THIS PAPER addresses the issues concerning human exposure to radiofrequency (RF) electromagnetic fields (EMF) as consumer-driven, wirelessly communicating systems are deployed in homes as part of the emerging Internet of things (IoT), likely to be adopted everywhere in the future (WHO 2010). The integrated energy network (or smart grid), for example, utilizes smart energy-delivery systems deployed in consumer residences that rely on bidirectional communications using existing telecommunications or newly developed (e.g., mesh) RF networks to constantly adapt and tune the delivery of energy to the consumer. However, even though public understanding and acceptance are critical for the adoption of these new technologies likely to be implemented by a host of companies (including electricity and other utility companies), members of the general public may be concerned about the potentially heightened levels of RF radiation in their home environment. Furthermore, the World Health Organization (WHO) identified in its international RF research agenda a need for measurement surveys to characterize population exposures to all RF sources, with a particular emphasis on new wireless technologies, including smart meters and other novel residential wireless communication systems (WHO 2010). Assessment of the RF emission levels of new wireless technologies in residential environments can address these concerns.
In the context of IoT, commonly installed RF-emitting devices in homes can be broadly categorized as devices for energy monitoring, devices for automatic light control, devices for heating or cooling, security systems, or smart meters. To the authors’ knowledge, only a limited number of studies have investigated the RF emissions from residential appliances other than communications devices, and those investigations have been predominantly of smart meters. In the United States, for example, two specific types of smart meters were investigated by the Electric Power Research Institute (EPRI 2010, 2011; Foster and Tell 2013; Tell et al. 2012a and b). Comprehensive studies on smart meters were performed in the United Kingdom by Peyman et al. (2017) and Qureshi et al. (2018) and in Australia by Girnara et al. (2011) and the Australian Radiation Protection and Nuclear Safety Agency (ARPANSA 2013). In general, the most important parameters to be considered for the assessment of smart metering devices were the output power of the device, the frequency of the emitted signal, the distance to the device, and the duty cycle of the device (i.e., the proportion of time the device actually transmits a signal).
The objective of this study was to develop a novel measurement method to characterize a wide array of in situ RF IoT devices, smart meters, and other sources of residential RF-EMF exposure using a wide range of technologies (wireless fidelity [Wi-Fi], long range [LoRa], Zigbee, Sigfox, general packet radio service [GPRS], etc.) and frequency bands (e.g., the industrial, scientific, and medical [ISM] 41 MHz, 433 MHz, 868 MHz, and 2,400 MHz bands), and to compare their emissions with levels of telecommunication and broadcasting signals present in the residence. For this, a new duty cycle assessment method is used incorporating the spectrogram mode of a spectrum analyzer, which allows a graphical overview of the variations in transmission frequency or signal amplitude over time. The proposed method was applied to a convenience sample of 10 residences in Belgium and France, resulting in a total of 55 devices characterized.
MATERIALS AND METHODS
Selection of residences
A convenience sample of 10 residences was selected (in Belgium and France) in which a relevant number of devices of the above-mentioned categories were present, i.e., energy monitoring, devices for automatic light control, devices for heating or cooling, security systems, or smart meters. In Table 1, the details of this sample are listed, including the number of devices per residence as well as wireless technologies that could be identified. Different smart meters (electricity, gas, and water) are highlighted. The measurements were performed during the period from April 2017 to November 2017.
The RF-EMF levels (i.e., the electric-field strength E in V m−1 or the power density S in W m−2) were assessed using both broadband and frequency-selective narrowband measurement equipment.
A broadband measurement consisted in measuring the total (i.e., within a large frequency span) electric-field value Ebb at a given position using a Narda NBM-550 field meter equipped with an EF0391 (dynamic range: 0.2–320 V m−1; frequency range: 100 kHz–3 GHz) or EF0691 probe (dynamic range: 0.35–650 V m−1; frequency range: 100 kHz–6 GHz) (Narda, San Diego, California, US). Although this type of measurement is useful to identify residential sources of RF-EMF (by holding the probe close to a suspected source) or locations of maximum exposure (in terms of electric-field strength), no frequency-specific information can be obtained. Hence, it is unable to identify the source’s emission frequencies, and the specific contribution of the source to the total electric-field strength remains uncertain.
For this, a spectrum analyzer setup is needed, which, in this case, consisted of a triaxial R&S TS-EMF isotropic antenna (dynamic range: 1 mV m−1–100 V m−1 for the frequency range 30 MHz–3 GHz) (Rhode and Schwarz, Munich, Germany) in combination with an R&S FSL6 spectrum analyzer (SA I; frequency range: 9 kHz–6 GHz) for narrowband measurements, or a PCD 8250 precision conical dipole antenna (dynamic range: 1.1 mV m−1–100 V m−1 for the frequency range 30 MHz–3 GHz) (Seibersdorf Laboratories, Seibersdorf, Austria) in combination with an R&S FSVA40 signal and spectrum analyzer (SA II; frequency range: 10 Hz–40 GHz). The measurement uncertainty of the considered setups was ±3 dB for (CENELEC 2008; Joseph et al. 2012a). This uncertainty represents the expanded uncertainty evaluated using a confidence interval of 95%.
Besides the emission levels and frequencies of the assessed RF-emitting device, a third important factor in the exposure assessment is the duty cycle (DC), i.e., the proportion of time the device actually transmits. To measure the DC, the R&S FSV30 signal and spectrum analyzer was equipped with firmware option FSV-K14, which enables the spectrogram mode. A spectrogram is a graphical overview of a measurement as a function of time and is obtained by capturing, at a certain speed (defined by the sweep time [SWT]), successive traces of either a part of the spectrum (i.e., in the frequency domain, defined by a certain frequency span) or of the time domain (i.e., with a frequency span of 0 Hz, or zero span mode), according to the objective. The former type is used, e.g., to detect frequency-hopping channels, and the latter is used to determine the DC of a noncontinuous signal.
A flowchart of the proposed procedure to assess the residential exposures to RF-EMF is shown in Fig. 1. First, a frequently used room of the residence (usually the living room) is scanned with the broadband probe to locate the maximum field level. At that location, a spectral survey is performed to identify continuously present RF signals, which are then measured more in detail. Finally, all RF-emitting devices (e.g., smart meters and IoT devices) present in the residence are characterized, which by the proposed method comprised three parts: determination of the transmission frequencies, measurement of the maximum emitted fields, and calculation of the duty cycle. All steps are explained in more detail in the following sections.
Spectral survey and assessment of continuously present signals
At the location of the highest electric-field level in the selected room, a spectral survey is performed, after which the relevant, continuously present signals are assessed more accurately, according to the measurement procedures proposed by Joseph et al. (2012a, 2012b, 2013) and Verloock et al. (2014). The considered signals are predominantly outdoor signals, such as telecommunications and radio downlink (DL) signals, and, if present, also Wi-Fi and cordless phone signals. For this part, SA I is used with the specific settings listed in Table 2.
This measurement gives a baseline to put in perspective the subsequent measurements of residential RF-emitting devices.
Characterization of residential RF-emitting devices
As many residential sources of RF-EMF do not transmit continuously, their signals are seldom detected in the spectral survey. In fact, the length and frequency of the signals depend on the specific use and/or transmission technology of these devices. As most of them do transmit at a fixed power (only advanced two-way communications devices—such as mobile phones—can make use of power control), it is sufficient to determine the maximum received power (Pmax) at certain distances from the device—which is then used to calculate the maximum electric-field strength (Emax) or power density (Smax) (Table 3)—as well as a typical DC in order to determine the time-averaged exposure level, which can be finally compared to exposure-limiting guidelines such as those issued by the International Commission on Non-Ionizing Radiation Protection (ICNIRP 1998) or the Federal Communications Commission (FCC 2001; IEEE 2005). In Table 3, an overview is given of the measured quantities and exposure metrics (and their relation) used in this study.
An inventory of the present IoT devices, smart meters, and other RF-emitting devices was created, and for each a defined set of procedures was performed (bottom right part of Fig. 1). First, using SA I, the frequencies of the RF signals transmitted by the device were determined. Specific SA settings for this step included a wide frequency span, a short SWT (i.e., a fast measurement, in order to capture short pulses), and the maximum hold mode (max hold) to retain all transmission frequencies. Using the same setup, the peak emitted electric-field values (Emax) (Table 3) were then measured at three different measuring distances, i.e., at 0.2 m, 0.5 m, and 1 m, defined as the distance between the surface of the device and the middle of the measurement probe.
The final step comprised the accurate determination of the DC, since for noncontinuous signals, Emax (which assumes DC = 100%) can result in a significant overestimation of the exposure. For this, a (large) number of subsequent time domain traces of the power within a certain frequency bandwidth were captured using the spectrogram mode of SA II. These traces were then analyzed to determine the total time the device actually transmitted (Ttransm) during the period of observation (Tobs), and the DC was calculated as
This measurement involved a zero frequency span setting (i.e., time domain measurement), a short SWT (i.e., high temporal resolution), a resolution bandwidth (RBW) at least as large as the signal bandwidth, and max hold mode.
In this study, three types of signals were observed: periodically (at a fixed interval) transmitted signals; arbitrarily transmitted signals, for which transmission depended on the occurrence of (random) events such as a change in temperature, a user interaction, etc.; and signals with a combination of a fixed and an arbitrarily transmitted active signal, e.g., in the case of a signal containing management and user data (e.g., transmissions by a wireless access point).
In the case of a periodically transmitted signal, both the duration of the periodically transmitted pulse (i.e. the pulse time Tpulse) as the period between pulses (i.e., the repetition time Trep) are defined. This results in a fixed duty cycle
which is valid independent of the observation time Tobs. Since at least two pulses should be correctly measured to determine Trep and thus DC, it requires Tobs > Trep.
In the case of a nonperiodically transmitted signal, neither the pulse time nor the period between two pulses are necessarily fixed or are easily defined. In this case, an action is defined (e.g., a push of a button) and the total signal transmission time when such an action occurs, Taction (= Ttransm), which can consist of multiple pulses of varying length Tpulse,i. Now, the DC is calculated as follows:
where the observation time Tobs corresponds to a defined period. For example, for comparison to RF safety guidelines issued by ICNIRP (ICNIRP 1998) or the FCC (FCC 2001, IEEE 2005), Tobs is defined as 6 min and 30 min, respectively.
Finally, in the case of a combined signal (e.g., Wi-Fi), the resulting DC is the sum of the periodic and nonperiodic signals. But it should be noted here that the DC of a Wi-Fi signal (i.e., the DC of the dominant Wi-Fi channels) was determined using the measurement method proposed by Joseph et al. (2013), and in this case, no distinction could be made between uplink (UL) and DL traffic, as both are present in the same frequency band.
Assessment of fields emitted by mobile phones
In addition to the assessment of continuously present RF signals and the characterization of residential (IoT) devices, the UL communication between a mobile phone and an outdoor telecommunication base station was also investigated to establish context. In each residence, at least one mobile phone measurement was performed, where the fields emitted by the phone were recorded at a distance of 0.5 m, handheld, and operational in either global system for mobile communications (GSM; voice call), universal mobile telecommunications system (UMTS; voice call or data transfer), or long-term evolution (LTE; data transfer) mode. For the mobile phone assessment, the duty cycle was assumed to be 100% during the entire observation time, except for GSM, which uses time division multiple access (TDMA) and has an inherent DCmax of 12.5%. In each case, DC may be overestimated.
Metric for comparison to exposure guidelines
Finally, to enable comparison with exposure limits issued by ICNIRP (or the FCC), the time-averaged electric-field strength Eavg is calculated using
with DC calculated for Tobs 6 min (ICNIRP) or 30 min (FCC, IEEE), and subsequently used to calculate the exposure ratio RS:
with Savg the time-averaged power density (Table 3) and Sref and Eref the ICNIRP (or FCC) general public reference levels for the power density and electric-field strength, respectively. RS indicates the number of times the measured power density is higher or lower than the power density reference level (or maximum permissible level). The closer RS is to 1, the closer the measured power density Savg is to the reference level, with the reference level being exceeded if RS is higher than 1.
Spectral survey and assessment of continuously present signals
To illustrate the first part of the proposed method (Fig. 1), the electromagnetic spectrum from 30 MHz to 3 GHz measured in the living room of residence 1 is shown in Fig. 2. The spectrum comprises LTE800 signals (UL and DL), signals in the 868 MHz ISM band (transmitted by smart home devices), GSM and UMTS900 (UL and DL) signals, and signals in the 2400 MHz ISM band (Wi-Fi, magnetron, etc.). Additionally, a 1.29 GHz signal was observed, probably transmitted by a surveillance or navigation system. However, only one component was detected and as it was not reproducible, it was disregarded. Next, narrowband measurements of the continuously present signals and of the Wi-Fi signal in the ISM 2400 MHz band were performed (Table 4). To determine the Wi-Fi exposure, both the duty cycle of the dominant channel (i.e., channel 11, with center frequency (CF) 2.462 GHz; DCch11 = 3.7%) (Joseph et al. 2013) and the worst-case duty cycle (DCWi-Fi = 100%) were used to determine the corresponding field level. In this case, all the measurements, including the cumulative exposure level of the considered signals (Ecum = 0.076 V m−1 or 0.370 V m−1 with DCWi-Fi = 100%), were well below the FCC and ICNIRP guidelines, with a maximum RS of 3.6 × 10−5.
Table 4 further lists the measurements performed in all 10 residences. On average, the cumulative exposure level in the residences was 0.225 V m−1 (0.497 V m−1 with DCWi-Fi = 100%), due to continuously present signals ranging from frequency modulation (FM; radio) at 100 MHz to Wi-Fi at 2400 MHz. The most frequently present signals were LTE800, GSM900, UMTS900, and Wi-Fi (at 2400 MHz), which were observed in all 10 residences. Of the three telecommunications signals, GSM900 was the most dominant (only in residence 10 [France] did LTE800 contribute more than GSM900). In fact, its exposure level was similar to that of Wi-Fi (average DCWi-Fi = 4.89%). When present, other telecommunications signals, such as GSM1800 (number of occurrences, n = 2), LTE1800 (n = 4), and UMTS2100 (n = 4), often contributed greatly to the total residential exposure. Also often present was digital enhanced cordless telecommunications (DECT; cordless phone) (n = 7), with an average of 0.135 V m−1 as a dominant contributor as well. In addition, in some cases, FM signals, digital radio and television (TV) signals, and signals in the ISM 868 MHz or 2,400 MHz (besides Wi-Fi) bands were detected, but their contributions were limited. On average, the exposure ratio was 5.5 × 10−6, while the maximum exposure was found in residence 6, with RS = 1.3 × 10−4 (and worst-case 1.9 × 10−3) due to the larger (in relation to the other residence) presence of ISM868, GSM900, DECT, UMTS2100, and Wi-Fi.
Characterization of residential RF-emitting devices
Example—smart electricity meter
In Belgium, where a smart meter pilot project is underway, smart electricity meters are usually networked to the central system of the energy supplier via a communications module (CoMo). Other smart meters (for water and gas) present at the same property connect into this CoMo using either a wired or a wireless link (e.g., via wireless meter bus [M-Bus] or Wi-Fi). In this section, a specific measurement of an electricity meter’s CoMo is described. In total, five wireless CoMos were assessed in this study, and all but one communicated with the grid through GPRS technology (similar to GSM). Fig. 3 presents the frequency spectrum of the CoMo signal, measured with SA I in max hold mode. The CoMo UL signal used three frequencies: 903.2 MHz, 904.2 MHz, and 908.0 MHz. Using a wide enough RBW to capture the three frequencies at once, the signal amplitude was subsequently measured as a function of time using the zero span spectrogram mode of SA II to obtain more detailed information about the rate of transmission and hence to determine the duty cycle. Part of the CoMo transmission as a function time is shown in Fig. 4. In theory, a CoMo should transmit once every 15 min, following the logging of the data. However, signal repetitions as fast as every 43 s were observed. Furthermore, each transmission consisted of a series of bursts sent over a 3.6 s interval; although in the DC calculation, a continuous signal was assumed. Combined with the communication technology’s inherent duty cycle of 1:8 (like GSM, GPRS uses TDMA), the CoMo’s theoretical DC was 0.05%, while the maximum observed DC was 1.05%.
Table 5 summarizes the measurements at 0.2 m from the 55 investigated devices: the maximum electric-field level, the 6-min-averaged duty cycle, and the comparison to the ICNIRP exposure guidelines. The measurements at 0.5 m were compared with the mobile phone UL measurements and the impact of the (6-min-averaged) duty cycle on the average exposure is shown in Fig. 5. In the following, the considered RF devices are described as a number of broad grouped categories (in Fig. 5 as well, but less broad).
- Smart meters. In the acquired sample, smart meters came in two categories: those that transmitted data to the central system of the utility company (all electricity meters and one water meter in a residence where no smart electricity meter was present), and those that transmitted their data in-house to a smart meter of the first category (all other smart gas and water meters). Both types were usually deployed in more remote locations in the residence such as the garage, storage room, or hallway.
- For indoor-outdoor communications (i.e., the first category), the electricity meters used GPRS (CF = 899–908 MHz; GSM900 UL band) and the water meter used Sigfox (CF = 868 MHz; ISM 868 MHz band). In theory, the electricity meter’s CoMo transmits once every 15 min (DC = 0.05%, including TDMA). However, the maximum duty cycle was 1.05% (including TDMA) in this sample. With Emax between 11 V m−1 and 20 V m−1 at 0.2 m, using the latter DC resulted in RS = 8 × 10−4 to 2.5 × 10−3, a higher exposure compared to the other smart meter results (Table 5), though still significantly lower than the ICNIRP limits. The water meter, on the other hand, transmitted only once per day a signal with Tpulse = 6.49 s (DC = 0.008%) making the field strength of the signal difficult to measure. For completeness, Emax at 0.5 m was 0.072 V m−1 with a single (random) electric-field component measured (RS = 2.4 × 10−10).
- Another electricity meter transmitted its data via a Wi-Fi back channel. In this case, both Emax and DC were slightly lower (7 V m−1 and 0.08%, respectively), and correspondingly the RS (1.1 × 10−5) was lower.
- For in-house communications, the smart meters in the sample used wireless M-Bus (CF = 869 MHz; ISM 868 MHz band) with a DC of 0.002% (one signal of Tpulse = 15–18 ms every 15 min). None of them were measured at 0.2 m, but at 0.5 m, Emax values were lower than 1 V m−1 and significantly below the emissions from the smart meter of the first category (Fig. 5).
- Smart home devices. In this residential sample, a number of devices could be characterized as smart home devices (e.g., weather station and temperature sensor, Philips Hue device, smart toothbrush, and motion sensor). Most of these devices were continuous, periodic transmitters, with a duty cycle on the order of 0.01% up to a few percent (weather station, DC = 0.31–2.90%; energy-monitoring plug and gateway, DC = 0.05%; heat alarm, DC = 0.02%; temperature sensor—with a user-defined DC = 23.76% [in theory, max 1% because LoRa]—Philips Hue gateway DC = 0.25%; one was a periodic transmitter when in use [toothbrush, with DC = 33.63%], and one [a motion sensor] actually transmitted continuously with DC = 100%). One other device detected changes in the environment to commence a certain action (thermostat, DC = 0.02%, with one signal during a 6 min interval). All smart home devices operated in the ISM bands: three devices in the 2,400 MHz band (energy-monitoring device, using Wi-Fi; motion sensor, CF = 2,450 MHz; and Philips Hue gateway, using Zigbee, with CF = 2,475 MHz), two weather stations in the 434 MHz band, and the others in the 868 MHz band. The peak electric-field strengths measured at 0.2 m ranged from 2 × 10−3 V m−1 (weather station receiver) to 5.1 V m−1 (Philips Hue gateway), with >1 V m−1 fields for five of the assessed devices (temperature sensor, Philips Hue gateway, motion sensor, thermostat, and smart toothbrush). Taking into account the 6 min duty cycles and the transmission frequencies, the highest exposures were found for the smart toothbrush (RS = 4.8 × 10−3), the temperature sensor (RS = 6.0 × 10−4), and the motion detector (RS = 5.0 × 10−4).
- Remote controls. Remote controls rely on user control and transmit at arbitrary moments. Their transmission frequencies are usually in the ISM 433 MHz and 868 MHz bands, with two exceptions: TV remotes working at Wi-Fi 2,400 MHz and 41 MHz. A single push of a remote control button defined the action, and the minimum observation time was the minimum time between two pushes (i.e., 0.6 μs, as timed by the investigators). Depending on the device, the transmitted signal was either continuous during the action (in this case, DCmax = 100%) or comprised one or multiple pulses (DCmax < 100%, unless Tpulse > 0.6 μs). The maximum field levels measured at 0.2 m were in the range 0.16–6.0 V m−1, the duty cycles for Tobs of 6 min (for comparison with ICNIRP guidelines) between 0.003% and 0.19%, and the maximum RS at 0.2 m was 1.5 × 10−5 (TV remote at 41 MHz).
- Bluetooth devices. One Bluetooth-connected computer mouse and two speakers were assessed during a (failed) pairing initialization process. In this case, the duty cycle was found to be 2.84%. Peak electric-field strengths varied around 0.4 V m−1, corresponding to an exposure ratio of approximately 1.5 × 10−6.
- Wireless access points. In this sample, all Wi-Fi cable modems and range extenders transmitted in the Wi-Fi 2,400 MHz band. Their duty cycles ranged between 2.46% and 15.80%, and with peak electric-field strengths of 2.75–12.51 V m−1, this resulted in exposure ratios of 1.0 × 10−4 to 1.2 × 10−4 at 0.2 m.
- Other. Other devices assessed included a doorbell transmitting at 868 MHz; two DECT cordless phones and a DECT base station (maximum RS of 2.2 × 10−3 at 0.2 m); a Wi-Fi printer; a walkie-talkie (PMR 446 MHz band) with a worst-case (i.e., during a 6 min call) RS of 0.53 at 0.2 m; two wireless (non-Bluetooth) computer mice transmitting in the ISM 2400 MHz band with a worst-case (i.e., 6 min use) RS of 4.3 × 10−3 at 0.2 m; and two pairs of baby monitors (separate parent and baby units), one using DECT (maximum RS = 1.9 × 10−3 for the parent unit), the other transmitting in the ISM 868 MHz band (maximum RS = 0.013 for the baby unit).
- Mobile phones. The UL signals of mobile phones were measured at one distance, 0.5 m. In Fig. 5, the maximum and time-averaged field levels are depicted. On average, the field levels of GSM UL communications were the highest (up to 11 V m−1), and for UMTS UL, they were the lowest (up to 2 V m−1). However, GSM has an inherent DC of 12.5% due to its TDMA structure. After taking this into account, the highest Eavg were found for LTE UL.
Potential impact on residential RF-EMF exposure
Residential RF emissions from a total of 55 devices (e.g., IoT devices and smart meters) were characterized by determining the transmission frequencies, peak emitted fields at various distances, and duty cycles. The emissions were compared to the ICNIRP guidelines for public RF-EMF exposure, as well as to the present exposure levels resulting from environmental sources (telecommunications and broadcasting signals) and emissions of mobile phones, in order to identify the potential impact on the residential RF exposure. When comparing Figs. 5a and b, one can see that mere comparison of the peak electric-field strength Emax may result in a wrong exposure ranking. Moreover, to further assess the potential impact on the exposure from a non-user-controlled device, the deployment location is highly important. For example, the highest field strengths at 0.5 m were measured for three CoMos (smart meters), but the resulting exposures (considering a 1% duty cycle) rank between wireless access points and GSM and LTE UL emissions, while their deployment out of sight of the residents ensures that the exposure potential remains limited.
The results obtained in the considered (convenience) sample of residences demonstrate that, in addition to the exposure due to environmental sources, wireless access points (due to their usual deployment in highly frequented rooms combined with a DC of several percent) and mobile phones and other personal communication devices (e.g., DECT cordless phones, walkie-talkies; due to their typically high emissions and use close to the body) will probably continue to represent the bulk of the residential exposure to RF-EMF in the smart home. A surprising addition to these dominant RF sources in this sample was the (albeit noncommercial) smart toothbrush (which may be characterized as a personal IoT device), due to its relatively high emissions, high duty cycles, and conditions of use (i.e., close to the body). Furthermore, monitoring devices such as motion sensors (with DC = 100%) and baby monitors (also high DCs) may additionally increase one’s residential RF exposure. Smart meters, on the other hand, and in particular communications modules wirelessly linked to the utility company’s central network, may contribute little to the RF exposure. Although field levels at 0.2 m reached as high as 20 V m−1, the potential for exposure is small given the rare transmissions and deployment in locations away from the residents.
Comparison to literature (smart meters)
Despite the fact that smart metering systems are not universal, the results obtained in this study are similar to those found in the literature. Girnara et al. (2011), Tell et al. (2012), and Peyman et al. (2017) found duty cycles typically lower than 1% for most smart meters and lower than 5% for heavily loaded smart meters. In the laboratory measurements of Peyman et al. (2017), a maximal power density of 15 mW m−2 (2.38 V m−1) was measured at a 0.5 m distance from the radiating smart meter. However, overall the maximum time-averaged exposure level was 6 μW m−2 (0.05 V m−1; measured at a distance of 0.3 m from a single smart meter acting as a wireless access point), and all of the exposure levels assessed at distances of 0.2 m and beyond around smart meters were well below the levels recommended by the regulatory guidelines such as the FCC (FCC 2001; IEEE 2005) and ICNIRP (1998).
It should perhaps be noted that the ICNIRP guidelines (i.e., the reference levels and averaging time, here Tobs) are currently being revised. However, there is no indication that the new guidelines would have any impact on our conclusions.
Furthermore, in comparison to the RF exposure from mobile phones and Wi-Fi networks, it was concluded by Peyman et al. (2017) that exposure from smart meters is lower due to their low duty cycle and the typically large distance to the human body in normal circumstances.
STRENGTHS AND LIMITATIONS
In this study, a wide range of RF-emitting residential devices were assessed (55 in total) using a wide range of RF communications technologies (Wi-Fi, LoRa, Zigbee, Sigfox, EnOcean, Bluetooth, etc.) for various purposes. To this aim, a novel measurement procedure was developed using the spectrum analyzer spectrogram mode to capture the signal amplitude in time and thus characterize the temporal structure of the emissions from a device.
During all of the described measurements, the investigators’ phones and laptops were turned off or in in-flight mode.
Two factors may have overestimated the exposure ratio RS. First, it was assumed that the device’s output power remained constant, and the peak electric-field strength was used to calculate the eventual exposure ratio. Second, in the determination of the duty cycle of a device, whenever burst signals occurred, their envelope (as captured by the spectrogram) was considered to calculate Tpulse or Taction. Doing so overestimated the DC and thus the exposure ratios RS of some of the considered devices.
Although some devices are meant to be used closer to the body, only whole-body exposure was considered here, using the peak incident power density at certain distances from the source in combination with the duty cycle. Both factors ensured a conservative approach to calculate the resulting exposure ratio.
It should be noted that measurements were also performed using the Narda broadband meter. This measurement probe is a handheld system that is very easy and practical to use. However, a large disadvantage is that the pulse length of an RF signal has to be long enough in comparison to the integration time of the system (270 ms) and that the signal level must be high enough (>0.2 V m−1) to enable its detection. Additionally, the contribution of different RF sources cannot be distinguished. Consequently, a broadband setup was not suitable to characterize the RF fields around the smart devices installed in houses, and its results were omitted from this paper.
Finally, it should also be noted that as the specific physical placement of RF sources is unique to the assessed environments, the measurements presented here represent a sample cross section in space and time and are not generalizable to a broader number of smart homes. However, the described results may illustrate the potential RF environment of the near future, in which everything is connected.
In this study, a novel measurement method was designed to characterize in situ residential RF emissions from emerging wireless solutions (e.g., IoT sources and smart meters) by determining the RF transmission frequency, the peak emitted fields at various distances, and the proportion of transmission time (i.e., duty cycle), for which the spectrogram mode of a spectrum analyzer was used. This method was applied to a convenience sample of 10 residences in Belgium and France containing, in total, 55 IoT devices, smart meters, and other RF-emitting devices. The measured emissions were also compared to present levels of telecommunications and broadcasting signals, emissions by a mobile phone using three current telecommunications technologies (GSM, UMTS, and LTE), as well as to the ICNIRP guidelines for general public RF-EMF exposure.
Overall, low to very low emissions were measured for nearly all of the devices, and it is concluded that, in addition to the continuous exposure due to environmental sources, when used, wireless access points and especially mobile phones and other personal communication devices (e.g., DECT cordless phones, walkie-talkies) will continue to represent the bulk of our exposure to radiofrequency electromagnetic fields in the smart home, due to their typically high emissions and use close to the body. However, RF-emitting devices with high duty cycles (e.g., in this sample, motion sensor, baby monitor, and an IoT toothbrush) may significantly increase the potential for exposure, especially when used or located close to the body. The potential impact on the exposure due to individual smart meters, on the other hand, and in particular due to the communications modules wirelessly linked to a utility company’s central network, is small, regardless of their emissions of up to 20 V m−1 at 0.2 m, given their rare transmissions and usual deployment away from the residents.
The authors would like to thank Robert Olsen (Washington State University) and Mike Silva (Enertech) for their valuable comments.
This work was supported by the Electric Power Research Institute (contract #0010007196). Ximena Vergara is an employee of the Electric Power Research Institute. S. Aerts is a postdoctoral fellow of the Research Foundation–Flanders (FWO-V).
Australian Radiation Protection and Nuclear Safety Agency. ARPANSA preliminary measurements of radiofrequency transmissions from a mesh radio smart meter. Yallambie, Australia: ARPANSA; Technical Report 163, ISSN 0157–1400; 2013.
European Committee for Electrotechnical Standardisation. Basic standard for the in situ measurement of electromagnetic field strength related to human exposure in the vicinity of base stations. Brussels, Belgium: CENELEC; TC 106x WG1 EN 50492; 2008.
Electric Power Research Institute. An investigation of radiofrequency fields associated with the Itron Smart Meter. Palo Alto, CA: EPRI; Technical Report; 2010.
Electric Power Research Institute. Characterization of radiofrequency emissions from two models of wireless smart meters. Palo Alto, CA: EPRI; Technical Report; 2011.
Federal Communications Commission. Evaluating compliance with FCC guidelines for human exposure to radiofrequency electromagnetic fields. Washington, DC: FCC; Tech. Rep. Suppl. C to OET Bulletin 65; 2001.
Foster KR, Tell RA. Radiofrequency energy exposure from the Trilliant smart meter. Health Phys 105:177–186; 2013. DOI 10.1097/HP.0b013e31828f5805.
Girnara K, Zombolas H, Wood A. AMI meter electromagnetic field survey. Keilor Park, Australia: EMC Technologies; Final Report Document Number M110736; 2011.
International Commission on Non-ionizing Radiation Protection. Guidelines for limiting exposure to time-varying electric, magnetic, and electromagnetic fields (up 300 GHz). Health Phys 74:494–522; 1998.
IEEE. IEEE standard for safety levels with respect to human exposure to radio frequency electromagnetic fields, 3 kHz to 300 GHz. New York: IEEE; C95.1-2005; 2005.
Joseph W, Verloock L, Goeminne F, Vermeeren G, Martens L. Assessment of RF exposures from emerging wireless communication technologies in different environments. Health Phys 102:161–172; 2012a. DOI 10.1097/HP.0b013e31822f8e39.
Joseph W, Verloock L, Goeminne F, Vermeeren G, Martens L. In situ LTE exposure of general public: characterization and extrapolation. Bioelectromagnetics 33:466–475; 2012b. DOI 10.1002/bem.21707.
Joseph W, Pareit D, Vermeeren G, Naudts D, Verloock L, Martens L, Moerman I. Determination of the duty cycle of WLAN for realistic radio frequency electromagntic field exposure assessment. Prog Biophys Mol Bio 111:30–36; 2013. DOI 10.1016/j.pbiomolbio.2012.10.002.
Peyman A, Addison D, Mee T, Goiceanu C, Maslanyj M, Mann S. Exposure to electromagnetic fields from smart utility meters in GB; part I: laboratory measurements. Bioelectromagnet 38:280–294; 2017. DOI 10.1002/bem.22044.
Qureshi MRA, Alfadhl Y, Chen X, Peyman A, Maslanyj M, Mann S. Assessment of exposure to radio frequency electromagnetic fields from smart utility meters in GB; part II: numerical assessment of induced SAR within the human body. Bioelectromagnet 39:200–216; 2018. DOI 10.1002/bem.22094.
Tell RA, Sias GG, Vasquez A, Sahl J, Turman JP, Kavet RI, Mezei G. Radiofrequency fields associated with the itron smart meter. Radiat Protect Dosim 151:1–13; 2012a. DOI 10.1093/rpd/ncr468.
Tell RA, Kavet RI, Mezei G. Characterization of radiofrequency field emissions from smart meters. J Expo Sci Environ Epidem 23:1–5; 2012b. DOI 10.1038/jes.2012.102.
Verloock L, Joseph W, Vermeeren G, Martens L. Procedure for assessment of general public exposure from WLAN in offices and in wireless sensor network testbed. Health Phys 98:628–638; 2010. DOI 10.1097/HP.0b013e3181c9f372.
Verloock L, Joseph W, Goeminne F, Martens L, Verlaek M, Constandt K. Assessment of radio frequency exposures in schools, homes, and public places in Belgium. Health Phys 107:503–513; 2014. DOI 10.1097/HP.0000000000000149.
Keywords:© 2019 by the Health Physics Society
exposure, radiofrequency; indoor exposure; public information; radiofrequency radiation