I had a strange situation when I measured the SOC-OCV curve of the battery

[Leon_Xp]

A strange situation occurred when I performed SOC-OCV testing on a LiFePO4 cell with a capacity of 105Ah using the incremental OCV method.

The test uses a current multiplier of 0.2C, a resting time of 2h, an ambient temperature of 25°C, a charge cutoff voltage of 3.7V, a cutoff current of 0.02C, and a discharge cutoff voltage of 2.5V. To ensure complete discharge, it is set to rest for 5min after discharge to 2.5V, and then discharge again to 2.5V using a current of 5A. This last process of low current discharge will Repeat three times.

In some discharging or charging sections, a non-monotonic variation of the voltage curve was observed. As shown in the attached figure, during discharge, the voltage appears to increase and then decrease. While in charging, the voltage decreases and then increases. The voltage even appears to be monotonically decreasing when the eighth charge is performed (the polarization establishment phase is excluded). According to normal perception, this should not happen. The battery was almost new and the charging and discharging equipment had not failed during operation, everything was well connected, but there was an average error of about 6mV in the measurement accuracy. However, through testing, this measurement error is not accidental, but is steadily present. Therefore, I think that the sensor error is not the cause of the error in the measurement results.

I have reviewed many literature sources and have not been able to find any relevant cases or explanations. I would like to ask for help if anyone has also encountered this situation or can give an explanation?

 

[HateGrid]

I haven't gone into this level of depth for charge analysis and haven't looked into the chemical reaction to the current pulse and then relaxation. However, with 6 mV of measurement error, I do wonder if some of what you are seeing is due to the noise distribution of your measurement circuit changing. I have seen similar issues due to cross-coupling of noise from undersized ADC capacitors / too high of a source impedance or operating the ADC at too high of a clock rate.

 

[Leon_Xp]

I haven't gone into this level of depth for charge analysis and haven't looked into the chemical reaction to the current pulse and then relaxation. However, with 6 mV of measurement error, I do wonder if some of what you are seeing is due to the noise distribution of your measurement circuit changing. I have seen similar issues due to cross-coupling of noise from undersized ADC capacitors / too high of a source impedance or operating the ADC at too high of a clock rate.

I'm guessing that this is happening because the voltage change in the 60% to 90% SOC stage is too small (the voltage plateau is too flat), causing the effect of sensor error on the measurement results to be magnified?

 

[Goboatingnow]

Looks like adc quantisation error to me

 

[HateGrid]

I'm guessing that this is happening because the voltage change in the 60% to 90% SOC stage is too small (the voltage plateau is too flat), causing the effect of sensor error on the measurement results to be magnified?

Correct, the signal that you are trying to measure (battery voltage) is being corrupted by the noise in the measurement system. Often people will just average the noise, but this assumes that the noise distribution is even which it may not be depending upon the source of the noise. The battery voltage changes slowly, so you can put a low-pass filter on the input (a simple RC filter will work) to remove high-frequency noise. Also, make sure your ADC circuit is completely stable.

With ADCs, you can do oversampling to help reduce the "stair-stepping" quantisation error.



See the following links for some good overviews:

• https://www.silabs.com/documents/public/application-notes/an118.pdf
• https://www.ti.com/lit/eb/slyy192/slyy192.pdf