The Nova’s memory error indicates a flaky signal, but which signal? The error manifests itself on address 40000(oct), bit 7, but only running MEMDIAGLONG.ASM. Running MEMDIAGSHORT.ASM and MEMADDRTEST.ASM on the full 32kW works fine. MEMDIAGLONG runs fine up to the 16k boundary which would imply a faulty memory chip with a pattern sensitive problem, but manually loading memory generates a parity error in any location for bit 7, including the first 16k, which would indicate a faulty bus latch.
(MEMDIAGSHORT.ASM, MEMADDRTEST.ASM and MEMDIAGLONG.ASM are diagnostics by of 11/27/79 by T.P. Povlick that are loaded using the front panel switches.)
The bit 7, address 40000 – 77777 (oct) error is now fixed by replacing the chip at row 2, column 7; IC U146. The 16kx1 memory IC is a MK4116N-4. (The Plessey board is 128kW consisting of 8 rows and 16 columns of 16kx1 chips.) Running MEMDIAGLONG.ASM on the full 32k now works fine. Since I currently have no floppy, disk or tape drive attached to the CPU, I am unable to run diagnostics on the full 128kW (256kB) address range.
I started the memory testing program MEMDIAGLONG at 13:50. The program runs 10 minutes per 1k of memory and was finished at 19:10 with no errors. Time running = 5h20m.