Just to bring some closure to this old thread, I finally picked up a
working TMS9980 cpu chip (after getting one faulty one off of eBay -- it
was even more dead than the one it was replacing). And it appears to work
properly in the 990/189 board, with the 9.3V supplied on the CPU socket.
Thanks,
Josh
On Thu, Nov 17, 2016 at 3:05 AM, Eric Smith <spacewar at gmail.com> wrote:
9.3V might actually work fine for a TMS9980, even
though it's below spec.
It's not going to damage the part, so it may be worth a try before
modifying the board for 12V to the CPU socket.
In NMOS digital parts that predate depletion loads, Vdd needs to be
significantly higher than the most positive logic level in order to bias
the enhancement nFET used for the loads (pull-ups). The Vdd voltage
doesn't have to have a precise value, but it needs to be somewhat more than
the FET gate threshold above the most positive logic level, and below the
breakdown voltage. The higher the Vdd voltage (below breakdown), the
faster the pullup will operate, so running below spec will reduce the
maximum speed at which the part will operate. This is also dependent on
temperature. The part is spec'd for operation over a fairly wide
temperature range (even if only "commercial" rated). Since the logic high
level is no more than 5.0V, and generally somewhat less, a Vdd of 9.3V is
probably more than adequate at room temperature, but may fail at
temperature extremes.
The MP9529 is a "selected" TMS9980. In most contexts, a "selected"
IC is
one that has been tested and found to meet specifications more stringent
than the normal specifications. However, in this case I think the MP9529
might actually be "selected" in the sense of being tested to *lower*
specifications than a standard TMS9980. It's unclear why they would want to
use the lower Vdd, except possibly to reduce power consumption.
With the introduction of depletion loads in later NMOS ICs, generally
starting around 1976, and becoming ubiquitous by 1980, the requirement for
a supply above +5V was eliminated. Similarly, by adding an on-chip
substrate bias generator, the need for an externally supplied substrate
bias voltage (Vbb, typically -5V) was removed.