[Coco] Any info on a mystery 256 color GIME mode?

[jdaggett at gate.net]

Roger

Considering the R&D Memo and have 512 capable and pallette ram of 256 colors 
seemed odd as 256 displayable colors can be done with 8 bits. 512 color with 9 bits. 
512 color can be done using RGB333. 256 color is done using RGB332. It is stupid 
to use lookup tables for 8-bit color unless you only plan to display 16 to 48 at aany 
given time. 

There could be a mode that uses the upper two bits of the pallette registers. The 
chip seems to have three DACs already for the RGB signal.  Internally they may 
already be 3-bit. To use just 2-bit color DACs, the lower bit of the DAC is set to zero 
through a 2:1 mux. In this case the mux is essentially a 2 input and gate. That is 
easy to do in silicon and does not take up much real estate. 

My quick guess based on the time frame the chip was designed and fabricated, the 
die uses either 0.75 micron or 1 micron process on either a 6 inch or 8 inch warfer. 
Best guess is a 6 inch wafer and 0.75 micron. So if the die is 0.25 inch square the 
real estate is there to put in some extra circuits. 

As I stated in my other email, there are some inconsistancies to me that are not 
totally clear. The yyyyyrgb format tends to lead me to believe that is for a composite 
signal. I would think that RGB 332 mode would be used if the CM8 is hooked up. 
Also if the mode ws intended for games, composite video to a TV is more likely the 
intention. 320x200 mode is not to much a stretch for a TV set. 160x200 is just fine. 
The CoCo1 and 2 maxed out at 256x192. Played games using a coco1 and a TV. 

If it does exist, the information needed is unknown as to how many steps are in the 
sequence to put the internal state machine in the 256 mode. 2? 3? 4? 5? Also what 
are they? Anything done will be just stabbing around in the dark. My first guess is to 
put the video in 320x200 or 160x200 mode. After that it is a poke in the dark!!!!
 
james

On 2 Dec 2003 at 15:36, Roger Taylor wrote:

> At 07:13 AM 12/3/2003 +1000, you wrote:
> >I was just wondering what your opinions are on my personal quest to
> >resolve the secret 256 color mode...is it real or is it fake? Read my
> >web page I have created documenting all the information I have
> >collected to date.
> >http://www.nickm.launch.net.au/ProjectArchive/256mode.html Nickolas
> >Marentes
> 
> My guess at the moment is that there's a hidden or undocumented switch
> in the GIME that enables the upper 2 palette bits, maybe enabling an
> intensity field of the usual xxrgbrgb color format.
> 
> If this secret mode is actually the same palette of 16 colors that can
> hold iirgbrgb values, then that's 4 intensities of an RGBRGB
> combination that yields one out of 64 possible colors.
> 
> Perhaps these upper 2 bits can only be active at a certain time, or by
> performing a "secret" process.
> 
> --
> 
> My big question is: is this 256-color mode a secret mode or a real
> mode that was being put into the CoCo 4 prototype?  I came across some
> CoCo write up, I think on one of the museum sites, that mentioned a
> CoCo 4 was being designed but never made it.
> 
> If the CoCo 3 guys snuck their photo into the ROM, then my guess is
> that somebody was either trying to sneak a 256-color mode in, or that
> there was actually going to be such a mode, since a 16-color mode just
> wouldn't cut it in a new CoCo model, especially when other computers
> at the time were doing 256 colors easily.
> 
> If this mode does exist in the CoCo 3, then we have 2 GIME chips to
> hack to death in order to discover it since the interviewed developers
> can't seem to recall anything useful to us.
> 
> We also have to ask ourselves, what amount of memory is required to
> support 256 colors at a time.  It seems to me that each pixel would
> have it's own byte or palette indice.  That's roughly 64,000 bytes for
> a 320x200 screen.  51,200 bytes for a 256x200 screen, etc.  It's easy
> to compute by multiplying the width times the height.
> 
> 
> -- 
> Coco mailing list
> Coco at maltedmedia.com
> http://five.pairlist.net/mailman/listinfo/coco