[music-dsp] 24-bit 96Khz

Michael Gogins gogins at pipeline.com
Mon Feb 12 18:51:39 EST 2001


My equipment is an M Audio Delta 66 sound card, which produces up to 96
kHz/24 bit output, going into a Mackie 1402-VLZ mixer, a Samson Server 240
amplifier, and Alesis Monitor One nearfield monitors or Sony WDM-6
headphones. The soundfiles are played from Cool Edit Pro which takes float
soundfiles and outputs through the Delta 66's ASIO driver. I'm sure that
somewhere in there, probably in Cool Edit Pro, the floats are converted to
ints for the ASIO driver.

I agree about the downsampling. I wonder if Cool Edit Pro does a simple
downsample on integral conversion ratios.

----- Original Message -----
From: Robin Whittle <rw at firstpr.com.au>
To: <music-dsp at shoko.calarts.edu>
Sent: Monday, February 12, 2001 4:42 AM
Subject: Re: [music-dsp] 24-bit 96Khz


> Michael Gogins wrote that with synthesised music (using Csound) that
> there was an audible difference between a "well engineered 44.1 kHz 16
> bit system and a well engineered 48 kHz float sample system".
>
> As far as I know, there is no DAC hardware which works with floating
> point numbers, so the latter system would consist of an integer DAC
> driven by some software which could accept floating point numbers.
>
>
> I assume you rendered the piece at two different sample rates 44.1 kHz
> and 48 kHz and stored the results in integer and floating point formats
> respectively.
>
> > This is a good test, because the signals, being Csound synthesis of
> > the same orc and score, are absolutely identical except for signal
> > bandwidth and DACs.
>
>
> This could point to problems in one or both sets of hardware, but
> another explanation might be that the complex algorithms produce audibly
> better results when run at a higher sample rate.
>
> I have written a ugen for Csound to downsample 2:1 with a FIR.  It takes
> two L R inputs (say with a-rate set to 88.2 kHz) and and spits out a
> single variable (also at 88.2 kHz) which consists on one cycle of the
> downsampled L and on the next the downsampled R.  By saving this as a
> mono file, but treating it at playback time as a stereo file, Csound can
> be made to render internally at twice the final sample rate, without
> mucking around with intermediate files and post-processing.  Saving as a
> .WAV file is a bit tricky, since the flags at the start will be set for
> Mono - but you can change one byte (I forget which, but it is easy to
> find out yourself) to make the file appear stereo to the player.  If
> saving as headerless 16 or 24 bit Ints, or as 32 bit floats, it is
> simply a matter of telling the player the file is stereo.
>
>
> The test would be to render a piece at 44.1 kHz and then at 88.2 kHz
> using some kind of down-sampling system to 44.1 kHz.  Then listen to
> them both on the one 44.1 kHz piece of hardware.
>
> An audible difference would show that it is best to synthesize digital
> audio at high sampling rates.   I think in general this would be *highly
> desirable*.
>
> What I was discussing in earlier messages was Analogue to Digital
> sampling of signals from the physical world, where, as far as I know,
> 44.1 kHz is perfectly adequate if done properly.
>
>   - Robin
>
> dupswapdrop -- the music-dsp mailing list and website: subscription info,
> FAQ, source code archive, list archive, book reviews, dsp links
> http://shoko.calarts.edu/musicdsp/
>


dupswapdrop -- the music-dsp mailing list and website: subscription info,
FAQ, source code archive, list archive, book reviews, dsp links
http://shoko.calarts.edu/musicdsp/




More information about the music-dsp mailing list