Trying to fix a DEC PDP-11/34 backplane - diagnosis

Parent Category: Articles Category: Trying to fix a DEC PDP-11/34 backplane Written by Administrator

"In the mid-1970s, DOAs were running at about 25 percent - one system out of four a customer received would be Dead On Arrival [...]
Quite honestly, we were much more sensitive to production targets than quality targets in those days because of explosive growth in customer demands. The natural human instinct is to push that stuff out and get the revenues.
It's also hard to train people properly when you're growing at 35 percent a year and have an enormous backlog of orders to meet. [...]
Customers were so anxious to get their hands on the DEC products that the level of negative feedback wasn't sufficient to make the company change its approach to quality until much later, in the early 1980s."

Tony Tynan, Digital Employee 1974-1991, from "Digital at Work"

Abstract

This article describes some attempts to repair a PDP-11/34 backplane.

The project isn’t finished yet, any help is still welcome!

pdp1134 total

The Problem

During initial start-up of a PDP-11/34a it came up that the NPG signal chain on the DD11-PK backplane was interrupted.

The NPG signal runs over the whole backplane and connects two pins in each slot. The signal is partly guided through tracks and partly through wire wrap.

pdp1134 backplane NPG path

The DD11-PK backplane consists of 6 * 9 flip chip sockets that are soldered onto a solid PCB from one side.
There are three connection layers between the socket pins:

  1. On the connector side, the pins are connected with printed circuit board tracks. They are completely inaccessible.
  2. On the pin side, some connections between pins are also done with printed circuit tracks.
  3. The socket pins are mainly connected by wire-wrap on the pin side.
 pdp1134 backplane total connectorside  pdp1134 backplane total pinside

 

 

An optical inspection showed that all pins of the NPG chain are correctly connected by wire-wrap. However, an electric measurement turned out that the resistance between the NPG pins wasn't nearly zero ohms, but one fluctuating between 0 and 10 kOhm.

The Diagnosis

At first, naturally the wire-wrap connections are the suspects. However, they are (allegedly) more reliable than solder connections, and on verification by measurement they didn’t had problems.

Finally I could find a pin that caused a greatly fluctuating resistance when moving it slightly. A macro photo revealed the problem:

pdp1134 bad pin org

(The magnification is about 10x, a pin’s diameter is 0,7 mm)

You can see that the solder does not coat the wire-wrap pin. The solder also has not formed a meniscus and therefore never was in contact with the pin. So the pin has a loose contact to the soldering joint and therefore no safe connection to the conductor path on the other side of the board.

This is a manufacturing failure. Apparently at production oxidized or contaminated (read: "dirty") pins were soldered in by DEC. Initially, these had probably electrical contact and passed the first tests, but then they failed over the course of time.

Another photo shows pins with different types of bad soldering joints:

pdp1134 good and bad pins

You can see:

  • The pin on the lower right and the one in the middle of the photo are probably not coated at all (anymore).
  • On the upper right of the photo you can see a pin which is only coated on one side.
  • Other soldering joints on the photo are looking good.

Is there still hope?

During work with the PDP11/34 it came up that the machine at first passed the initial tests (EXAM/DEPOSIT via console) as expected. Later on, after some of the cards had been plugged in and out to other sockets, the machine passed the tests less often and then not at all. After years of storage the bad soldering joints now failed at first start-up. Or with other words: The PDP-11 slowly but surely is crumbling away…

A quick estimation showed that many of the square pins – about 20% to 40% - are at least not coated on one side. Critical are those pins that are not coated on any of the four sides. Their count cannot be estimated. A DK11-PK backplane has 1944 pins … - a number that actually forbids any manual checkout operation. If only 1 % of all soldering joints have failed, this already means 19 sources of error!

By the way, the soldering joints of many pins are not visible at all because of the dense maze of wire-wrap:

pdp1134 backplane pins wires

With these symptoms you could really consider the backplane to be a total loss with good conscience and simply throw her away.

Or you take it sporty!