BQ34Z110: SOC estimation error

We have a product that uses BQ34Z110 to gauge a 12V, 5AH, SLA battery. This part has been in use for a long time (> 3years) and all of a sudden we're having issues with SOC estimation.

This is how the system is designed.

• Product displays battery status as bars based on SOC returned from the gas gauge IC.
• >85% SOC is considered as full battery
• <30% SOC is considered as low battery (only for issuing a warning to the user.
• The 'Cell BL Set Volt Threshold' parameter is set to 11.1V
• System automatically powers off when the battery voltage drops below 11.2V
• System can be turned ON only if the battery voltage rises above 11.3V or when operated from mains.

We have a burn in process for all the products manufactured. In particular, this product is taken through 4 cycles or discharge / charge starting with discharge. After each discharge, the batteries are charged for a minimum of 8 hours. Lately what we have observed is that as we progress in the burn-in the battery voltage and SOC goes out of synchronization. Results of first three discharge cycles are given below (of course, the first discharge cycle does not start from a fully charged condition).

As you can see, the SOC value corresponding to the terminal voltage increases as we proceed with the burn in. Due to this we are unable to guarantee a fixed usage time once the low battery warning is issued and in most cases the low battery message itself is not shown (as in cycle 3, because SOC is still >30 even when the unit is about to switch off).

Chart below shows the discharge characteristics from three cycles.

As you can see, the progression of battery voltage, current or power remains the same in all three iterations. However the SOC value stays at 100% for 1 hour 41 minutes for the second discharge cycle and 2 hours 58 minutes for the second cycle.

Charging characteristics are as below.

Rate of climb of SOC during first charging cycle appears quite ideal where 80% is achieved in about 4 hours (as we had characterized earlier). However the second charging cycle starts charging at a higher SOC level though it corresponds to the SOC at which the previous discharge cycle completed. Battery charger is based on BQ24450.

Few points with respect to history

• We have not made any changes in hardware or software to trigger this observation.
• Almost all units (>25) that we have produced in this batch is having this issue.
• Tried with batteries from two different manufacturers and batteries from different batches, while same results seen.
• Same golden image used for all experiments, again it has been the same since we started production 3 years ago.

Please though in some light to see what could be wrong.