RAMTESTing a Horizon

From: Robert Maxwell <RMaxwell_at_atlantissi.com>
Date: Fri Feb 27 13:14:57 2004

  Thanks to all those who provided insight in the "Smoke on the Horizon"
thread... You may be happy to hear that the machine is running now: all the
blown tantalums are replaced, as well as the regulators that may have
suffered due to untimely capacitive departures. I took warnings about the
-5V supplies to heart and replaced all the tantala (plural of tantalum?)
around the negative supplies, largely because the -5V regulators are powered
from the -16V unregulated line, and the worst failures happened on caps
across the +16V line.

  Now that the smoke has cleared, I've been RAMTESTing the 64K HRAM card,
and am a little distressed to note failures. These usually occur when the
program is run from fresh-from-startup RAM: once RAMTEST (3 or 5) has run
through a cycle (or bombed due to an error), subsequent tests run for hours
with no further problems. The failures don't seem to be bit-specific, but
they've occurred mostly in the highest 16K (particularly the top 8K) of
address space.

  Do I need to be concerned about this? Anyone know if the RAMTEST program
trips up on previously-unaccessed space? Or do the RAMs need a little
time/exercise to get to an internal temperature where they are stabler?
These are 16Kx1 AMD devices (circa 1982), and my debug work in the past has
been limited to SRAMs or higher-density DRAMs (I know, I know: I just
haven't lived until I've debugged core).

Regards,
Bob
Received on Fri Feb 27 2004 - 13:14:57 GMT