80-Bus News

  

November–December 1982, Volume 1, Issue 4

Page 31 of 51

board chip select, CS, did not float high as expected but remained resolutely at about the two volt level. This, despite the inclusion of a 100K pullup resistor on the clock board to ensure the CS line remained high. Taking in to account the loading effect of the high impedance meter used to measure this voltage, the approximate voltage on this line was guessed as being about the 2.5 volt level. Changing the PIO or the buffer made no difference, it remains a mystery where the leakage causing the line to be lower than expected is occurring.

However, accepting that this leakage does occur, and as this voltage is about the threshold of the board chip select buffer, which is a CMOS device with a threshold of about 2.5 volts, we thought that any spurious noise on this line, and there has to be some, was causing the CMOS buffer on the clock board to select the clock chip. This was confirmed with a ’scope looking at the select pin of the clock chip, after the input and power down buffers. A definite short glitch was noted on the select pin during a computer reset sequence. The cause, like the voltage leakage remains unknown. As the output lines of the PIO are directly connected to the clock chip which again is a MOS device, these are also susceptible to noise due their high impedance inputs. This appears to cause the clock chip write input to open during a computer reset sequence and so to write the rubbish on the data lines to the clock chip registers which were being addressed by the rubbish on the clock chip address lines, also directly connected to the PIO. We were relieved to cure the problem by the simple expedient of changing the 100K pullup resistor for a 22K pullup.

Unfortunately, this did not remove all the problems, the clock would now work correctly whilst the computer was left running, however, on power down and subsequent power up the clock was occasionally being corrupted. Having previously eliminated the power down and battery backup, this left the power up condition. This was confirmed as being the cause by powering down the computer with the clock connected, then unplugging the clock and powering the computer up again. Reconnecting the clock with the power on and then reading the clock, the clock worked correctly. The clock circuitry has a 100K pullup resistor in series with a 100nF capacitor to stop the CS line becoming active during power up. 100K and 100nF gives a time constant of 10mS which is the time allowed for the power supply rails to settle. This may be adequate for a switch mode power supply, as the clock at home had shown no trouble in this respect. However, the works computer was fitted with a large conventional power supply. Investigation of the power supply start up revealed that the rails did not become stable for at least 100mS, and during that time the state of the PIO was extremely indeterminate. It was obvious then that the 10mS allowed for stabilisation was inadequate for the power supply in use, and the 100nF capacitor was changed for a 2uF tantalum, giving a 200mS time constant. This cleared the remaining problem.

Both the problems noted above could do with further investigation, why were these problems not noted on the prototypes, and are the conclusions drawn above and the solutions adopted, the correct ones. Our aim at work was to make the clock work, not to conduct an investigation into the behaviour of PIOs and power supplies at reset and power up. Perhaps Gemini may like to make some comment in a future issue.

To sum up then, the Gemini GM822 represents a useful addon which is simple to construct and easily added to the system. A one evening project with a useful end result. Having cured the problems mentioned above, both clocks have behaved well. Taking the cost of the competing clock boards into account, the Gemini GM822 seems to be about par for the course in terms of value for money and facilities provided.

I didn’t like the software supplied with the clock so I set to and wrote some new routines. These used large chunks of the originals and in this case I can not credit them as I do not know who wrote them in the first place, however, thank you, whoever you are. The routines are listed elsewhere and are written for use under CP/M only, but are otherwise Nascom/​Gemini compatible although I have maintained port 1CH as the clock port. This could easily be changed on reassembly. The most important improvement is that they contain error checking of the input, making it impossible to program wrong data into the clock. However,


Page 31 of 51