80-Bus News


January–February 1983 · Volume 2 · Issue 1

Page 6 of 56

The Interrupt System of the Z80

by R. O’Farrell

The interrupt structure of the Z80 microprocessor chip is one of the least understood and at the same time potentially most powerful facilities provided by this i.c. My purpose in writing these notes is to survey the interrupt structures offered us, to encourage other users to try these out for themselves.

It should be remarked immediately that it while it is relatively easy to use one interrupt at a time, a complex set up is not easily obtained on a general purpose machine, although on a dedicated machine such as a process controller it is possible. It is very likely that a Nascom or Gemini system will be used as a development system, to develop software for a purpose built dedicated controller. This possibility underlies the following notes. Whether a dedicated controller will be used, or the system software of a Nascom/​Gemini modified to support multiple interrupts is a matter for the individuals concerned, but it is essential that each interrupt be set up and carefully debugged along the lines we will examine.

By way of an aside, let me remark that a dedicated controller is no great deal. Consider a minimum Z80 system. This could comprise a similar configuration to that shown in Section 9 of the Mostek manual – a Z80, an EPROM, some RAM and a PIO. It would also need an oscillator to provide a system clock. A minimum system of this nature would not need any buffering, and could possibly be expanded with another PIO or two before that became necessary. Such a system could be wired up on a prototype board. What would it do? The PIO would allow it to sample or switch up to 16 lines without getting involved in elaborate multiplexing circuitry. A reasonable assumption might be that it would treat eight of these as inputs, and the other eight as outputs. An example of what can be achieved in this way was published in Micropower, Vol 2, No. 1, where it was shown how a Nascom could control a washing machine. Another example of what a dedicated system might do is to control a dot matrix printer, such as the IMP or EPSON.

As computer enthusiasts, we are concerned always with the finer points of Z80 usage. We have two problems – to get the hardware to work and to get the software written and debugged. These two problems are complementary and depend to a certain extent one on the other. In these notes, I will deal only with the software side of matters and will assume that the hardware is taken for granted and assumed correct – that all signals are clean, bounce free, and of the correct voltage levels. This is not always so!

The first interrupt structure on the Z80 is the Non Maskable Interrupt, known familiarly as the NMI. Those familiar with the pinouts of the CPU chip will know that pin 17 is called /NMI. This is an input, negative edge triggered. That is, the transition from high level (normal condition) to low level tells the CPU that an NMI is required. The CPU examines the state of this pin at the end of every instruction, and if the pin is active, then it proceeds to service an NMI. There are a small number of conditions which will prevent the NMI being recognised, and I’ll mention these to get them out of the way. The first, and not very likely condition is if WAIT states are continually being requested the current instruction is prolonged and never reaches an end. In consequence the NMI test is never reached. Such a condition is hardly likely to arise in serious use of a Nascom or Gemini, as the dynamic memory would not be refreshed, and the program lost. It might happen in use of a Z80 based controller, where the machine was executing Wait states, waiting for a very slow peripheral to react – said peripheral having perhaps gone on the blink! The other failure to see an NMI is if /BUSRQ is active, i.e. if something like a DMA chip has control of the bus.

Page 6 of 56