How far should you trust any of those battery "gas-gauge" ICs, the circuits that drive the remaining-battery-life displays on portable devices? Today, it's something users only learn through experience. The problem lies in the way ICs have approached the job. Some integrate current into and out of the battery, while others measure terminal voltage. Neither approach alone is ideal.
Limitations on current integration arise from the difficulty of predicting battery self-discharge, which varies with ambient temperatures, how long the battery sits between uses, and whether it's fully discharged before recharging or not. The algorithms that current-integration gas-gauge ICs employ to accommodate these parameters may not reflect any given end-user's reality.
Voltage-measurement approaches fall short when battery voltage is measured while the unit it's powering is in use. That's because the batteries' internal IR drop reduces the observed voltage. Even if the battery were disconnected before the measurement, its terminal voltage would not immediately return to its real no-load state.
Texas Instruments has overcome these limitations by fusing both approaches in a new algorithm-based technology in its bq20z8x two-chip set (a gas gauge and an analog front end). TI's "Impedance Track" algorithm exploits longer periods of inactivity to recalibrate the "starting position" for state of charge. It also eliminates self-discharge effects and low current draw from pack electronics.
As a result, the IC continually updates full-charge capacity by comparing the state of charge before and after applying the system load. The algorithm then uses current integration to measure and maintain real-time information for the state of charge and impedance. The result is a remarkably accurate model for predicting remaining run time.
Initially targeting applications with multicell lithium-based battery packs, Impedance Track also can support other types of battery chemistries, such as nickel metal hydride (NiMH) or nickel cadmium (NiCd).
Pricing for the bq20z8x chip set is $4.35 in 1000-unit quantities.