Overcoming battery limitations to facilitate self-powered operation
Overcoming battery limitations to facilitate self-powered operation
Fast forward to a world with 1 trillion wirelessly connected devices in which pervasive computing impacts every aspect of our lives. Now imagine that each of those de vices operates on a battery that lasts an average of three years, which is very generous considering that most of today’s Internet of Things (IoT) devices have batteries with much shorter lives. In that world, we would be changing 1 billion batteries per day just to maintain the network of devices.
Setting aside for the moment the environmental impact of battery disposal at that scale, nobody wants to take on the battery maintenance problem.
Today, this is what limits the mass adoption of IoT solutions. It is why factories have not installed monitors on their 10,000 assets and why shipping companies do not embed real-time tracking in every package label. When you examine the power consumption of IoT devices over their lifetime, most of the energy is used for wireless communication; of that electricity, a large amount is spent on network synchronization rather than transmitting data.
This calls for better networking solutions to enable massive scales of devices and ultralow-power (ULP) radios to enable self-powered operation, eliminating the battery and, therefore, the maintenance problem.
Quantifying Receiver Performance
We focus on four main specifications for receiver performance: active power, sensitivity, data rate, and signal-to-interference ratio (SIR), also called adjacent channel rejection. These generally trade off with one another, but there is no one figure of merit that captures their relative impact across all types of receivers, frequencies, modulations, and so on. To make it easier to observe trends and tradeoffs, we concentrate on ULP receivers, which we will, somewhat arbitrarily, define as having an active power <100 µW. Active power is compared because you can always duty cycle a receiver to trade off the data rate with average power consumption. For example, if you turn off a receiver 50% of the time, the average power will be half the active power, and the average throughput will also be halved. In the limit, synchronization of the receiver and transmitter after they have been off for an extended period of time will add significant overhead and set a lower bound on the energy used for communication. It is worth noting that synchronization is more challenging as the number of devices scales, especially into the thousands. Sensitivity is a measure of the minimum required received signal strength to achieve a target bit error rate of, usually, 10−3 or a packet error rate of 10−2. It can be limited by the gain of the receive path, the type of detector used for demodulation, and the amount of noise added by the receiver. It typically trades off with active power, the data rate, and bandwidth, but, as we will see for some ULP receivers, this is not always the case. The data rate is often sacrificed for lower power and better sensitivity. For example, Bluetooth, Wi-Fi, and the narrowband IoT (NB-IoT) all support lower data rates in their standards via stronger error correction coding and data repetitions to extend their wireless range. Theoretically, the data rate trades off one to one with the received signal-to noise ratio for a fixed bit error rate, according to the Shannon channel capacity theorem . For this reason, we also compare the normalized sensitivity to a single data rate of 1 kb/s:
Snorm = S – 10log(data rate/1Kb/s).(1)
Finally, SIR has recently been considered in ULP receivers because the equipment often has energy detection receiver front ends that are known to be susceptible to interference. Especially considering deploying devices at massive scales and the increasingly crowded wireless spectrum, ULP receivers must be able to coexist with many different types of incumbent wireless signals.
Power Versus Sensitivity
We compiled a survey of ULP receivers published in top-tier circuits journals and conferences . Figure 1 shows the power-versus-sensitivity
(range) tradeoff for the 191 receivers published at the time of writing.
FIGURE 1: A survey of wireless receivers published in selected IEEE conferences and journals from 2005 to 2021 . dBm: decibels referenced to 1 mW.
With the exception of nanowatt receivers, an empirical line with a slope of –1 decade power per 20 dB of sensitivity bounds the performance, which can be interpreted as a constant figure of merit. Conveniently, this implies that receiver power and range scale together, assuming a path loss coefficient of 2; e.g., a 10× increase in power results in a 10× increase in range. In Figure 2, the sensitivity is normalized to a 1-kb/s data rate using (1), which reduces the spread in points, particularly for nanowatt receivers since they have relatively low data rates. These normalized points are compared to two groups with a constant figure of merit: 10× power/20-dB sensitivity and 10× power/10-dB sensitivity. The latter is more commonly used for energy detection front ends .
FIGURE 2: Sensitivity normalized using equation (1) and plotted with constant figure-of-merit lines for coherent and rectified-first receivers.
ULP Receiver Architectures
Several receiver architectures have been published in the literature; however, most ULP receivers leverage some variation of a passive
envelope detection radio-frequency (RF) front end, eliminating power-hungry RF blocks, such as low-noise amplifiers (LNAs) and RF local oscillators (LOs), as shown in Figure 3.
FIGURE 3: An energy detection receiver architecture with passive voltage boosting at RF and digital processing gain for improving sensitivity. Xform: transformer; ED: envelope detector; BB: baseband; Proc: processing.
Hybrid architectures have been demonstrated that, for example, add back an LNA for improved sensitivity and that include a passive mixer-first architecture incorporating an RF LO. The power of these RF components is >20 µW and often >100 µW; therefore, we are not considering them ULP. Exploring the architecture in Figure 3 further, passive transformers and matching networks are added in front of the envelope detector (ED) to reduce the noise bandwidth and improve the sensitivity by up to 20 dB, extending the wireless range . This passive voltage boosting performs better with a high RF ED input impedance, which is easier to achieve at lower frequencies; therefore, <10-nW receivers tend to be subgigahertz (Figure 4).
FIGURE 4: A comparison of power consumption and operating frequency.
However, ULP receivers at <100 µW have been demonstrated across a wide range of frequencies, up to millimeter-wave bands. Following the transformer, a passive envelope detector is used for down-conversion, which has a wide bandwidth; therefore, the amount of added noise can be high. This limits sensitivity to around –50 dB referenced to 1 mW. Baseband gain and filtering stages operate in the subthreshold, with a low bandwidth to keep the power minimal, resulting in a typical minimum detectable voltage in the 1–10-mV range. Finally, digital baseband processing typically consists of correlators to identify an on–off keying (OOK) wake-up sequence, cutting down on false detections and adding 5–15 dB of processing gain. Data rates for these receivers are less than 1 kb/s (Figure 5), limited by the speed and bandwidth of the subthreshold analog and digital baseband circuits.
FIGURE 5: ULP receivers tend to be practically limited to peak data rates of 1 Mb/s, and the data rate trades with the active power as expected.
Many ULP receivers suffer from poor performance in the presence of in-band interferers. This is highlighted in Figure 6, which plots the SIR for all 191 receiver publications.
FIGURE 6: Interference is a challenge for ULP receivers, with many not reporting a measured SIR. RFIC: IEEE Radio Frequency IC Symposium; JSSC: IEEE Journal of Solid-State Circuits; CW: continuous wave.
Note that, for ULP receivers, the SIR is either poor or not reported. The ED-first architecture is inherently susceptible to interference because of its wideband response. Recently reported ED-first receivers have addressed this with Manchester encoding  and two-tone modulation . Other solutions use a passive mixer-first approach to reduce the number of power-hungry RF components while adding some level of selectivity through down conversion and high-Q baseband filtering  and frequency hopping . Mixer-first solutions have demonstrated exceptional SIRs, with sub-milliwatt active power.
ULP Receiver Adoption
The modulation scheme plays an important role in the required specifications of a receiver and hence its power consumption.
Figure 7 shows that coherent communication [e.g., binary phase-shift keying, orthogonal frequency-division multiplexing (OFDM), and quadrature amplitude modulation] requires significantly higher power to demodulate.
FIGURE 7: ULP receivers exclusively use noncoherent architectures with modulation formats such as OOK, pulse-position modulation, and frequency-shift keying.
All modern wireless standards use some form of coherent modulation for better spectral efficiency. Noncoherent modulation, such as OOK, frequency-shift keying, and pulse position modulation, is used exclusively for ULP receivers. This creates a gap between standard-compliant radios and state-of-the-art ULP receivers. Wireless standards are being modified to incorporate ULP receivers as wake-up radios to reduce the energy spent on synchronization. The IEEE working group for Wi-Fi created the 802.11ba task group to investigate adding a wideband OOK message to the 802.11 base standard. The OOK message is embedded in a standard Wi-Fi packet and can be generated with existing transmitters only after a firmware update. It can be demodulated with a ULP companion receiver that has an active power of <100 µW , —more than 100× less power than a fully compliant Wi-Fi radio. A Bluetooth special interest group is also looking at adding a wake-up message to the next version of the standard. Hopefully, more details on this will become publicly available soon. The 3rd Generation Partnership Project introduced a wake-up message for the NB-IoT in release 15 of the cellular standard. The NB-IoT uses OFDM with 12 subcarriers and quaternary phase-shift keying modulation, which inherently is not low power to demodulate. In release 15, NB-IoT paging events are preceded by a wake-up signal, which is a unique correlation-based OFDM Zadoff–Chu sequence that somewhat simplifies receiver implementation, resulting in lower power but not yet ULP . The significant advantage of this is that it has the potential to be rolled out worldwide in all LTE cellular networks, with only software updates. These examples represent a shift in thinking inside wireless standard communities to address the power needs of the IoT, helping
to realize the adoption of trillions of self-powered devices.
- J. Proakis, Digital Communications, 5th ed. New York, NY, USA: McGraw-Hill, 2007.
- D. D. Wentzloff, “Low power radio survey,” Univ. of Michigan, Ann Arbor, MI, USA. Accessed: June 24, 2021. [Online]. Available: www.eecs.umich.edu/wics/low_power_radio_survey.html
- P. P. Wang and P. P. Mercier, “28.2 A 220μW–85dBm sensitivity BLE-compliant wake-up receiver achieving −60dB SIR via single-die multi-channel FBAR-based filtering and a 4-dimensional wake-up signature,” in Proc. IEEE Int. Solid-State Circuit Conf. (ISSCC), 2019, pp. 440–442.
- J. Moody et al., “Interference robust detector-first near-zero power wake-up receiver,” IEEE J. Solid-State Circuits, vol. 54, no. 8, pp. 2149–2162, 2019. doi: 10.1109/JSSC.2019.2912710.
- K. Huang et al., “A fully integrated 2.7µW–70.2dBm-sensitivity wake-up receiver with charge-domain analog frontend, −16.5dB-SIR, FEC and cryptographic checksum,” in Proc. IEEE Int. Solid-State Circuits Conf. (ISSCC), 2021, pp. 306–307.
- A. Dissanayake et al., “A multichannel, MEMS-less −99dBm 260nW bit-level duty cycled wakeup receiver,” in Proc. IEEE Symp. VLSI Circuits, 2020, pp. 1–2. doi:10.1109/VLSICircuits18222.2020.9162785.
- J. Im, H. Kim, and D. D. Wentzloff, “A 470µW −92.5dBm OOK/FSK receiver for IEEE 802.11 WiFi LP-WUR,” in Proc. IEEE 44th European Solid-State Circuits Conf. (ESSCIRC), 2018, pp. 302–305. doi: 10.1109/ESSCIRC.2018.8494331.
- A. Alghaihab, Y. Shi, J. Breiholz, H. Kim, B. H. Calhoun, and D. D. Wentzloff, “Enhanced interference rejection bluetooth lowenergy back-channel receiver with LO frequency hopping,” IEEE J. Solid-State Circuits, vol. 54, no. 7, pp. 2019–2027, July 2019. doi: 10.1109/JSSC.2019.2907160.
- R. Liu et al., “An 802.11ba 495μW -92.6dBmsensitivity blocker tolerant wake-up radio receiver fully integrated with wi-fi transceiver,” in Proc. IEEE Radio Freq. Integr. Circuits Symp. (RFIC), 2019, pp. 255–258.
- J. Im, H. Kim, and D. D. Wentzloff, “A 217µW −82dBm IEEE 802.11 wi-fi LP-WUR using a 3rd- harmonic passive mixer,” in Proc. IEEE Radio Freq. Integr. Circuits Symp. (RFIC), 2018, pp. 172–175.
- T. Odelberg et al., “A 2.1mW–109dBm NBIoT wake-up receiver,” in Proc. IEEE Radio Frequency Integrated Circuits Symp. (RFIC), 2021.
About The Authors
David D. Wentzloff (email@example.com) received his Ph.D. degree from the Massachusetts Institute of Technology, Cambridge, in 2007. He is a
professor of electrical engineering and computer science at the University of Michigan, Ann Arbor, Michigan, 48109, USA. In 2012, he cofounded
Everactive, a fabless Industrial Internet of Things company developing ultralow-power wireless systems on chip. His research interests include
radio-frequency ICs, with an emphasis on ultralow-power design.
Abdullah Alghaihab (firstname.lastname@example.org) received his Ph.D. degree in electrical engineering from the University of Michigan, Ann Arbor, in
2020. He is an assistant professor of electrical engineering at King Saud University, Riyadh, 14215, Saudi Arabia. His research interests include radio frequency/mixed-signal IC design for wireless communication and low power wireless transceivers.
Jaeho Im (email@example.com) received his PhD. degree in electrical engineering from the University of Michigan, Ann Arbor, in 2020, where
he was also a research fellow in electrical and computer engineering. He is with the Department of Military Science Technology, Republic of Korea Army Training and Doctrine Command, Seoul, 04064, South Korea. His research interests include radio-frequency/mixed-signal IC design for low-power
wireless transceivers and millimeter-wave transceivers and Internet of Things system architecture and design.
Omar Abdelatty (firstname.lastname@example.org) received his M.Sc. degree from Cairo University, Egypt, in 2015. He is pursuing his Ph.D. degree in electrical engineering at the University of Michigan, Ann Arbor, Michigan, 48109, USA. From 2012 to 2015, he was with the Cairo Circuits and Systems Laboratory, Cairo University. His research interests include the design of energy-efficient wireless charging and wireless connectivity circuits for
Internet of Things applications, self-powered wireless transceivers, low-power frequency synthesizers, and 5G millimeter-wave circuit design.
Trevor Odelberg (email@example.com) received his B.S. degree in electrical engineering from Purdue University, West Lafayette, Indiana, in
2017. He is pursuing his Ph.D. degree in electrical and computer engineering at the University of Michigan, Ann Arbor, Michigan, 48109, USA.
His research interests include radio-frequency ICs and systems for next-generation wireless technologies.