[rescue] Sun 3/160 and CADDStation 32 deskside

[Romain Dolbeau]

Le jeu. 14 août 2025 à 18:28, Dave McGuire via rescue
<rescue at sunhelp.org> a écrit :
>   One could easily fill the address space with some nice low-power, high-density surface mount SRAMs.

Not so easily, unfortunately; Sun-3 are just a bit too old or too
modern for that.
Too old because the MMU isn't built-in (like the '030 and later), too
modern because it does have an external, discrete component MMU and it
can't be ignored for timing reasons.

One would need to compute the timings based on the Physical Address
(PA) bits coming from the 2-levels MMU, so after the delay of 2
consecutive SRAMs. Possible, but slow - it would likely add at least a
cycle or two of waiting on each access. With DRAM, machines like the
3/60 (the one I examined in detail) save time by using low-order
untranslated address bits for RAS (available when the CPU assert /AS),
and then the PA bits for CAS - as they are only needed after a fairly
lengthy required delay between RAS and CAS.. That way they don't pay a
huge latency penalty for having the MMU. It also limits how many bits
can be used during RAS and therefore how much DRAM can be easily
shoehorned into the 3/60 design (a side-effect is also that byte
selection is done the "wrong" way CAS vs. RAS so 72-pins SIMMs aren't
directly usable as replacement for a set of four 30-pins SIMMs...).

Of course this assumes you're putting the SRAMs directly on the CPU
bus. However for an external board they would live on the P2 bus. From
the pinouts (available in the 3/160 schematics, not sure where/if
there is any available documentation of it), the P2 isn't just the
'020 bus after translation; signals include "P2.CAS-", "P2.ENDRAS-",
"P2.REFR-", which suggest some DRAM timings are handled by the CPU
board. And some signals are missing, as I don't see e.g. DSACK[01]-.

Finally, 5V SRAM of interesting size aren't very cheap. One could get
away with level shifters, but they also add costs and potentially
latency.

We've thought quite a bit on how to expand RAM easily/cheaply on Dan's
recreation of the 3/60, and it turned out to be a lot harder than what
we originally expected :-(

Cordially,

-- 
Romain Dolbeau