In the data sheet Maxim has a very large warning in bold letters that says this:

WARNING: Negative undershoots below -0.3V while the part is in battery-backed mode may cause loss of data.

Be aware that this is a huge caution. This chip appears to be VERY much more sensitive than other chips that typically are speced at more like -0.5V. I have also found that many other chips speced at the -0.5V type levels will actually realistically will work fine with -0.6V negative undershoot.

I do not know just how your test platform is built but this warrents taking a look at with a good oscilloscope.

You probably will need to ensure that the part of your program that is accepting the user input for adjusting the time is not exercising some of the system bus or control signals in some manner that leads to the there being negative undershoots on any signal line that is connected in common to the RTC. If so, it is likely that this is what is making your RTC fail. If when you finish adjustment the program goes to another part and re-writes a new time into the RTC any previously caused loss of data is purged and the thing starts to appear to work correctly.

With this type of problem it may very well be completely coincidental that the A and non-A flavor of the chips showed you the alternate behavior that you saw. With negative undershoot some chips will be more sensitive than others. Chips from different lots or date codes could even show this difference in sensitivity to the negative undershoot even if they had the same part number!!

Unless you carefully validate your circuit to ensure that there is no negative undershoot you can never be sure of the behavior of any chip on the board. If you have used the best practice type of measurements to ensure no negative undershoots and then tried the A and non-A comparison on multiple sets of chips then I would be willing to accept that there is a chip type difference here but not before that multiple comparison is done.

-----------------------
Now lets correct a term here:

I used the term "negative undershoot" here because that was how Maxim has the data sheet written for your RTC. However if you talk to any self respecting SI (signal integrity) engineer you will soon learn that the following terms are the correct ones.



When Maxim's data sheet referred to "negative overshoot" they were trying to refer to the portion of the signal that has transitioned to below the nominal negative signal level at ground which is fact correctly referred to as an overshoot because the signal has gone farther that it was meant to go.

Michael Karas