Wi-Fi HaLow for Long-Range and Low- Power Internet of Things: System on Chip Development and Performance Evaluation

Introduction

IEEE 802.11AH System Overview System Overview

For physical (PHY) layer transmission technology, IEEE 802.11ah, also referred to as WiFi HaLow, employs multiple-input multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM), which is suitable for ultra-high data rates and supports multiple bandwidths and streams to increase the data rate up to 347 Mb/s depending on the application. Furthermore, IEEE 802.11ah can provide wide-area WLAN services based on its propagation characteristics, which can extend the service coverage of the frequency band below 1 GHz [9]. Additionally, the coverage can be further extended using narrow bandwidth transmission modes and coding gain through the symbol repetition method. To improve transmission and energy efficiency, system specifications for the medium access control (MAC) layer have been designed to provide efficient packet structure, beacon signal, and protocol for IoT applications that demand long-range and low-power communications in wide area networks [10]. This structurally optimized energy-efficient specification has attracted attention as an alternative to Bluetooth. The IEEE 802.11ah standard specifies the minimum performance requirements for service coverage extending up to 1 km and for data rates of 100 kb/s or more. Thus, extended service coverage of up to 1-2 km can be supported compared to existing WLAN technologies, which have much lower coverage, and the network capacity can accommodate more than 8000 devices. Moreover, power-saving features potentially increase battery life up to 5–10 years. This enables IEEE 802.11ah to be used in applications such as smart grids, sensor networks, and machine-to-machine (M2M) communications that require low-power operations, as well as in applications that require high-speed data transmission over long distances, including cellular offloading and wide-area WLAN services.

Main Features of the Phy Layer

The PHY layer design of the 2, 4, 8, and 16 MHz bandwidth modes of IEEE 802.11ah are the 1/10 down-clocked versions of the 20, 40, 80, and 160 MHz modes of the IEEE 802.11ac standard. By decreasing the clock frequency, the overall operating frequency is lowered and the time interval is scaled up without any change in the system parameters. For example, the IEEE 802.11ah 2 MHz and IEEE 802.11ac 20 MHz bandwidth modes both have 64 subcarriers, of which 52 are data subcarriers and 4 are pilot subcarriers. Thus, the subcarrier spacing of IEEE 802.11ah is reduced to 1/10 of that of IEEE 802.11ac. In addition to these basic modes, IEEE 802.11ah introduces a 1 MHz band-

width mode to enhance frequency utilization and extend service coverage. In this mode, a modulation method is added that repeats the symbol twice and transmits it to achieve a 3 dB gain. Consequently, when comparing the minimum data rate modes of IEEE 802.11ah and IEEE 802.11g operating at the same transmit power in the 2.4 GHz industrial, scientific, and medical (ISM) band, we can obtain a 24.5 dB link budget gain. In the total link budget gain, the sub-1 GHz link budget gain is 18.5 dB (free space path loss gain of 8.5 dB + noise bandwidth gain of 10 dB), and the 1 MHz bandwidth mode gain is 6 dB (1/2 bandwidth gain of 3 dB + 2 × repetition coding gain of 3 dB) [11]. In IEEE 802.11ah-based short-range communication applications that utilize its advantages in terms of link budget, even if the transmit power is 0 dBm or lower, a wider service coverage than in conventional WLAN technologies can be ensured while enabling low-power operations. For this reason, the IEEE 802.11ah technology, which supports higher data rates compared to other communication technologies (Bluetooth, Zigbee, Sigfox, LoRa, etc.), is garnering attention as an IoT connectivity technology for high-capacity data transmission in low-power, low-cost broadband sensor network applications.

Main Features of the Mac Layer

As IEEE 802.11ah can accommodate thousands of stations in a wide band, enhanced distributed channel access (EOCA)-based channel access methods increase the possibility of degrading performance and energy efficiency because of the channel access contention between hidden nodes and peripheral devices. To alleviate this problem, the following technologies were designed for the IEEE 802.11ah MAC protocol to accommodate a large number of stations in a broadband network and to increase energy efficiency.

Comparison with Other Iot Systems

Table 1 shows a comparison between IEEE 802.11ah HaLow and alternative systems in terms of various aspects. The Bluetooth Low Energy (BLE) and Zigbee systems have limitations such as short radio reach, low data rates, and high power consumption. Even though several solutions are available in the IoT market, no one technology has dominated because of the diversity and complexity of wireless sensor network applications and user requirements [14].

WiFi technology has established a strong connected ecosystem, which is essential for smart device wireless interfaces, and it is used both in daily life and in numerous industries. WiFi technology is emerging as a strong candidate for IoT applications. Because it supports high data rates in fixed or low-speed mobile environments, it allows channels to be efficiently utilized through the media access contention of distributed nodes in unlicensed bands. This distributed channel access method also has the advantages of network scalability, maintenance and repair convenience, and price competitiveness, and it is widely used in homes, businesses, and hotspots. Accordingly, its application range is increasing [15].

The traditional WiFi technologies, such as IEEE 802.11a/b/g/n/ac, have been widely used in mobile devices and consumer electronics until recently, providing high-speed data communication services to terminals within tens of meters. Because the transmission distance is very limited, it is difficult to use these technologies for low-power, low-cost, and broadband IoT networks. For example, IEEE 802.11b/g/n operating at 2.4 GHz provides high-throughput services, especially in a home environment. However, it can accommodate only a few clients with large power consumption in a relatively small area.

Implementation

System Architecture Design

The chip is implemented in a 40 nm complementary metal oxide semiconductor (CMOS) process. Figure 1 shows the chip-level fabrication layout of the chip's implementation. The IEEE 802.11ah SoC implemented in this study is a single chip that integrates a digital baseband, wireless RF transceiver, analog-to-digital converter (ADC)/digital-to-analog converter (DAC), processor, power management unit (PMU), and memory with the external connection interfaces. It is fully compliant with the IEEE 802.11ah specifications, and it is prepared in accordance with the upcoming WiFi Alliance (WFA) extended range ah (ERah) certification. This chip provides data rates from 150 kb/s to 15 Mb/s using 1, 2, and 4 MHz channel bandwidths, and supports a range of uses from low-power sensors, such as low-speed sensors, to high-performance applications, such as high-speed and high-definition surveillance cameras. An ARM processor is embedded in the chip to provide high processing performance. It has a dual-core structure comprising a Cortex-MO-based communication processor and a Cortex-M3-based application processor. To accommodate various IoT applications, various peripheral device interfaces are supported. The device also offers enough memory to selectively support AP and station functions.

The RF transceiver is integrated with support for pre-power amplifiers, fractional-N synthesizers, and various commercially available external front-end module (FEM) devices, and has a design optimized for the sub-1 GHz band. The receiver includes an internal digital gain control stage that provides a low noise figure of 4 dB and a wide dynamic range of at least 100 dB, which supports 10 dB lower receiver sensitivity and 20 dB higher maximum input power level compared to standard requirements. The transmitter provides a gain control range of at least 30 dB and a maximum transmit power of 0 dBm.

Digital Baseband and Transceiver Design

The implemented IEEE 802.11ah chipset supports binary phase shift keying (BPSK), quadrature PSK (QPSK), 16-quadrature amplitude modulation (16-QAM), and 64-QAM modulation schemes in a single stream mode and provides a data rate of up to 15 Mb/s. It supports short guard intervals (up to 4 µs) and 1,2, and 4 MHz channel bandwidths in both short and long frame formats in the specified standard. A dual-mode single chip was implemented that selectively supports AP and station modes.

To improve the efficiency of broadband loT networks, RAW avoidance, TWT, and differentiated EDCA parameters were implemented. Furthermore, non-TIM operations, dynamic association ID (AID) assignment, and TWT were implemented to improve power consumption efficiency. A relay operation mode is also supported to extend service coverage. To expand the number of stations that can be accommodated in a single basic service set (BSS), the multicast AID authentication control standard is followed. Energy limited (EL) operations are sent and received at specific times, enabling low-cost, low-power loT applications.

The RF transceiver is designed with a direct conversion transceiver architecture and includes both the pre-power amplifier and the entire radio frontend circuit, including the low noise amplifier (LNA) and low dropout (LDO). Radio calibration interfaces, that is, transmit signal strength indicator (TSSI), transmit local oscillator feed through (LOFT), transmission and receiving IQ imbalance, low pass filter, phase locked loop, BIAS current, and sigma-delta ADC, are provided in the digital baseband.

In the digital baseband, after the training signal is transmitted to the transceiver, the signal input is analyzed through the feedback loop, and the calibration parameter is optimized. In the frequency bands from 750 to 950 MHz, the 1, 2, and 4 MHz channel bandwidths are supported. The receiver (RX) noise figure is less than 4 dB, and the RX gain range is at least 100 dB. The RX input intercept point 3 (IIP3) is -17 dBm at the LNA maximum gain. In the transmitter (TX) pre-amplifier, the maximum linear output power is 0 dBm, and the TX gain range is 30 dB. A 10-bit ADC/DAC is used, and a -34 dB error vector magnitude (EVM) at 0 dBm transmit power with 64-QAM modulation is ensured. The integrated RF phase locked loop (PLL) phase error is less than 0.7°.

The RF/analog transceiver operated in the 750 to 950 MHz frequency range, included a fully integrated preamplifier, and supported a 30 dB TX gain control range with linear TX output power up to 0 dBm. It also included an external amplifier, supporting transmission power up to 23 dBm. The RX had an input range of 105 dB from 5 to 110 dBm.

Challenges in Soc Design

In the SoC design, the most challenging aspect is to minimize the power consumption with small chip size. To support long-range and high-through-put applications, the SoC includes two CPU cores and enough internal memory. Because the other loT technologies support low data rate (e.g., less than 1 Mb/s) with narrow bandwidth, CPU and memory requirements can be minimized. However, IEEE 802.11ah supports data rates from 150 kb/s to 15 Mb/s, and then it requires higher computing power and more memory to accommodate various applications. Due to such stringent requirements, the size of the implemented SoC is 3.32 × 3.51 mm, which makes it larger than other loT chips but small enough to be accommodated into loT devices. The developed chip can support both AP and station mode, so the size can be reduced more if implemented only for station. The power consumption of the SoC is minimized with low-power design and a power saving mechanism. The receiver consumes 30 mA at 3.3 V, and the transmitter consumes 32 mA at 3.3 V for 0 dBm transmit power. Receiver sensitivity improvement and calibration features for RF impairments are also important in market competition. Moreover, the temperature sensor and battery voltage monitor circuit provide an interface to optimize performance according to the process, voltage, and temperature variations of the chip.

Performance Evaluation and Field Tests

Receiver Sensitivity

Receiver sensitivity was measured when the signal was generated by the vector signal generator (VSG) in a shielded room using the implemented WiFi HaLow chip. The VSG and the chip receiver port were connected via a cable whose loss was considered in the measurements. For all channel bandwidths and MCS modes supported by the chip, we measured the received signal strength, which guaranteed a packet error rate (PER) of 10 percent when transmitting 10,000 packets at a packet length of 256 bytes. According to the experimental results, the receiver sensitivity was -109 dBm in the 1 MHz bandwidth MCS10 at the lowest data rate of 150 kb/s. At the 4 MHz bandwidth MCS7 with a data rate of 13.5 Mb/s (64-QAM, 5/6 code rate), the receiver sensitivity was -81 dBm.

Outdoor Throughput

The throughputs of a single-MAC protocol data unit (MPDU) and aggregated-MPDU (A-MPDU) were measured, and they were compared in the outdoor point-to-point environment. An AP and a station were installed at a 3 m height from the ground, and packets were delivered from the AP to the station in full load condition. The transmit power measured for both devices was 16 dBm, and the antenna gain was 2 ± 1 dBi. Tests were performed by varying the distance between the two communication devices under line of sight (LoS) conditions from 0.5 to 1.5 km. As an example, the performance of the A-MPDU mode with 4 MHz bandwidth was 4.48 Mb/s at 0.5 km and 2.12 Mb/s at 1.5 km. For the single-MPDU mode with 4 MHz bandwidth, the measured throughput was 3.62 Mb/s at 0.5 km and 1.13 Mb/s at 1.5 km. These results indicate that transmission as an A-MPDU showed up to 1.87 times better throughput than that of the single MPDU.

Figure 2a shows a map with four different zones where the mobility test was performed. The four zones were classified by distance from the AP and the differences in the regional propagation environment. In this test, the AP was installed on a hill 10 m above the ground, and the station was placed on the front dashboard in a sport utility vehicle (SUV) moving along the four zones at 10 mph speed. Figure 2b shows the received signal strength indicator (RSSI) and throughput measurements in a 4 MHz bandwidth mode. Zone A represents the non-LoS condition, where the LoS was blocked by a hill between Zone A and the AP. The signal at Zone A was measured at a distance of 380–480 m from the AP. Zone B represents the LoS condition at 260–360 m from the AP. The signals at Zone C and D were measured at a distance of 360–750 m from the AP with the non-LoS and LoS conditions, respectively, depending on the movement path. When the signal was measured at Zone C, the SUV traveled away from the AP, and the SUV's body consequently blocked the LoS from the dashboard to the AP. The signal at Zone D was measured along the same path as that of Zone C, but the SUV traveled toward the AP, and therefore an LoS condition was established for Zone D. The User Datagram Protocol (UDP) throughput was measured while the station moved at 10 mph. The average throughputs over four different zones were measured as 0.9, 1.4, and 1.77 Mb/s in the 1, 2, and 4 MHz bandwidth modes, respectively. With such field tests, it can be shown that IEEE 802.11ah can be a promising solution for high-throughput IoT networks that cannot be realized with the other loT technologies.

Receiver sensitivity improvement and calibration features for RF impairments are also important in market competition. Moreover, the temperature sensor and battery voltage monitor circuit provide an interface to optimize performance according to the process, voltage, and temperature variations of the chip.

Indoor Throughput

The experiments for multi-station uplink performance evaluation were carried out in an indoor laboratory (21 m × 24 m) where multiple rooms were separated by partition walls, as shown in Fig. 3. In this laboratory, 10 stations were randomly placed, and an AP was installed in the room in the upper left corner. This study compared the performance of the stations at an instance when only one of 10 stations transmitted to the AP with the one when all 10 stations simultaneously transmitted. During the test, every station sent packets under its peak load condition. The peak load condition in a network throughput measurement tool was set to 5 Mb/s UDP packets, which mostly yielded a packet size of 1500 bytes. An adaptive modulation and coding function was enabled, and the channel bandwidth was set to 2 MHz. In the single-station case, the station (#8) achieved a throughput of 2.6 Mb/s. This is used as a reference for the performance of the simultaneous transmission case in the multi-station environment. When all stations transmitted their own packets simultaneously, 10 stations successfully sent throughputs of 328.6,275.3, 274.0,245.0,251.3,246.0,240.3,237.3,243.3, and 301.6 kb/s, respectively, on average. The aggregated throughput of 10 stations is equal to 2.643 Mb/s, which were obtained with multiple measurements.

Power Consumption

The developed IEEE 802.11ah chip supports the low-power operation modes that are specified by the IEEE 802.11ah standard, including the TIM, non-TIM, TWT, and wireless multimedia power save mode. Furthermore, the chip divides the power into six domains and controls each domain separately to reduce the power consumption by switching on only the necessary domains. When the system boots, it transitions between active, sleep, and deep sleep modes depending on the operation condition for low power consumption. In sleep mode, the clocks of all peripherals except the CPU, bus, memory, digital baseband, power management controller, and real-time clock are turned off, and only the clock oscillator is selectively turned on and off. In deep sleep mode, all circuits except the always-on block are turned off to save power, reducing power consumption to less than 20 µA. In the TIM mode, a station wakes up periodically and receives a beacon to send or receive data after identifying a channel access period. In the non-TIM mode introduced for loT sensor applications, the listening interval is negotiated to prevent excessive uncontrolled uplink channel access and improve energy efficiency.

The estimated power consumption and battery life of the chip is shown in Table 2. The prediction was achieved by applying the total consumed current measured on the chip to the calculation under two scenarios, the TIM and non-TIM modes of operation. Power consumption varies depending on the pattern of application data traffic, for example, the size and period of application data. To avoid such dependency of power consumption on the traffic pattern, we do not consider application data for power consumption assessment. In the TIM mode, a station monitors (receives and decodes) DTIM and TIM beacons to identify a channel access period. In the TIM mode, an AP sends a 27-byte short beacon frame with MCS10 in every beacon interval (i.e., a wakeup interval in Table 2). In the non-TIM mode, a station sends a 17-byte trigger packet with MCS10 and 0 dBm transmit power to an AP to confirm the presence of buffered data with the AP and receives an Acknowledgment (ACK).

In Table 2, “pre-active” denotes the time spent on configuring the transceiver to send or receive packets. “Active” denotes the actual time for receiving a beacon frame in the TIM mode, and transmitting a trigger frame and receiving its ACK in the non-TIM mode. The wakeup interval is considered to be extendable to a period that is proportional to listening intervals. The battery voltage (VBAT) is assumed to be 3.3 V, and the battery capacity value is 4200 mAh. Table 2 shows that when the wakeup interval is 10 s, a life of approximately 13 years is estimated, and when the wakeup interval is 1 s, a life of approximately 2.7 years is estimated.

Discussion

Wireless Channel and Channel Access

The sub-1 GHz band propagates to a relatively long distance due to its propagation characteristics, and it experiences interference due to the electromagnetic waves emitted from numerous electronic devices. Consequently, the channel quality is inevitably poor. In addition, as WiFi uses unlicensed bands, it is generally considered less stable than mobile communications. Accordingly, to be used in service areas closely related to human safety, properties, and life, such as smart grids, smart factories, and smart homes using loT, research on interference mitigation technology is required to ensure stable operations in the case of interference.

In the sub-1 GHz band, most loT devices follow the listen-before-talk channel access protocol, which checks before transmission whether other users are using the transmission channel. It transmits information only when the channel is not in use. Thus, the transmitting side transmits only when there is no signal in the channel, and the receiving side transmits an ACK signal to the transmitting side when it receives the signal. If another user is using the same frequency, it is difficult to find a chance to transmit, and even normal transmissions are difficult due to interference. In this case, the ACK signal does not reach the signal sender, leading it to retransmit the same signal several times, quickly depleting the battery and potentially reducing communication efficiency due to unnecessary signal transmissions.

Various Applications of Wifi Halow

Conclusion