StereoView raster sizes

[Thant Tessman]

In article <9008102119.AA04851 at chem.chem.ucsd.edu>, sdempsey at UCSD.EDU
(Steve Dempsey) writes:
> Recently I was looking at some stereo images on a 340VGX with StereoView 
> and I noticed that that they seemed slightly compressed along the vertical
> axis.  I ran a test program to generate some rasters of a known size and
> then measured the raster heights.  My initial assumption was that a raster
> with a height of 492 pixels and displayed with setmonitor(STR_RECT)
should have
> the same height as a raster of 984 pixels displayed with setmonitor(HZ60).
> 
> Instead, I found that the stereo raster was about 6% smaller than I expected.
> Now that I know the value of this factor I can adjust the aspect ratio
> of my projection transformations to compensate, but I was wondering if
> other people have seen this and if they measure the same shrinkage factor.

I assume it is because you have a Mitsubishi Diamondscan multisync 
monitor.  This monitor is a little too smart.  When the stereo stuff
was first done by StereoGraphics, they modified the monitors to double
the vertical displacement.  This made the pixels twice as high as they
were wide.  The proper aspect ratio for the window was 1280 in 'x' to 
492X2 in 'y'.

But since 40 scan lines were missing (used for vertical blank), the 
total picture height was slightly less than the non-stereo image.
SGI's stereo (and the newer StereoGraphics stereo) uses a Mitsubishi 
multisync monitor.  It does all adjustments digitally, and it 
automatically adjusts for a different number of scanlines by 
stretching the scanlines available to fit the area stored in its 
memory.

In other words, it stretches the 492 stereo scanlines to 
fit into the same place that 1024 scanlines fit in non-stereo
mode.  This makes the pixels higher by 2.08 instead of 2, or 
stretched vertically by about 4%.  This means that if you want
to compensate for it in a window that is 1280 by 492, the aspect
ratio should be 1280 to 1024, or 1.25.

Unfortunately, all this means that an image that is exactly correct
for the Mitsubishi is slightly incorrect for other monitors and
vice versa.  SGI's marketing wasn't sure that 4% incompatibility 
was a big enough concern to go with a custom modified (higher 
cost) monitor.  

Another solution is to adjust the monitor to be correct for 
stereo mode and slightly incorrect for the non-stereo mode.

Hope this helps.

> Steve Dempsey						       (619) 534-0208
> Dept. of Chemistry Computer Facility, B-014	   INTERNET:   sdempsey at ucsd.edu
> University of Calif. at San Diego		     BITNET:   sdempsey at ucsd
> La Jolla, CA 92093				       UUCP:   ucsd!sdempsey

thant