How to fine-tune TI impedance tracking battery fuel gauges for shallow discharge applications

TI's Impedance Track™ battery fuel gauge technology is a powerful adaptive algorithm that remembers how battery characteristics change over time. Combining this algorithm with the specific chemistry of the battery pack provides a very accurate state of charge (SOC) for the battery, extending the life of the battery pack.

However, updating information about the total chemical capacity (Qmax) of a battery requires certain conditions. This becomes more difficult at the extreme stable voltage state of a LiFePO4 battery (see Figure 1), especially if the battery cannot be fully discharged and allowed to rest for several hours. Figure 1 shows the typical open circuit voltage (OCV) characteristics versus depth of discharge (DOD) for LiCoO2 and LiFePO4 battery chemistries. This article focuses on the impedance tracking techniques of References 1 and 2.

Figure 1: Figure 1 DOD-based battery OCV measurement.

TI recommends using the Impedance Tracking 3 (IT3) algorithm for all LiFePO4 batteries. Improvements to IT3 over earlier Impedance Tracking algorithms include:

Better low temperature performance through better temperature compensation

More filtering to prevent SOC capacity jumps

Higher accuracy for non-ideal OCV readings of LiFePO4 batteries

Conservative remaining capacity estimates, and additional load selection configurations

The IT3 is included in TI's bq20z4x, bq20z6x, and bq27541-V200 fuel gauges (this list is not exhaustive).

Typical conditions for Qmax update

The impedance tracking algorithm defines Qmax as the total chemical capacity of the battery, which is calculated in milliampere hours (mAh). For a correct Qmax update, the following two conditions must be met:

1. Both OCV measurements must be made outside the rejection voltage range, based on the battery chemistry identity (ID) code determined by TI. OCV measurements can only be made on an idle battery (not charged or discharged for several hours).

Reference 3 lists some disqualified voltage ranges, some of which are shown in Table 1. We can see that for chemical ID code 100, OCV measurements are not allowed if any cell voltage is above 3737mV or below 3800mV. In effect, this is the “forbidden” range for the best accuracy of OCV measurements. Although the article gives the SOC percentage, the fuel gauge determines the disqualified range based on voltage only.

Table 1, extracted from Reference 3, lists the voltage ranges that fail based on the chemical properties updated with Qmax.

2. The minimum pass charge must be integrated by the fuel gauge. By default, it is 37% of the total battery capacity. For shallow discharge Qmax updates, this pass charge percentage can be reduced to 10%. This reduction comes at the expense of SOC accuracy, but is tolerable in systems where it is not possible to update Qmax otherwise.

Now that we understand the shallow discharge Qmax update requirement, let’s look at an example of data flash parameters that we need to modify in a lower capacity battery pack configuration. The default impedance tracking algorithm is based on a typical laptop battery pack that has two parallel groups of three series cells each, a 3s2p configuration. Each group has a 2200-mAh capacity, so the total capacity is 4400hAh. LiFePO4 cells have about half that capacity, so if they are used in a 3s1p configuration, the total battery pack capacity is 1100mAh. If a smaller capacity battery pack like this is used, specific data flash parameters need to be fine-tuned in TI’s fuel gauge evaluation software to obtain the best performance. The rest of this article describes this process.

Example calculation

Consider a 3s1p configuration battery pack using A123 Systems TM1100-mAh 18650 LiFePO4/Carbon Rod cells. The TI chemical ID code for this battery type is 404. This battery will be used in a storage system with a normal temperature of around 50°C. The discharge rate is 1C, and a 5-mΩ sense resistor is used for the fuel gauge for coulomb counting purposes.

As shown in Table 1, the OCV measurement failure voltage range for chemistry ID 404 is 3274mV (minimum, ~34% SOC) to 3351mV (maximum, ~93% SOC). Most LiFePO4 cells have a very wide failure voltage range (see chemistry ID 409 for comparison). However, depending on the cell characteristics, it is possible to find a higher minimum failure voltage for the shallow discharge Qmax update. For chemistry ID 404, it is possible to increase this value to 3322mV, allowing a shallow discharge Qmax update window of 3309 to 3322 mV (see Figure 2). Designers can use this mid-range low error window to implement data flash modifications. Since only the high and low failure voltage ranges can be set, the host system must ensure that no lower OCV measurements are made below 3309mV. (The OCV measurement error increases dramatically between 3274 and 3309 mV as the associated error grows.) Although only a 13-mV window is active at the lower OCV measurements (3322 – 3309 mV = 13 mV), it corresponds to a 70% to 64% SOC range.

LiFePO4 cells have a very long relaxation time, so we can increase the data flash parameter "OCV Wait Time" to 18000 seconds (5 hours). Since the normal operating temperature of the battery is increased, the parameter "Q Invalid Max Temperature" should be modified to 55°C. In addition, "Qmax Max Time" should be modified to 21600 seconds (6 hours).

Figure 2. SOC-related error for a 1-mV voltage error.

To reduce the Qmax pass charge from 37% to 10%, the DOD Max Capacity Error, Max Capacity Error, and Qmax Filter need to be modified as they all affect the reject time between OCV1 and OCV2 measurements. The Qmax Filter is a compensation factor that changes Qmax based on the pass charge.

These parameters are set to achieve a "maximum capacity error" of less than 1% based on measured pass charge, including the ADC maximum compensation error ("CC deadband"). However, some modifications to these values ​​are required to allow for shallow discharge Qmax updates.

Example 1 Qmax update timeout period

To achieve less than 1% cumulative error for a 10-mΩ sense resistor for a 1000-mAh battery, and a hardware-set 10-μV fixed “CC deadband,” the timeout period for Qmax updates is determined by the following:

10 μV/10 mΩ = 1-mA compensation current.

1000-mAh capacity × 1% allowable error = 10-mAh capacity error.

10-mAh capacitance error/1-mA compensation current = 10 hours.

Therefore, from start to finish, including rest time, there are only 10 hours available to complete a Qmax update. After the 10-hour timeout, once the fuel gauge takes its next correct OCV reading, the timer restarts.

Example 2 Data Flash Parameter Modification

In a design using an 1100-mAh battery with a 5-mΩ sense resistor, the timeout period for Qmax updates can be calculated using the same method:

10 μV/5 mΩ = 2-mA compensation current.

1100 mAh × 1% = 11 mAh。

11 mAh/2-mA compensation current = 5.5 hours.

In this case, the capacity error percentage needs to be relaxed to increase the Qmax timeout. Changing the "Maximum Capacity Error" (from the default value of 1%) to 3% gives:

1.1 Ah × 3% = 33 mAh

This will increase the Qmax failure time to:

33 mAh/2-mA capacity error = 16.5 hours.

The "DOD Capacity Error" needs to be set to 2 times the "Maximum Capacity Error", so you can change it to 6% (the default value is 2%).

The default value of 96 for the "Qmax filter" needs to be reduced proportionally, depending on the percentage of charge passing:

“Qmax filter” = 96/(37%/10%) = 96/3.7 = 26

Table 2 shows the typical data flash parameters in the fuel gauge evaluation software that must be modified to implement the shallow discharge Qmax update. These special parameters are protected (classified as "hidden") but can be unlocked by TI application personnel. The example battery pack used in this table is the battery pack described above, which is a 3s1p battery pack using A123 1100-mAh 18650 LiFePO4/carbon rod cells (chemical ID 404).

Table 2. Some protected data flash parameters that can be modified by TI application personnel depending on system usage.

(1. This parameter is important during the golden image process. If a standard 4.2-V lithium-ion battery is used and it is only charged to a 4.1V system level, the first Qmax update after the battery is charged to 4.2V is still necessary to meet the 90% capacity change requirement. Based on the chemical ID code set by the fuel gauge, the start and end point checks are performed for the capacity change of the specified battery capacity, i.e., the "design capacity", and the estimated DOD.

2. Wide temperature variations can cause errors when calculating Qmax. In systems that operate normally at high or low temperatures, it is necessary to modify this parameter.

Qmax update event

The following events describe a practical way to implement a Qmax update after the data flash parameters are changed as described in Examples 1 and 2.

1. A Qmax update should be initiated when the battery voltage is within the low correlation error window shown in Figure 2. The designer's own algorithm can be used to discharge/charge the battery into this range.

2. In this example, in order to enter the valid measurement range (chemical ID is 404), all cell voltages must be greater than or equal to 3309mV and less than or equal to 3322mV. If the cell voltage is just outside the valid range during regular discharge, another discharge or charge cycle must be started before the set "OCV wait time" of 18000 seconds. If after 6 hours and 10 minutes, all cell voltages are within the range of 3309 to 3322mV, a correct OCV measurement has been made.

3. The next step is to fully discharge the battery. Once the battery is fully charged (i.e. 100% SOC), it should rest for another 6 hours and 10 minutes before taking the second OCV measurement. After that, the Qmax value is updated. If charging takes about 2 hours, the timeout period needs to be at least 8 hours. From the calculation of the 16.5 hour timeout period in Example 2, we know that this is more than enough time, with an additional 8.5 hours of buffer time.

4. Issue a ResetCommand (0x41) to the fuel gauge when it is in the on mode to reset the OCV timer.

Table 3 shows the results obtained by cycling the cells as described using the example battery configuration.

Table 3 Results of full cycle and shallow charge Qmax update.

(From empty to full charge)

in conclusion

TI's impedance tracking technology is a very accurate algorithm for determining battery SOC through battery age. In some LiFePO4 battery applications, it is not possible to fully discharge the battery using a period of inactivity, so it is necessary to study a shallow discharge method for Qmax update. This article describes the factors to consider and data flash programming configuration to implement a shallow discharge Qmax update. Modifications to these parameters must be approved by TI application personnel based on system configuration and requirements.