26. No Logic In using PICs

I considered using PIC chips for memory logic, so they could select whether an address was for RAM or ROM. Programmatically this is not a problem, the code can be written to do this quite easily. The problem is that they cant do it fast enough.

Most Z80 instructions work over 4 clock cycles, T1, T2, T3 and T4. During T1, the address pins A0-A15 are set to the value of PC (Program Counter) and half a cycle later, MREQ (Memory Request) and RD (Read) pins go active. By the end of T2 the data is fetched from the data bus on D0-D7.

If the Z80 is running at 1MHz then each cycle is 1μs (1 micro second), so fetching an instruction takes at most 2μs.

Now consider a PIC microcontroller running at 4MHz. Again each instruction takes 4 cycles, so the instruction to read the pins on a PIC such as MOVF GPIO,W will take 1μs (4Mhz / 4 cycles), so at most we could only perform 2 instructions on the PIC while the Z80 is trying to fetch the next instruction or operand on the data bus. A branch instruction on the PIC takes 8 cycles, so we could not even fit a loop into the time frame.

The only it could work with a PIC is if the Z80 was running at a much slower clock rate. However this Z80 can go at 2.5MHz and the Z80B can go at 4MHz, and it'd be a shame not to run at full speed.

You can clock some of the PICs with an external crystal to 20MHz and the more expensive PICS to 60MHz and this would allow 5 or 15 instructions in 1μs respectively, but probably still not fast enough at higher Z80 speeds.

Time to bring out the classic TTL chips.