Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!munnari.oz.au!csc.anu.oz.au!csis!rackland
From: rackland@csis.dit.csiro.au (Ross Ackland)
Newsgroups: sci.electronics
Subject: Re: Audio from a VGA card
Message-ID: <1990Nov14.222052.7915@csis.dit.csiro.au>
Date: 14 Nov 90 22:20:52 GMT
Organization: CSIRO Div of Inf Tech
Lines: 47

Had an interesting reply to my original suggestion of generating audio 
using the video DAC's on a VGA card. Here it is in a slightly edited
form:


>Your idea is not bad at all.  I thought of doing the same thing with one of
>HP's hi-res video cards awhile back.  The biggest problem to overcome is the
>horizontal and vertical blanking intervals.  You will have to program the
>VGA timing registers so as to generate NO interval at all, lest you get
>a bad 60 Hz buzz and such.  I'm not sure if it can be done, but its worth a
>try.  Also, you should capacitively couple the signal into your amp.  Use about
>1000uf cap, positive side connected to VGA.  I recommend testing on a cheap
>amp before you try your home stereo.  Finally, be sure to DISCONNECT or
>TURN OFF your monitor when you enable the program.  You can very easily
>damage many monitors with out-of-spec timing.  You can try a sine wave at
>first to test it out.  The VGA dot clock is about 25 MHz (I think), so
>each pixel represents a 40 ns sample of sound.  You will need to use mode 0x13
>(320x200x256) to get 8 bits per pixel/sample.     Unfortunately, in this mode,
>scan lines are doubled and pixel rates are halved.  This results in only
>80ns * 320 *200 = 2.5 milliseconds of sound in one frame.  Thus, not only
>will you signals have to be periodic on a scan-line basis, but also on a
>frame basis (the ends of scan lines must approach each other in value, as
>well as the top and bottom of the frame).  
>
>Due to the limited amount of sound per per frame, and the
>inability to update the frame in 2.5 ms, I scrapped the effort and never
>built anything.
>
>Also note that VGA RAMDACS only give you 6 bits per gun, which is
>really a shame. 
>
>Finally, it is worthy of note that the original MacIntosh used
>a similar scheme to get sound and floppy motor drive signals.
>However, they only used one word of data per scanline (at the 
>end of each line) and gated that word to the sound DACs instead
>of the video DACs at the right moment.  Because the video ram 
>memory fetch still ran even during the blanking intervals, these
>did not affect the sound quality at all.  They still had the
>requirement that the last scanline sound data approach the
>first scanline sound data in value if you expect to avoid
>pops in the sound due to discontinuties.  Of course these pops
>would occur at a 60 Hz rate, and thus your filter might help
>with those too.
>
>Jim Christy
>HP Labs
>jchristy@hplren2.hpl.hp.com