Re: Another look at the Carte Blanche video

[Vladimir Ivanov <vladitx@XXXyahooXXX.com>]

On Wed, 23 May 2012, Alex Freed wrote:

> On 5/23/2012 7:34 AM, Vladimir Ivanov wrote:
> > 2048 entries holding RGB data (x9?) should look like this:
>
> My design doesn't use any BRAM except for the font "ROM" for video. A bit 
> more work needed, but even SRAM (loaded from a disk or initialization EPROM ) 
> could be used for a look-up table. We use less than 10% of the SRAM 
> bandwidth.

Good. A bit late answer, but I finally stole few minutes to generate 9-bit 
wide lookup table if you want to experiment. I am just a csa2 lurker these 
days.

http://www.softjunk.org/sw/apple2/ra2/files/screenshot/ntsc/bpp/tbl_ntsc9.h

9-bit wide shift register at 14 MHz fed with Apple II serial video. Goes 
through lookup table #0 .. #3 depending on phase relation with 3.58 MHz 
color carrier, i.e. 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3 ...

Each entry is R8G8B8 triplet, which you should reduce to R3G3B3 or 
similar to fit nicely in a single 2048x9 BRAM.


Example:
  Serial video is 00110011011001100....
  lookup table #0 with 001100110  -> output pixel 0
  lookup table #1 with 011001101  -> output pixel 1
  lookup table #2 with 110011011  -> output pixel 2
  lookup table #3 with 100110110  -> output pixel 3
  ... repeat for the whole line ...


And the few points of caution:
  1) I can't remember vector direction, so just try both, i.e. whether 
shifter is clocked left or right
  2) lookup entry gives middle pixel value for a 9-bit wide vector, like 
this "xxxxXxxxx", so allow for some spare pixels on left and right edge of 
screen
  3) HGR and DHGR have different start phases with respect to color 
carrier, so take this into account as well or simply do DHGR always

1) and 3) give total of 8 permutations to try, in the worst case.

Some excuses in advance if I messed up something.

  -- Vlad