We have designed in PCF2129AT RTC into one of our products. They were not long in use in the field when some of them lost their date. They had risen their OSF status bit to indicate oscillator stopped failure and then Linux driver refuses to return the date.
It is hard to establish what is the root cause but, nevertheless, I can replicate same kind of an issue at my desk. When I cool down the chip, it sometimes takes abnormally high current from the battery and turns on the OSF bit. This is one example of the events:
CH2 is battery voltage and CH1 voltage at PCF2129 VBAT pin. VDD is off. There is 1k resistor from battery to PCF2129 VBAT. It can be calculated from the picture that the chip took approximately 700uA current. However, the voltage didn't drop even near the 1.2V where the OSF bit should go up but it still did.
I have measured CLKOUT and verified that the oscillator really stops for a moment. The time and date is not corrupted and the chip continues normally after the event, only OSF being set.
There is 1uF at VBAT and 10nF at BBS. We use I2C.
What is going on? Any ideas, how to debug or fix this?