What's new
By:

DL-0921 / CPS-B-21 Reverse Engineering

WydD

WydD

Enthusiast
Joined
Mar 6, 2019
Messages
21
Reaction score
165
Location
France
Heyo!

I'm very pleased to share to the community my work on the CPS-B-21.
I have spent months analysing the die shot of this chip and I've produced a complete documentation and schematics.
DL-0921.png


Here's the overview image of the schematics


You can access the complete documentation here (and the source code): https://gitlab.com/loic.petit/cps2-reverse/-/tree/master/DLs/DL-0921
And you can access the online schematics viewer here: https://petitl.fr/cps2/DL-0921/

CPS2 Reverse Engineering Project
This work is part of the more global effort of recreating the schematics of the whole CPS2 including all the custom chips.
Currently you can find complete schematics for the A-board, the B-board and E-board, as well as the reverse of the DL-1727, DL-1827, DL-1927 and DL-2027 chips.
Project home

I hope you enjoy it!

WydD
 
Seriously awesome work, thanks for the time and dedication you put into this!
 
Great work! Was very surprised at the amount of pages = that's a lot of gates! No surprise it dies all the time!

The great thing with the C Chip is the C board is on a connector so it's no problem on pinouts etc allowing any chip(s) to be used.
The A chip does not have much choice and must fit on the pinout / space.
 
Great job! thanks for sharing.

For CPS-A-01, maybe using flexible PCBs could be used to connect the 4 sides for connections to an elevated replacement PCB?
 
pacoarcade said:
Great job! thanks for sharing.

For CPS-A-01, maybe using flexible PCBs could be used to connect the 4 sides for connections to an elevated replacement PCB?
Click to expand...
That was the same idea, that I had when he released the other CPS2 customs. flexy strips that connect to a bigger PCB on the top. Otherwise, is complicated to replace and I'm not considering using level transtalators...which will make things even more complicated.
 
Hammy said:
Great work! Was very surprised at the amount of pages = that's a lot of gates! No surprise it dies all the time!

The great thing with the C Chip is the C board is on a connector so it's no problem on pinouts etc allowing any chip(s) to be used.
The A chip does not have much choice and must fit on the pinout / space.
Click to expand...
Nope I actually have more than 1 solution for the CPS-A-01. I already know I can't implement a solution fitting in the OG QFP footprint.
 
This is excellent work, hopefully CPS-A will get looked at in due course...

I'm hoping to figure out from this, in terms of key programming, what the unknown registers do and how the 16x16 multiply block works, i've long suspected they configure it in some way.
A while back I discovered the unused /hblank pin 117, weird how Capcom basically ignored the hardware H/V blanking and went with software blanking. It can be combined with vblank and lutoeb to get a proper fix for the colour issues (on certain monitors) with 3wonders title screen etc.

WydD, i've got reverse eng'd schematics of various B and C boards, you're welcome to them if you have interest in adding CPS1 to your documentation project?
 
Last edited:
Hammy said:
Great work! Was very surprised at the amount of pages = that's a lot of gates! No surprise it dies all the time!
Click to expand...
69 pages... ~2000 components but a lot of them are compound gates (like 8 bit shift registers and such)... ~10000 wires (depending on how you count)


Apocalypse said:
If you reverse the CPS-A-01, would it be possible to use your work to create a modern replacement for the physical chip? All credits given of course.
Click to expand...
Yep, the schematics are provided as CC-BY, basically a long as you don't steal the credits I'm fine with it. And to everyone who is inspired by this: please let me know, I'd love to see this work being useful :)


rockbottom said:
I'm hoping to figure out from this, in terms of key programming, what the unknown registers do and how the 16x16 multiply block works, i've long suspected they configure it in some way.
A while back I discovered the unused /hblank pin 117, weird how Capcom basically ignored the hardware H/V blanking and went with software blanking. It can be combined with vblank and lutoeb to get a proper fix for the colour issues (on certain monitors) with 3wonders title screen etc.

WydD, i've got reverse eng'd schematics of various B and C boards, you're welcome to them if you have interest in adding CPS1 to your documentation project?
Click to expand...
For the first part: I tried to document as much as possible everything that I noted. You can see how the multiplier works with the previously unknown registers here https://gitlab.com/loic.petit/cps2-...scheme.md#multiplication-and-the-palette-lock and for the blanking, I'm not that knowledgeable on this. All I know is that the sync signal is combined with the two VSYNC/HSYNC on CPS1 and 2 and you have documentation and schematics on how VSYNC/HSYNC are generated.

Second part: I never considered adding CPS1 because the schematics were leaked. But I'll have to think about how to integrate those maybe :)
 
WydD said:
... You can see how the multiplier works with the previously unknown registers here ...
Click to expand...
So looking up "UAF386" (DSP "mega cell" multiplier) in the old Ricoh databook, the TCx inputs allow the use of either two's-complement or sign-magnitude inputs for the mult. So from this new schematic, games could choose mode via bit 0 of (previously unknown) "MULT-TC" reg. None ever did so evidently... but good to know I guess :D The CHECKx and RASTER reg's are new to me also.

Can't help but wonder how many hours this would have saved me back when I was figuring out the keys for the pre-b21 configs :P Still, great to now be able to confirm they're correct. :thumbup:
 
You must log in or register to reply here.
Back
Top