perm filename ECO.LOG[KL,SYS] blob sn#286316
filedate 1977-06-02 generic text, type C, neo UTF8
COMMENT ⊗ VALID 00002 PAGES
C REC PAGE DESCRIPTION
C00002 00002 1) Modified PAG board 4AF30. Added signal CON KI10 PAGING MODE L
1) Modified PAG board 4AF30. Added signal CON KI10 PAGING MODE L
to pin E43(7) on PAG3 page. This fixes the bug whereby every
page fault looked like "no PT DIR match" in KL paging mode.
JBR 5/10/76 0300
[This is most of REV 7B. I did not move the terminator resistor
as was indicated in the prints and I don't think it matters much.]
[I think this one is rev 9 or 9A]
2) Modified CON board 4AF35. Deleted CON MBOX WAIT H from pin FN1.
Deleted wire from E62(14) to E57(9), but left terminator on E57(9).
Added wire from E62(15) to E57(10) CON WRITE FM PAR L
from pin FN1 to E57(9) -CLK SBR CALL L
and from E53(4) to E57(11) -CON CLK B L.
Modified CTL board 4AF36. Deleted terminator on -CLK SBR CALL L.
Added twisted pair from 4A36F1 to 4F35N1 -CLK SBR CALL L.
This fixes the spurious FM parity error problem.
JBR 6/5/76 1900
Rev 10 [I think]
3) SCD TRAP REQ n is set by CON COND INSTR ABORT and SCD TRAP CYC n.
This is changed to CON COND INSTR ABORT and (SCD TRAP CYC n or SCD
TRAP REQ 2). If TRAP REQ n gets set but then a page fault happens on
the instruction fetch (before the NICOND dispatch can set TRAP CYC)
the TRAP REQ is lost.
Rev 10A [I think]
4) PHYS REF causes VMAX SEL 1 and VMAX SEL 2 to force the entire VMA to
be loaded. On MCL4 (center) add an and gate with inputs from MCL SPEC
MEM CYCLE and magic numbers bit 8 with output going to E56 pin 7.
When this mod is installed, the microcode has to be changed so that
the PHYS REF appears in the same microinstruction as VMA←AD.
5) Swapped resistors R7 and R8 on M8553 board (print CNT8 in the DTE20.
This is some more of Rev 7B. These resistors generated +3A and were
putting out about +1.6V. Apparently DEC noticed this a long time ago
and fixed it in the prints but not in their production line. So they
put out correct prints but incorrect boards.
6) From looking at the prints, here are all the differences I could find:
1) On page CSH2, the signal CACHE TO MB DONE L has been deleted from pin
5 of the 10109 in E74 at A3.
2) On page CSH5, the PAGE REFILL time chain has been changed from T9 up.
3) On page CSH7, the EBUS multiplexors read back different PAGE REFILL
time chain signal names.
4) On page CLK1 the signal DIAG CHANNEL CLK H (connector pin DU2) has been
added to pin 9 of the 10210 in E46 at B2.
5) On page CLK4 the signal CON MBOX WAIT on pin 14 of the 10121 in E31 at
B6 has been replaced with the output of an inverter whose input is
CRAM MEM 02 A H.
6) On page CLK4 the signal CLK PAGE ERROR H is generated and driven on
connector pin AN1 by a 10105 in E42 at C2. The signal is the or of
CLK PAGE FAIL EN H and CLK INSTR 1777 H. CLK INSTR 1777 H used to
drive connector pin AN1. This signal went to the CON board only and
the prints now use CLK PAGE ERROR H where CLK INSTR 1777 H was used.
7) On page CLK5, CLK INSTR 1777 H is now read back as diag read 100 bit 32
and CLK PAGE ERROR H is read back as diag read 103 bit 34.
8) On page MBX1, CCA HOLD ADR H is now the output of a flip flop whose input
comes from what used to be CCA HOLD ADR H. Also, CCA HOLD ADR FF L has
been renamed CCA HOLD ADR L.
9) On page MBX2, the signal MEM WR RQ L on pin 5 of the 10105 in E62 at A7
has been replaced by -MEM RD RQ L on connector pin EM2. MEM WR RQ L
with connector pin DR2 is still terminated.
10) On page PIC2 the signal -PI4 MR RESET L has been added to pin 4 of the
10109 in E37 at D7.
11) On PIC2 a buffered version of PI4 MR RESET L is wire ored with PI2 TIM1 H.
12) On PIC4 the signal name PI4 CONO DLY L has been attached to pi 2 of the
10131 in E83 at D8.
13) On PIC5 the signal -PI4 SYS CLR H has been replaced with -PI4 CONO DLY H
at pin 7 of the 10104 in E19 at D7 and at pin 7 of te 10104 in E20 at C5.
The explanation for this is as follows: (see KL.BUG item 8 for a description
of the bug that this fixes.)
Before Rev 8 it was possible to be executing microcode that would turn
off a PI channel that was trying to interrupt at the same time. The
PI5 TEST/PI5 EBUS REQ synchronizer would be latched up while
the EBOX had control of the EBUS and when it let go, PI5 EBUS PI GRANT
would be set and an interrupt would be started for that channel. Now,
however, if the EBOX does a CONO PI while it has the EBUS grabbed,
PI4 CONO DLY forces the synchronizer to reset causing it to first go
through the PI5 LOAD state and look at the new values of the PI requests
which may have changed as a result of the CONO PI.
14) On MBZ1 the signals CORE BUSY H and L are flip flops that are loaded from
where they used to be generated at D6. The signal on pin 6 of the 10118
in E48 at B8 has been renamed from -MEM START C L to -MBZ4 CORE BUSY A L.
15) On MBZ3 the signal -MEM START C L on pin 10 of the 10117 in E57 at B7
has been replaced with CORE BUSY L. The signal MR RESET A H has been
added to pin 12 of the 10109 in E60 at C2.
16) On MBZ4 the signal CORE BUSY A L has been renamed MBZ4 CORE BUSY A L
and the pin 11 output of the 10101 in E8 at A5 has been named
MBZ4 CORE BUSY A H.
17) On MBZ6 the signal on pin 10 of the 10117 in E24 at A3 has been renamed
from CORE BUSY A L to MBZ4 CORE BUSY A L.
18) All MTR EBUS signals have had 18 added to them. On MTR2 HI has been
replaced by -MTR RESET PLSD H at B7. The signal MTTR EBOX CNT EN L
is generated differently. See D6-D4.
19) On MTR3 the signal MTR RESET A L is connected to pin 3 of the 10101
in E11 at A5 and is terminated.
20) On MTR4 HI has been replaced with -MTR RESET A H on pin 12 of the 10113
in E46 at A2. MTR EBOX WAITING is produced differently. See A7-A5.
21) On MTR5 diag read 117 bit 25 reads back -MTR CONO MTR, H.
7) The meter board. Deleted flip flop in series with EBOX counter enabling
signal, added -CLK MB XFER into and gate producing MTR EBOX WAITING.
JBR 5/31/77 1700