PDP-11/84, PMI and Q-bus

[Johnny Billquist]

Okay, since this topic have become a subject of much discussion and some 
diverse opinions, I decided to really read the manuals to try to find 
the bottom of this all.

My main source of information here have been the "KDJ11-B CPU Module 
Users's Guide". EK-KDJ1B-UG-001

Now, to start with the position of the CPU card in an 11/84.

Chapter 7, section 7.2. Page 7-1:
"7.2 PMI INTERFACE
The PMI interface signals are defined as the PMI bus master signals, the 
PMI slave signals and the PMI Unibus adapter signals. These interface 
signals are assigned to the C and D rows of the backplane and are 
defined as the interconnect bus. The PMI interface signals on the C/D 
bus are normally assigned two pins to provide an interconnection between 
the slots. The KDJ11-B module is only assigned one pin and therefore its 
position in the backplane is critical. The LSI bus signals that are used 
with the PMI protocol use the A and B rows of the backplane defined as 
the LSI bus."

Now, if Pete is correct in that the PMI bus on the 11/84 really goes to 
both pins on all slots, then it should be okay to place the CPU in any 
slot. I haven't tried that, but I might when I have the time. I suspect 
he's right since otherwise I would have expected the CPU to be in slot 
3. But DEC could be doing some fancy wiring... :-)

The fact that the placement is critical is obvious if we talk about 
Q-bus systems, and I suspect the comment is written from that point of view.



As for wether Q-bus memory (or any other Q-bus peripherial) will work, 
I'll quote some signal descriptions.

Chapter 7, page 7-4. Table 7-3 PMI Unibus Adapter Signals
"Pin: CF1   Mnemonic: PUBSYS L   PMI Unibus System

In a Unibus system, PUBSYS L is asserted by the UBA to direct the 
KDJ11-B to follow PMI protocol for all data transfers, wether the PSSEL 
L is asserted or not. LSI-11 bus protocol is disabled for all PMI 
devices when PUBSYS L is asserted.

In an LSI-11 system, PUBSYS L is always negated. If PSSEL L is negated, 
the KDJ11-B follows LSI-11 protocol and the PMI memory then responds to 
the LSI-11 protocol by the LSI DMA devices."

Chapter 7, page 7-5. Table 7-4 LSI Bus Signals

"Pin: AF2   Mnemonic: BRPLY L    Reply

During PMI cycles, BRPLY L is asserted by the KDJ11-B and the PMI slave 
to prevent the next bus master from gaining control of the bus too soon. 
In a Unibus system, BRPLY L is asserted by the UBA as a slave response 
during the PMI DATOB cycle and interrupt DATI cycle.

    *** NOTE ***
The PMI memory slave modules in a Unibus system must have BRPLY L 
disabled at all times.

Pin: AH2   Mnemonic: BDIN L     Data Input

The BDIN L signal is only used in PMI Unibus systems during interrupt 
grant cycles. The KDJ11-B asserts BDIN L after it gates the interrupt 
priority, BDAL bits <3:0>, onto the bus. The UBA then latches the 
interrupt priority data using the leading edge of BDIN L.

Pin: AM2   Mnemonic: BIAKI L    Interrupt Acknowledge In
Pin: AN2   Mnemonic: BIAKO L    Interrupt Acknowledge Out

These signals are only used in PMI Unibus systems during the interrupt 
grant cycles. The KDJ11-B asserts the BIAKI L signal, and the UBDA 
acknowledges it by asserting one of the Unibus bus grant signals.

Pin: BB1   Mnemonic: BPOK H     Power OK

This signal is only used in PMI Unibus systems for the Unibus 
power-up/power-down protocol. This signal is asserted and negated by the 
UBA in response to the Unibus AC LO signal. The assertion of AC LO may 
be prolonged by the Unibus devices or the PMI memory during power-up."

I could go on describing more details on how these signals are used, 
since it's all described in the manual. Suffice to day that the KDJ11-B 
*knows* when it's in a Unibus system (i.e 11/84) and will not behave 
like a normal Qbus. First of all, it will always use PMI protocol to 
access memory, no matter what the memory might claim. Second, interrupt 
handling is done differently, since all interrupts are expected to be 
dealt with via the UBA.

There are also schematics for different signals in the manual, and you 
can clearly find the PUBSYS L signal included in some decoder path as a 
way to force a specific behaviour independent of other signals that 
might exist.

Now, can we now accept that it's not a Q-bus in the 11/84? :-)

	Johnny

-- 
Johnny Billquist                  || "I'm on a bus
                                   ||  on a psychedelic trip
email: bqt at softjar.se             ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol