[bit.listserv.emusic-l] PCM follow-up question

silver@ENIAC.SEAS.UPENN.EDU (Andy Silverman) (02/13/90)

Well, I have this Yamaha PTX-8 drum pad set, and they say that the sampling
involved is "12-bit companded PCM."  What exactly does this mean?  The sounds
that come out of the thing are so cheesy (even on an extra Waveform ROM
cartridge) that I never use them anymore, instead using to trigger a Roland
R-5 via MIDI.

Andy Silverman
silver@eniac.seas.upenn.edu

adamson@ITD.NRL.NAVY.MIL (Brian Adamson) (02/13/90)

     Companded probably means they are really using 8-bit A/D conversion
and then using some kind of dynamic range assignment rule supposedly
optimized for drum sounds to try to get the same dynamic range as
you would get if you had used a 12 bit A/D converter to start with.

        Although this may give you some added dynamic range (sort of), it
does not change the signal-to-noise ratio max that can be achieved
using only 8 bit conversion (approx 45-50 dB).  That may be part of
the problem.

        I posted a message a little while ago about linear samples
versus companded (compressed according to u-law, A-law, etc).

                        Hope this helps,

                        Brian Adamson

kg19+@ANDREW.CMU.EDU (Kurt A. Geisel) (02/13/90)

How does companding get more dynamic range out of PLAYBACK circuitry
that is only 8-bits wide?  By having a seperate overall amplitude
control channel?

- Kurt
Kurt Geisel                       SNAIL :
Carnegie Mellon University            65 Lambeth Dr.
ARPA : kg19+@andrew.cmu.edu           Pittsburgh, PA 15241
UUCP : uunet!nfsun!kgeisel  "We just need to short-circuit the continuum on a
BIX  : kgeisel               5 or 6 parsec level."  - Forbidden Planet

JEFF@UTCVM.BITNET (Jeffrey R Kell) (02/13/90)

On Mon, 12 Feb 90 20:18:35 EST Kurt A. Geisel said:
>How does companding get more dynamic range out of PLAYBACK circuitry
>that is only 8-bits wide?  By having a seperate overall amplitude
>control channel?

No, you encode/decode sample values on a non-linear scale, most likely
some exponential function.  If the A/D D/A conversions are strictly
limited to 8-bit you would have to do this on the analog side; for example
placing a dBx compander chip set on the analog side of the converters.
The magic is the non-linear signal treatment as opposed to simply reducing
resolution or padding the signal.

<Jeff>
******************************
* If the first person who answers the phone cannot answer your
* question, it's a bureaucracy.
******************************