A Low-Cost, Open-Sourced Platform for

High-Fidelity Characterization of Large WPT Coils

Gregory E. Moore1 , Usman M. Khan2 , Timmy Yang3 , Kedi Yan4 , Tri Nguyen5 , Shi Ming Kuang6 ,
Chase Whyte7 , Vaishnavi Ranganathan8 , Joshua R. Smith9
1,2,4,7,8,9
School of Electrical and Computer Engineering, University of Washington, USA
3,6,9
School of Computer Science and Engineering, University of Washington, USA,
5
School of Mechanical Engineering, San Diego State University, USA
1
[email protected], 2 [email protected], 3 [email protected], 4 [email protected], 5 [email protected],
6
[email protected], 7 [email protected], 8 [email protected], 9 [email protected]

Abstract—The Mostly Printed Field Characterization

System (MPFCS) is an accurate, inexpensive, and open-source
platform for high-fidelity, rapid, large volume WPT coil field
characterizations that provides a significant reduction in scan
duration in comparison to similar simulations.

Keywords—Wireless power transmission, Near-field Coil Char-

acterization, Open-source, Electromagnetic modeling, Magnetic
field measurement, MPCNC

I. I NTRODUCTION
The use of efficient near-field wireless power transfer Fig. 1: MPFCS with attached B-Field end effector character-
(WPT) using magnetic resonant coupling (MRC) is rapidly izing complex relay resonator.
normalizing. This new charging paradigm provides immense
benefits across a range of sectors, from low-power implanted
medical devices, to mid-power wireless consumer electronics negligible permeable structures, let alone those volumes con-
devoid of a single port, and extending to the high-power taining permeable or conductive structures, the computational
systems enabling wirelessly powered commuter busses. Each complexity quickly exceeds even high end servers.
of these energy replenishment technologies are fundamentally To enable the rapid prototyping demanded by the next
dependent on the characteristics of the field created between generation of WPT applications, we present the Mostly Printed
the inductively coupled primary transmitting (TX) and sec- Field Characterization System (MPFCS), Fig. 1: an inexpen-
ondary receiving (RX) coils. As such, characterizing the field sive, open-sourced, easily assembled, high-fidelity with large-
that couples these coils is of paramount importance to real volume power transfer efficiency (S21 ) and B-field character-
application of WPT advances. ization system built around the Mostly Printed Computer Nu-
WPT research has cleared initial validation and is aging merical Control (MPCNC) platform from V1 Engineering [1]
into specialized applications. Measurements of field profiles with end effector modifications to enable measurements from
required to pass maximum permissible exposure conservative tilted and rotated RX. The system operation is demonstrated
checks have become standard operating procedure. This can using two modes of operation: 2-port B-field volume charac-
lead to expensive prototype design testing at specialized facil- terization and a 2-port S-Parameter volume characterization.
ities. Alternatively, abstracted and idealized models of power The tested system covers a volume of (600 × 600 × 300)mm3
transmission have been supplemented by high-frequency sim- with a resolution less than 0.14mm, a tilt arc of 118◦ down to
ulations, such as those provided by ANSYS HFSS, for appli- 1◦ resolution, and pan arc of 180◦ down to 1◦ of resolution.
cation to specific TX and RX setups. However, barring sim- For movements of less than 10cm, the time per scan is
plification to abstract away complex interactions between the less than 4.38s providing a significant improvement in field
environment and the interfacing coils, these simulations require characterization rate by comparison to simulating the field at
immense computation time and eventually code validation for each point within a volume of interest.
regulatory approval. Additionally, in applications where the Omitting the cost of the computer running the PC based
coils are moving or arrayed within large 3D volumes with controller and 2-port Vector Network Analyzer, the system as
assembled and tested costs around 650USD, and with a B-field
978-1-7281-9633-6/21/$31.00 ©2021 IEEE probe addition approaches only 950USD. Costs are kept down
from customized solutions [2] by using easily assembled off 2) Tilt/Pan Head: MPFCS substitutes the MPCNC’s cutting
the shelf components wherever possible. head with a TPH and end effector, Fig. 2a. This provides the
user the capability to tilt and pan (i.e. rotate about the Z Tower
II. M ATERIALS AND M ETHODS axis) the mobile coil under test (CUT), as demanded by the
experiment. The TPH is operated by 2 servo motors (HITEC
A. Mechanical Subsystem RCD USA, Inc.), which provide 118◦ tilt and 180◦ pan, with
The MPFCS can be broken down into three major mechan- repeatable accuracy down ±0.5◦ .
ical subsystems: the MPCNC [1], the Z Tower Tilt/Pan Head 3) Head Tool Mounts: Three probe and calibration end ef-
(TPH) as the end effector, and a series of 3D Printed Mounts. fectors are featured in this paper, each designed in Autodesk’s
The latest mechanical designs and implementation details can Fusion 360 and subsequently 3D printed using ABS plastic.
be found at https://www.thingiverse.com/thing:4729725. a) LED Laser Calibration Mount: This end effector,
1) MPCNC: The MPCNC was developed as a low-cost, Fig. 2a, contains two lasers situated perpendicular to one-
open-source CNC system with an expandable sweep volume. another for calibrating with respect to the X, Y, and Z axes
Alternatively, however, it can serve as a high-fidelity volume movement, as well as the tilt and pan Fig. 2a. The two
sweeping tool due to the precise three dimensional control over mounted, focusable, 5mW red dot LED lasers can be powered
an end effector it provides. The Z Tower raises and lowers an by batteries mounted to the Z Tower or by the RAMBo 1.4
attached TPH and end effector up to the implemented height, directly. A bubble level mounted on the shelf holding the XY
while the XY rails provide precision planar motion control of planar directed laser allows for zeroing out the tilt prior to
the suspended Z Tower. The constraining frame and mobile calibrating the X, Y, and Z axes.
Z Tower are built from simple EMT Conduits, with the CAD b) B-field Characterization Mount: This end effector,
files for the 3D printed joint piece provided for free online Fig. 2c, is a cuboid with a hollowed out space for the Beehive
under a Creative Commons license. 100B magnetic field probe [4]. When the 100B is mounted,
The MPFCS has implemented the MPCNC with an EMT the B-field of the CUT may be converted directly from the S21
tubing frame that provides for a sweep volume of (600×600× measurements made by a connected vector network analyzer
300)mm3 ([X, Y, Z]), which allows for coils centered within (VNA). The 100B probe offers high spatial resolution (loop
the volume with radii up to 212mm to have their realized field, diameter = 3.81mm) for converting S21 measurements to B-
up to their critical coupling distance (CCD), characterized field measurements with calibrated accuracy up to 3.1GHz (1):
within the 300mm [3]. This space accommodates the character-
ization of even large, complex fields involving several arrayed 24547.1 S21
B= 10 20 (1)
coils and the use of large phantoms and substrate blocks to f0
accommodate a variety of complex application environments.
where, f0 is the test frequency and S21 is the power transferred
to the Beehive 100B load from an immobile CUT at f0 .
c) Mobile Universal Coil Mount: This head, Fig. 2b,
provides secure attachment between the TPH and a mobile,
suspended RX coil connected via cables routed along the
Z Tower to one of two VNA ports. This allows for the
characterization of the S21 within the volume swept by the
mobile RX coil with respect to the immobile TX Coil.

B. Electronic Subsystem
(a) (b) The MPFCS uses the RepRap Arduino Mega-compatible
Motherboard (RAMBo) 1.4 board as an integrated controller
and driver solution for controlling the stepper motors directing
the MPCNC XYZ movements, homing to XYZ endstops,
controlling the TPH stepper motors, and interfacing with a
computer via USB. For the implemented system, a HP 8753
two-port VNA provides calibrated measurements as requested
by the PC based controller. The VNA is capable of measuring
S and Z Parameters and transferring them to the PC based
controller. The PC based controller directs the operation of
(c) both the RAMBo 1.4 board and VNA, records the results.
Fig. 2: (a) Tilt/Pan Head with Mounted LED Laser Calibration
C. Software Subsystem
Mount. (b) Universal Coil Mount with a coil mounted consist-
ing of a 50mm spiral offset by 15mm from a 38.5mm loop. (c) The software is divided into RAMBo 1.4 based MPCNC
B-field Characterization Mount with Beehive 100B mounted controller firmware and full system coordinating PC based
 B. Mode 1 - Complex Field Characterization
To demonstrate the time efficiency of 2-port S-parameter
measurements in a volume on MPFCS in comparison to
simulation, a pair of biased repeaters (BR) were swept through
a volume 2x their CCD to measure S21 [X, Y, Z].
BRs are commonly used in WPT setups, because in a
series resonant circuit they allow for the doubling of the
coupling coefficient [6]. Both the TX BR, affixed to the floor
of the MPFCS, and the mobile RX BR, attached to the TPH,
consisted of a loop coil, with an outer diameter (OD) of
Fig. 3: Software Operation 38.5mm, offset by 15mm from a 50mm OD spiral and tuned
to be resonant at 13.56MHz. With a step size of 2.5mm, the
controller with a TkInter GUI for ease of access. The RAMBo resulting NE quadrant of the swept volume is shown in Fig. 5.
1.4 runs on modified open-sourced Marlin firmware [5], which When the two coils are within their mutual near-field, the S21
is optimized for the MPFCS. reduces due to over-coupling. A second region of attenuation
The MPFCS GUI coordinates the timing and intention of in S21 can be seen as a conical frustum sheet extending from
movements sent to the MPCNC stepper motors and receives the OD of the TX. On this plane, a portion of the flux through
measurements requested from the VNA, Fig. 3. After acquiring the RX cancels, thus reducing the measured S21 .
a measurement from a given location, the MPFCS GUI collates On a 32 core, 2.5GHz Intel Xeon CPU with 128GB of
the data recorded from the VNA with the RAMBo position DDR4 2400 RAM, simulating S21 between a simplified digital
information and plots results in real-time for the monitoring twin of the pair of these biased repeater coils at 13.56MHz,
of scans. The latest RAMBo 1.4 firmware, MPFCS GUI, and with 201 points spanning 2MHz in a balanced ANSYS HFSS
instructions for the operation of the MPFCS can be found at simulation, requires 23hrs 24min CPU Time (as reported by
https://github.com/usmank13/MPFCS. ANSYS HFSS) and 3hrs 11min per position interrogated.
Moreover, this duration does not include the time intensive task
impedance matching the coil for resonance at 13.56MHz. By
III. R ESULTS comparison, with step sizes of up to 10mm, the MPFCS takes
only 2.7s per movement and 0.21s per VNA measurement.
A. Calibration Measurements If measuring the S-parameter outputs, this results in a 2600x
MPFCS was evaluated for precision, reproducibility move- reduction in measurement duration.
ment. To assess these qualities, the LED Laser Calibration Demonstrating where the MPFCS system truly excels, the
Mount was first leveled with the horizontal surface on which same 50mm OD biased repeater is swept through the complex
the MPCNC is mounted using the incorporated bubble level. field resulting from a 2-D planar WPT array consisting of 19
To measure the precision of the system, the MPFCS was planar spiral coils, each with an outer diameter of 60mm, and
directed to step through 36 equally spaced points around the driven by a single centered coil [7], Fig. 6. While the MPFCS
trajectory of a circle with a 10mm diameter in the XY, XZ, still requires a constant duration per measurement location, a
and YZ planes as shown in Fig. 4. Marking these points with simulation of even a single location for a 2-port S-parameter
a fine pen, the planar error between the marked step endpoint volume characterization would tax even capable of servers,
and a circle of equal diameter centered at the same origin was and would likely require significant idealization in the design.
averaged over all 36 points in each plane [2]. The average
error was found to be 0.13mm, 0.135mm, and 0.068mm C. Mode 2 - Simple Loop B-Field Characterization
for the XY, XZ, and YZ planes respectively. Additionally, In contrast to characterizing 2-port S-parameter volume,
before proceeding to the endpoint of each step, the LED measuring the B-field of a single TX coil demands only a
Laser Calibration Mount would return to the circle origin. By single simulation run at the frequency of interest. To mimic the
marking this location over 36 return to origin movements, process completed by the MPFCS, a flux surface could then be
the reproducibility is then measured as the area encircled
by the grouping divided by the number of returns to origin,
resulting in a reproducibility error of ±0.025mm, ±0.029mm,
±0.028mm for the X, Y, and Z returns respectively. The TPH
resolution was determined by stepping the respective motors
by multiples of the manufacturer specified minimum step size
until a repeatable noticeable change in the laser projection at
the maximum separation dictated by the MPFCS volume was Fig. 4: The concentric calibration circles in the XZ, YZ, and
observed. The precision limit found to be ±0.5◦ , ±0.5◦ for XY planes demonstrating the precision in the reaching the 36
the tilt and pan respectively. programmed points and the return to origin location
 (a) (b)

Fig. 7: (a) Measured B-Field by the Beehive 100B from a

38.5mm Non-resonant Loop (b) The simulated B-Field from
Fig. 5: The S21 characterization of a volume where a biased a digital twin of the 38.5mm Non-resonant Loop.
repeater RX coil (inset top-right, but inverted from its sweep
then shown in Fig. 7a.
orientation) is swept at 2.5mm steps through the field resulting
Once the two S21 outputs were aligned, the system achieved
from a twin biased repeater TX coil.
0.77dB Normalized RMSE over the full matrix of measure-
ment locations, well within the 3dB quoted accuracy specified
for the Beehive 100B probes.
IV. D ISCUSSION
Although the MPCNC was intended to be used as a CNC,
by changing the end effector it becomes a versatile tool for
repeatedly executing any task across a volume. While this
work demonstrated high-precision volumetric characterization
of fields, the system could be repurposed for other types of
sensing (EMC Testing) and measurement (3D Scanning).
The high-fidelity measurements possible with the MPFCS
will enable statistical model development (e.g. physics driven
machine learning) for systems that are prohibitively complex
or impractical to model analytically. This can feedback into the
Fig. 6: The S21 characterization of a volume where a biased coil design and prototyping process as well as provide insights
repeater RX coil (inset bottom-right, but inverted from its into edge cases for future WPT application.
sweep orientation) is swept at 2mm steps through the complex
field resulting from a 2-D planar WPT array ACKNOWLEDGMENT
We thank the WPTC reviewers for their helpful feedback.
swept through the locations of interest in post-processing and This work was supported by University of Washington Institute
the average B-field through the surface could be calculated. for Neuroengineering and, in part, by awards from the National
For example, a simple 38.5 OD loop coil without impedance Science Foundation (EEC-1028725) and U.S. Department of
matching, Fig. 7b, requires 47min CPU Time and 22min Defense (W81XWH-20-1-0010).
Real Time, neglecting post-processing time. By comparison, R EFERENCES
sweeping the Beehive 100B at 2mm steps through a quarter
[1] “The mostly printed cnc,” 2019. [Online]. Available:
volume with extents equal to 2x the CCD requires more than https://docs.v1engineering.com/mpcnc/intro/
9hrs. This being said, as additional high permeability and [2] H. Han, R. Moritz, E. Oberacker, H. Waiczies, T. Niendorf, and
conductive structures, e.g. biological phantoms, are added to L. Winter, “Open source 3d multipurpose measurement system with
submillimetre fidelity and first application in magnetic resonance,”
the simulation environment the time efficiency of simulation Scientific Reports, vol. 7, no. 1, p. 13452, 2017. [Online].
rapidly deteriorates. Available: https://app.dimensions.ai/details/publication/pub.1092200010
To verify the ability of the MPFCS to effectively character- and https://www.nature.com/articles/s41598-017-13824-z.pdf
[3] R. Shadid and S. Noghanian, “A literature survey on wireless power
ize the B-field resulting from a given coil, a physical twin of transfer for biomedical devices,” International Journal of Antennas and
the simulated single loop coil above, which has been matched Propagation, vol. 2018, 2018.
to the simulated S11 at 13.56MHz and 0dBm was chosen as [4] “100 series emc probes,” 2020. [Online]. Available: https://www.beehive-
electronics.com/datasheets/100SeriesDatasheetCurrent.pdf
a computationally uncomplicated reference for comparing the [5] “Marlin firmware,” 2020. [Online]. Available: https://marlinfw.org/
measured B-field from the MPFCS with a simulated replica. [6] D. Ahn and S. Hong, “A study on magnetic field repeater in wireless
The 38.5mm CUT was centered at X, Y = (300mm, 300mm) power transfer,” IEEE Transactions on Industrial Electronics, vol. 60,
no. 1, pp. 360–371, 2012.
and affixed to the floor of the MPFCS and the Beehive 100B [7] X. Shi and J. R. Smith, “Reconfigurable and adaptive coupled relay
probe was then swept in 2mm steps through the quarter volume resonator platform for a moving receiver,” in 2019 International Workshop
up to the CCD. The resulting B-field at 13.56MHz, Eq. 1, is on Antenna Technology (iWAT), 2019, pp. 182–185.