🚀 go-pugleaf

In regards to Brian Moffet's article wrt to more colors...

I've done a couple of video tutorials before, but I'll do just the 
numbers for more colors this time...

The unasked question is, "What is acceptable graphics resolution?"
And the general answer seems to be "As good as current TV." And in
fact that is most likely true. You then might ask "How good is TV?"
And come up with a somewhat more obscure answer. 

Brian was looking to get 24 bitplanes for (sometimes called "true color")
his images. I used to work at the Image Processing Institute at USC and
one of the research projects that was done there was digital compression
of video images for television. One of the results was that for most of 
the people a 5 bit triple (R,G,B, 32,768 colors) was indistinguishable 
from an 8 bit triple (16M colors). With that in mind you might consider
another way in which the basic amiga architecture could be rewacked. 

Continue to use color registers, but keep them 16 bits wide, 5 bits each
of R, G, and B, plus on "transparent" bit for genlocking and such.

As far as memory accesses go though, if you wanted to enable a 16 bitplane
image (direct through the color DACs) that was 702 X 440 for NTSC overscan
then you are looking at 38,610 bytes per bitplane or 579,150 bytes for
a 15 bit screen, 617,760 for a picture with genlock bits in it. 

With 16 bits/pixel and 702 pixels per line, you need to fetch 1404 bytes
per scanline. Using a bitplane architecture might make this a bit tougher
but the bandwidth is the same. A scan line takes just 63.5 microseconds
or about 22,110,230 bytes/second to keep the video pipeline full. Of course
ideally you would like some time for the processor to get in there as well
so you might actually want to be able to suck data out of that memory at
44,220,470 bytes/second (1 CPU, 1 Video cycle interleaved). Now on byte
accesses that would mean you need to fetch data in 23 nanoseconds, if 
you had 16 bit wide memory that could be pumped up to 46 nanoseconds.
32 bit wide memory could let that slide to 92 nS. Now this also means
that your processor will need to get in and out in 92 nS, which I believe
a 33Mhz 68030 can do. Plus there are static rams available to do 92 ns
accesses on a regular basis so that's no problem. 

So what is the big deal? Actually nothing at all. That is the nice thing
about engineering, you can just crank the numbers through and then go
down to Fry's electronics and buy the parts. The only hitch is that the
cost of 90 nanosecond static rams, say a meg worth at least since each
screen will take 579K bytes, hmmm maybe you better make that 2 meg (the
address space is set aside for it). Will be about $50/chip and each
chip is 32K X 8 (Toshiba TC55328) so 2 Meg will be about $3,200 your
cost, say $7,000 will be contributed to the final system price. And that
is just for the Chip RAM!

I know, there are cheaper ways to do it, look at Targa boards etc. And
the only problem is that these systems don't have the "instant update"
type effect you've come to associate with an Amiga. Even the Sun 3/80 
with an 8 bit color option doesn't look "quick" until you plug in the 
graphics accellerator. 

The point that I've spent way to long making is that there is absolutely
no technical reason why any of the systems that any of a number of people
have fantisized about on the net can't be built. There are however monetary
constraints in that the implied assumption is "and I want it to cost the 
same price as my original Amiga." which is currently impossible. I guess
this is sort of a flame too in that if it is possible I'd rather see 
more stuff that said "Here is where I'd like to see the Amiga move toward,
and here are concrete suggestions for moving the architecture forward so
that 5 yrs from now when you can build it the software still works." or 
"This is what I'd like to see in a high end Amiga and this is what I'd
like to pay for it." Rather than, "Why don't you guys open your eyes and
realize that a 2 billion color display would sell zillions of Amigas." 


--Chuck McManis
uucp: {anywhere}!sun!cmcmanis   BIX: cmcmanis  ARPAnet: cmcmanis@sun.com
These opinions are my own and no one elses, but you knew that didn't you.
"A most excellent barbarian ... Genghis Kahn!"