MC146818

  This page is focussed on clearing the RTC on the 8573-P75. This system was made in Japan, with lots of Panasonic parts. I found this pinout on a Japanese spec sheet.

.Motorola MC146818AP or compatible

This is a rectangular 24-pin DIP chip, usually in a socket. Compatible chips are made by several manufacturers including Hitachi (HD146818AP) and Samsung (KS82C6818A). The number on the chip should end in 6818.

Although this chip is pin-compatible with the Dallas 1287/1287A, there is no built-in battery. This means that CMOS RAM can be cleared on this chip by just removing it from the socket for a few seconds and replacing it.

To reduce chances of damage to the chip, it is preferable instead to short 
pins 12 and 24 for a few seconds( with power off).

Here is a picture of the 6818 chip:

                             ___________
         1   -| *  U     |-  24  5 volts DC
         2   -|          |-  23
         3   -|          |-  22
         4   -|          |-  21
         5   -|          |-  20
         6   -|          |-  19
         7   -|          |-  18
         8   -|          |-  17
         9   -|          |-  16
         10  -|          |-  15
         11  -|          |-  14
 Ground  12  -|__________|-  13

Franc Zabkar wrote:
There are two clocks, a "hardware" clock and a "software" clock.

The hardware clock is maintained by an MC146818 compatible RTC which is operated at a crystal frequency of 32.768 kHz. This crystal is
usually packaged in a small tubular metal can. The RTC continuously
updates the date and time in CMOS RAM. The smallest increment is 1
sec.

DOS and Windows maintain a separate software clock derived from the
14.31818 MHz crystal (usually a rectangular 4-pin oscillator in a metal package). This clock is divided by 12 to produce a 1.19318 MHz clock which is then applied to a timer IC. Channel 0 of this timer IC is programmed to generate an interrupt (IRQ0) 18.2065 times per second
(= 1.19318M / 64K). Hence, each tick of the software clock corresponds
to 54.9 msec. The S/W time is stored in RAM as the total number of ticks since midnight. It is initialised from the h/w clock at bootup.

Since the two crystals may drift from their nominal values, it is possible that the S/W and H/W clocks may not correspond. The frequency of the 14.31818 MHz crystal may sometimes be trimmed for better accuracy by adjusting a small variable capacitor located nearby.

Another reason for the S/W clock lagging behind the H/W clock may be that IRQ0 is not being regularly serviced, despite it having the highest priority. Other devices may be demanding too much of the CPU's attention, thus causing some timer interrupts to be missed. Some people claim that buggy power management may also cause this.

This little QBASIC program reads the time in hh:mm:ss format:

 PRINT "The H/W time is ";

 FOR I = 4 TO 0 STEP -2
 OUT &H70,I
 PRINT HEX$(INP(&H71));" ";
 NEXT I

 PRINT
 PRINT "The S/W time is ";TIME$

I have verified that writing directly to the H/W clock in CMOS RAM does not affect the S/W clock until the next reboot. However, the TIME command at the DOS prompt adjusts both the the H/W and S/W clocks
simultaneously.