PCB ‘Get Well’ process

Inspecting the PCB’s revealed several transistors missing on some cards, but apart from the usual dust and grime they appeared to be in reasonable condition.  

Rather than simply replace the missing transistors and start troubleshooting, I decided to perform a ‘get well’ process on all the PCB’s.  This consisted of using a Digital Volt Meter (DVM) to test all of the diodes and transistors on each board. I discovered that most cards had diodes that need replacing, often several diodes, and the occasional bad transistor. 

After completing the ‘get well’ process, I powered up the machine.  The lights came on, but it would not respond to the font panel switches.  I started trouble shooting the system, but soon decided that a second ‘get well’ process was required as many components had failed as soon as power was applied.

“Get well” round two saw more diodes and resistors replaced than round one.  At the end of this stage I had replaced more than 2,000 diodes and several hundred transistors.  Of course it still wouldn’t respond to the front panel switches under power!

Troubleshooting again… I didn’t have an extender card, but DEC had very considerately brought most of the signals out to the backplane, so it was reasonably easy to find bad components on most cards, with the exception of the Accumulators and Memory buffer cards. These double height cards, particularly the Accumulator cards, are complex and perform multiple functions depending upon the instruction being executed.  

What I discovered during this process is that some batches of diodes used at manufacturing time were considerably worse than others.  I was faced with the choice of laboriously troubleshooting each of these diode problems one at a time, or simply replacing all the ‘bad batch’ diodes.

To be clear these diodes were all fine at the time of manufacture, it’s just that some batches performed better fifty years on than others.

Replacing all the diodes on cards that had demonstrated several diode failures considerably helped improve the machines status. It is a somewhat ‘brute force’ approach to troubleshooting, but the alternative was a prolonged ‘death by a thousand diodes’, and so out they went.

From this point on normal troubleshooting techniques yielded good results, with most solid faults able to be resolved within an hour or so.  There were many solid faults.

I eventually got the machine to a state where it would deposit data into memory, albeit reading the memory location gave variable results.  It was time to set up the memory.