Path: utzoo!mnetor!tmsoft!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!elroy.jpl.nasa.gov!jarthur!ucivax!orion.oac.uci.edu!ucsd!sdcc6!sdcc13!cpenrose
From: cpenrose@sdcc13.ucsd.edu (Christopher Penrose)
Newsgroups: comp.sys.next
Subject: Re: DAC
Summary: filter points may not be that important
Keywords: filters
Message-ID: <15007@sdcc6.ucsd.edu>
Date: 18 Dec 90 19:40:34 GMT
References: <16894@csli.Stanford.EDU> <108170005@hpcuhd.HP.COM>
Sender: news@sdcc6.ucsd.edu
Organization: University of California, San Diego
Lines: 49
Nntp-Posting-Host: sdcc13.ucsd.edu

In article <108170005@hpcuhd.HP.COM> edwardm@hpcuhd.HP.COM (Edward McClanahan) writes:

>Well, you could probably pump data to the DAC connected to the DSP chip
>at an arbitrary rate, but recall that an extremely steep low-pass filter
>is required before and/or after the D->A step.  This filter is difficult
>to design because it must try to avoid phase (and other) distortion near
>the Nyquist frequency (half the sampling frequency).  The 22KHz (Nyquist)
>frequency may be well beyond your hearing, but filters arn't perfect.  The
>low-pass filters in CD players (and the NeXT) tend to invert or shift the
>phase of frequencies near their "cutoff" point.  Also, those frequencies
>may be boosted or attenuated as well.  Suffice it to say that in order to
>minimize these effects, the low-pass filters are designed for set "cutoff"
>points.  The one in the NeXT apparently supports two points corresponding
>to the 22KHz and 44KHz sampling rates.

It would still be extremely useful to have the ability to convert signals
at arbitrary sampling rates.  First, we don't know what the exact cutoff 
points are on these filters.  Anyone have the frequency response of the 
NeXT DAC filters?  I have used MTU digisound-16 dacs that had a ceiling
sampling rate of 48KHz.  The filters, however, rolled off (-60db) at about
18.9KHz.  As disk and cpu resources are limited, I chose lower sampling rates
for my computer music pieces.  My piece "Lesion", was converted at 30KHz
using the 18.9KHz filter.  There was a small amount of high frequency aliasing
around 12KHz-15KHz, but I liked it in context.  No one else has noticed this
distortion.  A later piece: "CircusCircus" was converted at 36KHz.  I 
perceived no aliasing distortion and the frequency response spanned the
bounds of the filter.  

>On several Leo Kottke CDs, an audible ring at around 10KHz cannot be 
>avoided by playing them on any CD player I have tried (some costing over 
>$3000).  Alas, even albums arn't immune to this problem if the studio 
>used digital mastering (more and more common these days).

I don't want to get into the classic analog/digital debate, but each audio
medium clearly influences its output sound.  You may notice digital
phase cancellation, reinforcement, and aliasing, but I can't help hearing
poor channel seperation, wow & flutter, high frequency distortion, 
and noise associated with turntables.  The digital medium provides me, as
a composer, with a lot of control, and I'd like to be able exploit the
"limitations" of the medium if I choose.  Arbitrary sample rate conversion
would be cool.  If 2.0 doesn't support it, maybe I'll hack it myself. 

Again, anyone have the frequency response handy for the dac filter(s)?  As it
stands the converters are half a sample out of phase for stereo conversion.
But this seems to be nitpicking.  The conversion interface is the best
that I have ever used, and its sitting in my living room.

Christopher Penrose
jesus!penrose