80-Bus News


November–December 1982, Volume 1, Issue 4

Page 32 of 51

there is still room for improvement, as since writing these I have discovered that it is possible for a clock register to become programmed with 0FH as a legitimate character, caused by spurious pulses on the PIO lines. Under these circumstances the program goes into a permanent loop, thinking that the clock has just changed time. Some sort of time-out is required, where say, after having had 256 goes at reading the clock and failed, the routine returns the register contents regardless.

Having fitted a real time clock (RTC) some use has to be found to justify the clock. In the circles I move in, a couple of people have already found uses for the clock. Richard has modified his M80 assembler to include the time and date on printed listings, which can be very useful. David has written an elegant little utility called INDEX which gives a CP/M directory listing and allows a one line comment after each program entry. The date being added to the index created at the time of updating. I hope INDEX will see it’s way round a larger circle goon, although, as it requires the GM822 RTC to be hung on a Z80 PIO at port 1CH, it does make it a dedicated program, and therefore not universally usable. Note that in each instance port 1CH has been used, and it has been mutually agreed amongst ourselves that port 1CH will be the clock port. My own interest has been to hook the clock into a compiled Basic program.

As was mentioned earlier, the MM58174A clock exhibits a peculiar characteristic, in that it gains time if access is made to it over too long a time frame. I do not have the National data sheet to hand, but if I remember correctly, the clock must be accessed within a period of roughly 15mS from reading the first register to reading the last. This means that the access software has to be very fast. On the other hand, in instances where it is not the intention to display the clock on the screen, reading it back at its fastest increment of tenths of seconds, this is of little consequence. The software provided by Gemini certainly fails in this respect in that it wastes a lot of time within the read loop determining whether the clock has changed time, and if not, to convert the incoming nibble to ASCII and to save it in the temporary workspace. The obvious method to correct for this would be to copy the registers ‘as is’ to the temporary workspace using the tightest possible loop, perhaps using an INI instruction or even an INIR instruction if it could be contrived, then, having copied the registers to the temporary workspace, to sort them out later. I have not given this approach much thought, but it should be possible. The later National MM58174AN chip does not suffer from this problem, but is more expensive.

Making use of the clock to display the time continually. on the screen, whilst a nice idea, is not as easy as it sounds. That is, assuming the computer is to be used for other things apart from displaying the time. It is not difficult to write a program which will continually read the clock and place the result on the screen, Gemini have already provided the clock read routines, and a simple change to the software could be made to cause the routine restart itself instead of making the clock return to the routine which called it. This is a massive waste of time as the processor is 100% wrapped up in reading the clock. There is no time left for other things. As the clock is capable of creating interrupts, one method would be to use the interrupt output connected to the strobe input of the PIO to create an interrupt every time the clock changed state. The fastest interrupt rate on the MM58174 is one every half a second, this could be used to update the clock display every second, whilst causing minimum disruption to processing of other things. As it happens, the interrupt output of the GM822 was deliberately not returned to the strobe line of the PIO, as this could be the cause of some very strange software faults by the creation of undesired interrupts when used with disk systems (which must not be interrupted), or other interrupt driven auxiliaries elsewhere in the system, where the use of the EI and DI instructions could cause other problems. The interrupt output is connected to one of the output bits which means it can only be used when mode 3, the bit masked interrupt mode of the PIO is selected. This mode, whilst more flexible is also more difficult to implement.

This is an OCR’d version of the scanned page and likely contains recognition errors.

Page 32 of 51