IEEE SENSORS JOURNAL, VOL. XX, NO.

XX, XXXX 2023 1

Planar On-Silicon Inductor Design for

Electromagnetic Tracking
Aleksandr Sidun, Manish Srivastava, Kilian O’Donoghue, Herman Alexander Jaeger, Marco Cavaliere
Member, IEEE, Daniel O’Hare Member, IEEE, Christian van den Bosch, Pádraig Cantillon-Murphy Senior
Member, IEEE

Abstract— In this paper, a planar on-silicon inductor design for

electromagnetic tracking is presented. The design process has been
demonstrated for a planar sensing coil in a 65 nm on-silicon process.
The designed multilayer planar inductor configuration achieves SNR
of 23.6 dB and sensitivity of 202 µV/T/Hz. The designed inductor
satisfies system requirements for clinical use in electromagnetic
tracking applications. The current design has been verified on the
TSMC 65 nm process and exhibits a close correlation with simulation
results. Measured SNR of 22.1 dB and sensitivity of 198 µV/T/Hz at a
frequency of 20 kHz and magnetic flux density of 150 nT is achieved.
An improved lumped element model to account for multi-layer planar
inductor designs is also proposed.
Index Terms— CMOS, electromagnetic sensing, EMT, inductor, PCELL, SKILL programming language, virtual bronchoscopy.

I. I NTRODUCTION navigation without using a real-time X-ray. This is possible by

combining a pre-operative CT scan with the EMT data. One

E LECTROMAGNETIC tracking (EMT) is a critical tech-

nology for tracking and navigation in image-guided
interventions, when there is no line of sight to the target [1].
clinical example of EMT is for airway navigation [6]. When
the diameter of the bronchoscope exceeds the diameter of the
airways, navigation can be continued with EMT, where tracked
Clinical applications which currently exploit EMT for virtual medical instruments are advanced further into the lung under
navigation include bronchoscopy [2] [3], neurosurgery and virtual navigation. The basic overview of this system is the
ear, nose and throat surgery [4] [5]. In general, an EMT basis of most commercial EMT systems as shown in Fig. 1.
system relies on two main components: a field generator, which
generates a spatially dependent magnetic field, and a magnetic Airgap
sensor, which, by field measurement, can determine its relative Signal AC Field Tracking
position and orientation with respect to the field generator. One Generator Amplifier Generator Sensor Coil
of the two components is fixed, while the other is tracked.
Prior work [2] described the open-hardware electromagnetic
P&O Data Sensor
tracking system called Anser EMT which allows for virtual Algorithm
Demodulation
Acquisition Amplifier

This work has been submitted for review 13.03.2023. This work was
funded by the European Union through the ERC Consolidator Grant Fig. 1. Block diagram of the Anser EMT system.
Deep Field no.101002225.
A. Sidun is with the Tyndall National Institute, Cork, Ireland (e-mail: A signal generator circuit provides a reference sine wave
aleksandr.sidun@tyndall.ie)
M. Srivastava is with the Tyndall National Institute, Cork, Ireland (e-mail: for each emitter coil channel. The signal is amplified and
manish.srivastava@mcci.ie) transmitted through the field generator coil array on a printed
K. O’Donoghue is with the Tyndall National Institute, Cork, Ireland circuit board (PCB). A tracking sensor coil produces an induced
(e-mail: kilian.odonoghue@tyndall.ie)
D. O’Hare is with the Tyndall National Institute, Cork, Ireland (e-mail: alternating voltage due to the field generator. The sensed
daniel.ohare@tyndall.ie) signal is amplified and sampled. The resulting samples are
H. A. Jaeger is with the Tyndall National Institute, Cork, Ireland (e-mail: demodulated and processed through the position and orientation
alexander.jaeger@ucc.ie)
M. Cavaliere is with the Tyndall National Institute, Cork, Ireland (e-mail: algorithm to yield a position and orientation vector for the
m.cavaliere.ucc@gmail.com) tracking coil [2] [7]. The dimensions of existing discrete
C.vdBosch is with the Tyndall National Institute, Cork, Ireland (e-mail: inductive sensor coils places a constraint on the mechanical
c.vandenbosch@tyndall.ie)
P. Cantillon-Murphy is with the School of Engineering, University design of the instrument and the spatial resolution of the sensor.
College Cork, Cork, Ireland (e-mail: padraig@alum.mit.edu) The instrument must be typically rigid (non-flexible) for 10 mm
2 IEEE SENSORS JOURNAL, VOL. XX, NO. XX, XXXX 2023

and the sensor coil should consist of several hundred turns to magnetic field linking it changes. The temporal variations of
achieve the required sensitivity (Fig. 2). Otherwise, there is a the magnetic flux Φ(t) through the N turn inductor will induce
risk of a sensor lead breakage due to the instrument bending [8]. a voltage which follows:
dΦ(t)
e(t) = −N (1)
dt
H
where Φ = s B · dS is the magnetic field B linked by one
turn, assuming they all have the same area S hereby:
I
d
e(t) = −N B · dS (2)
dt s
The surface integral can be replaced by the discrete term
Fig. 2. A typical electromagnetic inductive sensor.
summation for numerical evaluation:
Moreover, the induced voltage from the sensor transmitted N
d X
through the long cable introduces additional noise and decreases e(t) = − B · Ai (3)
the accuracy of the system. These shortcomings with existing dt i=1
discrete inductive coil-based sensors makes an integrated on-
where N is the number of turns, Ai is the area of the ith
silicon solution very attractive. While this paper specifically
turn and B is the flux density at the centre of the coil. For the
describes the optimisation of the magnetic sensor coil, the
sensors that are small relative to the spatial variation of the
overall goal of this work is to implement the tracking sensor
magnetic flux density, the variation in flux density across the
coil, sensor amplifier, filter and ADC on a single die [9]. This
sensor cross-section is usually linear, so (3) may be simplified:
solution provides the following benefits:
• Significant physical constraint reduction (decreasing sen- N
dB(x, y, z, t) X
sor size from 5.5×0.5 mm to 1.5×0.5 mm); e(t) = − Ai (4)
dt i=1
• Reduced sensor noise and losses in the cable resulting in
accuracy improvement due to the single die implementa- Assume the magnetic field detected by the sensor is sinu-
tion; soidal (as is the case for the Anser EMT system) with amplitude
• Accuracy improvement, as the magnetic flux density point B0 :
approximation is better achieved with a small die than
with a long 10 mm coil. B(x, y, z, t) = B0 (x, y, z) sin(ωt) (5)
where ω = 2πf0 is the received field angular frequency in
II. P LANAR I NDUCTOR B ASICS
radians per second. Then (4) can be represented as:
A. Basic Planar Inductor Structure
N
The planar inductor is a two-port device in the shape of X
e(t) = −ωB0 (x, y, z) cos(ωt) Ai (6)
a rectangular spiral that winds from the edge to the center
i=1
and passes outside through the underpass layer, as depicted
in Fig. 3. The main geometric parameters of the inductor The resulting magnitude of the induced voltage Vs for the
indicated in Fig. 3 are: nominal magnetic flux B0 may be made explicit:
• Number of turns N ; N
X
• Width of trace w; Vs = 2πf0 B0 Ai = k s B 0 f 0 (7)
• Spacing between traces s; i=1
• Outer dimensions of the coil (width a and length b).
where f0 is the received signal frequency in Hz and ks is the
s w sensitivity of the sensor in Volts per Tesla per Hertz (V/T/Hz):
N
X
Port 1 ks = 2π Ai (8)
b
i=1
Port 2
Considering input signal to be sinusoidal, the root mean
square (RMS) value of the signal is:
Underpass
a q
layer Vs k s B0 f0
Vs 2 = √ = √ (9)
Fig. 3. The basic planar inductor structure. 2 2
Since the inductor is not ideal, the performance of the
In the present application, the planar inductor is an inductive coil will be limited by the thermal noise contributed by the
sensor of electromagnetic field that uses the principle of parasitic resistance of the coil trace Rs . According to the
electromagnetic induction. According to Faraday’s law of Johnson–Nyquist formula [10], the mean square noise voltage
induction, a voltage is induced across an inductor when the for a given bandwidth ∆f of the inductor can be expressed as:
SIDUN et al.: PLANAR ON-SILICON INDUCTOR DESIGN FOR ELECTROMAGNETIC TRACKING 3

Vn 2 = Sn (f )∆f = 4kB T Rs ∆f (10) s1 = a − N w − (N − 1)s (17)

where ∆f is the sampling bandwidth in Hz, Sn (f ) is the
noise voltage spectral density, kB is the Boltzmann constant and s2 = b − N w − (N − 1)s (18)
T is the absolute temperature in K. Thus, the signal-to-noise
ratio (SN R) of the inductor in is given by: q
g= s21 + s22 (19)
q
Vs 2 ks B0 f0 where s1 and s2 are the geometric parameters of the coil in
SN R = q =√ (11) µm, a and b are the width and length of the coil in µm, w is
8k B T Rs ∆f
Vn 2 the width of the trace in µm, s is the spacing between traces
In a standard applications, Q-factor is a very important in µm, N is the number of turns, g is the average diagonal
characteristic of the inductor because it refects resonating ability of the coil and t is the thickness of the trace in µm (Fig. 3).
of the inductor. In EM-sensing application, inductor is operating Coefficient values are: k1 = 9.21 · 10−7 , k2 = 2.54 · 104 ,
at a very low frequencies (10-100 kHz) and captures EM field k3 = 0.01016, k4 = 0.447 [13]. The resistance of the coil can
by Faraday’s law of electromagnetic induction and does not be calculated from the sheet resistance of the metal:
employ resonance. Therefore, Q-factor of the inductor is not l
considered in this work. Rs = Rsheet (20)
w
where Rsheet is the on-silicon sheet resistance of the metal
B. The Lumped Physical Model of a Planar Inductor in Ohm per square (Ω/□), l and w are the length and width of
A planar inductor is a complex structure with multiple the trace in µm, respectively. Oxide and substrate capacitances
physical parameters distributed over the device. However, a can be calculated from well-known formulae, where ϵox is the
planar inductor can be replaced by the equivalent lumped absolute permittivity of the oxide, tox is the oxide thickness
model, such as that proposed in [12] and shown in Fig. 4. and Csub is the substrate capacitance per square unit:

Cs 1 ϵox
Cox = lw (21)
2 tox
RS LS
1
CSi = lwCsub (22)
2
Cox Cox The series feedforward capacitance Cs which accounts for
the capacitance due to the overlap between the inductor spiral
RSi and the underpass layer, is given by:
CSi CSi RSi
1 wup wN ϵox
Cs = (23)
2 tox
Fig. 4. Lumped physical model of a spiral inductor.
where wup is the width of the underpass layer in µm and
tox is the oxide thickness between the inductor spiral and
This model requires the value of the resistance of the
underpass layer in µm. As follows from [14], the combined
trace Rs , the resistance of the substrate RSi , inductance Ls ,
impedance of Cox , CSi and RSi can be combined in Cp and
underpass capacitance Cs , oxide capacitance Cox and substrate
Rp :
capacitance CSi . For a rectangular planar coil inductance
1 RSi (Cox + CSi )2
(in µH) can be calculated by the formula [13]: Rp = 2 2 + 2
(24)
ω Cox RSi Cox

Ls = k1 A0 (A1 − A2 − A3 )+ 1 + ω 2 (Cox + CSi )CSi RSi

2
Cp = Cox 2 2 2 (25)
g t+w 1 + ω (Cox + CSi ) RSi
+k3 A0 (2 − 0.5) + k4 ] (12)
s1 + s2 s1 + s2 tsub
RSi = Dout −Din
(26)
2 0.5(l + 2 )wGsub
A0 = (s1 + s2 )N (13)
where Gsub = 1/Rsub is the conductivity of the silicon
2s1 s2 substrate. This capacitance is in parallel with the feedforward
A1 = log10 (14) capacitance Cs . The total equivalent capacitance of the inductor,
k2 (t + w)
can thus be calculated as:
s2 s1 + g
A2 = log10 ( ) (15)
s1 + s2 k2 Ctot = Cp + Cs (27)
s2 s2 + g The given formulae allows to simplify the equivalent model
A3 = log10 ( ) (16) of the inductor as shown in Fig. 5 [14].
s1 + s2 k2
4 IEEE SENSORS JOURNAL, VOL. XX, NO. XX, XXXX 2023

RS
Rp Ctot
LS

Fig. 5. Simplified lumped physical model of a planar inductor at low

frequencies.

III. I NDUCTOR O PTIMIZATION

On-chip magnetic sensor design requires a particular op- Fig. 7. Inductor design with tapering. [18]
timization approach for coil design which looks to optimise
SN R and differs significantly from typical power applications,
where inductance or power dissipation are the key performance centre of the coil. This allows for an increase in the number
indicators [11]. In this work, by considering amplifier gain of turns and sensitivity of the coil according to (8) without
in conjunction with the sensitivity of commercially available significant increase of the resistance.
sensors [8], a target SN R can be identified. As follows from
Section II, the performance of the sensor is a tradeoff between B. MATLAB Optimisation
two parameters: sensitivity and noise. The Signal-to-Noise A numerical optimisation of the sensitivity techniques
Ratio SNR combines both parameters and is the important discussed in Section III A was undertaken in Matlab (The
figure of merit for this work. Typical values for the sensitivities Mathworks, MA, USA). As SN R is the figure of merit of
of commercial sensors, such as the Aurora discrete coil sensors our design, Fig. 8 represents the optimization process for the
(Northern Digital Inc., Ontario, Canada) are 0.1-0.2 V/T/Hz [8]. SNR of the M9 ultra-thick metal coil 450×1200 µm size for
Since our solution contains a built-in amplifier, this sensitivity increasing numbers of turns.
specification can be achieved by amplifier gain. Thus, during
the optimization process, we seek for the best SN R considering
the sensitivity specifications of existing discrete coil solutions. 24 0.09

0.08
A. Sensitivity Improvement Techniques 22
One of the most popular techniques for improving the 0.07
sensitivity of a planar coil is a multilayer stacked design 20
0.06
(Fig. 6).
SNR [dB]

18 0.05

16 0.04
SNR (w = 2 m)
SNR (w = 5 m) 0.03
Metal 9 14 SNR (w = 8 m)
k s #(w = 2 m)
0.02
k s (w = 5 m)
12
k s (w = 8 m) 0.01
Metal 8
10 0
Metal 6 0 10 20 30 40 50
Metal 7 (underpass) Number of turns

Fig. 8. Single-layer inductor optimization for M9 on TSMC 65 nm

technology.
Fig. 6. Multilayer stacked inductor structure. [15]
Spacing between the turns is minimized (2 µm). Solid lines
This method allows to increase flux linkage by increasing represent SN R and dashed lines represent sensitivity for each
the effective cross-sectional area without increasing the area width of the trace. Considering achievable amplifier gain of
of the inductor on a die [15], [16]. Upper metal layers (which 60 dB, once design optimisation is complete, a target on-
are intended for power supply and top-level routing) in CMOS chip inductor sensitivity of 0.06 mV/T/Hz was determined
technology have less resistance than lower layers. Thus, using to achieve comparable accuracy to existing discrete sensor
lower metal layers may lead to a decrease in SN R. Another coils. An optimal configuration which meets this sensitivity
technique, a tapering design, described in [17], is shown in requirement with 21.2 dB SN R can be achieved with 40 turns,
Fig. 7. 2 µm width of trace and 2 µm spacing.
In tapering design technique, width/spacing gradually in- Figure 9 represents the effect of tapering for different initial
creases/decreases while the trace traverses from the edge to the widths of the trace and tapering coefficient kw defined as
SIDUN et al.: PLANAR ON-SILICON INDUCTOR DESIGN FOR ELECTROMAGNETIC TRACKING 5

(Cadence Design Systems, Inc., San Jose, California) enviro-

20
ment. A parameterized cell, or PCELL, is a programmable
design entity that allow users to create a parametric instance
18 in the integrated circuits (IC) design environment each time it
is placed. PCELLs provide the following advantages:
16 • Speed up layout data entry by eliminating the need to
create duplicate versions of the same functional part;
SNR [dB]

• Eliminate errors that can result in maintaining multiple

14
versions of the same cell;
• Manufacturing process independence;
12 • Easier creation of complex designs.
kw = 1 This PCELL can be written by using the interactive pro-
10 k w = 0.95 gramming language, SKILL (Cadence Design Systems, Inc.,
k w = 0.9 San Jose, California). As this work used many different input
w=2 m 5 m 8 m parameters to achieve an optimal solution, creating a PCELL
0 2 4 6 8 10 was an attractive opportunity to speed up the design process.
Number of turns

Fig. 9. Single-layer inductor design configurations for M9 with tapering B. Structure and Features of the Inductor PCELL
design.
The graphical user interface of the new inductor PCELL is
shown in Fig. 10. This PCELL supports the following features:
• Multilayer design (up to 8 layers in the stack);
follows:
• A different width, spacing and number of turns for each
wi+1 layer;
kw = (28)
wi • Custom via placement;
• Physical parameter estimation (R, L, C etc.);
where wi is the width of the ith turn in µm. From Fig. 9
we can observe that reducing the width of the trace does not In a multilayer planar coil design, the resistance of the
give a significant increase in sensitivity, but negatively affects via (which is used to connect metal layers) may become a
resistance, which leads to decreasing SN R. Tapering was not significant bottleneck.
pursued further in this application. In order to eliminate this problem, a custom via tuning option
Based on the design procedure mentioned above, three was added to PCELL. This option allowed the user to adjust the
inductor configurations have been obtained. The optimisation size of the via to minimize via resistance. The PCELL provides
has been performed based on the magnetic flux density estimates of SN R, sensitivity and physical parameters of the
of 150 nT and operating frequency 20 kHz to match the coil (like resistance, inductance and capacitance). In order to
Anser EMT system requirements [2]. Optimised inductor increase possible design features, the PCELL contains stitching
configurations are shown in Table I. and paralleling functions. Like other PCELLs, this inductor
PCELL satisfies all the limitations of the process design kit
TABLE I (PDK) rules of the 65 nm manufacturing process.
O PTIMAL I NDUCTOR C ONFIGURATIONS FOR 65 NM TSMC PROCESS
V. I MPLEMENTATION
Optimal Inductor Configurations The optimized design described in Section II was imple-
Coil Number Width, Spacing, SNR(a) , Sensitivity, mented on the TSMC 65nm process within the overall sensor
geometry of turns µm µm dB µV /T /Hz IC and the chip was mounted in a QFN package [19].
M9 38 2 2 21.1 77
M9/M8 38/28 2/5 2/0.4 22.2 133
M10/M9/M8 15/38/28 7/2/5 3/2/0.4 23.6 202.31 A. Impedance measurements
a At f=20kHz and B=150nT
Impedance measurements were performed using a HP4285A
Precision LCR Meter. Open and short correction have been
It can be seen that the best SN R and sensitivity can be used to account for test lead effects, as well as result averaging
achieved by the three-layer stacked inductor implementation. over 4 iterations (Fig. 11).
This configuration has been selected as a final solution for our Figure 12 shows the comparison between measured and
design. simulated impedance using the simplified inductor model. As
expected, it can be seen that the inductor exhibits completely
IV. I NDUCTOR PCELL D ESIGN resistive behaviour up to 1 MHz.
As can also be seen from Fig. 12, the simplified model
A. PCELL Design in Cadence® Virtuoso does not predict the impedance behaviour well, due to some
Following the preliminary numerical analysis described in important parasitics, which are not taken into account in the
Section III B, the planar inductor was specified in the Cadence model. These parasitics include the layer-to-layer capacitance
6 IEEE SENSORS JOURNAL, VOL. XX, NO. XX, XXXX 2023

2400
2200
2000
1800
1600

1400

Impedance [ ]
1200

1000

800

600

Measurement results
Proposed model fit
Simplified model fit [C.P.Yue et al]
400
5 6 7
10 10 10
Frequency [Hz]

Fig. 12. Inductor impedance measurements

IC Package Socket

C9-10 Cbond

C10-8 Cpack

C8-9

Fig. 13. Updated model for the multilayer stacked inductor.

Fig. 10. Graphical user interface of the inductor PCELL. as well as the capacitance of the bondwires and socket leads
(Fig. 13).
In order to fit the measured impedance, a modified model
has been introduced (Fig. 14). Capacitors C9−10 , C8−9 , C10−8
represent the corresponding layer-to-layer capacitances which
can be approximately calculated using the packing factor of
each layer:

ki + ki+1
Cij = Aplate (29)
2
where Aplate is the area of the inductor and k is the packing
factor of the layer of the inductor which can be calculated as
follows:
wi
ki = (30)
wi + si
where wi and si is the width and spacing (in µm) between
the traces, respectively. The terms Cbond and Cpack , which
represent the capacitance of the bondwires and packaging,
Fig. 11. Test setup for impedance measurements
can be calculated as parallel capacitances. The updated model
which includes all parasitic capacitances is depicted in Fig. 14.
It is worth emphasising that C9−10 , C8−9 and C10−8 represent
SIDUN et al.: PLANAR ON-SILICON INDUCTOR DESIGN FOR ELECTROMAGNETIC TRACKING 7

process-specific capacitances which may be analytically calcu- A Helmholtz coil is a co-axial pair of identical coils, wired in
lated while Cbond and Cpack will depend on the measurement series to produce a uniform field in a small cylindrical volume
setup for testing. For simplicity, the described capacitors which between the two coils. The radius of the coils is equal to the
couple across Ls can be combined in a single term Cm : inter-coil spacing, and the field is effectively uniform (within
1%) in a cylindrical volume, co-axial with the coils.
C10−9 C8−9 The coil is controlled by a National Instruments USB-6343
Cm = + C8−10 + Cbond + Cpack (31) data acquisition unit with DAC and ADC both running at
C10−9 + C8−9
400 kS/s, driving a sine wave into an LT1210 power amplifier.
The driver circuit also includes a bank of relay-switched
capacitors in series with the coil, creating a tunable resonant
RS
series LC circuit, whose performance is controlled through a
RSi Ctot current sensing resistor.
C10-9 For a desired field frequency and strength, the software
LS C8-10 Cbond Cpack
C8-9 controls the appropriate relays to select the nearest available
capacitance value to drive the coil near resonance, outputs
the frequency of interest to the amplifier at an initial default
Fig. 14. Full updated model for the multilayer stacked inductor.
amplitude, and adjusts this amplitude according to the sensed
Therefore, the simplified inductor model for multilayer current to generate a field at the desired strength. The total
stacked inductor takes the form: (Fig. 15) harmonic distortion (T HD) of the coil current with reference
to the desired frequency is also monitored through the software.

RS
45
RSi Ctot
LS Cm 40

35
Fig. 15. Updated low frequency model for the multilayer stacked inductor.
30
SNR [dB]

25

B. SNR and sensitivity measurements 20

The key metric for this design was SNR, as it represents both
15
noise and sensitivity of the inductor. SN R has been measured
in conjunction with the analogue on-chip front-end and an 10
Theoretical SNR Measured SNR
off-chip 2nd order Butterworth filter to prevent aliasing. The B = 0.15uT B = 0.15uT
5
data acquisition has been performed using a NIDAQ USB-6343 B = 0.5uT B = 0.5uT
(National Instruments, Austin, USA). The signal was fed into B = 1uT B = 1uT
0
the inductor by a three-dimensional Ferronato BH300HF-3-B 0 10 20 30 40 50
Frequency [kHz]
Helmholtz coil, which is optimised for generating fields with
frequencies from DC up to 300 kHz and field strengths up to Fig. 17. Signal-to-Noise ratio of the system.
150 µT (Fig. 16).
The results of the measurements are depicted in Fig. 17.
Since the analogue front-end contributes to the noise voltage,
it is worth taking it into account. The
p measured input-referred
noise of the
√ analogue front-end ( Sn(AF E) (f )) is equal to
1.25 nV / Hz. Hence, the root mean square input referred
noise voltage of the analogue front-end for a given bandwidth
∆f is: q q
Vn(AF E),in 2 = Sn(AF E) (f )∆f (32)

where Sn(AF E) (f ) is the voltage noise density of the analogue

front-end. Thus, the SN R takes the form:
q
Vs 2
SN R = q q (33)
Vn 2 + Vn(AF E),in 2
Fig. 16. Test setup for SNR measurements. Using (11), we can obtain:
8 IEEE SENSORS JOURNAL, VOL. XX, NO. XX, XXXX 2023

R EFERENCES
k s B0 f0
SN R = √ √ (34) [1] A. Sorriento et al., ”Optical and electromagnetic tracking systems for
2∆f ( 4kB T Rs + Vn(AF E),in 2 ) biomedical applications: a critical review on potentialities and limitations,”
IEEE Rev Biomed Eng.
It is worth noting that the SN R calculations are obtained for [2] H. A. Jaeger et al., ”Anser EMT: the first open-source electromagnetic
tracking platform for image-guided interventions” , International Journal
the bandwidth of 100 Hz which corresponds to the refresh rate of Computer Assisted Radiology and Surgery, vol. 12, no. 6, pp. 1059-
of the system (Fig. 17). 1067, Jun 2017.
As can be seen from Fig. 17, the SN R of the inductor [3] W. F. W. Kappelle et al., ”Electromagnetic-guided placement of na-
soduodenal feeding tubes versus endoscopic placement: a randomized,
provides good matching with the theoretical predictions. multicenter trial,” Gastrointest Endosc, vol. 87, no. 1, pp. 110-118, Jan
The manufactured chip containing prposed inductor structure 2018.
is depicted in Fig. 18. [4] K. A. Khan, P. Nardelli, A. Jaeger, C. O’Shea, P. Cantillon-Murphy, and
M. P. Kennedy, ”Navigational Bronchoscopy for Early Lung Cancer: A
Road to Therapy,” Advances in Therapy, vol. 33, no. 4, pp. 580-596,
Apr 2016.
[5] F. Kral, E. J. Puschban, H. Riechelmann, F. Pedross, and W. Freysinger,
”Optical and electromagnetic tracking for navigated surgery of the sinuses
and frontal skull base,” Rhinology, vol. 49, no. 3, pp. 364-8, Aug 2011.
[6] B. Wood et al. ”Navigation with Electromagnetic Tracking for Interven-
tional Radiology Procedures: A Feasibility Study,” Journal of Vascular
and Interventional Radiology, vol. 16, Issue 4, pp. 493-505,2005.
[7] M. Cavaliere, O. McVeigh, H. A. Jaeger, S. Hinds, K. O’Donoghue
and P. Cantillon-Murphy, ”Inductive Sensor Design for Electromagnetic
Tracking in Image Guided Interventions,” in IEEE Sensors Journal, vol.
20, no. 15, pp. 8623-8630, 1 Aug.1, 2020.
[8] Northern Digital Inc. Aurora 5DOF Sensor 0.5 mm x 8 mm, Part
Number: 610099. Accessed on: Dec. 3, 2021. [Online]. Available:
Fig. 18. Die photo of the proposed inductor structure. https://www.ndigital.com/medical/products/tools-andsensors/
[9] M. Srivastava et al., “Electromagnetic Tracking - On-Chip real-time
tracking for surgery,” U.K. Patent 2218643.1, filed on 12 December
2022.
VI. C ONCLUSION [10] B. Razavi, ”Fundamentals of Microelectronics”, John Wiley and Sons,
2006.
In this paper the design and optimisation process for a [11] J. Burghartz, B. Rejaei, ”On the design of RF spiral inductors on silicon,”
IEEE Transactions on Electron Devices, vol. 50, no. 3, pp. 718-729,
planar inductor for time-varying magnetic field sensing in March 2003.
magnetic tracking applications has been presented. Optimal [12] K. B. Ashby, W. C. Finley, J. J. Bastek, S. Moinian, and I. A. Koullias,
multi-layer inductor configurations for the TSMC 65 nm process ”High Q Inductors for Wireless Applications in a Complementary Silicon
Bipolar Process,” (in English), Proceedings of the 1994 Bipolar/Bicmos
are represented in Table I. Table II shows the overall comparison Circuits and Technology Meeting, pp. 179-182, 1994.
between the calculated and measured parameters of the inductor. [13] F. E. Terman, Radio Engineers’ Handbook. New York and London:
It can be seen that the calculations are closely matching with McGraw-Hill Book Company Inc., 1943, p. 1019.
[14] C. P. Yue and S. S. Wong, ”On-chip spiral inductors with patterned ground
the experimental results. A modified planar inductor lumped- shields for Si-based RF IC’s,” IEEE Journal of Solid-State Circuits, vol.
element model has been proposed to account for multi-layer 33, no. 5, pp. 743-752, May 1998.
stacked designs and real-world instrumentation effects. [15] P. Park, C. S. Kim, M. Y. Park, S. Do Kim, and H. K. Yu, ”Variable
inductance multilayer inductor with MOSFET switch control,” IEEE
Electron Device Letters, vol. 25, no. 3, pp. 144-146, Mar 2004.
TABLE II [16] B. V. Deevi and N. B. Rao, ”Multilayer Grown High-Q On-chip Inductor
M EASURED INDUCTOR PERFORMANCE for RF Applications”, 2015 3rd International Conference on Signal
Processing, Communication and Networking (ICSCN), 2015.
[17] J. A. Power, S. C. Kelly, E. C. Griffith, and M. O’Neill, ”An investigation
Coil parameters estimation of on-chip spiral inductors on a 0.6 mu m BiCMOS technology for
Coil Calculation Measurement RF applications”, ICMTS 1999: Proceedings of the 1999 International
parameter result result Conference on Microelectronic Test Structures, pp. 18-23, 1999.
Sensitivity, µV /T /Hz 202.3 197.9 [18] F. Passos, M. H. Fino, and E. R. Moreno ”Fully Analytical Charac-
SNR(a) , dB 23.6 22.1 terization of the Series Inductance of Tapered Integrated Inductors”,
Inductance, µH 7.7 8.07 International Journal of Electronics and Telecommunications, March
Resistance, Ω 1276 1317 2014.
a At f=20kHz and B=150nT [19] Design Requirements - Quad No-Lead Staggered and Inline Multi-
Row Packages (with optional thermal enhancements). Accessed on:
Nov. 18, 2022. [Online]. Available: https://www.jedec.org/standards-
documents/docs/dg-419d

VII. ACKNOWLEDGEMENT
The authors would like to acknowledge the support of
Science Foundation Ireland (SFI) and the Microelectronic
Circuits Centre Ireland (MCCI). Views and opinions expressed
are however those of the author(s) only and do not necessarily
reflect those of the European Union or the European Research
Council Executive Agency. Neither the European Union nor
the granting authority can be held responsible for them.