[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: RGB to VGA on a IIGS (again)
mdj wrote:
On Jun 25, 8:05 pm, "Michael J. Mahon" <mjma...@aol.com> wrote:
I love my AppleColor Composites too, but I'm finding my eyes are
starting to struggle with long exposure to anything less that pristine
text. Whether this is a function of getting old, or a side effect of
now spending so much time in front of higher quality displays I'm not
sure :-)
I used a monochrome monitor for programming for years, with a 10" TV
for color, but a few years back, I set up an AppleColor Composite and
it was good enough that I never bothered to set up a monochrome monitor
in addition.
It helps that I keep my monitor behind my "open faced" Apple //e, so
when I'm in position to type, my outstretched arm can barely touch
the monitor screen more than two feet away. At that working distance,
the dot structure of the shadow mask is not distracting.
My physical layout is similar, but as I'm naturally far-sighted I can
see the effect. It's a little bit like when audiophiles talk of
'listening fatigue' ;-)
Ah--there's the difference. For me, that distance falls into the
"too far, and yet too close" zone of my bifocals, so I see it with
a little pleasant blurring. ;-)
80 column text tends to look a little odd, with alternate scanlines
interpreting the pixel widths slightly differently. Readable, but not
particularly clear.
This is a result an interaction between the monitor's sampling of
what it thinks is an interlaced video signal, but isn't.
I figured it'd be something like that, or alternatively it's that the
upsampling algorithm is very simplistic and doesn't cope at all well
with the Apple II's slightly odd clock rate.
Actually, some multiple of the color subcarrier frequency is a quite
rational choice for sampling frequency, but that doesn't match the usual
640- or 720-pixel value.
The active time of the Apple II video line is 40/1.0205 microseconds,
while the active time of a standard NTSC signal is 52.6 microseconds,
for an NTSC-to-Apple II ratio of about 1.34:1.
So the number of NTSC horizontal pixels corresponding to 280 Apple
pixels is about 375, and to 560 Apple pixels, 750. The equivalent
resolution for 640 pixel SHR mode is 858 pixels. So none of the
"usual" VGA-derived resolutions correspond well with Apple resolutions.
The result is a "beat" between the sampling frequency and the Apple
pixel clock. For a 640-pixel NTSC sampling, the sampler "lines up"
with an Apple pixel about every 10 "280-pixel" pixels, or about
every 20 "560-pixel" pixels. This should result in a quite visible
pattern of vertical stripes with those spatial frequencies.
There is a *very* minor amout of this present on my display.
Your monitor is doing a good job of supressing it.
The fact that your sampling alignment changes from even to odd lines
is the tipoff that it's interlace (or the absence of interlace)
related.
"Possibly" :-) The effect seems bound to Apple scanlines rather than
underlying video scanlines, so I figured it's related to limited luma
processing causing the shift in areas where there is only one lit
pixel, like vertical lines in the characters, a sort form of
simplistic anti-aliasing in the upscaling algorithm which is probably
quite effective for "video" signals, but less so for computer imagery.
I'd have to hook it back up and again and study it more closely to be
sure.
I see what you mean--this is a likely explanation with digital
processing. I was putting too much responsibility on the sampling
and not enough on the post-processing.
I think this problem will plague us indefinitely until either somebody
designs a custom solution, or physical display resolution becomes so
damn high than the effects disappear into the 'noise'
Right. "Smart" upsampling to 1080p should do pretty well. I haven't
tried that yet...
-michael
NadaNet file server for Apple II computers!
Home page: http://members.aol.com/MJMahon/
"The wastebasket is our most important design
tool--and it's seriously underused."