0.154 [Roberto Fresca]
WIP:
- 0.154: Roberto Fresca added 'Ming Xing 100 (Star 100)' (Sang Ho 199?). Mostly fixed Star 100 colors (fg tiles are 4bpp, not 3bpp) and improved bg color. Fixed rightmost reels being cut off [Alex Jackson]. Use a Chinese title for Star 100 since that is all it shows on the title screen [Justin Kerk].
- 3rd June 2014: Roberto Fresca - Finaly found the attributes RAM for each reel and worked it to get perfect graphics. So the video circuitry finally has 5 (FIVE) layers! About the 1Ch-1Dh commands, I was right. They are just part of a RAMDAC initialization (thanks Angelo for the tip!), creating and dinamically modifying a 256-entries 6-6-6 RGB palette. Even when through attr RAM the tiles use bits 4-5-6-7 for colors, I can't find a proper way to map the color codes. Just displaced the color codes in the gfxdecode start to get the title girl ok, and most of the 1st gfx bank colors ok (it's a temporary hack till can understand the palette format/bank/access). BTW, the game is working properly. The only issue is the lack of accurate colors.
- 24th May 2014: Roberto Fresca - Here is another dark hardware manufactured by Sang Ho (bankrupt approx in 2010), silkscreened 'SANGHO PM-003 (VER-B2)'. It's a Z80 based system, with 1x KC89C72 (AY-3-8910 compatible), a couple of Altera CPLD's (one identified as EPM7032LC44-15), 1x unknown custom IC marked 0453LK003, and 1x (unused) OKI 6295 for ADPCM samples system. I spent some hours on it, and got some preliminary response. Seems that the system has at least two video RAMs that should render to different tilemaps, and finally merge for the final output. EDIT #1: Just identify the whole second video RAM, hooked it to a new tilemap, and cooked all together. Also adjusted the 1st graphics bank and tile codes accordingly. Still preliminary, but at least now we know for sure the game involved *IS* 'Star 100'. EDIT #2: The girls appeared (also the reel symbols), but obviously with 8 pixel lines instead of 32. The second tilemap is not accurate, since is supporting 8x8 pixels tiles, when the second graphics bank (that contains the girls and reels symbols) has tiles with a size of 8x32 pixels. This is why you can see the tiles lines 'clipped'. Even when this is an improvement, I'm forced to break the second tilemap in different pieces, giving 8x32 pixels support. I'm strongly think that this video circuitry starts to looks close to the Golden Star video hardware. EDIT #3: Tried setting an accurate second tilemap for these 8x32 tiles. Still misaligned, but it's a start. Palette seems to be RAM based, placed at 0xF300, but I couldn't find an accurate way to decode it, yet. Also the program sets a sort of initialization through commands using ports 0x1C-0x1D, writting the index or address through port 0x1C, and then writting 3 consecutive values to 0x1D. The range starts with index 0x00 and ends with 0xFF, what made me think that maybe the program is writting a 256 colors palette based on RGB values. Need more investigation.
