Digital I\O is easy. You could get 100% accurate 320bit 96MHz sigital audio
if you had the bandwidth. I am not aware of any issues wihin Windows. Almost
all consumer soundcards use an AC97 codec which is limited to 48kHz so
everything must be resampled.

Sorry I may have misworded it, my technical speak is a little poor. A
Soundblaster card using 16bit DAC made with professional production
techniques (you really *NEED* 4 layer PCB's for high quality audio PCBs)
performs overall as a 13bit DAC due to design issues.

Sure putting it external helps, but it's still real hard. It's not outside
noise that's a problem, it's internal noise. As I say, soundcards are
generally desinged with very expensive EMI minimisation packages run by
extremely experienced engineers.

Sorry, I don't mean to discourage you. I too had had such an idea.

Finally, this doesn't mean it's impossible, just very hard.

Tom.

-----Original Message-----
From: Simon Nield [mailto:simon.nield@QUANTEL.COM]
Sent: Tuesday, October 31, 2000 9:47 PM
To: PICLIST@MITVMA.MIT.EDU
Subject: Re: [OT]: Muso's of the world unite! ;)


Tom,
agree with a lot of what you said, but some clarification:

>1) Good luck ever getting 24\96.

There are pro soundcards available that do this with digital IO. The HW for
Digidesign's protools
for example. I think there may be some 'issues' with 24bit audio in Windows
itself, but that would
not stop you accessing 24bit audio with your own application.

>sophisticated EMI reduction software) and one of they're "16-bit"
soundcards
>can only achieve 13-bits dynamic range.

S/N+THD is the problem here actually, not dynamic range. (signal to noise
plus total harmonic
distortion)

>Why cause it's REAL hard and REAL
>expensive to design even a 16\48 soundcard let alone 24-96.

The digital bits are trivial, the hard/ impossible bit is coping with the
noise inside a PC... the
sensible soultion is to not waste money trying and do the conversion
externally.

>2) You're not leaving any DSP headroom. You always have more effects
>headroom then master headroom. This is why 20\48kHz is a common format.
>You'd never burn a 16\44.1 master to a CD, you always downsample first. So
>that the effect that moves a 45kHz signal down to 38kHz has something to
>work on. A 20\48 converter is prob. more manageable and well suited to CD
>mastering. 24\96 is more for DAT (24\48). You'd need a REAL good ear to
>notice the lack of >24kHz artifacts in a 48 compared to a 96 recording.

down-sampling from 48KHz will cause an increase in noise. 48KHz sampling
exists as it is a nice
integer multiple of common broadcast / film frame rates (24, 25 and 30 Hz)
and is >> 2x20KHz so the
anti-aliasing and reconstruction filters are not horrifically complex. 96KHz
either lets you use
even simpler filters, which means less filtering artifacts, or you can
increase the bandwidth.

More dynamic range is an issue when processing signals though. Even 24bit
data / 56bit accumulator
is limiting in a fixed point processor - IIR filters can end up with very
small coefficients,
effectively limiting filter accuracy. Using a reasonably low-spec PC you
should be able to do some
pretty cool real-time (with some latency) effects using much easier floating
point code.

Regards,
Simon

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.