Apur-ee.105
net.general,NET.general,net.periphs
utzoo!duke!cincy!pur-ee!mike
Fri Jun 19 16:40:15 1981
DMC hardware fix
The DEC dmc11 interprocessor link board M8200 has some hardware bugs which
showed up in the process of bringing up our local network on our first vax
11780.  The problem manifests itself as a bushang condition when running a
DMC with other DMA devices transfering further down the bus from it (further
from the UBA or processor).  The problem has three seperate causes - 1. a
race condition in the NPG recognition circuit, 2. an RC delay network in the
NPG circuit which cannot handle back-to-back NPGs on the bus, and 3. an NPR
request flag controlled by the micromachine internal to the DMC which does
not drop fast enough after receiving NPG (as compared to the usual random
logic arbitration circuit).
       To cure these problems, refer to page D15, sheet 16 of 22 of the
engineering drawings for the DMC11 (Dec print set order # MP00076).  You
will find a resistor, R31, a capacitor, C110, and a diode, D4 on the left
hand side of the page.  Remove these components.  In their place insert a
7404 inverter chip in the spare chip location E130 in the upper left hand
corner of the board.  Run a wire from pin 1 of this 7404 to the former
input side of the RC delay circuit you just remover (the lefthand hole the
resistor was in).  Run another wire from the righthand resistor hole to pin
8 of the 7404 you just inserted.  Connect pins 2 to 3, 4 to 5, 6 to 13, and
12 to 11, and 10 to 9.  Connect pin 14 of the 7404 to +5volts and pin 7 to
ground (use the leads of bypass cap E100 as connection points).
       Remove the 7408 in location E105 and replace it with a 74s08.
Remove the 7400 in location E106 and replace it with a 74s00.  This
completes the modification.  We have had one unit out of 14 where this
failed to cure the problem, or the timing is rather critical and apparently
varies enough with different chips elsewhere in the circuit to cause
problems on a few boards.  We simply left that board on an 11/45 where the
faster bus ensures that (statistically speaking) it works reliably.
       The only solution to this problem without hacking at the hardware is
to put the DMC at the end of the bus, but this makes it the lowest priority
dma device, which is not good for a dmc, and limits you to one dmc per UBA.

                                       Mike Marsh
                                       pur-ee!mike

-----------------------------------------------------------------
gopher://quux.org/ conversion by John Goerzen <[email protected]>
of http://communication.ucsd.edu/A-News/


This Usenet Oldnews Archive
article may be copied and distributed freely, provided:

1. There is no money collected for the text(s) of the articles.

2. The following notice remains appended to each copy:

The Usenet Oldnews Archive: Compilation Copyright (C) 1981, 1996
Bruce Jones, Henry Spencer, David Wiseman.