Self testing PETs
by Leonard Tramiel
One of the events in the year or so I spent working on the PET that sticks out in my mind was testing the first batch of boards. It felt surreal, even at the time, to take a board out of a box, plug in a power supply, monitor, and keyboard and have a computer that was unavailable only a short time before.
The first batch of PET boards used an ... unusual RAM chip made by MOS Technology. Besides being an design that was a hybrid of static and dynamic RAMs, with none of the advantages of either. the chips had a high percentage of bad devices. The factory testing didn't find bad chips but ones that did work were pretty reliable.
When the first batch of boards came in from the manufacturer, most of the machines didn't work correctly. One of the features of Microsoft BASIC is a memory test that does a simple check of memory and reports the number of bytes that were found to be available. In these 8K PETS that number should have been 7167. Most of the machines would come up but display the wrong number. We wrote a program that read the number printed on the screen, computed what memory location failed, tested that address to determine which bit failed and indicated which chip was bad. So we would plug in a tape drive, load up the program, and the machine would diagnose itself and display the chip that needed to be replaced. Not bad for 1977.