message 5 of 7


In article <13051@accuvax.nwu.edu>, mitel!spock!meier@uunet.uu.net
(Rolf Meier) writes:

> In article <12978@accuvax.nwu.edu> goldstein@delni.enet.dec.com (Fred
> R. Goldstein) writes:

> >bits as it desires, preserving the audio content.  If you call between
> >North America and Europe, the network MUST change speech and PCM audio
> >because Europe and North America use different PCM standards!  They're
> >mutually unintelligible, though both are 64 kbps PCM.  Similarly, the
> >network MUST NOT change a clear channel (data).

> Actually, if you decoded ulaw with an Alaw decoder, or vice versa, the
> difference is practically inaudible compared to the use of the proper
> decoder.  However, the conversion is made anyway, in order to meet the
> quantization requirements.

While it is true that the mu-law and A-law encoding/decoding curves
are very similar, the actual digital representation of the signals is
quite different, requiring code conversion to be intelligible.

Mu-law and A-law codecs both use a quasi-logarithmic transfer
function, to obtain optimal signal-to-noise ratios over a wide dynamic
range.  The quasi-log characteristic is achieved by breaking a
non-linear transfer function into a series of linear "chords", with
each chord consisting of several equal-sized steps.  The step size is
doubled for each successive chord (the piecewise approximated curve is
symmetrical about zero). Thus, for a given full-scale value, signals
closer to zero are encoded with greater precision than would be
obtained with a linear code.  The resulting encoding gives nearly
equal stepsize (when measured in dB) for signals within the encoding
range.  The dynamic range of the mu-law codec is approximately 72 dB,
which compares well to the 42 dB range of a linear 8-bit code (seven
bits plus sign).

The mu-law function provides eight chords, of 16 steps each, while for
some reason, the European A-law standard has a first chord of 32
steps, and six remaining chords of 16 steps.  Mu-law provides better
S/N over the full range, while A-law gives reduced distortion at low
levels.

These differences are almost inaudible, but the standards threw in a
big monkey wrench.  Mu-law encoding could be called "bit-inverted
sign-magnitude", where "positive full scale"= 10000000 "positive zero"
= 11111111 "negative zero" = 01111111 "negative full scale"= 00000000

A-law inverts alternate bits, to give:
"positive full scale"= 10101010
"positive zero"      = 11010101
"negative zero"      = 01010101
"negative full scale"= 00101010

I guarantee you, this WILL be noticed!  Still, as far as standards are
concerned, I guess we "telecom types" don't have it as bad as some
other technical fields...


      Paul M. Elliott      Optilink Corporation     (707) 795-9444
               {uunet, pyramid, tekbspa}!optilink!elliott