PI>>Telephone companies chose 8,000 8-bit samples/second to give an upper
PI>>frequency response of ~4000 Hertz (Nyquist limit).

PI> Yes (8K 8bits = 64Kbits/s only in Europe, but in my country too), but remme
PI>that you must apply the mu-law (or the a-law) to get a good result, applying
PI>this compression you'll get the same result as using 12 bits.

PI>bye SET
- - - - - - - - - - - - - - - -
After all the useful suggestions, my consideration has migrated to the
OKI MM5570 ADPCM CoDec. This chip has compression and a TouchTone DTMF
generator, Voice-operated TX & RX mute/process, various ringing &
call-up tones, and onboard output amplifiers.
The OKI spec sheet says:
"These CODECs convert 300 Hz to 3400 Hz voiceband analog signals into
either 32 Kbps to 2048 Kbps ADPCM or 64 Kbps to 2048 Kbps PCM serial
data."
1> I take that info to specify the input range of sampling rates with the output
being  somewhere in the 4-8 Kbps range.  Correct?

2> What are the comparative merits between A-law & u-law compression schemes?

3> I must send this data thru Async RS-232 port.
The spec sheet says "Transmit/Receive synchronous operation only."
If I transmit 8Kbps compressed-voice PCM serial data at RS-232 Async
port speed of 57600 baud into a FIFO buffered receive port, shouldn't
that be capable of maintaining sufficient sync stream flow for the CoDec
to reproduce correct speech data integrity?

Many DANDY helpful answers heard from all over the world...Keep 'em
comin'!

TIA,
John
sirejohn@bbs-la.com