IoT Wireless Sensors and the Problem of Short Battery Life

Wireless sensors provide great insight in applications like monitoring environmental conditions or industrial plants and machinery. Because they are simple to install, they can be deployed in a multitude of situations. In coming years, we will see an explosion of new uses for wireless sensors as the “Internet of Things,” or “IoT,” is widely deployed. But one of the factors that most limits the use of wireless sensors is their limited ability to do the job for a reasonable amount of time. When a wireless sensor’s operation is fully dependent on a battery, and the battery is depleted, it becomes just a piece of junk.

If you are designing battery-operated wireless sensors, you face numerous challenges in ensuring your devices operate for a reasonable amount of time. The typical approach is to use energy for just the required activity, then put the device in low-power-use mode. The operation of a wireless sensor can be segmented in a series of activities, each one requiring a certain level of power for a certain amount of time. The most common activities:

• Waking up, taking a measurement and processing data into a message
• Powering up the RF power amplifier, transmitting the message, and powering the RF PA down again
• In bidirectional sensors (transmit and receive): waking up, powering up the receiver, receiving, processing data, acting on a message, and powering back down

It is easy to see that multiple actions play a role in discharging the battery.

The simplest way to increase the battery life is to use a bigger battery, a battery with higher capacity. Nevertheless, your customers are likely to expect their sensors to be small and to offer high performance (so they can send lots of data and have local intelligence/data crunching capability). Clearly, your customer expectations are diametrically opposed to the easiest way to solve the issue of short battery life.

How do engineers estimate battery life?

As a design engineer, you need to start making compromises and find the balance between battery size and the wireless sensor’s functionality to get the best performance from a small battery with a sufficiently long time interval between battery replacements.

The optimization process starts by understanding the energy requirements. Gathering data about energy usage is the first step to characterizing device performance.

A battery has a defined amount of energy, specified in Watt hours (Wh) and capacity, specified in amp hours (Ah). If you know how much power is required to operate your device, you can calculate the battery life.

Battery life (hours) = Battery capacity (Wh) / Average power drain (W)

The battery’s energy is also the product of its voltage rating (V) and capacity (Ah). The voltage rating is a midpoint value on the battery’s discharge curve empirically determined to correctly relate the battery’s energy and capacity. Based on this, battery life can also be determined by the formula:

Battery life (hours) = Battery capacity (Ah) / Average current drain (A)

However, when the device is in real operation, the battery life is typically shorter than the number you calculated. The most common comment is: “the battery quality is bad.” Representatives for big battery brands will offer detailed specifications and explain that among batteries of the same type, it is common to have capacity variations of 5 to 10 percent. But even using conservative battery capacity estimates, battery life typically falls short. The device dies before it is expected to. Why does this happen? Did we correctly estimate energy usage? Probably not. Let’s explore the problem.

The complexity of measuring dynamic current drain

In battery-powered devices like wireless sensors, to save energy the device sub-circuits are active only when required. Engineers design the device to spend most of its time in a sleep mode with minimum current drain. During sleep mode, only the real-time clock operates. The unit then wakes up periodically to perform measurements. The acquired data is then transmitted to a receiving node.

Figure 1: Current levels during the three main states of a wireless sensor

Table 1

The different operating modes result in a current drain that spans a wide dynamic range from sub-µA to 100 mA, which is a ratio on the order of 1:1,000,000.

Traditional measurement techniques and their limitations

A well-known method for measuring current is to use the ammeter function of a DMM. The accuracy of current measurements made with modern digital DMMs looks good, but specifications are defined for fixed ranges and relatively static signal levels, which isn’t exactly the situation on a wireless sensor due to its dynamic current drain.

The DMM is connected in series between battery and device to measure the current. From time to time we see some reading instabilities due to the sensor’s active cycle or even the transmit mode. We know that DMMs have multiple ranges, and with auto range it should be able to select the most appropriate range and give the best accuracy. However, DMMs aren’t ideal. The auto range takes time to change range and settle the measurement results. Time to auto-range is often 10 to 100 ms, longer than transmission or active modes times. For this reason, the auto-range function needs to be disabled and the user needs to manually choose the most appropriate range.

The DMM makes measurements by inserting a shunt in the circuit and measuring the voltage drop across it. Normally to measure low current, you choose a low range based on a shunt with high resistance; to measure high current you choose a high range based on a low-resistance shunt. The voltage drop is also called burden voltage. Due to this voltage drop, not all the battery voltage reaches the wireless sensor. Most accurate low ranges for sleep current measurements have burden voltage during current peaks that may even cause the device to reset. Practically, we end up compromising and using a high current range that keeps the device operating during current peaks. This compromise enables us to handle peak current and measure the sleep current, but at a high price. As the offset error is specified on range full scale, it heavily impacts measurements on low current levels. Its error contribution can be 0.005% error on 100 mA range = 5 µA, which is a 50% error on 10 µA or 500% error on a 1-µA current level. This current level is where the device spends most of its time, so this error has a huge impact on the battery life estimation.

After measuring the sensor’s low current level during sleep mode, we have to measure the active and transmission pulses. Measurements need to include both the current level and the time the sensor spends at that level. Oscilloscopes are excellent tools for measuring signals changing over time. However, we need to measure current in the 10’s of mA level, and current probes do not do a good job there due to their limited sensitivity and their drift. Good clamp probes have 2.5-mArms noise, and the zero compensation procedure needs to be repeated often. Current probes measure the electric field over a wire, so the trick to increase sensitivity is to pass the same wire multiple times so we multiply the magnetic field – this multiplies the current readout, enabling us to measure the current a bit better. With this approach, we can capture the current pulse of the activity and the transmission time. Even within the activity and transmission, the current changes levels: it looks like a train of high and low levels. To properly calculate the average current the waveform needs to be exported and all the measured points need to be integrated to get the average value.

Oscilloscopes do a good job of capturing a single burst. However, the measurements are more complex if we want to verify how many times the sensor activates in a timeframe and how often it sends out a TX burst. Oscilloscopes can easily do a good job with measurements taken over the short term, but sensors may have operational cycles of minutes or hours, which can be complex to capture and measure.

Measurement innovations

The Keysight N6781A source/measure unit (SMU) for battery drain analysis overcomes the limitations of traditional measurements with two innovations: seamless current ranging and long-term gap-free data logging. The SMU is a module that can be used with the Keysight N6700 low-profile modular power system or N6705 DC power analyzer.

The seamless current ranging is a patented technology that enables the SMU to change the measurement range while keeping the output voltage stable without any dropout due to ranging. This feature enables you to measure the peaks with high current ranges and measure the sleep current with the 1-mA FS range, which has 100 nA of offset error. This low offset error (100-nA offset error is 10% at 1 µA or 1% at 10 µA), orders of magnitude better than a traditional DMM.

Figure 2: The Keysight N6781A SMU allows accurate measurements across dynamic current levels.

The seamless current ranging is combined with two digitizers to measure voltage and current with simultaneous sampling at 200 kSa/s (5-µs time resolution). Digitized measurements can be captured over 2 seconds and displayed with full time resolution and proportionally longer time with lower resolution. However, for long-term measurements, the internal data logger in the Keysight N6705B modular DC power analyzer integrates the 200-kSa/s measurements over a user-specified integration period (20 µs to 60 seconds) without losing any samples between the integration periods. As the data logger is gap-free, all the samples fall in one integration period or in the next one — no samples are lost. With the data logger, engineers can now measure the current and energy drain performance of a wireless sensor for up to 1000 hours of operation.

Figure 3: Data logger: all the samples are integrated in consecutive sample periods. No samples are lost. For every sample period, min and max values are also available.

Figure 4: Recorded current drain over 200 seconds of operation provides new insight into a device’s dynamic current drain.

Measuring the sleep current is just a matter of placing the markers and directly reading out the values provided. The measurement in Figure 4 is made with a single acquisition over a long period of time; we get the complete picture of the current drain as well as an accurate measurement of the sleep current at 599 nA.

With pan and zoom capability, it’s possible to look at the current level and time spent at every power level. Details that traditional measurement tools do not see can now be identified and measured. A clear example is the trailing pulses identified by “???” in Figure 4. The software revealed this surprise: the device drain polsed energy at ~90 µA peaks for 500 ms for an average current of 3.3 µA. When we add this current drain to the 599 nA sleep current, it moves to 730 nA, 22% higher current than we expected. This type of surprise can be one of the reasons for underestimating energy requirements and delivering a shorter battery life than anticipated.

In wireless sensor power optimization, engineers get great value by understanding the details. Knowing how much energy it takes to send out a single packet of information is very important when balancing user experience against battery drain and answering questions such as “should I send information once every second, every 5 seconds or every 10 seconds?” Engineers can accurately estimate the battery drain impact of any firmware change and validate it in a reasonable time with real measurements.

Joule measurements made easy

Joules are useful in battery life estimation, as every activity has a defined amount of energy. We can also compare device performance using Joules/transmitted bits. But engineers rarely use Joules because they need to be calculated from voltage, current and time.

Figure 5: Using Keysight 14585A software, you can measure energy directly in Joules.

With the Keysight 14585A control and analysis software, energy in Joules can be measured directly. For example, you might measure the energy consumed by transmitting a packet (see Fig.5) captured with a triggered measurement. This is one benefit of having two digitizers for voltage and current with simultaneous sampling that enable point-by-point power measurements. Joules can be easily read out as a value between the markers, and designers can go a step further by defining Joules/transmitted bit.

Summary
Engineers who design IoT battery-powered devices use advanced power management techniques to conserve battery life. Traditional measurement techniques are complex, time consuming and don’t deliver the measurement accuracy required to optimize and validate battery drain, and often this causes engineers to underestimate the power required to operate the device. The Keysight SMUs for battery drain analysis enable accurate current drain analysis with one picture that provides a complete and detailed current and energy drain analysis. Post-analysis software simplifies the engineer’s job by offering visibility into details never seen before.

With Keysight’s latest introduction of the N6785A SMUs for battery drain, these capabilities are now available up to 80 W and from nA to 8 A. The new SMUs are used in multiple applications from smartphone and tablet testing to automotive ECU and IoT wireless sensors and chipsets. www.keysight.com/find/N6781A-EU