Application of the Maximum Power Extrapolation Procedure for Human Exposure Assessment to 5G Millimeter Waves: Challenges and Possible Solutions.

This paper describes an investigation on the application of the Maximum Power Extrapolation (MPE) technique on a fully operational Fixed Wireless Access (FWA) FR2-band 5G gNB. The data was acquired in [27.1-27.3] GHz band using a network scanner over nearly 10 minutes periods to allow a statistical analysis and an accurate estimation of the role of each contribution to the total uncertainty, including the fading affecting the 5G FR2 reference signal. The results show that the level of the electromagnetic field is well below the limits imposed by Italian legislation. However the goal of the paper is more fundamental, and shows an approach that can be used to identify the critical elements of the measurement set-up, suggesting where to concentrate efforts to improve the measurement procedure. In particular, the uncertainty budget highlights three contributions, (i.e. estimation of the traffic beam level, of the probe response and of the 5G FR2 reference signal) that deserve further investigations.


I. INTRODUCTION
Field level measurement for human exposure assessment is an important issue in the communications engineering community.
Research on biological effects of radiofrequencies is a very active field of research. The results of the researches are analyzed by the International Commission on Non Ionizing Radiation Protection (ICNIRP) [1], that gives guidelines on The associate editor coordinating the review of this manuscript and approving it for publication was Muhammad Zubair . limiting exposure to electromagnetic fields for the protection of humans exposed to radiofrequency electromagnetic fields up to 300 GHz.
ICNIRP recently published updated guidelines, in terms of basic restrictions and reference levels of many different quantities (Specific Absorption Rate (SAR), absorbed energy density, incident field, incident power density, etc.) depending on the frequency range, on the observation time and on the exposed people. Such guidelines were assessed by ICNIRP according to the review of available literature on biological and health effects related to temperature rise. These indications are the basis for national regulations, that can be also more restrictive.
Loosely speaking the guidelines require observation of the field over an adequate time window to obtain an estimate of the energy associated with the electromagnetic field. 1 The observation time generally ranges from 6 to 30 minutes [1].
The observation window can be shortened using MPE [2] techniques. Such techniques allow to estimate the maximum field level from data acquired in a short time window assuming that all the resources of the communication system are assigned to a single user. Consequently, they give an upper bound of the field level in the measurement position. The value obtained from MPE procedures must subsequently be multiplied by an appropriate correction factor taking into account the stochastic nature of the communication in order to obtain a realistic value of the field level [3], [4], [5], [6], [7], [8].
While many measurement campaigns for validation of MPE techniques in FR1 band are discussed in the open literature, at the best knowledge of the authors only a few papers present measurements in Frequency Range 2 (FR2) [19], [20], [21], [22], [23].
Indeed, the deployment of FR2 was slow, both due to the complexity of millimeter wave technology, and to the absence until now of a ''killer application'' that represents the trigger for the deployment of this technology.
FWA is an efficient alternative to wired connections with data rate competitive with Fiber-To-The-Home (FTTH) solutions without the costs associated with the last mile involved in fixed access deployments. There is therefore a great interest in FWA systems as key technology able to reduce 1 International guidelines also require the spatial average of the field, which is not the focus of this article.  the digital divide in rural communities helping to reach the ambitious goal of the European Digital Agenda program, i.e. providing the entire population with the possibility of accessing the Network through high-speed connectivity services.
The current step of the roll-out phase of 5G networks in Italy is involving BSs working in FR2 band, mainly for FWA applications. This gave us the chance to carry out measurements in full operational conditions. The aim of this paper is to describe the results of the MPE technique to estimate the maximum field level radiated by the gNB.
The description of 5G frame is present in many books and papers, besides the 3rd Generation Partnership Project (3GPP) documentation [24], and will not be repeated here. A simple introduction limited to the concepts useful for the MPE technique is discussed for example in [12] and [14].
This paper describes an investigation on the application of the MPE technique on a fully operational FWA FR2-band 5G gNB.
The results show that the level of the electromagnetic field is well below the limits imposed by Italian legislation. However the goal of the paper is more fundamental, and shows VOLUME 10, 2022 an approach that can be used to identify the critical elements of the measurement set-up, suggesting where to concentrate efforts to improve the measurement procedure.
The paper is organized as follows: in Section II a quick overview of MPE for 5G is given. Section III is devoted to the description of the measurement campaign, whereas in Section IV the MPE is applied to the measurements. Conclusions are drawn in Section V.

II. THE MAXIMUM POWER EXTRAPOLATION TECHNIQUE FOR 5G
The goal of the Maximum Power Extrapolation procedure is the estimation of the maximum field level [V/m] that could be obtained at the measurement point.
Although methods that allow to avoid the use of a reference signal have been recently proposed [14], standard MPE procedure requires the estimation of a reference quantity that is a cell specific signal always ''on air'' and transmitted at maximum power. The standard reference signal adopted in 5G MPE is related to Synchronization Signal/Physical Broadcast CHannel (SS/PBCH) [12]. In particular, we selected the average field level of the Resource Elements (REs) of the highest Synchronization Signal Block (SSB) received in the measurement position (in the following E SSB ). SSBs are periodically transmitted using broadcast beams (Fig. 1). Instruments usually give the power level at the instrument input connector P SSB . From the knowledge of this value, of Antenna Factor (AF) [25] and of the power losses α of the cable connecting the antenna to the receiver, it is possible to obtain the field level: wherein Z in is the input impedance of the instrument. The maximum field level E max 5G can be estimated as: (2) wherein N SC is the number of subcarriers in the whole frequency channel, F TDC is the duty cycle when TDD multiplexing is used and F beam is the ratio between the traffic beam and the broadcast beam in the measurement direction.
In fact, most of radio resources are used for user traffic data, that are transmitted using traffic beams (Fig. 1). Since traffic beams have higher gain compared to broadcast beams, the REs of the user traffic data have higher power compared to the ones used in SSBs. F beam is the boosting factor taking into account such gain difference [12].
F beam can be obtained from measurement, observing traffic data by waterfall reconstruction [10] in a sufficiently long time interval, or using a User Equipment (UE) able to force the traffic toward the measurement position [12]. Furthermore, an estimation of F beam can also be obtained by the  Broadcast Envelope Radiation Pattern (BERP) and the Traffic Envelope Radiation Pattern (TERP) [26]. 2 The above described MPE procedure is based on the structure of the signal frame. The differences between FR1 and FR2 frames involve mainly the numerology and the number of SSBs in Burst, and do not impact on the MPE procedure. Consequently, it can be applied without modifications to FR2.

III. MEASUREMENTS
In this Section the relevant information related to the measurement campaign is given, as well as a statistical analysis of the acquired data.

A. MEASUREMENT SITE
The measurement site is placed in Guidonia Montecelio, a medium-size town ( 87000 inhabitants) north-east of Rome (Italy). The FR2 gNB (see Fig. 2) was on a two floors building, at an height of approximately 14.5 m. The measurement position was at almost 100 meters from the gNB. The communication was in Line of Sight (LOS) condition.
The measurement area is typically suburban with a building structure consisting mainly of low-rise houses (see Fig 3).

B. MEASUREMENT SET UP
The measurement set up was based on a Mobile Network Scanner [27]. The measurement chain is shown in Fig. 4 and consists of:

C. ACQUISITION AND DATA ANALYSIS
Before discussing the application of MPE procedure in next Section, it is useful to enter into some details regarding the acquired data.
The network scanner used during the measurement campaign gives some parameters directly related to the power of the REs of the SSB. One of them is SSS-RePower, that is the average power of resource elements of Secondary Synchronization Signal (SSS) in the SS/PBCH block. This is a quantity suitable for the MPE procedure. Accordingly, we use the SSS-RePower of the SSB received at maximum power as reference value. Two different acquisitions were carried out: first 378 samples were acquired over a period of 654 s; then other 355 samples were acquired over a period of 607 s. Both sample sequences were not uniformly acquired, with a time step less than 5 s.
The area served by the gNB is divided into three sectors. In Fig. 6 the SSS-RePower of the sector in which the measurement position is placed is shown as a function of time (vertical axis) for all the SSBs of 5G frame (horizontal axis). We can note that the SSB are organized in groups of four. The SSB associated to the broadcast beam pointing toward the measurement position is easily identifiable as being #20, due to the higher average level of the REs associated to it. This column contains the data that will be object of analysis.
The SSS-RePower [dB] measured samples of the 20 th SSB are plotted in Fig. 7 for both data sets. The plot shows quite a large variation of the measured values.
The variation of the reference signal used in MPE is not a new phenomenon, and is subject of an active research [28], [29]. However, due to the absence of 5G FR2 gNB, the studies have been carried out mainly on 4G and 5G in FR1. Consequently, in the following the data will be analyzed in detail.
Considering the data in dB, the CDF of transformation Z = (x −x)/σ , x being SSS-RePower samples,x andσ the estimated average and standard deviation respectively (summarized in Table 1), is shown in Fig. 8  This result poses a number of questions. First of all, are these variations related to a variation of the incident field level or to uncertainties of the measurement chain? This requires a detailed analysis of the contribution to uncertainty of the whole measurement channel to identify possible causes of large random variations. Such an analysis will be carried out in the next Section. Indeed, mismatch uncertainty associated to cables working at high frequency can give a fast variation of the measured signals if the cables are long compared to the wavelength, can be bent and have bad phase-stability. However, during the measurements the cables were fixed, and in any case the mismatch uncertainty turns out to be quite small. Loosely speaking, the analysis of the measurement chain will show an uncertainty level incompatible with the large variations of the signal. Our conclusion is that they reflect a phenomenon associated to the incident field.
A further interesting observation is that the distribution is Gaussian in logarithmic scale, i.e. the natural values follow a Log-normal distribution. As well known, Log-normal distributions approximate the product of many independent random contributions.
Variation of the signal versus time is a phenomenon known as fading [30]. Usually, fading is divided in ''fast fading'' and ''slow fading''. The former is caused by fast phase variations of the incident signal in presence of multipath, and has a Rayleigh distribution when there is no LOS path, or a Rice distribution when it is present. Slow fading is caused by 'macroscopic' variations of the environment geometry, e.g. due to the variation of the percentage of urbanization and typology of the buildings in different areas of the town, and follows a Log-normal distribution. Accordingly, the distribution of measured samples seems to match a slow fading-like phenomenon, that however does not have simple physical explanations in the conditions of the measurements carried out.

IV. EVALUATION OF MAXIMUM FIELD
In this Section the MPE is applied to measured data described in previous Section. The procedure is explained step by step, and the uncertainty of each contribution is calculated in last subsection.

A. PARAMETERS REQUIRED FOR THE MPE PROCEDURE
As a preliminary step, some information on the signal object of measurement was obtained from both the operator and the network scanner.
The licensed band assigned to the mobile phone operator is [27.1-27.3] GHz, divided into two equal sub-bands, used in carrier aggregation.
The numerology adopted in the frame is µ = 3. Accordingly the subcarrier spacing is 15 × 2 µ = 120 KHz. From these data it is possible to estimate the number od subcarriers. Due to carrier aggregation, the total bandwidth is 200 MHz, leading to a total number of subcarriers equal to: The Gain (2.6 dB) indicated by the vendor [31] allows to evaluate the Antenna Factor AF (at center band) as: wherein η = 377 ł is the free space impedance, and Z 0 = 50 is the input impedance of the downconverter. The value of the F TDC is standard, and equal to F TDC = 0.743.
The value of power losses of the cable 4 was provided by the vendor of measurement equipment, and is equal to α = −3.2 dB.
Estimation of the F beam represents a critical point. As previously noted, F beam can be estimated by measurements with a quite good accuracy. However, this requires a Scalar Spectrum Analyzer [10] or a Signal Analyzer [12] able to cover FR2. Such equipment was not available during the measurement campaign. The remaining option is to use information on the radiation patterns of the broadcast and traffic beams. However, these data were not available, and information was limited to the maximum gain of the patterns of traffic and broadcast beams. In order to obtain an estimate of F beam we made an educated guess on the patterns. First of all, the connection is in LOS at a distance (almost 100 m) sufficient to accept the far field condition. We assume that the traffic beam is able to point its maximum toward the measurement position. Regarding the broadcast beams, we suppose that the set of beams used to cover the sector overlap at 3dB. This scheme gives a value of F beam ranging from 9 dB to 12 dB (see Fig. 9). Accordingly. we consider a rectangular distribution of F beam from 9 to 12 dB with an estimated average value of 10.5 dB.

B. FIELD LEVEL FROM MPE PROCEDURE
At this point, we have all the parameters required by Eq. 1 and 2.
Regarding Eq. 1, the mean is estimated in Watt for both data sets as: and the corresponding field level of SSB is Finally, the best estimate of maximum mean value of the electric field can be extrapolated from Eq. 2 as It is interesting also to note that the best estimate ofĒ max 5G is only slightly higher than the maximum field value measured using Channel Power measurements, reported in [22], equal to 0.07 V/m.

C. UNCERTAINTY BUDGET
The uncertainty budget is reported in Table 3 [32]. In the following a short explanation on how the estimation of some contributions has been evaluated is reported. When possible, the information provided by vendors was used, in particular for the scanner [27], the downconverter [33] and the antenna [31]. The other uncertainty sources were considered as follows.

1) COAXIAL CABLE
Information on the cable attenuation was provided by the vendor.
The presence of the coaxial cable gives a further contribution to the uncertainty budget due to the presence of multiple reflections between the two terminations of the cable. This contribution is called 'cable mismatch uncertainty' in the uncertainty budget. Due to the random value of the phase in multiple reflections process, the output voltage amplitude ranges in: wherein α is the attenuation of the cable (provided by the vendor), | ant | is the reflection coefficient of the antenna [31] and | dwc | is the reflection coefficient at the input of the downconverter. The mismatch uncertainty has a U shape, and the standard uncertainty is obtained by dividing M by √ 2. The value of the reflection coefficient at the input of the downconverter is indicated in [33] in terms of VSWR (VSWR ≤ 2.4). Therefore the reflection coefficient is The quantities involved in the mismatch uncertainty of the coaxial cable are reported in Table 2.
2) gNB ANTENNA-F beam As discussed in section IV.A, supposing that the set of beams used to cover the sector overlap at 3 dB, F beam value ranges from 9 to 12 dB with rectangular distribution.

3) INCIDENT FIELD FADING
The standard deviation was discussed in Section III.C, and the values are summarized in Table 1 for both data sets. The standard deviation of the mean is obtained dividing by the square root of the number of samples, giving 0.19 dB and 0.21 dB respectively.

4) EXPANSION FACTOR
Since in MPE we are interested in an upperbound of the field, we consider a one tail test of the Gaussian distribution, VOLUME 10, 2022 obtaining an expansion factor (right tail) of 2.33 and an expanded right tail uncertainty with 99% of probability equal to 3.68 dB.
Accordingly the average field level is with 0.99 probability. 3 As a last observation, it must be noted that the result of the MPE is a strong upperbound that supposes that all the resources of the communication system are used by a single user placed in the measurement point. Consequently, the MPE value must be reduced by a correction factor, as recalled in the Introduction, in order to have a realistic value of the field level. This corrected value can be compared to the limits imposed by the legislation. However, the results obtained in this work show that the field level is well below both the Italian legislation (40 V/m) and the ICNIRP reference levels without applying any corrective factor.

V. CONCLUSION
In this paper the MPE methodology is applied to a FR2 signal. The results show that E max 5G ≤ 0.13 V/m with a probability of 99%. This rather low level of field is not completely unexpected, considering that the current state of technology poses significant limits on the generators working in FR2.
Even if the procedure has been developed in the framework of a research project, it is suitable for automatic implementation, including the uncertainty budget estimation. During the measurement only one base station station was active in FR2. However, the measurement equipment is fully able to collect data on all the gNBs eventually present in the measurement site.
The analysis of the uncertainty budget also shows the critical points of the procedure. In particular, F beam estimation represents the most challenging point. The absence of detailed information on traffic and broadcast patterns and on the uncertainty of maximum value of the gain of broadcast and traffic beams forces toward some educated guesses on the pattern. As a consequence, the largest single contribution to the uncertainty budget turns out to be related to F beam term. This problem, discussed also in [12] and [14] pushes toward the use of techniques able to measure the actual F beam value in the measurement session to reduce the uncertainty related to this factor.
It is also worth noting that a further important contribution to the uncertainty budget is related to the probe antenna. In fact, antennas specifically designed for accurate field level measurements in FR2 are not yet on the market and the available antennas often do not pay the required attention to uncertainty evaluation, providing uncertainty ranges wider than necessary. The development of probes specifically designed for measurement in the FR2 band would allow a significant reduction in measurement uncertainty.
A further interesting result of this paper regards the lognormal distribution of the measured quantity. The knowledge of the sampling distribution allows to apply the theory of small samples, offering a rigorous mathematical framework to reduce the number of measurements.
Finally, it is understood that the approach followed in the paper, and based on a rigorous estimation of the uncertainty budget, also allows a rigorous method to compare the different implementations of the MPE technique proposed in literature. Comparison of the specific measurement set-up used in the paper with different MPE technical solutions, based on scalar or vector spectrum analyzers, is out of the scope of this paper, but will be considered in future contributions. company) to provide the experimental data and for their invaluable help in analyzing the measured data. CHIARA LODOVISI received the Ph.D. degree in engineering electronics from the University of Rome Tor Vergata, Italy. She worked for five years, as a RF Engineer Consultant for H3G mobile operator. For five years, she worked on optical communications, study, and implementation of submarine and satellite optical links and radio over fiber. She is currently a Researcher with CNIT, Italy, and the University of Rome Tor Vergata. Her research interests include 5G networks, health risk assessment of 5G communications, interoperability over fiber between TETRA/LTE systems, and 5G networks.
LUCA CHIARAVIGLIO (Senior Member, IEEE) received the Ph.D. degree in telecommunication and electronics engineering from the Politecnico di Torino, Italy. He is currently an Associate Professor with the University of Rome ''Tor Vergata'', Italy. He has coauthored more than 150 articles published in international journals, books, and conferences. His current research interests include 5G networks, optimization applied to telecommunication networks, electromagnetic fields, and health risks assessment of 5G communications. He received the Best Paper Award from the IEEE Vehicular Technology Conference (VTC Spring), in 2020 and 2016, and the Conference on Innovation in Clouds, Internet and Networks (ICIN), in 2018, all of them appearing as the first author. Some of his papers are listed as the Best Readings on Green Communications by the IEEE. Moreover, he has been recognized as an Author in the Top 1% Most Highly Cited Papers in the Information and Communication Technology (ICT) field worldwide and top 2% world scientists according to the 2021 update of the science-wide author databases of standardized citation indicators.
FULVIO SCHETTINO (Senior Member, IEEE) received the Laurea degree (Hons.) and the Ph.D. degree in electronic engineering from the University of Naples, Naples, Italy. He is currently an Associate Professor with the University of Cassino and Southern Lazio, Cassino, Italy. He is a member of the ELEDIA@UniCAS Research Laboratory, the ICEMmB-National Interuniversity Research Center on the Interactions between Electromagnetic Fields and Biosystems, the Italian Electromagnetic Society (SIEM), and the National Interuniversity Consortium for Telecommunication (CNIT). His current research interests include numerical electromagnetics, regularization methods, the connections between electromagnetism and information theory, the analysis, synthesis and characterization of antennas in complex environments, antennas and propagation for 5G, and energetic applications of microwaves. Open Access funding provided by 'Università degli Studi di Cassino e del Lazio Meridionale' within the CRUI CARE Agreement