The End of ‘PLAGUE’   &   a NEW Board.

The Final Definitive Memory Plague Note

In the last issue we promised that we would collect together all thoughts about ‘Memory Plague’, and publish them. We are indebted to many members of the IMNC for writing to us detailing various symptoms and possible causes. For more recent members of the INMC let us first define ‘Memory Plague’: this was a euphemism coined to cover the small percentage of Nascom Series 1 memory cards (that’s the one with the four EPROM sockets) that proved to be unreliable through causes not otherwise due to faulty or slow chips. ‘Plague’ soon became apparent after the release of the Series 1 card, and the symptoms are unreliabilty when working machine code programs in the expansion memory. ‘Plague’ is not usually revealed by Tiny Basic or the memory test programs as these represent data stored in memory and not op-code. ‘Plague’ usually affects the op-code fetch (M1) cycle as the timing is more critical, and results in the misreading of the op-code byte, causing programs to crash. The probable cause is noise generated on the Nascom 1 busses, although this has never been conclusively proved. A number of ‘cures’ were published, which consisted of basically three things:

1. Gridding the back of the memory card to reduce power supply noise.
2. Pulling up the outputs of the RAMs to increase operational speed.
3. Adding a time constant in the form of a capacitor to the RAM pullups to damp any noise on the output buffer.

On a Nascom 1 these work. On a Nascom 2, step 3 represents overkill, and can cause poor operation at 4MHz, the time constant being calculated for 2MHz.

Over that intervening period other possible causes have come to light. One is the 33R damping resistors in series with the address and CAS lines are too low to be properly effective. Another possible is the DBDR signal coming off too early. Yet another is the MREQ signal (applied to IC31) arriving too late, related to the system clock which latches it.

With the advent of Nascom 2, which is buffered on board, bus noise has been much reduced, and plague is almost nonexistent. However, Nascom 2 has revealed that the memory card will not run reliably at 4MHz without ‘WAIT states’.

We therefore now recommend the following:

Nascom 1

1. For each row of logic chips, grid the back of the pcb, connecting the +5 volt rail from the termination on the logic half of the pcb to the +5 volt termination of the 100n decoupiIng capacitor immediately opposite. For each row of logic chips connect the 0 volt rail likewise; where a row is adjacent to pin 12 of each of the EPROM sockets, incorporate this into the 0 volt grid.
2. Change R7 – 14 and R17 to 68R, and RAS links LK1 and LK2 should also be 68R resistors.
3. Where the Basic ROM is fitted, add a 150pF capacitor from pin 1 to 0 volts of IC9 on the buffer board, this will slightly increase MREQ, and thereby increase DBDR.