Micro­power

I was a little concerned about the problem of access time for the VDU RAM so the next step was to make absolutely sure that this problem was minimised. The most critical cycle, timing-wise, is the read cycle because data is read in on the falling edge of the last clock of the cycle (T3). With the write cycle, data is valid rom the falling edge of T1 until slightly afterthe rising edge of T3 and may be written anywhere between the falling edge of T2 and the falling edge of T3. Hence as long as the falling edge of the T3 clock is before the rising edge of the access slot, there is less chance of a bad read. At 2 Mhz this is all well and good as the falling edge of the T3 clock can be placed about 375ns into the access period. This gives an access time somewhere in the region of 375 to 400ns. With the 4 Mhz clock, things begin to look dodgy because, with the T2 clock falling edge just on the falling edge of the access slot, there is only 250ns before the falling edge of T3. So access time to the VDU RAM must be better than 250ns in order to allow time for the address bus multiplexors to change the address bus over and let the RAM settle down. By placing the falling edge of the 4 Mhz on the falling edge of the access period, you are relying on the delay time of the gates to maintain a high output long enough to terminate the wait state. This is bad practice as the main clock only needs to be delayed by a slow gate and the wait state may not be terminated.

Now it became obvious that Snowdinger 2 would be completely different to Snowdinger, due mainly to using the same circuit for both 2 and 4 Mhz. Using the 1 Mhz clock ORed with VRAM a gated signal is produced for any access to the VDU RAM, ie VRAM is synchronised to the 1 Mhz clock. Finding a wait signal for the Z80 that would work at 2 and 4 Mhz was some what more difficult. However the final waveform is the result of ANDing the 2 Mhz with the 4 Mhz clock. For the Z80 to use this wait signal, both the clocks need to have there phases changed in order to sample the wait gate at the correct times. These phase changes are accomplished as follows:

1. The 2 Mhz clock is shifted back quarter of it’s cycle by exclusive ORing it with the 4 Mhz clock.
2. The 4 Mhz clock is shifted back quarter of it’s cycle by exclusive ORing it with the 8 Mhz clock

The reason for shifting the 4 Mhz clock is, so as to keep it’s falling edge away from the falling edge of the wait gate signal, hence avoiding any static hazards from slow gates, etc. This has the disadvantage of reducing the access time for the RAM by about 62.5ns, but makes the circuit more reliable. Figure 1 shows all of the waveforms needed to construct the circuit, however I soon discovered that this circuit will only work for a Nascom 1. A quick look with a 100 Mhz oscilloscope confirmed that Nascom 2s have different phase clocks, Figure 2 shows diagrams of the 4,2,1 and the serialiser load clocks for a Nascom 2.

The Nascom 2 uses a 74LS193, presettable 4-bit binary up/down counter. This circuits clock is positive edge triggered, hence all the clock outputs have different phases. At first sight this presents no problems, apply the theory to the new waveforms and design a new circuit. Unfortunately, changing the phases of the clocks also changes the position of the load