Sridharan Vengadaraman, marketing brand manager – EMEAI, Keysight Technologies, talks about how better measurements enable systemic optimisations of IoT ecosystems
Forecasters suggest there will be 50 billion Internet of Things (IoT) devices by 2020. What we’re not told is how many people will be employed to change the batteries in all these ‘things’. While the logistics experts work on that issue, IoT device designers are doing their bit by trying to cut the power consumption of their designs.
The typical approach to conserving battery energy is to segment the operation of an IoT device into discrete activities, each of which requires a specific amount of power for a specified duration. When the device is inactive, it is put into a lower-power mode.
Achieving the right balance between battery size and device functionality takes a deeper knowledge of battery life and current drain than traditional measurement techniques can provide. Why is this?
Estimated versus actual battery life
A battery stores an amount of energy specified in Watt hours (Wh), and has a current delivery capacity specified in amp hours (Ah). If you know how much power is required to operate your device, you can estimate the battery life in two ways.
The first is to divide the battery capacity (in Wh) by the average power usage (in W). The battery’s energy storage capacity is also the product of its voltage rating (V) and current delivery capacity (Ah). The voltage rating is a midpoint value on the battery’s discharge curve, which relates the battery’s energy to its current delivery capacity. This means that the second way to estimate the battery life is by dividing its current delivery capacity (Ah) by the average current drain (A).
Measuring dynamic current drain
Designers must characterise the energy requirements of a device in each operating mode, and the current drain and duration of each mode. This enables them to make systemic trade-offs, because once they know, for example, how much energy it takes to transmit a packet of information they can decide whether to send a packet once a second or once a minute to achieve an acceptable user experience.
Part of the measurement challenge is that the currents drawn in an IoT device can vary from sub-μA during sleep mode to 100 mA during transmit mode, a ratio of 1:1,000,000.
Measuring current drain with a digital multimeter
You can measure current drain using a digital multimeter (DMM), but this approach is limited since DMMs work best at fixed ranges on relatively static signal levels. The dynamic current drain of an IoT device can lead to unstable readings on the DMM as the device switches between modes. Even auto-ranging DMMs take time, perhaps 10 ms to 100 ms, to change range and settle their measurement results, which may be longer than the duration of an active mode. If you use a DMM, therefore, it makes sense to set the range yourself.
A DMM is also limited in that it makes measurements by inserting a shunt in the circuit and measuring the voltage drop across it. To measure low currents, you choose a low range based on a shunt with a high resistance. To measure high currents, you choose a high range based on a low-resistance shunt. However, the voltage drop across the shunt means that the IoT device isn’t supplied at the full voltage of the battery. This means that in sleep current measurements, it is possible for the battery voltage to drop so low during current peaks that the device resets.
The solution is to use a high current range that keeps the device operating during current peaks. This lets you handle peak currents and measure sleep currents, but at a price: as the offset error is specified at the full scale of each range, it can have a large impact on measurements on small currents.
Another approach to low-current measurements
One response to this issue is Keysight’s N6781A source/measure unit (SMU) for battery-drain analysis, for use with the Keysight N6700 low-profile modular power system (for ATE) or N6705 DC power analyser (for R&D).
It offers seamless current ranging, so it can change its measurement range while keeping the output voltage stable. It measures peak currents with high-current ranges and sleep currents with a 1 mA full-scale range, which has 100 nA of offset error. This offset error represents 10 per cent of a 1 μA signal or 1 per cent of a 10 μA signal, much less than for a DMM.
Measuring active and transmission pulses with an oscilloscope
Along with sleep-state measurements, designers need to measure the current draw and duration of the active modes of an IoT device. Oscilloscopes are good at measuring signals that change over time, but the active modes of an IoT device may draw tens of milliamps. Handling this means using clamp probes, but these suffer from around 2.5 mArms of noise and need frequent zero compensation. Current probes do not do a good job of measuring these levels due to their limited sensitivity and drift.
Since current probes measure the electric field of a wire, their sensitivity can be increased by passing the same wire through the probe many times to strengthen the field. This enables users to capture the current draw of an IoT device when active, although because the current switches between high and low levels during this phase, calculating the true average power means exporting the waveform and integrating its measurements.
Using a device current waveform analyser
Keysight’s CX3300 Device Current Waveform Analyser has a low-noise design and ultra-wideband current sensing so users can visualise previously un-measurable current waveforms. It supports current sensors that can range from 100 pA to 10 A, with a 1 GSa/s sampling rate, 200 MHz bandwidth, 14/16 bit dynamic range and 256 Mpts memory depth on a two- or four-channel model.
IoT devices are becoming increasingly complex, and so are having to work harder to manage their energy use by implementing power-management strategies such as sleep modes. Truly understanding a device’s current profile therefore means analysing it in the nA range.
Designers have used oscilloscopes with current probes for this, as discussed, but a dedicated instrument, such as the CX3300, can offer the following characteristics:
- High sensitivity to measure very low current, for example, in sleep mode
- High dynamic range to measure transitions between sleep and active modes
- Wide frequency range, starting from DC
- Accurate measurement without the core saturation effect of a current probe
- True visualisation of the current waveform, not an averaged current
Forecasting the battery life of an IoT device is more challenging than it looks. Making accurate measurements of currents that vary by 1,000,000:1 is a complex but necessary part of doing so. Tools such as DMMs, scopes and dedicated waveform analysers can help make these measurements, so long as their shortcomings are recognised and accounted for. Dedicated measurement tools can ease the process of predicting battery life, and also offer insights, such as Joule measurements, that can aid in architectural decision-making and systemic trade-offs.