Estimating Battery Life - A Crucial Attribute in Embedded Product Success
09 October 2023
Figure 1: The Qoitech Otii Arc Pro is a compact and portable unit (Source: Qoitech)
Battery-powered devices are becoming increasingly ubiquitous. Powering an embedded system from a primary or rechargeable battery offers heightened design flexibility, user convenience and application portability.
With continued development in Internet of Things (IoT) and Industrial Internet of Things (IIoT), plus their combination with machine learning (ML) at the edge computing trends, we expect more and more functionality from battery-powered devices. The burning question is, how long will battery reserves last?
Changing batteries is inconvenient for consumers and will prove very expensive for large-scale industrial IoT deployments. It is therefore paramount to prolong battery life through dynamic energy consumption profiling, using this data to select the most suitable battery chemistry and accurately forecast the battery lifespan. In the following text, the techniques and tools that are now available to accurately estimate an embedded design's battery capacity will be outlined.
Battery life has become an essential aspect of modern living. From plugging in your electric vehicle or scooter to ensuring your smartwatch or fitness tracker will last the duration of an exercise class, nobody likes a device that has to be kept charging throughout the day. It's the same for industrial and commercial battery-powered hardware, like IoT sensors. A fitness tracker running out of charge halfway through training is inconvenient, but for an industrial sensor, it could have far more serious consequences - halting production and resulting in considerable financial costs being accrued. Overall, battery life is directly linked to a product's success and its brand credentials. Increasingly, having a low-power mindset is becoming a necessity for every embedded developer. Acquiring the knowledge to implement hardware and software techniques that lower the average power consumption is now crucial. Also, prolonging battery life is important from a sustainability perspective - something that many organisations are very aware of in terms of their corporate image. A recent EU report, entitled ‘European Infrastructure Powering the Internet of Things’, illustrated the extent of the challenge, noting that by 2025 up to 78 million batteries will be disposed of globally on a daily basis.
The low-power mindset
Acquiring a low-power mindset requires embedded developers to approach their applications holistically. There may not be any quick fixes to lowering the average power consumption profile to prolong battery life. Typically, this is achieved through several iterative steps, each contributing a slight improvement. The following points highlight some critical topics embedded developers should investigate. These are:
• Identifying power consumption - What circuit functions are consuming energy, plus what is the relative timing or phasing with respect to other parts? Such questions should start the investigation process and are at the heart of a low-power mindset. The host microcontroller will have a significant influence on the device's power consumption profile and usually manages the power to attached peripherals. What sleep modes does the microcontroller have, and how may they be used? There is always a delicate balance between deep sleep modes and application responsiveness. In consumer applications, for instance, slow wake-up could result in user frustration.
• Managing the application duty cycle - How frequently does the application need to take a reading or instigate an action? For a smart thermostat, measuring the temperature every 30s may be more than sufficient. Other more complex sensors might require a faster mode of operation. How often the battery-powered device is active (and consuming higher current levels) rather than sleeping will significantly impact the average power consumption profile. Shutting down peripheral functions and sensors while the microcontroller processes data also lowers the consumption profile.
• Scheduling tasks to avoid current peaks - With the information gathered from the points discussed above, it may be possible to change the scheduling of software/hardware tasks to prevent current peaks. Not only do current peaks impact on the average power consumption, but also the battery's performance. High currents may lower the battery's state of charge quicker, thus reducing battery life.
• Selecting an energy source - The costs of replacing primary cell batteries limit their usefulness for most applications. Most battery-powered applications make use of rechargeable batteries that can be recharged in-situ. Energy harvesting techniques are increasingly popular for recharging batteries from ambient energy sources (such as solar, wind, vibration and RF). Also, for some applications where the average current profile is relatively low, replacing the battery with a supercapacitor or ultracapacitor offers a viable approach.
• Selecting the right battery - It is important to investigate the battery's characteristics besides the embedded development factors highlighted previously. Batteries come in all shapes and sizes and rely on different chemistries, so checking the key parameters provided within their datasheets is vital. It should be ensured that the battery's ideal discharge profile suits the application's duty cycle and consumption attributes. However, matching the battery to the application is challenging. Any embedded system has a more dynamic consumption profile than a static discharge curve. Consequently, a more informed approach is typically needed.
• Measuring current consumption - Perhaps the most vital aspect of profiling the power consumption of an embedded system is current measurement. A digital multimeter is a practical measurement instrument, but in this context a more dynamic response, measurement range and resolution accuracy will typically be needed. For example, the current of a battery-powered, wireless-connected IoT sensor may range from a fraction of a µA in sleep to tens of mA while in operation, with a dynamic range of 50:1. Such highly dynamic characteristics dictate the use of specialist test equipment.
Figure 2: UART debug messages from the DUT are displayed alongside real-time current measurements with the Otii Pro software (Source: Qoitech)
Profiling embedded system current consumption
The Qoitech Otii Arc Pro and Otii Ace Pro range of power profiling units are specifically designed to measure the power consumption of battery-powered embedded systems and work with the Otii Pro power analyser software. The Qoitech Otii Arc Pro (see Figure 1) features a power profiler, source measurement unit, DC energy analyser and power supply. The combined hardware and software set-up can measure, analyse and record real-time current measurements with 5nA resolution accuracy at a 4kS/s sampling rate.
The Otii Arc Pro power supply outputs from 0.5VDC to 5VDC up to a maximum current of 5A, and the measurement process does not impose any burden voltage on the load. Integration into the embedded system under test is achieved using Otii's UART interface or GPIO pins. Developer debug messages sent from the device under test (DUT) are synchronised with real-time current measurements and displayed in the debug screen area (as illustrated in Figure 2). Engineers can accurately display breakpoint and watchpoint level consumption analysis with this approach.
The Qoitech Otii Ace Pro provides similar features to the Arc Pro, but increases the maximum output voltage to 25VDC in 1mV steps and provides a current measurement range from nA to 5A. Measurement resolution is within 0.4nA margins and the sampling rate is configurable up to 50kS/s.
Battery characterisation and estimation
Earlier in this article, we highlighted the importance of understanding the battery's discharge characteristics and how this is challenging to achieve for any dynamic embedded system. An optional addition to the Qoitech Otii software solution is the Otii Battery Toolbox.
As embedded systems become more complex, with increasing commercial and environmental pressures to prolong battery life, it’s necessary to take a more scientific angle when selecting a suitable battery. Rather than simply starting with a battery capacity and dividing it by the average consumption to estimate the likely battery life, sophisticated tools like the ones just mentioned are enabling a more objective strategy. The Otii Battery Toolbox has 3 features for evaluating battery life based on the discharge profile. It adds more functionality beyond the basic battery estimator provided within the standard Otii software.
As a battery discharges, its output voltage reduces (with the exact amount of voltage reduction depending on its chemistry). Likewise, the embedded system will continue to operate down to a point where brownouts disrupt system operation.
Figure 3: Initial steps of the battery profiler in operation, highlighting the dynamic nature of the DUT current consumption across 4 different supply voltages (Source: Qoitech)
The first stage of determining the battery life with the Otii Battery Toolbox involves analysing device performance over a range of battery voltages, starting at the highest battery voltage down to the lowest voltage from which the device will operate without error. The finer the measurements, the more accurate the results. The Toolbox software can reduce the Otii’s output in, say, 100mV steps between the defined battery limits. The initial battery choice will guide the range of voltages selected, but it may differ from the battery type once the results have been completed. For example, a LiPo battery has an output voltage of 4.2VDC when fully charged to 3.0VDC, beyond which it should not be used.
Figure 3 illustrates the Toolbox profile of a device operating over a range of supply voltages, from 3.7VDC to 4.5VDC in 200/300mV steps. The upper-right corner of the image displays the minimum, peak and average current consumptions for each voltage range. The battery profiler options, at the bottom centre of the screen, allow the entry of the 2 discharge current and duration values - high for active mode and low for sleep/inactive mode. The cut-off voltage and/or the number of iterations are also set.
In the next stage of the profiling operation, the DUT is removed from the Otii unit, and the first battery candidate is plugged in. The profiling software switches between the high and low values for the duration of the test until either the battery voltage reaches the cut-off voltage, or the limit of repetitions is reached. By storing the battery profile, it can be used by the Otii to emulate the battery. This way, the Otii output emulates the profiled battery output voltage and internal resistance to facilitate further testing. This process takes the guesswork out of selecting a suitable battery, yielding a far more realistic and representative picture of a device battery's actual performance.
Real-world battery profiling
With an increasing focus on battery-powered devices across a broad cross section of industry sectors, prolonging battery life is of paramount importance to contemporary embedded system implementations. The need for a low-power mindset is key, along with the support of next generation effective tools.
Contact Details and Archive...