Micro­power

3) Another hardware solution. The display logic accesses the V.D.U. RAM once per microsecond – it is just possible to arrange for two accesses in every microsecond, and hand over alternate ones to the processor when it needs them. Wait state logic can then be used to synchronize the processor with these time slots.

Now all these are perfectly feasible solutions, but each has its own peculiar advantages and disadvantages. Approach (1) for example would need a new version of Nas-sys, and would meet with displeasure from all those who like the flexibility of memory-mapped display (no more Basic pokes to screen for example).

Approach (2) involves the processor in waits of up to 48 microseconds, and also has some wrinkles associated with the fact that the first character displayed on a line is actually being accessed 2 microseconds earlier. This approach formed the basis of a brave but unfortunately over-simple attempt in issue 3 of INMC 80 – problems such as this cannot be tackled just with Boolean algebra. The best tools are sheets of squared paper on which waveform diagrams can be drawn and redrawn, an the detailed information provided by the chip manufacturers on how to make Z-80s Jump through hoops.

A solution to approach (3) is presented here. It too has its disadvantages: early Nascom 1’s may have V.D.U. RAMs which are too slow to permit two accesses per microsecond. As it is, the modification is so simple that it is worth trying anyway – if the result produces spurious display misbehaviour then your RAMs are too slow, in which case have a go at approach (2) (if you do, be sure that you completely understand the display mechanism, prepare careful timing diagrams, and study the Z-80 manual on T-states and access cycles),

Another disadvantage is that the Z-80 will no longer be able to execute code from the V.D.U. RAM – which is a daft thing to want to do anyway. (This quirk occurs because, for reasons best known to the chip designers, M1 cycles – code accesses – have a different timing requirement to all others. This already creates all sorts of problems on Nascoms, such as the need for wait-states to access EPROMS on 4MHz machines.)

The advantages of approach (3) are, however, persuasive. Unlike (2), the maximum wait of the processor is one microsecond – undetectable in practice. It is extremely simple – there is little more logic involved than was required for the “snowplough”. The 2MHz and 4MHz versions are similar enough to permit a speed selection switch, for those of you whose reflexes are not up to landing a moon-rocket at 4 MHz, and don’t want to modify software.

The waveform diagrams contain all of the theory of operation. One very important point to realise is that the processor clock is the inverse of that used by an unmodified Nascom – there is an extra inverter in the signal path. This is essential to ensure that the “wait state” generator is sampled by the Z-80 at the only