What's new
By:
Filters

Effort to get the GP9001 VDP chip, part of Toaplan V2 hardware, decapped

6t8k

6t8k

Student
Joined
Oct 3, 2019
Messages
31
Reaction score
31
Location
Germany
Over the past few months, in a thread at shmups forum, we've been analyzing the cause why some Toaplan V2 PCBs are not recognized by the OSSC and show artifacts with some other ADC products.
Essentially, with certain games, the GP9001 VDP outputs a broken Hsync signal, and in the course, I developed a simple patch that initializes the GP9001 VDP with configuration from another game, thereby eliminating that quirk.

Dogyuun and Batsugun however use two GP9001 working in tandem, and the patch cannot be applied in an analoguous manner there: it leads to graphical errors, probably because it interferes with the teamwork between the two chips.

In a nutshell, instead of burning dozens and dozens of EPROMs trying different initialization vectors, stabbing in the dark, just to potentially arrive at the conclusion that it's not possible or to arrive at an imperfect (and as such unacceptable) result, I figured that the best way by far would be to (eventually) decap the GP9001 – since that would have to be done anyway at some point for projects like the MiSTer and MAME.


I don't have a direct contact, but I know some people who could potentially get something going or know others who could.

In 2012, Dr Decapitator (now aka. caps0ff, who in 2016 also did sound-related chips for Vimana, Teki Paki, Fire Shark and Ghox), managed to dump a set of SNES DSPs, which are a crucial component of certain game cartridges, for byuu et al., the author of bsnes (article).

Coincidentally, byuu (now called Near) is currently trying to get the two SNES PPU (= Picture Processing Unit) chips scanned.
The magnification factor of the existing scans by John McMaster is not high enough to trace all the details.

Supposing the SNES PPU effort could pick up pace medium-term, I wonder if there are some synergy effects that could be leveraged?
 
Last edited:
If you need a chip, I have a non-working Snow Bros 2 I'd donate to the cause.

Related, not related, byuu renamed to near recently. And then somewhat confusingly renamed some of their emulators to byuu...
 
@rewrite: That's cool, can you exclude that it's not the GP9001 that's broken?

As for byuu → Near: byuu is a new emulator (frontend), and "higan and bsnes development will continue as normal". Given Near's rationale, I don't think this or the online handle change has to do with the SNES PPU.
 
6t8k said:
@rewrite: That's cool, can you exclude that it's not the GP9001 that's broken?
Click to expand...
I suppose I can't, actually, ha.

And I didn't think the name change was related, just relaying that their name changed, and that it's confusing for people who are used to byuu being the person for it to suddenly be the frontend's name instead.
 
Contacted Near, caps0ff, and another person who could potentially be aiding a decapping effort about this.

Still not entirely sure on what I've let myself in for - at this time I still only have a vague understanding about what exactly it entails to arrive at an HDL (Verilog/VHDL) representation, based e.g. on some time and effort estimations for different ICs like here. Then we'll probably have to either code a suitable computer vision algorithm to extract the logic, or go for a crowdsourced effort and manually or semi-automatically retrace the die. But I am determined! :D

Even if for some reason we don't make it until the HDL representation in one smooth go, that wouldn't be a catastrophe. The photograph would be there for a start, and as a last resort, others could then take it on in the future. One step at a time, the first step is what counts!

I guess for now it's time to wait for what arises from above-mentioned contactings. When we have more information on the chances of success and monetary scope, I'll look into options for a crowdsourced pot.


By the way, there are different four-digit numbers on the fourth and last line of the GP9001's imprint. Not aware of any difference those would make; they're likely just batch numbers, but I suppose it wouldn't hurt to decap at least two different ones while we're at it, no? But that decision would also somewhat depend on the additional cost involved, I suppose.

So far I've seen these (most certainly very incomplete):

#game(s)
9044Tatsujin Oh
9150V-V, Fixeight, Mahou Daisakusen, Tatsujin Oh
9152Dogyuun, Fixeight
9232Dogyuun
9235Knuckle Bash
9335Batsugun
 
These are some expensive games. Knuckle Bash being the least expensive. I’d be will to send mine in, if we could raise a little money so I’m not out the whole cost of the PCB.
 
So far no reply from anybody who could decap, but I'm in the process of drawing more attention to this. My means are fairly limited though, so if y'all know somebody who knows somebody who knows somebody – you get the idea. If you want it to happen, by all means, make some noise ^^

@xtrasmiley as long as the GP9001 is fine, we can use faulty PCBs, lifting the burden a little.

A big thanks to everyone who kindly offered to help so far! :)
 
Should I start a pcb fund so we have a donor board? Happy to split the cost on the one guy's post a few before this offering to send his faulty Knuckle Bash. Anyone want to throw in the other half?
 
I’m willing to donate and kick in a few bucks, just not the full cost of my PCB. Also, mine is fully working, but of course if we can track down a faulty one first that would be better!
 
@dewmansnk: thanks for these suggestions!

I focus on universities and people who are more "on our side" / more grassroots so to say, but still have a reputation and the means to produce very good results (like caps0ff for example). Decapsulation and imaging can be quite expensive.

However, I also know that I don't know unless I ask, so I requested a quote from all four companies.

Personally I'm not comfortable creating a hardware and/or monetary fund before we have secured a viable avenue – of course it never hurts to scout for low priced or broken PCBs in the meantime, and if you want to create a fund already, then of course more power to you! :)
 
6t8k said:
@dewmansnk: thanks for these suggestions!

I focus on universities and people who are more "on our side" / more grassroots so to say, but still have a reputation and the means to produce very good results (like caps0ff for example). Decapsulation and imaging can be quite expensive.

However, I also know that I don't know unless I ask, so I requested a quote from all four companies.

Personally I'm not comfortable creating a hardware and/or monetary fund before we have secured a viable avenue – of course it never hurts to scout for low priced or broken PCBs in the meantime, and if you want to create a fund already, then of course more power to you! :)
Click to expand...
Yeah an insider would be ideal as it would keep costs down and you know they are more committed to the ideals of hardware preservation. To your earlier point, creating the HDL representation from the decapped images presents a second challenge, but one that is more easily crowd sourced. I believe even a few of those decapping places offered services for HDL recreation from the chips. This might start hitting gray areas of what is considered protected IP and what isn't, but if you get a dialogue opened and decapping prices seem reasonable, maybe quoting some HDL recreating could be in the cards too. Anyways, great to hear that you are sending out feelers for quotes, curious as to how many zeros that quote will have in it.
 
Two out of four companies have responded by now, I figure it'll stay that way. More or less quoting the responses:

Company A:
  • Decapsulation, imaging of the top metal layer + bond wire assembly: $600.00
  • Planar layer-by-layer deprocessing with high-magnification, high-resolution image-stitched mosaics of each metal layer down to silicon: $10,000.00
    (estimated for 250nm technology, three metal layers and w/o scanning electron microscopy)
Company B:
  • Deprocessing of the die and detailed photographs of each layer: $10,000 - $20,000
  • Netlist extraction (which I did not require, but I guess since I explained my intent as thoroughly as I could, it was nevertheless estimated): 5-10x that price

eh
 
This is my first time reading this thread, and my first response was going to be "decapping and imaging is the easy and cheap part", because delayering and tracing into netlist is extremely time intensive and therefore expensive.... but you got that answer!
 
6t8k said:
Two out of four companies have responded by now, I figure it'll stay that way. More or less quoting the responses:

Company A:
  • Decapsulation, imaging of the top metal layer + bond wire assembly: $600.00
  • Planar layer-by-layer deprocessing with high-magnification, high-resolution image-stitched mosaics of each metal layer down to silicon: $10,000.00
    (estimated for 250nm technology, three metal layers and w/o scanning electron microscopy)
Company B:
  • Deprocessing of the die and detailed photographs of each layer: $10,000 - $20,000
  • Netlist extraction (which I did not require, but I guess since I explained my intent as thoroughly as I could, it was nevertheless estimated): 5-10x that price

eh
Click to expand...
OK, so $10,600 plus one chip. Not too bad, I mean, we could do a GFM or something, but then the question is, is this one issue worth the money when there are so many other customs we could do instead?

I mean, 10k seems like a lot, but if every PCB collector/player/MiSTer fan chipped in like $20-50, we could raise it!
 
We have to find 100 of me and you and maybe him out there. As much as I want this, I think there are other areas where the money might be better spent with respect to preservation.

With that said, in for a go fund me. Never set one up, but happy to take the lead.
 
You must log in or register to reply here.
Back
Top