Wednesday, November 29, 2017

TGP status update

ROM status

In a previous post we talked about capturing TGP ROMs and later discussed a project to digitize them. However, as digitization proceeded, the quality of digitization was going down. After a bit of sleuthing, we noticed we had overlooked something.

First, one thing that had been noticed was that some chips have this interesting alternating bit appearance (ex: 315-5677a):

That is, the bits on the left of the column pairs are lighter than the ones on the right. Wasn't sure what to make of this, but most of the bits were resolvable, so didn't care too much.

Anyway, here's 315-5571 at 20x:

And here's 315-5677 at 20x:

Other than some color differences, at first these look identical. But let's put them together:

Aha! There was a slight process shrink. This turns out to be enough that it slightly reduced the accuracy of recovering some of the ROMs. Specifically, MB86233 is the large process, and MB86234 is the smaller process.

That aside, these dies are actually quite similar.

 And here's a MB86234 (315-5677):

At first these look identical, but there are some small differences, such as upper left layout. Here's MB86233 (315-5571):

While MB86234 (315-5677) looks like this;

Upper right also has some noticeable differences.

Moving forward

Following this, MB86234 were cleaned and re-shot at much higher resolution. This was a bit tricky as it made image stitching a lot harder, but most of the issues have now been resolved.

While some of the 20x MB86234 were fine, for example (315-5679A):

Others we couldn't get as nice an image at 20x (315-5679B):

But looked a lot better at  100x (315-5679B):

Although the contrast may not be as good, more detail is available, such as bits appearing as a + instead of a white blob. This allows, for example, differentiating a dust blob from a bit.

We are currently checking over results to see if the new results can be captured with CV or should be given to the typing monkeys. Most likely though we'll generate a CV reference and compare it to the crowdsourced result. We are also evaluating if any of the existing captured sets need to be re-done due to suspected errors. Hopefully though we can simply filter out inconsistent tiles and just recheck those.

Die ID

One interesting tidbit is comparing die IDs. For example, here is 315-5677:

Here is a table documenting the relationship between the various parts:

Sega #PartDecap #Die IDProcess
315-5571MB8623314MB 86233-002 Large
315-5572MB8623315MB 86233-005Large
315-5573MB8623316MB 86233-004Large
315-5677MB86234211MB 86234-004Small
315-5678MB86234212MB 86234-002Small
315-5679AMB86234213MB 86234-004Small
315-5677AMB86234215MB 86234-005Small
315-5679BMB86234216MB 86234-005Small

In particular, there are duplicate IDs! Haven't looked over these yet to see if the ROMs are identical.

Finally, where are the other Sega 315-5XXX parts? Stay tuned for a post in the near future.


We have collected sufficient data to document the TGP ROMs, but need to spend some time sifting through it to determine next steps. Stay tuned on the MW forum for announcements from Monkey administrators.

Enjoy this post? Please support us on Patreon! Note: with the Indiegogo campaign over we unfortunately don't currently have a way to accept one time donations.

No comments:

Post a Comment