Dealing with Backplanes

Parent Category: Stories Category: My PDP-11/44 Written by Administrator

After I got the 11/44 and my RL02 running, the CPU backplane was full. The KD11-Z backplane has only one hex and one quad SPC slot empty, but I planned to use an SMD disc, and SDI disc with UDA50 controller and DEUNA network.

So I had to get the 2nd backplane in my 11/44 connected to the first one. This was a 9-slot DD1-DK standard backplane. I had to build a system, which spans over two backplanes.

Mounting Backplanes

DECs packing philosophy is very modular: backplanes can be added to mounting boxes, as long as there are free power supply resources and mounting space. Backplanes are either of generic type, with a max of 9 quad and hex SPC (Small Peripheral Connector) slots, or they come prewired for older device controllers, which consist of many logic cards, or a mix of both.

As example see an empty 4-slot backplane and a DV11 Communication Multiplexer, occupying a 9-slot backplane:



But all backplanes can be connected together with standard UNIBUS cables (DEC name is “BC11”). Since these can be severals meters in length, connected backplanes can reside in different mounting boxes. (According to Gordon Bell, case and module packing costs make the half of a computers total costs, so let’s value DEC’s box+backplane approach as sophisticated engineering!)


Using the additional backplane was not very intuitive the first time. While the backplanes slots appear as homogenous array, they have a sophisticated  internal structure.



At first I needed a UNIBUS connector to connect both backplanes. I had no M9202, but four 10-feet cables, and needed a short 2-5 inch cable. One of my long cables was damaged, so I cut off one connector and soldered it to the other one in a short distance. It was no fun: the flex print plastic shields melted immediately as I touched them with the solder iron, and the cable wires loosed their alignment. Finally it looked ugly, but did its job.



Never mount a terminator in a “Modified Unibus Slot”!

I connected both backplanes with my jumper cable on row A (right), and powered on: “?22 CPU hung”. Oh well, I missed the UNIBUS terminator. I eagerly plugged the M9302 terminator next to the jumper ... power failure!

On the M9302 is a big red warning:



I ignored this because I have a 11/44 (and not an 11/03 or ‘34) but began to suspect that they were talking to me. Indeed: The only free Standard Unibus slot you can plug the terminator into is the last slot in row A/B, all others are ... Modified Unibus Slots. Well, just my pride was hurted.


I put the M9302 in the last slot, powered on: CPU did not respond. In DECs documentation. and in news groups I often read about the interrupt handshake signals: BR, BG, NPR, NPG and GRANT and ACK and such things. I found that empty backplane slots (and I had 8 of them!) have to be bridged: At first you have to plug G727 GRANT CONTINUITY cards in row D of every empty slot, something I could accept. But I had not enough G727 cards (I bought a rather filled up machine!), so the usual search+buy+pay+wait cycle began once more.



Finally my backplane was full of G727, I powered on ... “?22 CPU hung” again. Aaargh ... but wait: G727 closes the Bus REQUEST GRANT, but there is also this NON PROCESSOR GRANT signal.


Documentation said: I had to close the “NPG jumper” on those backplane slots, which do not contain a card with active data transmission (DMA). The four cards I wanted to plug in were all DMA capable (disc controllers and network adapter). By todays standards “Closing a jumper” means you would expect a printed circuit board with a little 2 pin header, which has to be short cut with a little rectangular jumper plug. In the 1970’s this means, you have to get your wire-wrap tool (you have one, don’t you?) and wrap a short piece of wire between two wrap posts. All this is done on the backside (Pin-Side) of the backplane, so I turned my 11/44 upsize down and removed the bottom cover.



Grid computing

What a sight: An unstructured monstruous array of 9 slots * 6 rows * 2 sides * 18 contacts = 1944 pins! DECs four-dimensional pin-coordinates-nomenclatur (as in  “Ground is on pin A 12 T 1”) should have warned me.



To check the NPG jumper I had to find row C, Pins A1 and B1 in each slot. I made a survey map and pin rulers in 1:1 scale, and began counting. I found that 4 rows were without NPG jumpers, which was luck because I planned to put just four cards into the backplane: I could use the slots where NPG already was open. So no wire wrapping was needed (I would have done it without wire-wrap tool): After I plugged my four cards into the four NPG-open slots, the CPU started again!

NPG jumper

Later I learned how to make an NPG jumper with modern tools. You can plug a standard female pin header onto the wire wrap pins, the distance between two diagonally arranged wire-wrap-pins is just  2/10 inch = 200 mils (see a DIL-EEPROM as reference).




What I learned:

  • First read, then think!
Download this file (unibus backplane dd11-dk.png)unibus backplane dd11-dk.png[Pin arrangement on an UNIBUS DD11-DK backplane, backside]265 kB