Editing PDP-11/70 Troubleshooting Guide

<templatestyles src="Template:StyledTable/styles.css" />
[[File:DEC PDP-11-70 (Living Computer Museum).jpg|thumb|right|220px|PDP-11/70 at the Living Computer Museum (Joe Mabel, CC BY-SA 3.0)]]
Fault-finding on the '''[[PDP-11/70]]''', from chapter 5 of DEC's ''PDP-11/70 Maintenance and Installation Manual'' (EK-11070-MM-002) and DEC Canada's ''1170 Maintenance Service Guide'' (KA-K1170-MG-003).<ref name="mm5" /><ref name="msg" /> Start with the power system: [[PDP-11/70 Power Supply Service]] gives every regulator's test point and limits.

== DEC's approach ==
DEC isolates a fault to a subsystem (CPU, cache, memory management, Unibus map, a Unibus device, main memory, a Massbus controller or device) with the on-line diagnostics, DEC/X11 or the 11/70 subsystem diagnostic, then uses the stand-alone diagnostic for that subsystem to find the failing module. It discourages component-level repair of 11/70 modules on site except where the fault is obvious and quick to fix.<ref name="mm5" /> For a machine that cannot load diagnostics, the M9301-YC, M9301-YH or M9312 bootstrap contains a short diagnostic of the instructions used by the loader, the first 28K of memory, memory management and the Unibus map, and half or all of the cache can be forced to miss.<ref name="mm5" />

When the console works but booting fails:<ref name="mm512" />
* The bootstrap halts inside its diagnostic: look up the halt address in the ROM listing to find the failing instruction.
* The machine loops: halt it and examine the PC. If the PC is meaningless, load each trap vector with its own address plus 2 and the following word with 0, and boot again; the machine then halts at vector plus 4, showing which trap occurred (004 CPU errors, 010 illegal instructions, 024 power fail, 114 parity, 244 floating point, 250 memory management, and the Unibus device vectors).
* It halts with the console PARITY indicator on: examine the cache error register and the other cache registers.
* Memory does not work at all: use toggle-in routines held in the memory management PARs instead of memory.

== Diagnostics ==
Each 11/70 diagnostic assumes the earlier ones have passed; run them in this order:<ref name="msgd" />
{| class="wikitable styled-table" style="width:70%; text-align:center;"
|+ '''Diagnostic order (KA-K1170-MG-003, pp. 3 to 4)'''
! Program !! Tests
|-
| EKBA, EKBB || CPU parts 1 and 2
|-
| EKBC, EKBD || Cache parts 1 and 2
|-
| EKBE || Memory management
|-
| EKBF || Unibus map
|-
| EKBG || Power fail
|-
| EQKC || 11/70 instruction exerciser
|-
| EMJA, EMKA || MJ11 and MK11/MJ11 memory
|-
| ERHA || RH70
|-
| EFPA, EFPB || FP11-C parts 1 and 2
|}
<ref name="msgd" />
DEC Canada warns that EKBB and EKBD also write on any write-enabled disc drive; spin down or write-protect every customer and diagnostic pack, or mount a scratch pack if the feature is wanted. EKBB reports "NO BR5 DEVICES" when no RH70 is fitted; that is not an error.<ref name="msgd" />

== Working without cache ==
Many cache faults stop the diagnostics loading, because the cache is part of the hard core. Deposit 14 (octal) in the cache control register at 17 777 746 to force misses in both groups, and start with CONTINUE, not START, because START and power-up clear the register. This disables the address memory (M8143) and data memory (M8144) boards but not the cache control (M8142) or data paths (M8145).<ref name="mm513" /> Examining the cache error address register (17 777 740 and 17 777 742) twice should show its own address on the second examine, a quick test that the address multiplexer and paths can pass ones.<ref name="mm513" /> DEC's manual gives short test loops to toggle into the PARs for these cases.<ref name="mm513" />

== Quick checks from the service guide ==
=== Data path check ===
Deposit all ones and then all zeros in location 0 and examine them; deposit 070707 in location 0, load address 17000000 and examine. Reading 070707 through the Unibus map shows the data and address paths are good; repeat in reverse.<ref name="msg145" />

=== CPU voltage margins ===
All CPU and RH70 logic should run without errors with the +5 V regulators anywhere from 4.75 to 5.25 V. Set all the CPU (and used RH70) +5 V regulators to 4.75 V and run EQKC for half an hour, then DEC/X11 for half an hour; then repeat at 5.25 V. If errors appear, move the regulators back one at a time to find the module concerned.<ref name="msg145" />

=== Timing margins ===
With a jumper from F13J1 to E21T1, set R162 on the M8139 timing board for 250 ns at D13T2 and run EQKC, then 140 ns and run it again, then return to 150 ns and remove the jumper.<ref name="msgpm" />

== Known faults from DEC's tech tips ==
{| class="wikitable styled-table" style="width:100%; text-align:left;"
|+ '''Selected field changes and tips (KA-K1170-MG-003, pp. 20 to 23)'''
! Symptom !! Cause and fix
|-
| Intermittent data parity errors || Green bypass capacitors on M7984 modules burn up; replace M7984s with burnt capacitors
|-
| Transient data parity errors in other core stacks on the same side || National Semiconductor NS7520 sense amplifiers on G116 modules generate spikes; replace G116s fitted with them
|-
| Unreliable connections on over-the-top cables || H224 cables damaged by inserting and removing neighbouring modules; replace or re-crimp
|-
| SACK timeouts, trap 0s and Unibus timeouts || 74123s from some vendors in the M8136 timeout logic not clearing properly; replace E15 and E25 with Texas Instruments 74123s (DEC 19-10436)
|-
| Poor power distribution || Replace power harness 7009540 with 7011051 (FCO PDP1170-R1)
|-
| H775-D breaker trips during power fail (MK11) || M8160 revision update (FCO M8160-R4)
|}
<ref name="msg20" />

== Power faults ==
If the machine is dead, check in order: the 861 indicators and breakers, the H7420 power lamps and CB1, every regulator lamp, the two LEDs on each 5411086 (both lit when AC LO and DC LO are not asserted), and the voltages at the backplane. One regulator lamp out points to that regulator; a whole group out points to its H7420 or the 861.<ref name="mm433" /> Details are on [[PDP-11/70 Power Supply Service]].

== Related pages ==
* [[PDP-11/70]]
* [[PDP-11/70 Power Supply Service]]
* [[PDP-11/70 General Maintenance]]
* [[PDP-11/70 Documentation]]

== References ==
<references>
<ref name="mm5">Digital Equipment Corporation, ''PDP-11/70 Maintenance and Installation Manual'', EK-11070-MM-002, May 1979, pp. 5-1 to 5-2 ([[:File:PDP-11-70 Maintenance and Installation Manual (EK-11070-MM-002).pdf]]).</ref>
<ref name="mm512">DEC, EK-11070-MM-002, pp. 5-4 to 5-6, sections 5.1.1 and 5.1.2 and Figure 5-1.</ref>
<ref name="mm513">DEC, EK-11070-MM-002, pp. 5-6 to 5-8, section 5.1.3, toggle-in routines.</ref>
<ref name="mm433">DEC, EK-11070-MM-002, pp. 4-104 and 4-106, H7420 fault isolation.</ref>
<ref name="msg">Digital Equipment of Canada, ''1170 Maintenance Service Guide'', KA-K1170-MG-003, April 1988 ([[:File:PDP-11-70 Maintenance Service Guide (KA-K1170-MG-003).pdf]]).</ref>
<ref name="msgd">DEC Canada, KA-K1170-MG-003, pp. 3 to 4, diagnostic order and set-up notes.</ref>
<ref name="msg20">DEC Canada, KA-K1170-MG-003, pp. 20 to 23, field change orders, tech tips and "should've been FCOs".</ref>
<ref name="msgpm">DEC Canada, KA-K1170-MG-003, pp. 126 to 130, annual PM, step 6.</ref>
<ref name="msg145">DEC Canada, KA-K1170-MG-003, p. 145, CPU voltage margins and data path check.</ref>
</references>

[[Category:Digital Equipment Corporation]]
[[Category:Troubleshooting Guides]]