A Multi-Frequency 3D Printed Hand Phantom for Electromagnetic Measurements

Abstract

It has been shown that the presence of a hand holding a wireless handset (cell phone) can influence antenna efficiency and the measurement of specific absorption rate (SAR) and electromagnetic compatibility. Head phantoms, used in handset compliance testing to estimate SAR in the head, have achieved low cost and multi-frequency use. Head phantoms typically consist of a thin plastic shell, open on the top, holding a tissue simulating fluid. The specific simulant fluid used is determined by the radio frequency of the test. IEC 62209–1 has recipes, using safe nontoxic materials, for all the required frequency bands. Thus, head phantoms can be reused at different frequencies simply by changing the tissue simulating fluid. However, standards have not adopted the use of hand phantoms because SAR limits in limbs are less restrictive than the head, the tissue depth in a hand is insufficient to make accurate measurements with current electric field probes, and the cost of a solid hand phantom is limited to a single frequency band. Our goal was to determine whether 3D printing techniques would allow the construction of a hand phantom with the same utility as existing head phantoms. We developed this phantom based on computer simulations to determine how much human anatomy needed to be included in the phantom to obtain results consistent with actual use. Electric field scans of a handset alone, and held by the hand phantom, were performed. Comparison of handset scans using the phantom and human subjects was planned, but not performed due to Covid-19 restrictions and subsequent changes in priorities. We feel a fluid-filled 3D printed hand phantom is viable and practical. The 3D print files are available on GitHub.

I. Introduction

The use of hand phantoms for electromagnetic measurements has been debated for years. Standards IEC/IEEE 62209–1528:2020 [1] and IEC 62209–1:2016 [2] have detailed sections explaining why they do not use hand phantoms for compliance testing. During normal operation, the head and hand are in the near-field of the handset and both absorb radio frequency (RF) energy. The specific absorption rate (SAR) limit for extremities, such as the hand, is higher than the limit for the head [3,4]. Numerical and experimental studies have shown that the SAR in the hand at the power levels used by typical handsets is not expected to exceed those limits [5–10]. Therefore, SAR measurement in the hand is not addressed by exposure compliance standards.

Balzano et al. [11] and Kuster et al. [12] reported that the presence of the hand either decreased or had no significant effect on the head SAR; however numerical results by Meyer et al. [13] showed one case with 7 % increase in head SAR due to the hand. Li and Douglas et al. [14,15] reported use of a hand phantom decreased head SAR and increased standard deviation. Beard et al. [16] reported human hands changed the position and magnitude of the peak electric field measured above a handset. Further, they found use of an individual’s left or right hand had no significant effect on the change in magnitude.

In spite of their lack of acceptance in SAR compliance standards, computational hand models and solid hand phantoms have found acceptance in the handset design process. Their use is required for antenna design and optimization, particularly for 5G handsets [17–19].

The hand phantom described in this paper is not intended for the measurement of SAR in the hand; it is intended to simulate the effect of the hand on the emissions of the handset. A phantom of this type could be used to simulate the effect of the hand for handset design optimization, compliance testing or for hearing aid compatibility testing [20].

The Specific Anthropomorphic Mannequin, referred to as the SAM head phantom, used for handset compliance testing [1] consists of an open fluid-filled plastic shell. It has several attractive features: it is light weight, it can be manufactured for a reasonable cost, and it can be used at any frequency simply by changing the fluid in the phantom. IEC 62209–1 [2] has simulant fluid recipes for all handset bands from 300 to 5800 MHz which use safe nontoxic materials. The SAM phantom is used to estimate SAR in the head; therefore it must be open so that electric field probes can be immersed in the fluid simulating head tissue. Because the purpose of the hand phantom is to simulate the effect of the hand on the RF radiation pattern of the handset, and not to measure SAR in the hand, then it can be a totally enclosed shell. This also means the hand phantom can be placed in any orientation without spilling the simulant fluid.

It has clearly been shown that the hand has a significant effect on the radiation pattern from a handset. While other hand phantoms have been proposed [15,21] we felt the low cost and multi-frequency utility of a 3D printed hand phantom was worth investigating.

II. Simulation Materials and Methods

The 3D printed phantom we designed is, of necessity, a homogeneous phantom because it is one fluid-filled chamber. First, we performed simulations to estimate the effect of a homogeneous full-arm phantom versus a heterogeneous full-arm. Then we performed simulations to determine how much of the upper-arm anatomy could be eliminated from the homogeneous phantom without altering the phantom’s effect on the electric field of the handset. Simulations were confirmed with two different handset models. All simulations were performed with XFdtd [22] 3D electromagnetic simulation software; Matlab [23] was used for analysis and data plotting.

A. Anatomical Model

The anatomical model used was the 4D XCAT female full-body phantom [24] which was developed to provide virtual patients for medical imaging research. We extracted the heterogeneous left arm which was 834 mm from finger-tip to shoulder. The heterogeneous model arm was separated into five tissue types: skin, bone, muscle, fat, and blood - where blood includes all vasculature. Electrical tissue properties [25] are shown in Table 1. The 834 mm heterogeneous arm model was converted to a homogeneous model with an insulating surface. The homogeneous phantom’s electrical properties come from IEC 62209–1 [2] and are also shown in Table 1. The upper portions of the 834 mm homogeneous arm model were progressively removed until we produced a model that yielded almost the same simulation results as the 834 mm homogeneous arm model while still being large enough to hold a handset. The final hand model was 270 mm in length, see Figure 1.

Table 1.

Electrical properties of materials at 900 and 1800 MHz.

Tissue	σ (S/m) 900 MHz	ε 900 MHz	σ (S/m) 1800 MHz	ε 1800 MHz
Skin	0.87	41.4	1.18	38.87
Bone	0.14	12.5	0.27	11.78
Muscle	0.94	55.0	1.34	53.55
Fat	0.05	5.46	0.08	5.35
Blood and vasculature	1.54	61.36	2.04	59.37
Tissue simulant fluid	0.97	41.5	1.4	40.0
Phantom shell	0	3.7	0	3.7

Figure 1.

Arm models holding a handset model. Panel A is the 834 mm heterogeneous arm model. Panel B is the 834 mm homogeneous arm model with insulating surface. Panel C is the final hand model which is the homogeneous arm model in B truncated to 270 mm.

B. Handset Models

The simulations were done with two different handset models. The models were the GSM/UMTS mobile handset and the Neo Free Runner mobile handset [26]. Both handsets are dual band, and the simulations were performed at 900 and 1800 MHz. These models are freely available and well validated.

C. 3D Printed Phantom

A print file was created using the 270 mm model from the simulations. A shell was created around the model which, because of the irregular surface, varied in thickness from 1.5 mm to 2.5 mm with the average thickness being 2.2 mm. The internal volume of the hand phantom is approximately 350 ml.

The phantom was printed by an EOS P 396 selective laser sintering system at 120 μm per layer using PA 2200 nylon-12 (relative permittivity 3.8 and loss tangent 0.05). The phantom was printed in two parts so that the unsintered nylon particles could be removed before assembly. The two parts are the hand shell and a wrist plug which includes two 6 mm hose connections to fill and drain the phantom. Because the phantom has no flat surfaces, a holder was needed to keep it steady during measurements. A holder was printed and an alignment pin was added to the back of the hand, see Figure 2.

Figure 2.

Components of the 3D printed hand phantom. A shows the hand shell. B shows the wrist plug. C shows the holder.

III. Laboratory Materials and Methods

A. Handsets

The handsets used in the simulations are older models and were not available for use in laboratory testing. We used two different handsets in the lab, i.e., Samsung Galaxy S9 and S21 Ultra smartphones, both of which support the 900 and 1800 MHz bands. To ensure the RF link remained active for an entire scan, both handsets ran the Android Application, Magic iPerf [27], which continuously tested the instant link speed with our base station emulator through UDP streams over the established cellular connection. The handsets were connected to our Valid8 base station emulator [28].

B. Scanning

Scans were performed with a DASY5 Neo and ER3DV6 miniature electric field probe [29] as seen in Figure 3. Each handset’s electric field magnitude was measured at 897 and 1747 MHz. As was done in the simulations, the electric field was measured on a one cm grid 15 mm above the face of the handset. Data were transferred to Matlab [23] for analysis and data plotting. Scans were done with both phones at both frequencies for three configurations:

• Phone alone, no phantom

• Phone and 3D printed hand phantom

• Phone and flat (rectangular) phantom

Figure 3.

Handset and 3D printed hand phantom in position for scan. E-field probe is visible at the top-left of the picture. Tubes to fill and drain the hand phantom are connected to the wrist plug.

C. Phantoms

The 3D printed hand phantom was filled with tissue simulating fluid using the recipes from IEC 62209–1 [2] for 900 and 1800 MHz. When scanning at a particular frequency was finished, the simulant fluid was drained into its storage container, the phantom was flushed with water for an hour and then flushed with dry air for an hour. Figure 4 diagrams the plumbing connections to the 3D printed hand phantom.

Figure 4.

The valves (clamps) in the fill and drain lines are both open when filling the hand phantom with simulant fluid, flushing it with water, or drying it with air. The hand phantom is filled with simulant fluid until it begins to emerge from the drain line, then both valves are closed. With both valves closed the hand phantom can be placed in any orientation without loss of simulant fluid.

The flat (rectangular) phantom was a plastic box, 1.6 mm thick, with an internal volume of approximately 500 ml (12.7 × 17.7 × 2.2 cm). The flat phantom was filled with the appropriate tissue simulating fluid and the handset was centered on its top surface during the scan.

IV. Simulation Results

A handset model was placed in the palm of the hand model and the simulation computed the electric field at centimeter spacing in an 8 by 15 cm grid (120 total points) parallel to, and 15 mm above the face of the handset. The 15 mm measurement distance is based on the requirement of ANSI C63.19. The three anatomical models of concern are: the 834 mm heterogeneous arm model, which we considered the most realistic; the 834 mm homogeneous model and the final 270 mm hand model. Figure 5 shows the simulations for the GSM/UMTS handset model and the three anatomical models. Plots of the other simulations are similar but are not included in this paper for brevity.

Figure 5.

Electric field (volts/meter) simulation results for the GSM/UMTS handset model and the three anatomical models.

Table 2 shows the descriptive statistics for all the simulations. For each combination of handset and frequency, the anatomical models were compared using the normalized root mean square deviation between corresponding scan points.

$$\sqrt{\frac{{\sum }_1^N{\left(E_{Ai}-E_{Bi}\right)}^2}{N}}$$

Where: N = number of scan points

• E_Ai = Electric field value in scan A,i’th point

• E_Bi = Electric field value in scan B,i’th point

Table 2.

Statistics of electric field simulation results.

Handset model	Frequency (MHz)	Anatomical model	Max (V/m)	Mean (V/m)	Standard Deviation (V/m)
GSM/UMTS	900	Heterogeneous 834	10.467	4.940	2.258
		Homogeneous 834	11.316	5.377	2.371
		Homogeneous 270	11.385	5.368	2.413
	1800	Heterogeneous 834	8.311	3.710	1.737
		Homogeneous 834	10.027	4.419	2.045
		Homogeneous 270	10.075	4.421	2.049
Neo Free Runner	900	Heterogeneous 834	3.048	1.226	0.736
		Homogeneous 834	3.631	1.261	0.797
		Homogeneous 270	3.587	1.234	0.818
	1800	Heterogeneous 834	0.840	0.221	0.179
		Homogeneous 834	0.900	0.242	0.191
		Homogeneous 270	0.893	0.241	0.189

Table 3 shows the result of these comparisons.

Table 3.

Root Mean Square Deviation (RMSD) between different model combinations.

Handset model	Frequency (MHz)	Anatomical models	RMSD (V/m)
GSM/UMTS	900	Heterogeneous 834 – Homogeneous 834	0.5930
		Heterogeneous 834 – Homogeneous 270	0.5835
		Homogeneous 834 – Homogeneous 270	0.1154
	1800	Heterogeneous 834 – Homogeneous 834	0.9085
		Heterogeneous 834 – Homogeneous 270	0.9272
		Homogeneous 834 – Homogeneous 270	0.0694
Neo Free Runner	900	Heterogeneous 834 – Homogeneous 834	0.2360
		Heterogeneous 834 – Homogeneous 270	0.2161
		Homogeneous 834 – Homogeneous 270	0.0903
	1800	Heterogeneous 834 – Homogeneous 834	0.0355
		Heterogeneous 834 – Homogeneous 270	0.0356
		Homogeneous 834 – Homogeneous 270	0.0048

V. Experimental Results

Because the handsets scanned in the laboratory were larger than the models used in the simulation it was necessary to increase the number of scan points. A 12 × 21 cm grid was used instead of the 8 by 15 cm grid resulting in 252 points instead of 120 points. Figure 6 shows the scan results for the scans for the S9 handset for both frequencies and the three configurations. Plots of the other results are similar but are not included in this paper for brevity.

Figure 6.

Electric field (volts/meter) measurements for the Samsung S09 handset in three configurations.

Table 4 shows the descriptive statistics for all the scans. The scans were compared using the normalized root mean square deviation between corresponding scan points, as shown in Table 5.

Table 4.

Statistics of laboratory scan results.

Handset model	Frequency (MHz)	Phantom	Max (V/m)	Mean (V/m)	Standard Deviation (V/m)
S9	897	None	86.33	18.64	13.90
		Hand	78.23	17.30	13.22
		Flat	47.00	11.04	8.18
	1747	None	54.90	14.77	10.15
		Hand	53.03	13.89	10.30
		Flat	39.10	8.27	6.62
S21 Ultra	897	None	77.60	16.09	13.19
		Hand	76.13	15.66	13.51
		Flat	44.8	9.86	7.57
	1747	None	68.23	17.57	14.02
		Hand	74.27	18.93	15.41
		Flat	40.60	8.26	7.69

Table 5.

Root Mean Square Deviation (RMSD) between different phantom combinations.

Handset model	Frequency (MHz)	Phantoms	RMSD (V/m)
S9	897	None ↔ Hand	2.869
		None ↔ Flat	10.447
		Hand ↔ Flat	9.101
	1747	None ↔ Hand	2.567
		None ↔ Flat	8.423
		Hand ↔ Flat	7.803
S21 Ultra	897	None ↔ Hand	1.721
		None ↔ Flat	9.105
		Hand ↔ Flat	8.891
	1747	None ↔ Hand	2.933
		None ↔ Flat	11.741
		Hand ↔ Flat	13.603
S9 ↔ S21 Ultra	897	Hand	5.565
S9 ↔ S21 Ultra	1747	Hand	8.366
S9	897 ↔ 1747	Hand	8.772
S21 Ultra	897 ↔ 1747	Hand	10.579

VI. Discussion

We knew the simulated handsets represented outdated models that would not be available for laboratory testing. However, they were the most realistic computer models of handsets available, and we did not intend a quantitative comparison of the simulation results to the laboratory measurements. Our intent was to compare scans where human subjects held the handsets to scans where the handsets were held by the hand phantom. Unfortunately, the Covid-19 pandemic caused the cancellation of our approved human use protocol. Consequently, we can make no definitive statements about the ability of our 3D printed hand phantom to accurately mimic the interaction of a human hand with a handset.

The tissue simulant fluids specified in IEC62209–1 are for the head and are formulated to produce conservative SAR results for compliance testing. In other words, they absorb more RF energy than typical tissue. As such, we would expect the energy absorbed by the phantom to reduce the electric field measured above the handset. Table 4 shows the maximum and mean E-field are reduced from the no-phantom case for seven of the eight cases of the handset with phantom. This is particularly true for the flat phantom because its flat surface puts it in more intimate contact with the flat back surface of the handset. The one case where the maximum and mean increase could be explained by the hand phantom detuning the 1747 MHz antenna to shift the lobes of the antenna pattern [13,16]. As such, the experimental results comply with expectations.

The hand phantom was easy to use. It was easy to fill using a large plastic syringe to inject simulant fluid into the input tube. Draining was accomplished by injecting air into the input tube, placing the drain tube in the storage container while holding the drain port of the hand phantom below the input port. Cleaning was easily accomplished by running water through the phantom for an hour followed by air to dry the interior of the hand phantom. Thus the hand phantom could be changed from one frequency band to another in about two hours.

VII. Conclusion

Our simulations show a hand phantom need not include the arm. This reduction in size requirement makes local manufacture possible and affordable. We have shown that use of a 3D printed shell type phantom is practical and offers several advantages. Unlike an open tank type phantom, a fully enclosed shell phantom can be held in any orientation without spilling the simulant fluid. Like a tank phantom, it can be used for different frequency bands simply by changing the simulant fluid.

While we have shown it is practical to use a 3D printed hand phantom, further research is required to quantitively show such a phantom accurately mimics the interaction of the human hand with a handset.

Biographies

Brian B. Beard received the B.S. in electrical engineering from the U.S. Air Force Academy, USAFA, CO, in 1973; the M.B.A. from the University of West Florida, Pensacola, FL, in 1989; the M.S. in biomedical engineering from Vanderbilt University, Nashville, TN, in 1993; and the Ph.D. in biomedical engineering from Vanderbilt University in 1995. He flew for the Air Force from 1973 to 1979, after which he worked for the Air Proving Grounds at Eglin A.F.B., FL, in the millimeterwave systems division. He designed radar systems to track both ground and air targets. His Ph.D. thesis work was based on instrumentation for in vivo measurement of cardiac dynamics. He is currently the deputy director of the Division of Biomedical Physics at the U.S. Food and Drug Administration’s Center for Devices and Radiological Health, Silver Spring, MD. He is an author on over a dozen papers on medical instrumentation and specific absorption rate (SAR) measurement. His current research interests include RF dosimetry and auditory prosthetics. Dr. Beard is a contributor to IEEE Standard 1528 and ASTM F2504.

Maria I. Iacono received a Master in biomedical engineering in 2007 and a Ph.D. in biomedical engineering in 2011 from the Polytechnic of Milan, Italy. In 2010 she was awarded a fellowship from the Rocca Foundation at Massachusetts Institute of Technology (MIT) to join the Martinos Center for Biomedical Imaging (Harvard-MIT HST Massachusetts General Hospital). From 2012 to 2018 she was with the U.S. Food and Drug Administration with both research and regulatory duties. Since 2018 she has worked in industry as Director of Regulatory Affairs. She is currently the author of over 18 peer-reviewed journal articles in high impact factor scientific journals, including Nature Scientific Reports, Neuroimage, Frontiers in Physiology, and PlosOne, and a number of conference proceedings. Her research interests include medical imaging, image processing, and computational modeling of virtual patients and medical devices.

Joshua W. Guag received the Bachelor’s degree in Biological Resource Engineering from the University of Maryland in 2008 and the M.S. degree in Electrical Engineering from George Washington University in 2013. He is a research engineer in the Center for Devices and Radiological Health of the Food and Drug Administration in Silver Spring, MD.

Yongkang Liu received the Ph.D. in Electrical and Computer Engineering from the University of Waterloo, Waterloo, ON, Canada in 2013. He is currently a staff fellow working with the Center for Devices and Radiological Health (CDRH), U.S. Food and Drug Administration (FDA). Before joining FDA, he worked at the National Institute of Standards and Technology (NIST), Gaithersburg, MD on wireless research in vertical sector applications, e.g., process/automation control and industrial robotics. His current research is focused on investigating novel uses of wireless technology (5G/Wi-Fi/Bluetooth) in healthcare applications and test methods to promote safe and effective use of wireless technology in medical devices.