Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!cs.utexas.edu!uunet!garfield!leif!andrew
From: andrew@kean.mun.ca
Newsgroups: sci.electronics
Subject: Re: RS-232 for voice
Message-ID: <9185@kean.mun.ca>
Date: 29 Jul 89 18:42:33 GMT
Organization: Computing Services, Memorial University
Lines: 32

Hmm.  Lets assume that we can get away with somewhat poor quality.

Okay... Let's assume 8 data bits giving a 48 dB dynamic range ( 20 * log 
(2^number of bits) ).

Next assumtion: highest serial rate _routinely_ available for RS-232 
transmission: 19.2k baud = 1.92k bytes/sec.

Next assumtion: the setup will be along the lines of:

   Microphone ---> amp ---> ADC ---> computer ---> serial link =====
  ======== serial link ---> computer ---> DAC ---> amp ---> speaker.


Nyquist says then that the highest relevant frequency that can be sent with a 
sample rate of 1.92 kHz is 960 Hz.  Just possibly passable... but still pretty 
terrible for a human voice.  Probably still could be understood, though.

This could be fiddled with, of course... using 5 data bits (at 5N1 and 19.2 k 
baud), you should still be able to get a 30 dB dynamic range and a 1371 Hz 
bandwidth... figures which *in theory* should give quite reasonable results.
(transmission time/sample = 365 us).



Question: The theory seems to work... but does nature agree??? Anyone have 
experience with applications?


-Andrew.

Bitnet: andrew@kean.mun.ca