BQ27426: Test conditions regarding the discharge curve and the corresponding value of SOC

Yes. Please perform a full charge and discharge with a constant current, after you ran a learning cycle.

www.youtube.com/watch

Thanks for your reply.

We have completed the gold learning and are now doing tests using the FS files generated by the gold learning.
Do you mean that we need to do a full charge and discharge on the test device as well, before doing the test?

As long as the test device is similar in age (and identical in make/model) to the cell you used for the Golden Image (including a learning cycle), it's not necessary to o full change and discharge before the test. Please make sure that you start with a relaxed cell before running the test so that the gauge has the correct cell state.

Thanks for your reply.
Does relaxed cell mean that we need to take a fully charged battery, leave it for 2 hours and then do the SOC change test?
Also, what is the purpose of this require of relaxed cell? Is this described in the user manual?

When we tested the SOC with a battery that had not been relax, the SOC value was change per 1% (meaning the test was PASS) using 1A current and continuous discharge.
However, when the test was carried out with a 10 minute rest period for every 12 mAh discharged, there was a jump from 24% to 0%, is there a correlation between this and not relaxed cell?

The purpose for the relaxed cell is that the gauge needs to determine cells state (depth of discharge) as a reference point for the gauging algorithm. This is ideally done with a true open circuit voltage measurement. This is also a big reason, why resetting the gauge is not recommended during regular use because the algorithm relies on correct cell state. It continuously monitors measurements and it will update cell state whenever conditions are right. However, if you run a test, then it's critical that the cell state is correct, otherwise the test will show reduced accuracy until the conditions are right for the gauge to auto-correct.

It's not directly described in the user manual but it's part of the impedance tracking algorithm. See the TI training for Impedance Tracking info: 

https://training.ti.com/impedance-track-benefits

www.ti.com/.../slyp819

Thanks for you reply。

When accuracy testing, it can perform full charge, full discharge and relax, which has met the accuracy testing requirements.

However, if the battery needs to be replaced during use, or if Fuelgauge is abnormally reset, it may not be able to perform full charge, full discharge and relax in the application scenario, how to ensure that Fuelgauge gets the accurate battery status in this scenario?

In mass production, it takes a lot of time to perform full charge, full discharge and relax, which affects productivity, is there any solution?

If the gauge is reset, then the host uC can detect this with ITPOR (which indicates that the gauge was reset). If ITPOR = 1, load the golden image followed by restoring the state and Ra subclasses. You can periodically (e.g. once a week or once every 10 cycles) save both the Ra and state subclasses on the host system persistent memory.

If the battery is replaced, the host uC should write the golden image to the gauge. If the battery is a new one, then it should behave close to what the golden image describes and the gauge will auto-correct in regular use.

Thanks for you reply

If Fuelgauge reset, the Fs file saved in nv needs to be loaded. If the battery is replaced, the original Fs file needs to be loaded. Is this understanding correct?

Does 10 cycles mean 10 complete charges <--> discharges?

Do I only need to save data for the RA table and Gas Gauging->State, the two subclass IDs?

Your understanding is correct.

10 cycles is just a rule of thumb. The idea is to take snapshots of cell parameters often enough to catch changes in parameters (due to age and/or cycles). These usually don't change fast (as in every hour or day or even week) so it's not necessary to read the state and Ra classes from the gauge often. To answer this scientifically, you'd have to examine how the cell ages for your use cases but as this is significant amount of work and analysis for you, it's usually easier to just save the state and Ra once a while (as in e.g. every 10 cycles, which should be more than enough to not miss significant degradation in Qmax and/or Ra.