Intermittent Keyboard Lockup
The machine was up and running 2K microcode, but it was soon clear that the keyboard was not right. Most keys seemed to work OK but occasionally the machine would lockup after a key was entered and appeared on the display. No further keys would be recognised and PRIME would be necessary to recover.
The fault appeared to be intermittent, which always makes troubleshooting more difficult. With more observation it was clear that while most keys did not provoke this problem, the Zero key would almost always do so - it would be registered and then cause a lockup. While this made the calculator pretty useless, it also provided a specific hook to use for troubleshooting.
Capturing Keypresses
Keypresses are captured and managed by a pretty complex piece of logic on the 6184 board. Much of the complexity arises from Wang’s decision to interface a very wide range of peripherals through the keyboard interface. This created a need to allow many devices to place data into the keyboard registers and also required that the keyboard could be disabled so that keystrokes did not interfere with peripheral data transfers. The 6184 board was instrumented with the Logic Analyzer of the MSO to look for the cause of the lockups.
These complicated captures shows the correct capture of keystrokes ‘7’ and ‘1’. The yellow trace shows the voltage on line Z2 as the key is pressed - this voltage goes from a small negative level to a positive level when the keyboard is disabled. The actual positive level depends upon how many signal lines are connected to Z2 by the encoding matrix in the keyboard. In the case of key ‘7’ the voltage is 1.37v, while in the case of key ‘1’it is 1.05v.
In both cases this voltage develops the LOCK signal (light blue trace) by discharging a delay capacitor (pink trace). It turns out that the level of the yellow signal will alter the rate of discharge of the delay capacitor, in turn altering the time at which the light blue LOCK signal crosses threshold and changes the state of the logical LOCK signal shown on trace D2.
Key zero turns out to be unique in that it has only one connection in the encoding matrix, all other keys have two or more. This in turn means that the yellow Z2 voltage reaches only 0.98v, the delay discharge is slowest and the LOCK signals do not trigger until too late, confusing the key processing logic and causing a lockup.
The deeper cause appeared to be that transistor Q3 on the 6184 board had lost gain and was not conducting sufficiently when driven by the signal from the zero key.
The ideal fix would be to remove the pair of Q2 and Q3 for characterisation on the curve tracer. This would hopefully show a relevant difference and also allow a replacement to vbe chosen since these transistors only have Wang part numbers with no cross-reference available.
No curve tracer is currently available so this fix was deferred.
Repair
An interim fix to avoid removing many components was to short out one of the string of five diodes that reduced the drive to Q3.
This gave a satisfactory result and the two cases at each end of the timing range were shown to work correctly. The drive is perhaps a little too great, with the timing intervals a bit shorter than necessary, but there appears to be an adequate margin and so this will be left until a curve tracer is available to help design a beter fix.
Key capture fixed after identifying a subtle analog-digital problem