Considerations for Battery Fuel Monitoring in Battery Backup Applications

Accurate battery fuel monitoring in battery backup systems requires special considerations. Using TI's battery fuel gauge with Impedance Track™ technology offers the distinct advantage of not requiring the battery pack to be fully discharged to perform an auto-learn (calculate the charge) as the battery ages. This article discusses different implementation methods and techniques to achieve a proper battery auto-learn cycle in backup applications. Additionally, we review a case study of capacity and impedance changes in an aging battery pack.

TI's impedance tracking algorithm uses the battery's voltage, current and impedance measurements to accurately calculate the remaining battery capacity and run time of the battery pack. The most accurate battery fuel monitoring requires the correct selection of the specific battery chemistry. For this article, there are six different categories of chemistry, each with several options.

The key issues when determining battery aging in a battery backup system are (1) the maximum chemical capacity of the battery (Qmax), which is measured in milliamp-hours (mAh), and (2) the actual measured impedance of the battery (R_a table value), which will determine the actual battery run time based on load and temperature.

Most notably, high temperatures will adversely affect Qmax and internal battery impedance. Charging and storing batteries at low voltages (between 3.9V and 4.1V for a standard 4.2-V battery) will extend their life, but at the expense of shorter run time.

Some previous battery fuel gauge technologies required that the battery be fully discharged to update capacity information. Impedance tracking technology eliminates this full discharge requirement and instead uses two relaxation voltage measurement points to update Qmax. In the default firmware, these voltage measurements are typically taken around a 40% change in the battery state of charge (SOC). With TI's improved firmware, this SOC range can be reduced to 10% for "shallow" discharge. Reducing the SOC range for Qmax updates affects the accuracy of the battery fuel gauge; the more SOC range used, the higher the accuracy.

Depending on the battery chemistry, it is necessary to perform two relaxation voltage measurements within the specified voltage range.

Table 1 is an excerpt from that document. As shown in the table, if the chemistry ID is 0100, the Qmax-update voltage measurement is not allowed between 3737 and 3800mV because the voltage profile for that SOC is flat. This disqualified voltage range is based on measuring the battery voltage drop after at least an hour of rest. During discharges greater than C/10 loads, impedance measurements and updates are performed. (The “C-rate” rating is based on battery capacity. If a 3s2p pack has a design capacity of 4400mAh, the C/10 rate is 440mA. In this case, the safe rate is 500mA.)

To store the varying resistance for different SOC values, 15 grid points are used. If one grid point is recalculated, all subsequent grid points must be modified accordingly. Discharging for more than 500 seconds is required to avoid transient effects and distortion of the resistance value.

How to Start a Qmax Battery Auto-Learn Cycle
TI has evaluation software that displays the status and allows control of the Impedance Tracking fuel gauge parameters. After confirming that the battery voltage is outside the disqualified range, a “reset” command can be sent to the fuel gauge, setting the R_DIS bit and clearing the VOK bit. After the fuel gauge has completed a correct OCV measurement, the R_DIS bit will be cleared. Now, the battery can be charged or discharged, which will set the VOK bit within a few seconds. With the firmware set for a shallow SOC change of 10%, the charge/discharge can be allowed to change the SOC by at least 15%. After stopping the charge/discharge cycle, allow the battery to discharge (to a fully depleted state for up to 5 hours) to outside the disqualified voltage range. The VOK bit should clear, indicating that the second valid OCV measurement has been performed and the Qmax update has been successfully completed.

Table 1. Unqualified Qmax-Updated voltage ranges based on battery chemistry

The following two examples illustrate different system implementations of a battery backup system.

Example 1 Passive Battery Discharge
In this configuration, the active current of the fuel gauge chipset (~375 μA) can be used to discharge the battery for a longer period of time. Depending on the capacity of the battery pack, this can be several months. The fuel gauge can be programmed to remain in active mode continuously by setting the SLEEP bit of the "Operation Cfg A" register to 0. Another method is to set /PRES GPI using a fixed bit (NR=0) set in the "Operation Cfg B" data flash register.

With the firmware modified for a shallow discharge (e.g. 20%) for Qmax update, the battery pack is allowed to discharge to 75% of its capacity and then the battery is recharged to full capacity. The Qmax parameter is updated accordingly. Note that only the Qmax value is updated during this cycle and not the battery impedance (R_a table value). We assume that a few hours of sleep are allowed at the end of the charge to make a second relaxation voltage measurement.

Example 2: Active Battery Discharge
In this configuration, the system's discharge resistors can be used to actively discharge the battery. This should be controlled by a host processor either internal to the battery pack or external to the system. As mentioned previously, impedance grid point updates require a discharge current of more than 500 seconds C/10.

Even if the 10% minimum discharge requirement applies to the Qmax update, the battery pack should ideally be discharged with two impedance grid point updates. These occur during discharge at approximately 11% SOC intervals (i.e. 89%, 78%, 63%, 52%, etc.). In this case, a discharge from 100% to 75% charge is sufficient. If the battery charge is stored at 80% SOC for durability reasons, two impedance grid point updates will occur within 25% discharge.

A correct Qmax update occurs only after two consecutive relaxed voltage measurements separated by a charge or discharge are complete (assuming both measurements are outside the disqualified voltage range for the specified chemistry ID). Therefore, after the pack has been actively discharged to 75% of its capacity, several hours of rest are required, depending on the SOC. (Depending on the battery chemistry, a half-charged state may require up to 3.5 hours, while a fully discharged state may require up to 5 hours.)

Case Study
A Microsun Technologies 3s4p 8.8-Ah battery pack with many LGDS218650 cells using the bq20z80 chipset manufactured in June 2006 was used to study the effects of long-term storage. The pack was stored at about 45% capacity at room temperature for two years without charge and discharge cycles. The important parameters are the Qmax change and the cell impedance change, as well as the accuracy of the remaining capacity and run time calculations. The estimated self-discharge of these cells is less than 4% per year.

A constant resistance load of 3Ω was used to discharge the battery pack (equivalent to approximately a 3.5A discharge rate). The Qmax change and impedance value change are shown in Table 2 (next page) and Figure 1, respectively. On average, Qmax decreased by 3% while the cell impedance increased by 35%. Consistent with these cell variations, the initial discharge cycle after a two-year rest period was more than 99% accurate; specifically, 67 mAh of capacity was reported when the termination voltage was reached (67 mAh/8819 Qmax = 0.00761, or 0.761% error).

Figure 1. Battery impedance changing over time

Table 2 Qmax and battery impedance values ​​of sample battery pack before and after discharge