Path: utzoo!utgpu!attcan!uunet!husc6!uwvax!rutgers!apple!amdahl!oliveb!felix!trwrb!arenberg
From: arenberg@trwrb.UUCP (Jeff Arenberg)
Newsgroups: comp.sys.next
Subject: Re: NeXT sound
Message-ID: <829@csed-pyramid.trwrb.UUCP>
Date: 3 Nov 88 18:11:28 GMT
References: <361.236F0D6C@mailcom.FIDONET.ORG>
Reply-To: arenberg@csed-pyramid.UUCP (Jeff Arenberg)
Organization: TRW EDS, Redondo Beach, CA
Lines: 28

In article <361.236F0D6C@mailcom.FIDONET.ORG> postmaster@mailcom.FIDONET.ORG (Bernard Aboba) writes:
>The big advantage of putting the CODEC on the motherboard, is I suppose 
>the ability to have a voice message delivered to another workstation, 
>input via a microphone.  However, this eats up network bandwidth like 
>crazy, and makes very little sense when an electronic mail message would 

I'm curious why so many posters have problems with real time voice
communications over the network.  As I understand it, the system uses an 8
bit mu-law encoding method, sampled at 8 kHz.  This translates to a
maximum of 64 kbps for each transmitter.  On a standard ethernet at say 5
mbps you could have about 39 simultaneous two way conversations.  Of
course, this would take up the whole net bandwidth, but you shouldn't
assume that everyone will talk at once.  Also, you can certainly get away
with a 4 bit mu-law at 4 kHz sampling without lossing too much
understandability and get four times more conversations in, if you are
really desperate.

For those who wish to know, in the mu-law method the data sent represents
some logrithmically scaled change in the current output level, rather than
using the data as an absolute value of output level.  This allows for a
significantly greater dynamic range, but limits the ability to accurately
track fast, large changes in sound level.

Jeff Arenberg
-------------------------------------------------------------
UUCP : ( ucbvax, uscvax ) !trwrb!csed-pyramid!arenberg
GEnie: shifty
-------------------------------------------------------------