Convert 720A to 720C

The Wang 720C is the final iteration of the 700 series and had the capability to interface with the most sophisticated peripherals including memory expansion and disk storage. Most of the extended capability was in the microcode, but some hardware changes were required.

Three of the eleven logic cards have revised versions for the 720C:

5930 card supports four extra restart addresses for “Run-Learn” mode
5975 card has revised Step mode logic
5976 card performs extra IOB decoding for advanced peripherals

Jan van de Veen of the Small Wang Museum very kindly assisted in obtaining these three cards so that the 720A could be converted to a 720C, allowing the fullest set of microcode and software to be run.

Card Installation and Backplane Wiring.

The updated cards have two new backplane signals, these are routed with edge connector pads that are unused in the -A and -B models so cards and backplanes are all compatible, regardless of combination. Therefore, the 720A backplane could be modified to support the 720C cards, knowing that the 720A cards would still work.

Two wires must be added to the backplane:

one connection between the 5930 and 5975 cards, using otherwise unused pins
one connection from keyboard to 5930, for Run-Learn mode.

Keyboard

The keyboard installed in the 720A was found to have wiring to support Run-Learn mode, and was therefore intended for use in a -C machine despite being installed in an -A chassis. This suggested that the keyboard may have been replaced at some stage, although the service sticker inside the cover only recorded a tape head adjustment in March 1971. No date stamp could be found on the keyboard, this would have been an interesting comparison with the 1970 date on the chassis.

Therefore, the keyboard did not require any changes were required, but a wire had to be added to the keyboard connector harness to bring the Run-Learn signal to the backplane and the 5930 card.

Startup and Troubleshooting

Setting RUN mode and PRIMEing did not bring up a working calculator. A few nixies showed faint glow, this is a very general and non-specific symptom meaning only that microcode is not running correctly, but little more than that. Since the 720 was previously running OK with 720A cards, the fault must have been introduced by one of the three 720C cards.

Trying the other hardware restart modes (Verify, SetPC, Record) was very useful because SetPC gave a stable and reliable display of all zeros in the X and Y registers. The display is entirely microcode driven and each nixie tube must be addressed in correct sequence for the correct period if a stable display is to be produced. The stable all-zeros display showed that the logic engine could support fairly complex microcode, including

reading and writing RAM
incrementing counters
branching on counter overflow
maintaining pointers to digit data in RAM

This narrowed the problem area to the 5930 card which deals with microcode ROM addressing, including conditional branches according to machine state and flags. Perhaps some conditionals were working correctly support the all-zeros display but a less commonly-used flag or condition was failing, causing more complex microcode to fail.

Testing Conditionals with 2-word ROM

Diagnosis with 2-word ROM (opens in new tab)

Testing microinstructions (opens in new tab)

The 2-word ROM is designed to assist with diagnosing specific microinstruction failures. A test microcode word can be set up on the switches and executed in a 1- or 2-instruction loop while relevant logic signals are probed to test if the correct signal sequence is seen.

In this case the ROM was set up to execute each of the conditional address branches and the decoding of each instruction was monitored and tested. 1970s era equipment was very adequate for this task: an HP545A Logic Probe to follow signals and a 546A Logic Pulser to inject test pulses. This rapidly showed that microcode branches for

key-pressed, ALU carry, ALU zero, ALU overflow, S-flags 1 and 3 were correctly decoded
saved-carry, Q-flag, S-flags 0 and 2 were not decoded or executed

The four defective instructions each used one gate of an SN7401 quad-NAND, which was found to have failed such that it was unable to drive any of its outputs to an active-low state. Failure of this one chip was full explanation of the failure of four microinstructions.

The SN7401 was removed, confirmed as defective and replaced with a NOS chip.

Successful Conversion to 720C

The 720 now came to life as a 720C and passed the Wang RAM test over 200 registers.

It’s possible that other faults could lurk on the two upgrade cards that are concerned with IO functions, this will be tested as peripherals or peripheral emulators are brought into use in future.