OK, this yellow stick is a "special backup" nice for tests only
Yes, the problem can be in the memory protection circuit, around Z2. Just a little explanation how its works :
The RAM 5101 and CMOS (Z1, Z2) are powered by VAB source (the battery @ ~3.5 volts). Their functions are the following:
- Z1 (4081) act like a big AND GATE between three signals : a basic chip select from the decoding, and two others signals Q1 and Q2.
Q1 and Q2 are protections at power-up and power-down.
Their role is to disable the RAM access (and especially write), when the power become unstable. This prevent spurious access at startup and stop.
Q1 and Q2 are only single pulses at power change, but remain stable after.
- The "zener group" (CR38, CR37, R156, R157 and Q5) is a voltage threshold detector. It detect when the +5 volts drop (at power-off) or rise (at power-on), and then trigger the 4528.
-This 4528 circuit is a double monostable: It generate a delayed pulse on Q1 and Q2. The monostables are separate, one is used with positive edge, and the second is used with negative edge. It's simple, one is used for power-on, one is used for power-down. The duration of the pulses is determined by the couples C31/R159 and C32/R158.
What to check ?
As you have a voltmeter and another board, you can compare some signals.
First, check that the return-protect diode CR26 is good. When power is off, the +5 volts line should be 0 ! if you get ~3.5 v, then the battery power all the board (and also the ram protection as mentioned above - so detection could be inefficient).
Check the power voltage at Z1 and Z2 (pins 8 / 16), should be VAB (~3.5 volts).
Check the voltage at the 4528 triggers entries (pins 4 and 11, when power is on, and off) - could show a problem at the "zener group".
Check the stability of the Q1 and Q2 signals. This can show a problem with Z2 or R/C couples.