Gas Gauging: Art or Science?
Garry Elder, Systems Manager, Texas Instruments
As battery "gas gauge" integrated circuit (IC) technology for portable applications has evolved over the past 10 years, the level of sophistication continues to rapidly increase, and calculating battery performance or accuracy remains a highly complex task. The ever-increasing variety of cell types makes it difficult to apply a single scientific formula to determine an accurate state of charge (SOC) in a wide range of portable applications " from cell phones to notebook computers. Accurate calculation of SOC and run-time data in portable Li-Ion batteries traditionally measures usage stimuli, charge/discharge activity and temperature factors in relation to voltage levels, coulomb change and number of cycles. This gas gauge methodology requires interpretation of how the cells react between the initiation of actions and the corresponding results of those actions to provide SOC and runtime data. In addition, other methods might analyze data by modeling the battery. However, each method has differing elements and resolutions. Each element introduces possible errors into the equation, and the inaccurate resolution ultimately affects the battery's potential long-term maximum performance.
Typical models used in today's gas gauge ICs consist of rate, temperature, Δcoulombs and voltage factors to produce a mAh delta value from the last ΔSOC calculation or an actual SOC. The key issue, in addition to the model's accuracy, is that the model is very inflexible, if flexible at all, once the battery leaves the production floor. The modeling technique, although based on scientific principles and analytical data, is still an "art form" when applied to consumer portable applications. A battery's initial SOC setting, which was first determined from the original model's "pre-use" properties, and the battery's actual SOC after a few months of regular use can be quite different. As a result, consumer battery pack designers attempt to implement more sophisticated static model techniques and analytical data acquisition systems, while increasing power consumption, solution cost and development cost as highlighted in Figure 1.
Figure 1: The Proportional Increase and Decrease of Key Budgets to improve current static modeling techniques vs. dynamic Modeling Technology
To bring the "art" of gas gauging back into the science realm, while driving consumer product design in the right direction, the engineer needs a way to determine SOC, maximum available charge and run time data over the entire life of a battery. This article illustrates a recently introduced method that benefits consumer portable Li-Ion batteries using Texas Instruments' patented Impedance Track technology -- bringing science rather than art to the forefront of gas gauging technology.
Where are we with modeling today?
Today there are several methods to modeling that can be mixed-and-matched for the benefit of different applications, cell types, applications and usage environments. The two main categories for types of modeling used in gas gauging today are a) arithmetic and b) tabular. Each has associated factors such as discharge rate (ΔQ/Δt), temperature, voltage and lifetime ΔSOC usage of the battery to form the model to produce a mAh value of SOC.
Both methods require typical cell and/or battery characterization data which is not generally available from the cell suppliers. This analytical data is required to be generated to optimize the performance of the gas gauge model, ensuring the longest run-time of the battery-powered appliance without the risk of loss of data or uncontrolled system shutdown. This data, needed for improved model accuracy, can be quite extensive, taking between five and 50 cycles (or one to 14 days), over a range of discharge/charge rates and temperatures of the cells. This raw data has to then be analyzed to produce the appropriate table(s), co-efficients and other key information to support the models.
The type, resolution, desired accuracy and range of the modeling method chosen drives the gas gauging hardware requirements for performance range, accuracy, stability, memory density, computing speed and several other smaller factors. This in turn drives power consumption of the battery electronics and the overall cost of development and production including the electronics. The fundamental measurement accuracy of the current, time, voltage and temperature parameters needed depends on the model used.
|| Acceptable Error
||2 V to 4.5 V (per cell)
||< ±0.5%="" or="" 3="" ma="">
||10 mA to 8 A
||250 ms to 1 hr
||0°C to +70°C
||-20±C to +85±C
Table 1: Accuracy Levels of Parameter Measurement
The error factors above are only the fundamental parameter measurements. The total-error is a product of both parameter measurement accuracy and model accuracy. Model accuracy is affected by several factors. One of the major factors is battery degradation or wear out, sometimes called 'aging'. However, the degradation rate can vary with time, so aging is a loose term for degradation. Battery degradation can be affected by the cumulative amount of energy exchange, rate-of-energy exchange, temperature and voltage. Combinations of these factors can further increase the degradation rate.
The overall error of the model in addition to the parameter accuracy is hard to determine precisely due to the variances in battery degradation, model types, model variables and model resolution. This is where today's techniques for modeling is an art rather than a science. That's because these variances are often predicted and incorporated into the model compensations based on experience and gathered data.
The Dynamic Modeling Revolution
To improve gas gauging accuracy, there are several paths that can be taken. For several years now, hardware platforms suitable for portable applications have been available at reasonable cost providing high measurement performance and low quiescent power consumption. Further improvements down this path are becoming limited due to development cost versus solution cost versus technology tradeoffs.
The next major path to bring a significant improvement in performance is dynamic model solutions. To this end, the measurement of impedance of the battery is used in Impedance Track technology. This is common in state of health (SOH) systems that are used in advance power systems (APS), un-interruptible power supply (UPS) or other static batteries. However, these applications use very sophisticated, and hence expensive, systems. The true challenge is bringing this technology to the consumer portable application space in a cost-effective manner and Impedance Track technology has achieved that.
The use of the impedance of the battery is the most accurate passive measurement to determine the actual SOC. However, this does bring further challenges than just measuring impedance as it varies based on charge/discharge level, temperature and cycling.
Figure 2: Impedance Variation Due to the Battery's Cycling
When impedance variations of the battery can be predicted, then the accuracy of the gas gauging can be increased according to an understanding of the relationship of impedance to the actual maximum chemical capacity (QMAX) of the battery at that time, the cell voltage and the depth of discharge (DOD). This can then be compensated for discharge/charge rate and temperature to provide very accurate SOC value or runtime predications. Figure 3 shows the DOD versus cell voltage for four of the leading cell suppliers of Li-Ion cells. Note they are very similar except at the end of discharge; this is the area of most concern to the user and is where the static modeling techniques focus.
Figure 3: DOD vs. Voltage for a Range of Cell Suppliers
Today's models are not updated post battery production, so just adding the use of impedance is one-step forward to improving the gauging system. This is a significant advance but it is not a revolution. The revolution is completed by making the model adaptable to its application throughout the battery's existence. The impedance compensations are stored in tables that are easily created during the development stage of the battery production and are automatically updated as the gas gauge learns new information throughout its life.
The development of the initial tables is controlled by the model itself, with a single discharge cycle from full at a rate of between C/2 and C/5 at a nominal temperature. The table can then be used in all solutions for the same battery configuration, knowing that each time the battery is used in the field, its usage model and impedance affects will be updated.
Gathering suitable data is usually an engineering-intensive task, from both the skill level of the person involved and the equipment needed. The comparisons between data gathering and the measurement of the initial Full Charge Capacity and Remaining Capacity are shown in Table 2.
|| Today's Static Models
|| Dynamic Models
| Data gathering cycle
|| Typical application discharge rate (~C/2) @ typical application temp
Max discharge rate (~C) @ maximum operating temperature
|| ΔV/Δt < 1μv/s="" then="" discharge="" at="" between="" c/2="" and="" c/5="" at="" room="" temperature="" then="" δv/δt="">< 1μv/s="">
| Typical number of cycles
| Total cycling time
||~2.5 to 5.5 hrs
| Equipment used
|| Controlled Load and Power supply. Temperature chamber. Data logging PC.
|| Suitable fixed load (can be P rather than C if desired)
| Engineering skill level to perform the cycling
|| Medium: Needs to setup the appropriate equipment to run the variety of tests correctly to keep the cycling time down
|| Low: Needs a load attached to a full battery after a rest, and then once complete, waiting for the battery to rest
| Engineering skill level to perform data analysis
|| High: Cycling data requires formatting and the use of data analysis tools such as Microsoft Excel or MathCAD or data returned to the Gas Gauge maker for analysis
|| None: Model handles the data analysis and storage internally
Table 2: Battery Development Comparison of Static Model Technology vs. Dynamic Model Technology
Once the QMAX of the battery is known, then the compensated maximum usable capacity can be determined. The Smart Battery Data standard determines that full charge capacity (FCC) is the maximum capacity rated at C/5 or P/5. With QMAX available, then SOC can be determined by measurement of the voltage and temperature of the battery, similar to the Tabular model method.
As cell-to-cell variations exist, the data collected during the development analysis needs to ensure that the variations are accounted for or can be cancelled out through battery degradation compensation. However, due to the model updating nature of dynamic modeling, once the model is updated, a higher level of SOC accuracy is achieved.
|| Dynamic Model
| Setting Initial SOC
|| Charge to full and then discharge to empty
|| Not required. Tables contain the data based on voltage of the battery
|| Not required. Tables contain the data based on impedance of the battery
| Model updating actions
|| ΔV/Δt < 1="" μv/s="" then="" δsoc="" of="" 5%="" then="" δv/δt="">< 1="" μv/s="">
Table 3: Static Model Technology vs. Dynamic Model Technology
The benefits of this new modeling technique can be categorized in three areas:
Battery Development Phase
The ease of development of the battery electronics, and especially the gathering of data to tune the models to the nature of the cell and application, has been reduced by over two-thirds. The equipment needed has been vastly simplified and the required engineering skill level needed to perform this task has been significantly reduced.
Battery Production Phase
One of the most time and cost intensive elements to the production of a smart battery pack is the learning or capacity calibration of the electronics to the actual cells. With dynamic modeling technology, this phase is now not required. Even for those battery makers who choose an alternate route, the Impedance Track technology enables a very high level of accuracy with no effort, equipment, time or resources associated with a learning cycle.
The top care about to a portable system user is end-of-discharge inaccuracy to avoid an unexpected shutdown of the system risking loss of data or loss of transmission of critical data. With dynamic modeling technology, the reported versus actual state of charge always converges as the battery is discharged, so the reporting accuracy becomes more critical at low capacity states the reported state of charge is also becoming more accurate.
An additional benefit of this convergence and its accuracy allows a 'Reserve Capacity' feature to be enabled. This is a minimum-required capacity past where the reported RemainingCapacity( ) is zero is stored to ensure a smooth and risk-free shutdown even if the shutdown or save-to-disk activities are triggered at RemainingCapacity( ) = 0.
Figure 4: Convergence of Reported vs. Actual Capacity including Reserved Capacity
This data indicates the level of accuracy improvement over today's static model technology and the level of accuracy attainable with dynamic modeling technology. These accuracy benefits provide opportunities for increased run time, especially as the battery degrades.
Overall, today's hardware technology for gas gauging has reached a cost/performance plateau, with modeling techniques and accuracy lagging behind. The introduction of the dynamic modeling used in Impedance Track technology puts the development ball back in the hardware court as the modeling has not only caught up but is pushing the limits of the hardware technology with great benefits to the battery developer, battery producer and battery user.
Garry Elder is a Systems Manager at Texas Instruments. He received his Bachelor of Engineering in Electronic and Computer Engineering with honors from Bolton Institute. Garry has more than ten years experience in the battery management field. He started at Benchmarq and is now with TI.