A Bluetooth Low Energy Implantable Glucose Monitoring
System
Mai Ali, Lutfi Albasha, and Hasan Al-Nashash
Department of Electrical Engineering
American University of Sharjah
Sharjah,UAE
g00035464@aus.edu
Abstract—This paper presents a novel framework on a remotely
powered implantable glucose monitoring system. The RF signal
interaction with the biological tissue under investigation has been
characterized. The implantable unit is remotely powered by an
external unit through inductive coupling. The novelty of this
work resides in the design of two efficient class E Power
Amplifiers (PAs) with power efficiency of 76% and 73% for the
external power unit and the implantable unit respectively. The
implantable wireless transmitter utilizes low power
communication protocol, Bluetooth low energy, to transmit the
measured glucose data to patient’s mobile phone or PDA.
Keywords— Bluetooth low energy transmitter, Class E power
amplifier, Biotelemetry applications.
I. INTRODUCTION
Glucose monitoring is evolving from random measurements
taken over long time spans to regular monitoring. The current
state of the art is a meter that estimates the concentration of
glucose in a drop of blood obtained by a patient from his or
her fingertip or forearm. Those meters provide a measure of
the amount of blood glucose concentration at the time of
measurement but information such as the rate of change of
glucose concentration is not provided .
Glucose sensors are currently in the early development
phase. To broaden their deployment, enhancements are
needed in reliability, comfort, ease of use, and integration with
other technologies. Most of published work on implantable
glucose monitors is custom made. To optimize device power
consumption and minimize it’s dimensions; different radio
frequencies have been used.
Existing glucose monitoring systems are large in size
because they use discrete circuit components and are powered
by batteries.
This work addresses the aforementioned problems as
follows: firstly, utilizing low power system and circuit
techniques in designing the implantable unit wireless
transmitter. Secondly, Inductive power transfer will be used to
supply power to the implantable unit. This will not only
eliminate the need for periodical battery change but also result
in a reduced implant size.
Making glucose monitoring a daily life routine was the
motivation behind the selection of the receiver. Bluetooth
technology, and soon to follow Bluetooth low energy, is
available in almost all cellular phones and PDAs. Hence, one
of those daily used electronic devices can be utilized as a
receiver.
In this paper we present a general glucose monitoring
system description. We propose a telemetry link to power the
implantable unit, and suggest a communication standard for
data transfer. We also present the architecture of implantable
system that features an enhanced design of the RF transmitter.
Fig. 1. System Architecture
II. SYSTEM ARCHITECTURE
The diagram of the proposed glucose monitoring system is
shown in Fig. 1 above. The power will be supplied inductively
to the implantable unit. The external powering unit generates a
high frequency magnetic field in an external coil, through
which the power is inductively transferred to an internal coil
connected to the implantable system. The frequency at which
the power will be transferred is 13.56 MHz. This frequency
was chosen due to several reasons including its reduced tissue
absorption and its world wide availability as an ISM
frequency. The external powering unit will be fixed to the
patient body right outside the implantation side. The internal
coil picks up the radiated signal which will then be rectified
and passed to a voltage regulator to generate a steady voltage
supply for the implantable unit.
At sensor level, in order to define any waveform; sampling
must be done at a frequency that represents the full
dimensions of the waveform. High fidelity monitoring
sufficient to follow the intrinsic blood glucose dynamics of all
diabetic subjects requires a Nyquist sampling period of almost
10 minutes [1]. The implantable unit needs to be left for few
minutes in order for the accurate measurement to be taken.
After the glucose measurement is obtained, it will then be
transmitted to the nearby patient’s cellular phone through
Bluetooth low energy protocol. This communication protocol
was chosen for this design due to its reduced complexity, low
power consumption, and ability to support very low duty cycle
978-2-87487-022-4 © 2011 EuMA 10-13 October 2011, Manchester, UK
Proceedings of the 41st European Microwave Conference
1265
operation, which is the case in the majority of sensor
applications.
In the following, the interaction between the RF signal and
the biological tissue is characterized followed by the design
description of the main system units.
A. RF Signal-Biological Tissue Interaction
The factors that determine the RF signal power attenuation
in a biomedical implant are frequency, permittivity,
conductivity and permeability. It is desirable to keep the
transmitted power as low as possible while ensuring that the
detected signals are able to faithfully reproduce the required
information. This interaction can be viewed in two aspects,
energy absorption, where the guideline for the maximum EMF
exposure are defined in terms of Specific Absorption Rate
(SAR), and signal attenuation due to prorogation through the
biological tissue.
TABLE I. SKIN DIELECTRIC PROPERTIES AT 2.4 GHZ, 13.56 MHz
Property Value at 2.4 GHz Value at 13.56 MHz
Effective
conductivity
(o
c]]
)
0.7 S/m 0.23802 S/m
permittivity (e) 50×8.8S × 1u
-12
F/m 285.25× 8.8S ×1u
-12
F/m
Permeability
(p)
4n × 1u
-7
H/m 4n × 1u
-7
H/m
Attenuation
constant (o)
13.3 Np/m 1.3225 Np/m
The implantable transmitter will be placed in the
subcutaneous tissue directly underneath the skin. Thus the
biological tissue would be modelled by a single layer i.e. skin.
The average skin thickness of humans varies between 0.5 mm
up to 4 mm in palms and soles. For example if the implantable
unit is to be placed in patient’s thigh, which is quite an
appropriate place for implantation since it is less exposed than
other limbs like hands and arms, therefore, less interaction is
likely to occur with the surrounding environment that might
lead to implantable device displacement or damage. The skin
thickness at the thigh of an average human is typically in the
range of 1.55 mm. The dielectric properties of the skin have
been calculated for the two frequencies used in this design as
shown in Table I above. They were then used to determine the
two the RF operation safety limits as well as signal attenuation
due to propagation through the biological tissue [2].
B. Implantable Unit Class E PA
Compared to the conventional linear class operation of PAs,
amplifiers with displacement of peak drain/collector voltage
and current can attain much higher efficiency. They feature
reduced complexity compared to other classes of switching
mode PAs, where class D PAs require input and output baluns
and harmonic termination is required for class F PAs.
In order to compensate for the signal attenuation discussed
previously; the PA needs to be designed for a specific power
gain. Therefore, it is more useful to present the RF signal
attenuation in terms of power units.
Fig. 2. Power attenuation at 2.4 GHz
At 1.55 mm, the RF signal power is attenuated by 16.7 dB
as shown in Fig. 2. The PA needs to have a power gain large
enough to cancel this attenuation or at least compensate for
most of it.
Fig. 3. Implantable unit class E PA schematic
However, there is an inherent trade-off between the RF PA
frequency response, which dictates its power gain, and its
efficiency. Maximum frequency response is attained when
biasing the PA in the strong inversion region; in contrast,
maximum efficiency is achieved when the PA is operated in a
weak inversion regime. It has been proven that the optimum
bias point for RF PAs falls in the moderate inversion region
[3]. For this reason the PA was biased in the onset of strong
inversion so that the power gain requirement can be met while
still operating at high efficiency. The schematic of the class E
PA is shown in Fig. 3 above.
C. External Power Unit
The external power unit consists of a crystal oscillator that
generates 13.56 MHz carrier, a power MOSFET driver, and a
class E PA. The generated carrier is passed to a power
MOSFET driver to prevent the PA from loading the oscillator.
A class E PA is used to amplify the carrier signal and supply
the power inductively to the implantable system through an
external coil.
The power attenuation at 13.56 MHz at distance of 1.55m
through the skin was found to be 7.213 dB. If the implantable
unit is to be supplied with 10 dBm power, then the external
power unit PA needs to have a power gain of 17 dB assuming
perfect coils alignment and no loss is encountered in the
voltage rectification and regulation units represented by the
RF frontend block in Fig. 1.
Fig. 4. External power unit class E PA schematic
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01
-22
-21
-20
-19
-18
-17
-16
-15
Skin Thickness (m)
P
o
w
e
r
A
t
te
n
u
a
t
io
n
(
d
B
)
x = 0.001551
y = -16.7
1266
The schematic of the external class E PA is shown in Fig. 4,
the same design procedure in [4] was followed. However, the
PA bias point was pushed to the lower boundary of the strong
inversion region to meet the specified power gain requirement.
D. Bluetooth low energy transmitter
Fig. 5. Bluetooth low energy transmitter system diagram
The system diagram used to generate the Bluetooth low
energy signal is illustrated in Fig. 5. Bluetooth low energy
uses GFSK modulation in which the binary bit stream is
passed through a Gaussian low pass filter that has a pulse
shaping coefficient of 0.5. Accordingly, for a binary data rate
of 1.0 Mbps the filter's -3 dB Bandwidth is 500 kHz.
Bluetooth low energy uses a larger modulation index of 0.5
compared to 0.35 of the classic Bluetooth. Consequently,
Bluetooth low energy channels are spaced 2 MHz apart, rather
than 1 MHz apart as in classic Bluetooth technology. The
frequency hopping mechanism for the 40 channels ranging
from 2.4 GHz to 2.48 GHz described in the Bluetooth low
energy physical layer specifications is implemented [5]. The
mapping array of the Digital to real convertor in Fig. 5 is set
(0, 39) to hop over the 40 different offset frequencies.
III. SIMULATION AND RESULTS
First the two PA designs were simulated to validate their
performance in terms of switching characteristics, power
efficiency, gain and linearity. Then, Bluetooth low energy Tx
was simulated to confirm its compliance with the Bluetooth
low energy standards.
A. Circuit Simulation
The results shown below are for the 2.4 GHz implantable
class E PA. The first step to validate the class E PA
functionality is to verify its switching pattern. This design
used 0.35 μm BiCMOS library available under AWR software.
Fig. 6. Output Voltage ( ∆ ) Vs Output Current ( □ )
The voltage and current waveforms at the transistor’s drain
terminal are shown in Fig. 6. The voltage and current
waveforms experience partial elapse thus perfect switching is
not achieved. Consequently the PA efficiency will be less than
100% as shown in Fig. 7 below:
Fig. 7. DC-to-RF efficiency (DC-RF) %
The second corner in the PA performance is the power gain.
The gain versus the input power is shown in Fig. 8 below. The
1-dB compression point at which the gain is reduced by 1 dB
is illustrated as well.
Fig. 8. Power gain
Fig. 9: AM –AM (□), AM-PM( ) and gain (∆)
As can be seen in Fig. 9, the gain is constant until the point
where it intersects with the AM-AM curve after which gain
compression takes place. The AM-PM can be calculated using
(1) and referring to Fig .9 as given below:
AH - PH (Jcgrccs¡JB) =
∆(Pbosc)
∆(omplituJc)
13.69-11.97
6.918-18.11
= -u.1S°¡uB (1)
It is desirable to keep this value as low as possible, such that
the amplitude compression does not result in any phase
alteration .
Fig. 10: Intermodualtion distortion
The principle of two-tone test is illustrated in Fig. 10. The
difference between the fundamental and IM3 power levels is
0 0.3 0.6 0.833
Time (ns)
Vo vs Io
0
2
4
6
8
10
V
o
(
V
)
0
30
60
90
120
150
I
o
(
m
A
)
p2
p1
-40 -20 0 20 30
Input power (dBm)
0
20
40
60
80
D
C
-
R
F
(
%
)
p1
10.03 dBm
68.96
22.14 dBm
75.92
-40 -20 0 20 30
Input power (dBm)
-20
-10
0
10
20
G
a
i
n
(
d
B
)
p1
-38.84 dBm
14.02 dB
2.972 dBm
13.02 dB
-40 -20 0 20 30
Input power (dBm)
-30
-20
-10
0
10
20
G
a
i
n
(
d
B
)
0
3
6
9
12
15
P
h
a
s
e
(
d
e
g
)
p3
p2
p1
8.648 dBm
18.11 dBm
-7.065 dBm
6.918 dBm
-7.092 dBm
13.69 Deg
8.671 dBm
11.97 Deg
8.646 dBm -7.009 dBm
2.2 2.3 2.4 2.5 2.6 2.7
Frequency (GHz)
-100
-80
-60
-40
-20
0
P
o
u
t
(
d
B
m
)
p1
2.35 GHz
-84.93 dBm 2.5 GHz
-87.24 dBm
2.4 GHz
-9.984 dBm
2.45 GHz
-9.967 dBm
p1: Freq =2.4 GHz
Pwr = -24 dBm
-24 dBm, at, 2.4 GHz
1267
74 dBc. From the plot, the third-order intercept point (IP3) is
determined. The IIP3 can be calculated from Fig. 10 using (2):
IIPS(JBm) = P
ìn
(JBm) +
(Pout
Fund.
-Pout
IM3
)
2
(2)
From Fig 10, P
in
is -24 dBm, while fundemntal-IM3
difference is 74 dBc. Direct application in (2) gives a value of
13.4 dBm for IIP3. This agrees with the relationship that
relates the 1-dB point to the IIP3 given by (3) below:
IIPS(JBm) = 1 - JB (JBm) + 9.6 JB (3)
It is worth mentioning that linearity is not a requirement for
this class E PA design. The Bluetooth low energy signal is
frequency modulated; hence no data is contained in the signal
amplitude or phase. Consequently gain compression and AM
to PM conversion do not spoil the information signal.
TABLE II. CLASS E PAs PERFORMANCE CHARACTERISTICS
Parameter Implantable PA External PA
Frequency (MHz) 2400 13.56
VDD (V) 3.3 3.3
Gate bias voltage (V) 1.17 1.36
Current consumption (mA) 18.9 143
DC to RF efficiency (%) 75.92 72.6
Power gain (dB) 14.02 17.79
AM-PM (deg/dB) -0.15 -0.737
P-1dB (dBm) 2.9 -8.55
∆IM3 (dBc) 74 47.9
IIP3 (dBm) 13.34 2.59
The performance characteristics of the implantable and the
external class E PAs is summarized in Table II above. The
SAR limits for the implantable and the external PAs were
found to be 29.5 mW/Kg and 54 mW/Kg respectively which
comply with RF safety limits mentioned previously.
B. System Simulation
Fig. 11. Transmitter test bench
For the purpose of simulating the complete implantable
transmitter, incorporating the PA and Bluetooth low energy
transmitter and reflecting skin attenuation effect, the
transmitter test bench in Fig. 11 above was implemented.
Fig. 12. Output spectrum after modulator ( ∆
) and PA ( □
)
The power spectrum of the signal at the output of the
modulator and the PA is shown in Fig. 12. The difference
between the signals’ power levels represents the implantable
PA power gain which agrees with the results illustrated in
Table II.
Within the ISM band the transmitter shall pass a spectrum
mask, given in Table III. The spectrum shall comply with 20
dB bandwidth definition in Federal Communication
Commission (FCC).
TABLE III. BLUETOOTH LOW ENERGY TX SPECTRUM MASK
Frequency offset Bluetooth low
energy specs.
This design
± 500 kHz -20 dBc -22dBc
2MHz (|H -N| = 2) -20 dBm -66 dBm
3MHz or greater (|H -N| ¸ S) -30 dBm -87 dBm
The transmitted power was measured in a 100 kHz
bandwidth. This measurement requires the Tx to transmit
pseudo random data pattern while the frequency hopping is
switched off. Table III shows the specified Bluetooth low
energy Tx power requirements and the results obtained in this
design. All the three power requirements were met.
IV. CONCLUSION
The system architecture of glucose monitoring system was
presented. The implantable system was inductively powered
by an external power unit which operates at 13.56 MHz. Two
efficient class E PAs were designed for the external power and
implantable units. Due to its low power requirements,
Bluetooth low energy was utilized as the communication
protcol in th implantable tranmsitter.
Future work includes completing the circuit design of the
implantable transmitter and developing a prototype to perform
real measurements. Further reduction in the implantable PA
current consumption is needed.
V. REFERENCES
[1] D. Gough, K. Kertuz-Delgado, and T. Bremer, “Frequency
characterization of blood glucose dynamics”, Ann. Biomed.
Eng., vol. 31, no. 1, pp. 91–97, 2003.
[2] M. Ali, “low Power Wireless Subcutaneous Transmitter”, M.S.
Thesis, Dept Elec. Eng., American University of Sharjah, UAE,
2011.
[3] A. Shameli, P. Heydari, “A Novel Power Optimization
Technique for Ultra-Low Power RFICs,” Proc. Int. symps. Low
Power Electronics and Design, Tegernsee, pp. 274-279, 2006.
[4] M. Ali, L. Albasha, H. Alnashash, “A system study of a
wireless subcutaneous transmitter”, Int. Symp. on Mechatronics
and its Applications, Sharjah, April, 2010.
[5] (Dec. 26, 2009), Bluetooth SIG, Bluetooth specification
version 4.0 [Online]. Available: http://www.bluetooth.com
D R
1
2
3
1
2
3
1
2 3
4
R D
F l
1 2
3
4
BER
SRC MEAS
TP ID=TP5 TP ID=TP6
Frequency Hopping
Channel Generation
Mechanism
Rate = 1Mbps.
from2.40 GHz to 2.48 GHz
Constant channel @ 2.44 GHz
NO Frequency Hopping
Bluetooth Low energy GFSK Modulation
Frequency Hopping
in steps of 2 MHz.
AWGN
PA
Attenuator
2.352 2.372 2.392 2.412 2.432 2.452
Frequency (GHz)
-300
-200
-100
0
100
P
o
w
e
r
(
d
B
m
)
2.402 GHz
-13.64 dBm
2.402 GHz
0.9622 dBm
1268
