[Coco] CoCo 3 FPGA?

[L. Curtis Boyle]

On Tue, 31 Jul 2007 13:07:36 -0600, <jdaggett at gate.net> wrote:

Would it not be better then to just do 8 bit color, with a 24 bit palette  
(perhaps latched registers to set them all)? If you can pick you own  
colors from a 24 bit palette (original VGA only had 18 bit), then you  
could do 320x200x256 with only 64K of RAM (and 640x200x16 for the same).  
It would be easier to implement the code too, not having to worry about  
1.5 bytes for a palette definition, etc.). You could do a 160x200x256 in  
32K, which would match the original Coco 3 "mystery mode", except you  
would have way more colors to play with.


> On 31 Jul 2007 at 12:54, Joel Ewy wrote:
>
>> Why not a real 12-bit 4096-color mode as a less memory and (possibly)
>> CPU intensive alternative to 15- or 16-bit modes?
> *****************
>
> Memory wise 12 bit color will require as much memory as 16 bit unless you
> compress and eliminate the unused nibble. The issue really is not how  
> much
> memory needed for a scren but how fast that memory can be written.
>
> IF  you have 640x480 with 16 bit color, 2 bytes per pixel. You need  
> 614,400
> bytes of info. The 6809 takes a minimum of ten machine cycles to do a
> double byte load and store for a transfer. At 80nS machine cycle, 12.5MHz
> buss speed, to transfer a full screen will require ~24.6 milliseconds.
> Compared to the current Coco's 640x200 at 2 bit color of ~9.26  
> milliseconds
> for double byte load and store. Thee concern is time. With a verticle  
> refresh
> of 59.94 Hz, that means each screen needs ~16.68 milliseconds to refresh.
>
> For 16 bit color the refresh time needs to be reduced or the data  
> transfer
> into video memory needs to be speed up. To meet 640x480 16 bit color to
> write a full screen would need a buss speed of at least 20MHz. With the
> current setup of the CPU/GIME chip IDMA transfer would require 25nS
> dram max. To get that you need to either have SRAM or SDRAM.
>
> Finally you need to  have the software in SECB and/or Nitros9 su pport  
> that.
> Having the hardware doing it is one thing. If there is no software to  
> take
> advantage of it is then not worth the hardware effort.
>
> It is possible to get 20 MHz buss speeds with an FPGA. It is very much
> possible to expand many functions of the GIME chip. WHo is on board to
> upgrade software to take advantage of the  new features?
>
>
> james