Service Note #4   The Memory Board Problem and the Fix.

Author:  Tom Droege            
Date:    May 23, 2002

The problem with the Memory Board that caused it to sometimes read junk, 
was found to be due to an incorrect reset sequence.  As a result, the 
memory board read register could be reset to something other than 
zero.  Often one off.  This caused a low byte - high byte switch of the 
next word problem.  This happened depending on whether the CMD REG, CABLE 
RST REG, and the CABLE ACTIVE REG happened to end up in the right state on 
turn on.

Their was confusion on how to program the board since Merle engaged in some 
creative digital logic design which does not show up on the drawing.  So 
when I wrote the code, I did not understand Merle's comments on how to 
properly code it.  If you look at drawing "Cable Stuff" you will find a 244 
in the middle of the page.  What is not shown is that the upper bank of 
four drivers is enabled by the Q output of the CMD REG being low.  The 
lower bank of four drivers is enables by the -Q output of CMD REG being 
low.  Thus if the CMD REG is not in the right state, the reset signals sent 
by out &h300 I/O commands do not make it to do their job.

The correct sequence to reset the Memory board address register is thus:

OUT &H300, &H80          'set the CMD REG so that OUTs will send pulses 
                         '   through the lower path
OUT &H300, &H05          'reset the mar and the CABLE RST REG
OUT &H300, &H00          'set the CMD REG so that signals from the cable 
                         '   make it to the logic
P% = 7                   'set up to send a reset MAR through the path 
                         '   down the cable
Call Pulset (       )    'call the QBasic routine that sends the reset 
                         '   MAR signal down the cable by
                         '   sending the appropriate code to the stamp

With this scheme, the Memory MAR is reset twice.  A good thing to do, I 
think since we have not been reseting it often enough for several years.  I 
thing the reset through the Stamp could be left out if one wants to speed 
up the reset slightly.

Note that the Memory should be reset as above before starting a CCD read 
out sequence.  It should also be reset as above before reading out the 
Memory card to disk.