Simple Calibration Circuit Maximizes Accuracy in Li-Ion Battery Management Systems Design Note 471 Jon Munson Introduction In Li-Ion battery systems it is important to match the charge condition of each cell to maximize pack performance and longevity. Cell life improves by avoiding both deep discharge and overcharge, so typical systems strive for operation between 20% and 80% states of charge (SOC). Detection and correction of charge imbalances assures that all cells track within the desired SOC window, preventing premature aging of some cells that could compromise the entire pack capacity. Highly accurate measurements are required to determine SOC with Li-Ion cells due to their exceptionally flat discharge characteristics, particularly with the lower voltage chemistries (see example in Figure 1). 4.0 1 2 LTC6802IG-2 35 GPIO1 VREG VTEMP1 V– 1M 3 4 31 LT1461AIS8-3.3 7 VIN SHDN GND 8 VOUT 6 5 28 26 2.2μF 1μF OTHER PINS NOT SHOWN FOR CLARITY DN471 F02 Figure 2. LT1461 as an External Calibration Source for an LTC6802 Li-Ion Battery Monitor Accounting for the Error Sources Fundamentally, there are several key characteristics that comprise an overall accuracy specification: 3.5 3.0 VOLTAGE (V) Si2351DS 2.5 • Quantization error of the ADC 2.0 • Initial accuracy of the ADC (or calibration reference) 1.5 • Variation from channel to channel 1.0 25°C, 2.3A 0°C, 2.3A –20°C, 2.3A 0.5 0 0 20 40 60 CAPACITY (%) 80 100 DN471 F01 Figure 1. Discharge Characteristics of 3.3V Li-Ion Cell Although the popular LTC ®6802 Battery Stack Monitor offers high accuracy analog-to-digital conversion, some applications demand accuracy that is only attainable with a dedicated voltage reference IC. The LT®1461 is especially suited as a high performance calibration source, available in the small SO-8 package. Figure 2 illustrates this configuration. The calibration reference is measured with an ADC channel normally intended for temperature measurement. A programmable I/O bit controls power to the reference. 10/09/471 • Variation with temperature • Hysteresis effects, primarily that of the soldering process • Variation with operating time (long-term drift) The maximum specified error in the data sheet for the LTC6802IG-2 includes the first four items and is ±0.22%; about ±7mV when measuring 3.3V, the most demanding region of the discharge curve. The spec budgets ±3.3mV (±0.1%) as the maximum variation over the –40°C to 85°C operating temperature. Since the differential nonlinearity (DNL) of the ADC is about ±0.3 LSB, the quantization error contribution is about ±0.8 LSB, or ±1.2mV. L, LT, LTC, LTM, Linear Technology and the Linear logo are registered trademarks of Linear Technology Corporation. All other trademarks are the property of their respective owners. and normalizes all ADC readings with the same computed correction factor. The tolerance and drift of the reference and channel-to-channel variations are left uncorrected but the net uncertainty would be improved by almost a factor of 2, to ±6.2mV. Typical channel-to-channel variation is minimal, under ±1mV, leaving about ±1.5mV for trim resolution and accuracy in the IC manufacturing process. Thermal hysteresis is specified as 100ppm, and an additional approximately ±0.1% error may develop from the shift of the printedcircuit soldering process. A slightly more complex technique (method 2) involves storage of a single correction factor that accounts for the true reference voltage as measured with high accuracy test-fixture instrumentation. This then eliminates the initial error of the LT1461, improving the overall accuracy to ±4.1mV, nearly a 3× total improvement. Projected typical long-term drift is under 60ppm/√khr. If the practical vehicle battery system active life cycle is targeted at 5khr (about 15 years or 150,000 miles), an uncertainty of around ±0.5mV could develop. This is a relatively small contribution to total error. While small, there is still some channel-to-channel variation that can be calibrated out with a method that uses more initial test-fixture measurements (method 3). This is similar to method 2, but with high accuracy measurements of every channel taken (including the reference) and the saving of individual correction factors for each. This further reduces the error to ±3.1mV (almost a 4× total improvement). The LT1461AIS8-3.3 voltage reference IC has an output tolerance of ±0.04% and less than ±1.2mV of change over temperature with its exemplary 3ppm/°C worst-case stability. The LT1461 exhibits a long-term drift of under 60ppm/√kHr and thermal hysteresis of 75ppm. Solder reflow shift is expected to be under 250ppm (±0.8mV). Since a significant portion of the LTC6802 ADC error accumulates after the initial delivery of the IC, an external calibration technique improves accuracy in a finished product. Conclusion A precision voltage reference, such as the LT1461, can improve the accuracy of an LTC6802-based battery management system to about ±3mV worst-case. The reference is a simple addition to the highly integrated LTC6802 Li-Ion monitoring solution, thanks to the spare general-purpose ADC channels available. The low operating current of the LT1461 voltage reference also makes it ideal for this and other battery-powered applications. Examining Calibration Strategies There are a number of options to improve system accuracy, at the expense of additional complexity. With the simple circuit of Figure 2, several options are available that take advantage of the external calibration reference. Accuracy projections of several methods are tabulated in Table 1 and described below. References “Battery Stack Monitor Extends Life of Li-Ion Batteries in Hybrid Electric Vehicles,” Linear Technology Magazine, Volume 19, Number 1, March 2009, page 1. The simplest scheme (method 1) involves no local memory or measurements at production. This method takes readings of the nominal 3.300V calibration voltage periodically Table 1. Accuracy of Calibration Methods Described for 3.3V Measurements EXTERNAL CALIBRATION METHOD QUANTI- FACTORY SOLDERING CHANNEL THERMAL THERMAL (ALL TOLERANCES SHOWN IN ±mV) ZATION TRIM SHIFT MATCH VARIATION HYSTERESIS LONGTERM TOTAL DRIFT ERROR LTC6802 Without External Calibration 1.2 1.5 3.3 1.0 3.3 0.3 0.5 11.1 1: Calibration with LT1461, No Stored Information 1.2 1.3 0.8 1.0 1.2 0.2 0.5 6.2 2: Calibration with LT1461, Store Calibration Values for Reference Voltage 1.2 - - 1.0 1.2 0.2 0.5 4.1 3: Calibration with LT1461, Store Calibration Values for Reference Voltage and Each Input 1.2 - - - 1.2 0.2 0.5 3.1 Data Sheet Download www.linear.com For applications help, call (408) 432-1900, Ext. 2020 Linear Technology Corporation dn471f LT/TP 1009 116K • PRINTED IN THE USA FAX: (408) 434-0507 ● www.linear.com © LINEAR TECHNOLOGY CORPORATION 2009 1630 McCarthy Blvd., Milpitas, CA 95035-7417 (408) 432-1900 ●