[rescue] sun 3/60 update

[Dan Moisa]

There are severe limitations to the pre-inbuilt MMU, but most might show up
when you have way more RAM available than what the machines were designed
for. If we got the full firmware code, ideally for both the 3/60 and the
3/80, It would be technically possible to dump the 3/60 MMU and use the
inbuilt 68030 one and have a Frankenstein 3/70 or something. Tom is really
getting deep into the design and understanding how it works, let's see
where it goes.

This is my usual plea for more schematics. I also have a broken 3/80 I
would love to fix and release back into the wild. I'm pinging SUN folks on
LinkedIn whenever I find them on manuals, docs, etc, to limited success. I
refuse to believe that it's all lost, rather that it's in a basement
somewhere and we just need to show good intentions.

The PAL8L14 is indeed replaceable by discrete components and by two GALs, I
need to clean up the documentation since the links aren't in the right
place right now. Check out "Can we get rid of the PALs" in the main FAQ.
I'll look at why it's not in there decompiled and add it in.

The speed issue isn't as bad as mentioned and specific to the ram refresh
circuit and the ras-before-cas (or the other way around) so it can be
isolated. After the main issue of adding more RAM is out of the way, I'll
try to see how fast I can make it. I'm very sure large sections can be
replaced with modern CPLDs, again in the queue after some more essential
stuff.

For rare stuff like LS2518, eBay is the necessary evil. I'm pretty sure I
bought out the stock on a few of those parts. This is what I want to
replace first from the original. It's also fun to see who our competitors
are, in this case a bunch of PALs are sought after in the arcade world. A
ton of 68k resources come from the Amiga crowd which if anything is overly
active :)

On Tue, May 27, 2025, 10:16 Romain Dolbeau <romain at dolbeau.org> wrote:

> Le mar. 27 mai 2025 à 05:35, Dan Moisa via rescue <rescue at sunhelp.org> a
> écrit :
> > Alright folks here it is, courtesy of long weekend:
> > https://github.com/54weasels/sun3_60/
>
> Thank you so much for sharing this!
>
> Somehow, seeing the schematics in modern Kicad makes it feel even more
> complex and daunting than the original drawings... it's no longer
> possible to blame bad readability. incredible work to replicate those
> so nicely!
>
> I'd never realized that the "Sun-style" MMU on those machines was done
> in discrete logic and PALs... somehow from all the '030-era systems
> with gate-array custom chips, I expected to see that and wondered
> where you sourced them from... I understand now why they chose to do
> the sun3x for the '030 and use the built-in MMU - cheaper and smaller
> than having all of that on the board.
>
> That's a lot of PALs to program, as well. I was wondering about
> consolidating, but one of the links you have for the 3/60 mentions
> some timing issues for the PAL used in the DRAM refresh circuitry :-(
> Isn't the rare PAL8L14 just combinatorial ? At some real estate cost,
> it could be replaced by a pair of smaller PALs, couldn't it? (if it
> had registers, those many outputs might have dependency on one
> another, but not in a purely combinatorial PAL if I understand PAL
> equations correctly). That would double the load on the input lines,
> but I presume (quite ignorantly) that by replacing them with modern
> GAL the issue is mitigated. Also, it doesn't seem to have been
> decompiled in the PAL directory, is that an oversight or is there some
> issue that makes it impossible to do?
>
> It's a shame 5V stuff is disappearing, and some massively popular
> chips are no longer made. Even the ubiquitous Z8530 was retired
> recently :-( Maintaining/making more of those is going to get harder
> and harder. I wonder how you did manage to source all that stuff, some
> seems to be quite rare (e.g. LS2518).
>
> Cordially,
>
> --
> Romain Dolbeau
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sunhelp.org/pipermail/rescue_sunhelp.org/attachments/20250527/d19a4627/attachment.html>