Sensors and Actuators A 232 (2015) 359367

Contents lists available at ScienceDirect

Sensors and Actuators A: Physical

journal homepage: www.elsevier.com/locate/sna

A low power bioimpedance module for wearable systems

Stefano Rossi a,b, , Marco Pessione a , Valeria Radicioni a , Giovanni Baglione a ,
Monica Vatteroni b , Paolo Dario b , Luigi Della Torre a
a
Advanced System Technology (AST), STMicroelectronics, Via Olivetti 2, 20864 Agrate Brianza, Italy
b
The Biorobotics Institute, Scuola Superiore SantAnna, Viale R. Piaggio 34, 56026 Pontedera, Italy

a r t i c l e i n f o a b s t r a c t

Article history: In the emerging eld of wearable systems for remote monitoring of physiological parameters, the mea-
Received 30 September 2014 surement of bioimpedance has the potential to provide many useful information. On the other hand, in
Received in revised form 5 May 2015 this scenario, an optimization of power consumption of the circuit is crucial. A low power architecture for
Accepted 5 May 2015
the measurement of bioimpedance was identied in this work. It reduces the consumption in the most
Available online 12 May 2015
critical blocks of the system: the current driver, the signal sensing and the demodulator. The device was
Keywords:
prototyped and electrically characterized. The compromise between power consumption reduction and
Bioimpedance the increase in electrical noise was analysed and an effective signal processing technique was developed,
Wearable system showing that it is possible to achieve a signal to noise ratio good enough to enable applications like res-
Remote monitoring piration monitoring (breathing rate and amplitude) or cardiac output estimation. Preliminary tests on
Voltage to current converter healthy subjects showed a good correlation with spirometer, for breathing monitoring, and with Doppler
Impedance measurement echocardiography, for cardiac output. Thanks to the good functionality and the reduced current con-
Impedance cardiography sumption (750 A at 2.8 V power supply was obtained with a discrete-components implementation) the
module resulted suitable for the integration in wearable devices for remote monitoring of physiological
parameters, or other low power applications.
2015 Elsevier B.V. All rights reserved.

1. Introduction On the other hand, in these particular applications, an optimiza-

The term bioimpedance is commonly used to refer to the
impedance of the human body or a part of it. The measurement of
this parameter was employed in multiple elds, ranging from the
assessment of body composition [1] or of intra-thoracic uid status
[2,3], to respiration, cardiac output and peripheral blood ow [4]
monitoring.
Even if this technique is still not completely established in
clinical practice, its versatility, simplicity and low cost make it
very interesting also for the integration into wearable systems for
remote monitoring of physiological parameters [5]. In this inno-
vative and emerging scenario bioimpedance has been suited, for
example, for respiration monitoring or for uid status estimation,
both in research projects [6] and commercially available devices [7].
Selected papers presented at EUROSENSORS 2014, the XXVIII edition of the
conference series, Brescia, Italy, September 710, 2014.
Corresponding author at: Advanced System Technology (AST), STMicroelectron-
ics, Via Olivetti 2, 20864 Agrate Brianza, Italy. Tel.: +39 0396037249.
E-mail addresses: stefano.rossi@st.com (S. Rossi), marco.pessione@st.com
(M. Pessione), valeria.radicioni@st.com (V. Radicioni), gio.baglione@gmail.com
(G. Baglione), monica.vatteroni@sssup.it (M. Vatteroni), paolo.dario@sssup.it
(P. Dario), luigi.dellatorre@st.com (L. Della Torre).
tion of the power consumption of the circuit is crucial.
Hardware for bioimpedance measurement is sometimes based
on the impedance converter AD5933 (Analog Devices, Inc.) [8,9],
which represents a simple solution. Anyway, it is not optimized in
terms of current consumption (10 mA) and, furthermore, it requires
additional external circuitry for bioimpedance measurement [10].
This causes an increase in power consumption, cost and complex-
ity. A reduction of power consumption was obtained reducing the
duty-cycle [11] of the measurements, or through application spe-
cic integrated circuit (ASIC) design [12]. The rst approach is not
compliant with scenarios requiring a continuous monitoring of
the signal, e.g. breathing or cardiac output monitoring; the sec-
ond guarantees very low consumption leveraging the advantages
of full-custom design, but it is a very expensive solution.
The typically used architecture is shown in Fig. 1. The impedance
is measured by injecting a xed amplitude alternate current (typ-
ically sinusoidal) inside the body through an electrodes pair,
and reading back the voltage drop from pickup electrodes. Being
I cos(t) the injected current, the read back voltage has the form:
v(t) = Mz (t)I cos[t + z (t)] (1)
in which Mz (t) and z (t) are the module and phase of the
bioimpedance, Z(t). For this reason, after a rst gain stage,

http://dx.doi.org/10.1016/j.sna.2015.05.004
0924-4247/ 2015 Elsevier B.V. All rights reserved.
360 S. Rossi et al. / Sensors and Actuators A 232 (2015) 359367

applications to test the suitability of the impedance measurement

system in real scenario; a proper processing scheme for the extrac-
tion of these parameters was designed and is presented in Section
3. Results of tests on healthy volunteers are presented in Section 4.

2. Hardware overview

The block diagram of the proposed solution is shown in Fig. 3.

The driving current originates from a 50 kHz square wave, which
is ltered with a second order low-pass lter to generate a quasi-
Fig. 1. Block diagram of a typical bioimpedance device using analog demodulators. sinusoidal voltage waveform. A VI converter is used to drive the
current in the body. The demodulator is moved before the IA; there-
demodulation is needed to extract the information related to the fore, the IA works on the baseband demodulated signal and the
impedance, in the form of real and imaginary part or module bandwidth requirements for this block can be dramatically reduced
and phase components. Finally, analog or digital ltering can be with important benets in power consumption, cost and perfor-
used to divide the static and time dependent components of the mance of the component. Nevertheless the VI converter and the
impedance, typically referred as Z0 and Z. demodulator, which are the blocks still operating at the working
A rst critical element of this kind of architecture, in terms of frequency, are optimized as described below (Sections 2.1 and 2.2)
power consumption, is the instrumentation amplier (IA): since the to reduce power requirements. Analog ltering is used to separate
typical working frequency for a bioimpedance system ranges from and amplify the AC coupled component Z of the bioimpedance
tens of kilo-hertz to mega-hertz, the bandwidth of the IA needs to be before the analog to digital conversion.
quite large. This usually means an increase in power consumption
and also in cost, especially in discrete components design. 2.1. Current driver
The same concept applies also to the current driver: to have
small errors it is important to obtain a very high output impedance: The voltage to current converter is based on a modied How-
the higher the output impedance, the lower the dependence of land topology (Fig. 4), where we inserted a capacitor in the positive
the current amplitude on the load. Since the load is constituted feedback loop.
by electrodes to tissue impedance, it can vary a lot with frequency, The capacitor adds a pole to the output impedance expression;
electrode size and type, contact quality and so on. Unfortunately, such a pole can be designed to give, together with the amplier
the output impedance of common voltage to current converters (VI pole, a couple of complex and conjugate poles providing a peak at
converters) is very high as long as the operational amplier can be the corresponding frequency (Fig. 5).
approximated as ideal, but quickly decreases with the increase in
frequency, because of the limited gain-bandwidth product (GBP) of
the operational amplier. Therefore, a very high GBP is needed also
for a 50 kHz working frequency [13], resulting in high current con-
sumption. To relax the requirements for the current driver output
impedance, current sensing is sometimes used (Fig. 2), but in this
case two IAs are needed.
Demodulation is another critical block: it can be implemented
with standard demodulators (as in Fig. 1), synchronous sampling
[14] or gain and phase detection [15], but all these approaches
require at least two channels, resulting in an increase in consump-
tion, cost and size.
To solve all these issues, a novel hardware architecture to per-
form bioimpedance measurement is suggested in this work. The
proposed changes are related to system architecture (Section 2) as
well as to VI converter (Section 2.1) and demodulation (Section 2.2).
The proposed bioimpedance module was prototyped and charac-
terized. Breathing and cardiac output estimation were selected as
Fig. 3. Block diagram for the proposed bioimpedance device.

Fig. 2. Block diagram of a bioimpedance device using current sense resistor to mea-
sure the actual injected current. Fig. 4. Schematic of the VI converter.
 S. Rossi et al. / Sensors and Actuators A 232 (2015) 359367
Fig. 5. Output impedance of the VI converter versus frequency.
In this way the dominant pole of the op-amp, which in gen-
eral constitutes the deviation from the ideal behaviour causing the
reduction of the output impedance, is directly used for the syn-
thesis of the needed circuit properties. High output impedance at
the working frequency can be obtained with a narrow bandwidth
amplier, with benets in power consumption. In this design a
TSV6390 (STMicroelectronics) operational amplier is used (cur-
rent consumption 55 A), the theoretical output impedance at the
working frequency results higher than 1 M when the circuit is
driven by a pure sine-wave. When the circuit is driven by a ltered
square wave, as in this work, the output impedance is very high
only for the fundamental harmonic, while it is lower for the other
harmonics. This determines an increase in the current depend-
ence on the load, which can be modelled with a theoretical output
impedance of about 200 k, fully compliant with the application,
as discussed in Section 5.
2.2. Demodulator
The demodulator is realized through AC coupled buffers, with
high input impedance and high-pass transfer function, and sam-
ple and holds. Each buffer is used to sense the voltage of one of
the sensing electrodes; this voltage is then sampled at the car-
rier frequency, in order to obtain a baseband signal proportional to
Fig. 6. Calibration procedure to nd the module of the impedance.
361
the impedance seen between the sensing electrode and the inter-
nal voltage reference of the circuit. The IA performs the difference
between the two baseband signals and provides an output propor-
tional to the bioimpedance Z(t). Given the operational amplier,
the use of unity gain buffers allows reaching the maximum band-
width; therefore, the gain-bandwidth product of the amplier can
be minimized with benets in power consumption. In our design
the TSV630 (STMicroelectronics) operational amplier was chosen
for the buffers (current consumption 55 A, gain-bandwidth prod-
uct 860 kHz) and the LT1789 (Linear Technology Corporation) was
selected as IA (current consumption 67 A).
The sampling signal is provided by the micro-controller. Accord-
ingly to Eq. (1), in order to measure the module of the impedance,
the system shall be able to sample the peaks of the sinusoidal
voltage drop on Z(t). This is obtained through an automatic self-
calibration performed by the system at the beginning of the
acquisition (Fig. 6). Starting from an arbitrary phase shift phi0
between the sampling pulse and the square wave from which the
quasi-sinusoidal current is generated, the value of the impedance
M is acquired, then the phase shift phi is changed of a xed quan-
tity step and a new value of the impedance New M is acquired and
compared with M, in order to update the value of M accordingly
to whichever is the greater. The procedure is repeated until the
maximum M is found. This value corresponds to the peak of the
acquired voltage expressed by Eq. (1), and represents the module
of the impedance. Furthermore, since the delay introduced by the
low-pass lter and the voltage to current converter in the current
driving chain is known, the nal phase shift between the square
wave and the sampling pulse obtained from the procedure allows
estimating the impedance phase.
Since the phase shift is quantized, it is possible that the sampling
pulse, instead of sampling exactly the peak of the voltage, samples
a point close to it, causing an error in the reconstruction of the
module of the impedance:
   
 z = R Mz ej(z z )
M z
= Mz cos z 
in which M z is the measured module, which results an approxima-
tion of the actual module Mz because of the wrong estimation  z of
the actual phase z . It is worth to highlight that, since the error on
the module depends on the cosine of the error on the phase, very
good accuracy on the module is still obtained even with a quite
rough quantization on the phase shift steps: with a phase error of
5 degrees, the error on the module is only 0.4%. Therefore, it is not
needed to continuously run the calibration: since the time depend-
ent changes of the impedance phase z (t) are small, z (t) can be
362 S. Rossi et al. / Sensors and Actuators A 232 (2015) 359367
approximated with a constant with negligible error on the module,
and once the right phase shift for the sampling pulse is calculated, it
can be used to continuously sample the module of the impedance.
With this approach minimum consumption is required, since
only one chain with a sample and hold is used and the accuracy
of the module measurement is good. This is important because the
module is the main information used in many important applica-
tions, like breathing monitoring [16] or impedance cardiography
both for cardiac output estimation [17] or for uid status assess-
ment [2]. For other applications, like body composition, also the
phase is needed [1], but this kind of measurements are less interest-
ing in terms of a continuous monitoring through a wearable system,
because only a few measurements per day are needed. In any case,
a phase estimation of the impedance is obtained with the proposed
method and, furthermore, this architecture is very exible and it is
also possible to use a synchronous sampling approach [14] to obtain
the real or imaginary part of the impedance.
3. Data acquisition system and processing overview
The hardware architecture described in Section 2 is compliant
with many bioimpedance applications. Our target, in this work, is
to demonstrate its suitability in trans-thoracic applications, in par-
ticular for breathing monitoring and cardiac output extraction in
wearable devices.
The bioimpedance module was integrated in a device also
including an analog front-end for electrocardiogram (ECG), a 3-axis
accelerometer for activity and body posture estimation, and a Blue-
tooth transceiver. Bioimpedance and ECG signals were acquired
through the analog to digital converter embedded into the micro-
controller (STM32F1, STMicroelectronics) and transmitted to a PC
for signal processing and parameters extraction. A detailed descrip-
tion of the complete device is outside the scope of this paper,
which focuses on the bioimpedance front-end. It is mentioned
here because it was used as the data acquisition system for a
deeper evaluation of the bioimpedance module, allowing both elec-
trical characterization and data acquisition on healthy subjects.
In terms of safety, it was tested accordingly to the IEC 60601
by a notied body and it obtained the CE marking before the
execution of the tests reported in Section 4. All the algorithms
were developed for an easy integration into the on board micro-
controller; their computational cost was controlled, avoiding, for
example, frequency-domain analysis or other resource consuming
approaches.
3.1. Breathing extraction
The rst step in the processing of the raw Z signal for breath-
ing analysis is a down-sampling from the original 256 Hz sampling
frequency to 32 Hz. This is possible because the typical spectral
components of the respiratory activity have an extremely low
frequency (well below 1 Hz), and by decreasing the number of oper-
ations required for each sample, we can achieve a longer battery
life. A tight bandpass ltering is then performed to isolate the spec-
tral components of interest. This is accomplished by combining a
low-pass IIR lter with a 3 dB cut-off frequency of 0.27 Hz with a
computationally efcient high-pass FIR lter with a 3 dB cut-off
frequency of 0.06 Hz. This attenuates possible noise components
and completely removes any type of baseline wandering. After-
wards the resulting signal is smoothed through a moving window
average lter to attenuate any possible high-frequency component
still present.
The ToddAndrews peak detection algorithm [18] is then
applied to identify the occurrences of peaks and troughs in the sig-
nal. Each sequence of trough-peak-trough detected is considered a
valid breath only if its peak respects specic requirements based
on its amplitude and on its distance with the adjacent troughs.
Finally, three different techniques are used to detect noise
affecting the signal:
Electrode detachment detection
Z0-based identication of saturation or no current injection
Very small breath amplitude due to poor contact quality
If at least one of these techniques marks the signal as unreliable, the
resulting breathing rate is not computed. If instead the signal qual-
ity is good, the respiratory rate is easily estimated as the number
of breaths detected in the last 60 s.
3.2. Cardiac output extraction
Cardiac output estimation is based on the analysis of the Z
waveform and of its rst derivative dZ/dt [17,19]. The typical range
for Z amplitude is 0.10.2  [17]. With so small variations noise
becomes an important issue to deal with, especially when the
excitation current is decreased to reduce the power consumption:
lower current means lower input voltage in the read-back circuit,
and this results in a poorer signal to noise ratio (SNR). Typical val-
ues reported in literature [19] and used by commercial device like
Niccomo (medis. Medizinische Messtechnik GmbH) for the exci-
tation current are in the range 14 mA. As mentioned in Section
2.1, a current of 100 A is used in this work; therefore, with such
a decrease in current level, a low SNR is unavoidable. Furthermore,
as discussed in Section 5, since the overall architecture is optimized
to minimize the current consumption, an increased level of noise
is unavoidable.
Taking into account these considerations, the digital signal
processing for cardiac output estimation was designed. The goal
was to recover a SNR good enough to allow cardiac output estima-
tion, according to methodologies widely presented in literature. In
particular, after a rst bandpass ltering stage (3 dB bandwidth:
0.76 Hz) which reduces the noise outside the bandwidth of the
signal (both due to breathing and electronics), ensemble averag-
ing [20] is used to improve the signal quality by reducing the noise
inside the signal bandwidth. The signal periods corresponding to
cardiac cycles are identied through the ECG R-peaks position,
then, being pi (n) the period i of the signal, the sample k of the
ensemble averaged period p is computed as:
1
N
p(k)
= pi (k)
N
i=1
where N is the number of periods used for the average.
The strong effectiveness of this approach compared with sim-
pler solutions, like, for example, averaging of consecutive samples,
is related to the nature of the noise. It is well known that averag-
ing N random samples xi results in the strongest reduction of the
variance when the samples are uncorrelated, since it is:

1 1  1 
N N
Var xi = Var (xi ) + 2 Cov xi xj .
N N2 N
i=1 i=1 i=
/ j
As described in Section 4.1 and shown in Figs. 8 and 9, the noise
is not uncorrelated: its power spectral density is not white and
the autocorrelation function shows that low value of correlation,
below 0.2, can be obtained only between samples with a distance in
time of at least 1 s. In this scenario, simple averaging of consecutive
samples results in a small decrease in noise power. On the other
hand, ensemble averaging returns the average between samples
 S. Rossi et al. / Sensors and Actuators A 232 (2015) 359367 363

which are far enough to present a weak correlation, resulting in a
stronger decrease in noise power.
The averaged period is then processed by applying the deriva-
tive and identifying its maximum as well as the points commonly
referred as B and X [17,19] which corresponds to the rst deriva-
tive zero crossing and the rst local minimum after the derivative
maximum. The stroke volume (SV) is calculated using the formula
described in [21,19] and cardiac output (CO) is computed as
CO = SV HR
where HR is the heart rate.
4. Experimental results
For the evaluation of the designed bioimpedance module two
complementary paths were followed. First of all, to evaluate
the general functioning and to quantitatively asses the per-
formances, the module was characterized measuring relevant
electrical parameters. Then, in order to demonstrate its suitabil-
ity in real applications, it was used to measure breathing rate and
amplitude and cardiac output. Below the procedures followed for
each one of these tests are reported, as well as the obtained results.
An example of an acquired raw signal is reported in Fig. 7.
of the current source between 164 k and 257 k (mean 213 k,
standard deviation 43 k).
To verify the gain behaviour a variable load was connected to
the sensing terminals of the circuit. The current was forced into the
variable load through two 330  resistors (i.e. each sensing termi-
nal was also connected to one injecting terminal through a 330 
resistor). The variable load was changed between 0  and 120 
and was measured, the procedure was repeated on ve samples.
The gain resulted between 81.2 and 85.6 (mean value 83.1, standard
deviation 1.83), with a linearity error of 0.1 % or lower on a full scale
of 120 .
Noise is another important parameter, as remarked in Section
3.2. Since electrical noise is a stochastic process, it is interesting
to characterize it through representative statistics, like autocorre-
lation function and power spectral density. For this purpose, the
noise was measured on ten acquisitions, each ten minutes long.
The sensing input terminals were both short-circuited to the ref-
erence, which is available on one of the injection terminals (Fig. 3).
The input referred power spectral density is reported in Fig. 8,
while the time autocorrelation of one of the noise acquisitions is
shown in Fig. 9. The input referred noise resulted about 5 Vrms
corresponding to 0.05 rms .

4.1. Electrical characterization

The accuracy of the device gain is inuenced by many sources of

error, like accuracy of current amplitude, output impedance of the
voltage to current converter, sample and hold losses and distortion,
resistors tolerances and ADC non linearity.
To verify the current forcing sub-circuit, a variable load was con-
nected between the injecting terminals. It consisted in a xed 1 k
resistor in-series with a variable resistor. The voltage drop on the
xed resistor was measured with an oscilloscope (MSO8104A, Agi-
lent Technology) for different values of the variable resistor. The
use of this setup minimizes the impact of the error introduced
by the oscilloscope, since it is the same for all the measurement.
Loads between about 1 k and 7 k were considered. The test was
executed on ve different samples. The current amplitude on the
1 k resistor resulted between 91.3 A and 99 A (mean 96.2 A,
standard deviation 3.14 A). On the 7 k resistors, the reduction of
the current was between 2.28 % and 3.59 % (mean 2.85 %, standard
deviation 0.58 %). This can be modelled with an output impedance
Fig. 8. Power spectral density of the input referred voltage noise.

Fig. 7. Raw Z impedance signal. Fig. 9. Autocorrelation function of the input referred voltage noise.
364 S. Rossi et al. / Sensors and Actuators A 232 (2015) 359367

The total current consumption of the bioimpedance analog

front-end was about 750 A, with 2.8 V power supply. It was mea-
sured as average current from power supply, so it includes also the
mean value of the rectied signal current (quasi-sinusoidal, ampli-
tude 100 A) and all the buffers used to generate the references for
the circuit.

4.2. Breathing measurements

In order to check the behaviour of the device for breathing

acquisition, electrodes were placed on the chest, horizontally. The
injecting electrodes were placed externally, the sensing ones inter-
nally. The distance between the sensing electrodes was 8 cm, while
4 cm were used between the sensing and the nearest injecting.
Standard ECG electrodes were used and connected to the device
through cables. The subject, a 35 years old man, height 175 cm,
weight 60 kg, was asked to sit and to breath normally. Then he was
asked to change the breathing amplitude with different breathing
depths up to his maximum inspiration and expiration capability.
Fig. 11. Relationship between the amplitude of breathings measured from the
Air ow during breathing was also measured through a spirome- impedance and the volume measured by the spirometer. Regression line is showed,
ter (Fig. 10), in order to have a reference. The test was repeated in the coefcient of determination resulted R2 = 0.96.
three sessions, in different days, in order to take into account also
intra-subject repeatability of the measurement.
The linear correlation between impedance and spirometer was
observed and reported in literature since 70 s [16]. The capability
of our device to reproduce this result was investigated, in order to
verify its functioning in breathing monitoring applications. Fig. 11
shows the relationship between impedance and spirometer data in
the executed tests, together with the regression line. The coefcient
of determination resulted R2 = 0.96.

4.3. Cardiac output measurements

The second application which was selected for a preliminary

on the eld verication of the bioimpedance module was the
measurement of SV and CO.
A comparison with a different technique for the estimation of
these parameters was performed. Doppler echocardiography was
selected, for its non invasive nature. The measurements were done
on four healthy volunteers in Ferrarotto Hospital (Catania, Italy)
and in Cisanello Hospital (Pisa, Italy). Bioimpedance was acquired
with our device, in tetrapolar conguration, with standard ECG Fig. 12. Raw data, zoom on the cardiac component.

electrodes placed on the left side of the neck (pick-up electrode

close to the shoulder) and of the chest (pick up electrode at the
level of xiphoid process). The measurement was performed imme-
diately after the Doppler echocardiography, in order to avoid any
interference between the position of the electrodes and the place-
ment of the echographic probe. The subject was sitting during the
acquisition, the impedance signal was recorded and the processing
scheme described in Section 3.2 was used. The algorithm allowed
a strong reduction the noise on the signal: the measured value of
0.05 rms was decreased of about 10 times every 10 periods used
for the average. Figs. 12 and 13 shows examples of data before and
after the processing, 8 periods were used for the average.
The nal SV and CO results are summarized in Table 1.

5. Discussion

The current consumption value of 750 A is fully compliant with

Fig. 10. Comparison between the raw impedance signal (blue line) and the spirom-
eter signal (red line). (For interpretation of the references to color in this gure
legend, the reader is referred to the web version of the article.)
the integration of the circuit in a wearable device for remote patient
monitoring. It refers to the whole circuit of Fig. 3 with the excep-
tion of the microcontroller, which usually has many functions in an
embedded system and which is needed also for other approaches
like, for examples, the ones presented in Section 1. The consump-
tion results dramatically lower with respect to designs based on the
AD5933 (current consumption 10 mA, but it also requires external
 S. Rossi et al. / Sensors and Actuators A 232 (2015) 359367
Fig. 13. Filtered and averaged periods (8 periods average) of the raw bioimpedance
signal (up) with derivative (down). The localized B point, maximum and X point are
marked in the derivative signal.
Table 1
Results of comparison with Doppler echocardiography
Subject Echocardiography/Bioimpedance
Age Weight [kg] SV [ml] HR [bpm]
26 83 90.6 / 100 54 / 52
29 70 75 / 72 56 / 68
28 67 56 / 55 91 / 98
62 104 83 / 66 59 / 63
circuitry) [8]. This result is obtained when the device is work-
ing continuously; when the module is employed in applications
which does not need continuous monitoring, the overall power con-
sumption can be further reduced through duty-cycle control of the
measurements in a way similar to [11]. Furthermore, this result was
obtained with a discrete components design. This approach was
chosen because faster and cheaper to prototype and test. Anyway
the architectural improvements here proposed make the obtained
results comparable even with some full-custom designs, like for
example [22], even if full-custom approaches usually can benet
from a higher level of optimization of single blocks and from the use
of techniques like switched capacitors. The same concepts demon-
strated here could be successfully employed in integrated circuit
technology, obtaining a further reduction of power, when consid-
erations related to application (for example implantable devices)
or to costs scaling suggested such an approach.
Together with the low power feature, the device exhibits very
good functionality. In particular it is shown how to nd a compro-
mise between power reduction and noise. An increase in the noise
(we measured 0.05 rms ) is unavoidable: it is caused by the pres-
ence of active components (the input buffers) before the rst high
gain stage, constituted by the IA. Furthermore these components,
for their low power characteristics, are not optimized in terms of
noise and, nally, the use of sample and hold to demodulate is not
the best option in terms of noise: using standard analog demodula-
tion by multiplication for a coherent sine-wave, for example, results
in a frequency shift of the low frequency noise outside the base-
band, improving the signal quality. With sample and hold approach,
instead, the Flicker noise is still present inside the band occupied
by the demodulated signal. For all these reasons the input referred
voltage noise reaches the value of 5 Vrms . To express the noise as
impedance noise, anyway, the division by the injected current is
needed, and this is another crucial aspect: using a higher injected
current would result in a lower impedance noise with strong ben-
ets in SNR.
365
Anyway it is shown that the presence of noise can be mit-
igated with effective signal processing strategies; therefore, the
architectural choices determining the power reduction are fully
compliant with an excellent performance level. In breathing moni-
toring, indeed, the signal amplitude (typically above 1 ) provides
an acceptable SNR. For cardiac output, the SNR was dramati-
cally improved by ensemble averaging, which resulted an effective
processing scheme for the proposed architecture. The measure-
ments presented in Sections 4.2 and 4.3 demonstrates good
correlation with a spirometer, accordingly to what reported in lit-
erature [16] (for breathing) and with a Doppler echocardiography
(for cardiac output). In the last cardiac output measurement a sig-
nicantly lower value was found, anyway, since a gold standard
for cardiac output does not exist, comparison between different
techniques usually tolerates errors up to 2030% [21].
The results obtained in breathing monitoring and cardiac output
estimation constitute an important verication of the suitability
of the bioimpedance circuit in real scenarios and demonstrate,
together with the electrical characterization, the proper function-
ing of the proposed hardware architecture. Of course more tests are
needed in these elds, both to have a statistically signicant data set
and to explore the impact of movement artefacts and of electrodes
to skin interface in continuous monitoring tasks. On the other hand,
to correctly proceed in this direction, it is not enough to consider the
bioimpedance circuit: aspects like the embodiment of the selected
CO [l]
wearable system, in particular electrodes technology and design,
4.9 / 5.2 as well as their coupling with the body, are crucial in determin-
4.3 / 4.8 ing the robustness of a wearable device to movement artefacts.
5.1 / 5.4 The same device can perform in very different ways in presence
5.7 / 4.2
of movement, if coupled to the body through standard electrodes
and cables, or through smart textiles, or through a small patch,
etc. Cancellation algorithms (for example using inertial sensors
to estimate the movements) can also be implemented. Since this
work focuses on the bioimpedance front-end, which can be inte-
grated in different wearable systems, it was decided to postpone all
the aspects which require such a system level approach to future
works. From the circuital point of view, an important parameter
which must be considered to predict the robustness and reliability
of the device versus the electrodes technology and movement arte-
facts is the output impedance of voltage to current converter. This
parameter, indeed, determines the accuracy of the current when
the load impedance changes, and movements, together with elec-
trodes technology and with their interface to skin, cause exactly a
change in the impedance of the load. Typical values for the electrode
to tissue impedance, when working at tens of kilo-hertz, like in this
case, are reported to be in the range of a few hundreds ohm up to
one kilo-ohm [23]. In order to take into account the higher level
of drift which can arise in the eld of wearable devices, the device
was tested for load impedances up to 7 k, and showed an output
impedance in the range of 200 k, which guarantees an accuracy
below 1% on the current amplitude when the contact impedance is
into the range reported in literature, and well below 5% in all the
considered range.
Breathing and cardiac output monitoring applications were
selected for the preliminary evaluation of the bioimpedance mod-
ule because they were identied as interesting and as the most
challenging for wearable systems. Thanks to the very good results
obtained, it is reasonable to expect that the presented device can be
successfully suited also in other applications, like basal impedance
assessment for body composition analysis or thoracic uid evalua-
tion. These measurements typically do not require a continuous
sampling of the impedance, so they enable further power con-
sumption reduction through duty-cycle management and are less
sensitive to noise, because of the possibility to average a very
high number of samples (e. g. a window of some seconds of
signal).
366 S. Rossi et al. / Sensors and Actuators A 232 (2015) 359367
6. Conclusion
A low power architecture for the measurement of bioimpedance
was identied. It reduces the consumption in the most critical
blocks of the system, which are current driver, signal sensing
and demodulation. The device was prototyped, electrically char-
acterized and preliminarily tested in real bio-signal acquisition
tasks, like respiration analysis and cardiac output estimation. It
showed a good general functionality. The optimization of power
consumption required a compromise with an increased noise level,
nevertheless it was shown that this issue can be managed with
proper signal processing which enabled the use of the device also
in very noise-sensitive applications, like cardiac output estimation.
Preliminary healthy subjects tests were compared with spirome-
ter for breathing and with Doppler echocardiography for cardiac
output and demonstrated very good correlation. These tests com-
pleted the validation of the design concept. The device resulted
perfectly compliant with the integration in wearable devices for
remote monitoring of physiological parameters, or for other low
power applications. Other aspects, like the reliability of the mea-
surements in the presence of movement artefacts, as well as a
characterization on a larger number of subjects, must be performed
after having exactly dened the wearable device embodiment and
the electrodes technology; therefore they are not part of this study
and will be explored in future works.
Acknowledgements
The authors desire to thank Ferdinando De Negri, MD and Mat-
teo Bartalucci, MD, from the Department of Emergency Medicine,
Azienda Ospedaliera Universitaria Pisana, Pisa, Italy, who acquired
the Doppler echocardiography for the rst subject and Massimil-
iano Mul, MD and Luca Costanzo, MD, from the Clinical Division
of Cardilology, Ferrarotto Hospital-Policlinico, University of Cata-
nia, Catania, Italy who acquired the Doppler echocardiography for
the following three subjects. Thanks a lot to all of them for their
availability and the interest showed in our activity.
References
[1] U.G. Kyle, I. Bosaeus, A.D.D. Lorenzo, P. Deurenberg, M. Elia, J.M. Gomez, B.L.
Heitmann, L. Kent-Smith, J.-C. Melchior, M. Pirlich, H. Scharfetter, A.M. Schols,
C. Pichard, Bioelectrical impedance analysis Part I: Review of principles and
methods, Clin. Nutr. 23 (2004) 12261243.
[2] L. Wang, Fundamentals of intrathoracic impedance monitoring in heart failure,
Am. J. Cardiol. 99 (10 Suppl.) (2007) S3S10.
[3] M. Shochat, G. Charach, S. Meyler, M. Kazatzker, M. Mosseri, A. Frimerman,
P. Rabinovich, A. Shotan, S. Meisel, Internal thoracic impedance monitoring: a
novel method for the preclinical detection of acute heart failure, Cardiovasc.
Revasc. Med. 7 (1) (2006) 4145.
[4] D.-W. Kim, Detection of physiological events by impedance, Yonsei Med. J. 30
(1) (1989) 111.
[5] X.-F. Teng, Y.-T. Zhang, C.C.Y. Poon, P. Bonato, Wearable medical systems for
p-health, IEEE Rev. Biomed. Eng. 1 (2008) 6274.
[6] C.W. Mundt, K.N. Montgomery, U.E. Udoh, V.N. Barker, G.C. Thonier, A.M. Tellier,
R.D. Ricks, R.B. Darling, Y.D. Cagle, N.A. Cabrol, S.J. Ruoss, J.L. Swain, J.W. Hines,
G.T.A. Kovacs, A multiparameter wearable physiologic monitoring system for
space and terrestrial applications, IEEE Trans. Inf. Technol. Biomed. 9 (3) (2005)
382391.
[7] I. Anand, A. Doan, K. Ma, J. Toth, K. Geyen, S. Otterness, N. Chakravarthy, R. Katra,
I. Libbus, Monitoring changes in uid status with a wireless multisensor mon-
itor: results from the uid removal during adherent renal monitoring (farm)
study, Congest. Heart Fail. 18 (1) (2012) 3236.
[8] J. Ferreira, F. Seoane, K. Lindecrantz, AD5933-based electrical bioimpedance
spectrometer. Towards textile-enabled applications, in: Proceedings of the
Annual International Conference of the IEEE Engineering in Medicine and Biol-
ogy Society, EMBS, 2011, pp. 32823285.
[9] C. Margo, J. Katrib, M. Nadi, A. Rouane, A four-electrode low frequency
impedance spectroscopy measurement system using the AD5933 measure-
ment chip, Physiol. Meas. 34 (4) (2013) 391405.
[10] F. Seoane, J. Ferreira, J. Sanchz, R. Brags, An analog front-end enables electrical
impedance spectroscopy system on-chip for biomedical applications, Physiol.
Meas. 29 (6) (2008) S267S278.
[11] P. Bognez-Franco, L. Nescolarde, C. Glvez-Montn, R. Brags, J. Rosell-Ferrer,
An implantable bioimpedance monitor using 2.45 GHz band for telemetry,
Physiol. Meas. 34 (1) (2013) 116.
[12] M. Min, T. Parve, V. Kukk, A. Kuhlberg, An implantable analyzer of bio-
impedance dynamics: Mixed signal approach, IEEE Trans. Instrum. Meas. 51
(4) (2002) 674678.
[13] A. Tucker, R. Fox, R. Sadleir, Biocompatible, high precision, wideband, improved
Howland current source with lead-lag compensation, IEEE Trans. Biomed. Circ.
Syst. 7 (1) (2013) 6370.
[14] R. Palls-Areny, J.G. Webster, Bioelectric impedance measurements using syn-
chronous sampling, IEEE Trans. Biomed. Eng. 40 (8) (1993) 824829.
[15] Y. Yang, J. Wang, G. Yu, F. Niu, P. He, Design and preliminary evaluation of a
portable device for the measurement of bioimpedance spectroscopy, Physiol.
Meas. 27 (12) (2006) 12931310.
[16] L.E. Baker, Applications of the impedance technique to the respiratory system,
IEEE Eng. Med. Biol. Mag. 8 (1) (1989) 5052.
[17] R.P. Patterson, Fundamentals of impedance cardiography, IEEE Eng. Med. Biol.
Mag. 8 (1) (1989) 3538.
[18] B. Todd, D. Andrews, The identication of peaks in physiological signals, Com-
put. Biomed. Res. 32 (4) (1999) 322335.
[19] D.P. Bernstein, Impedance cardiography: pulsatile blood ow and the bio-
physical and electrodynamic basis for the stroke volume equations, J. Electr.
Bioimped. 1 (2010) 217.
[20] S. Everson, W. Lovallo, G. Pincomb, P. Kizakevich, M. Wilson, Validation of an
ensemble-averaged impedance cardiogram for estimation of stroke volume,
Proc. Ann. Conf. Eng. Med. Biol. 13 (Pt 2) (1991) 801802.
[21] D.P. Bernstein, H.J.M. Lemmens, Stroke volume equation for impedance cardio-
graphy, Med. Biol. Eng. Comput. 43 (4) (2005) 443450.
[22] J. Ramos, J. Ausn, A. Lorido, F. Redondo, J. Duque-Carrillo, A wireless, compact,
and scalable bioimpedance measurement system for energy-efcient multi-
channel body sensor solutions, J. Phys.: Conf. Ser. 434 (2013) 1.
[23] P. Riu, J. Rosell, A. Lozano, R. Pallas-Areny, Multi-frequency static imaging in
electrical impedance tomography: Part 1 instrumentation requirements, Med.
Biol. Eng. Comput. 33 (6) (1995) 784792.
Biographies
Stefano Rossi was born in Siena, Italy in 1979 and received the Laurea degree (hon-
ors) in Electronic Engineering from University of Pisa, Italy, in 2005. After one year
spent as Research Assistant at the CRIM Laboratory of Scuola Superiore SantAnna,
Pisa, in 2006 he joined STMicroelectronics, Milano, Italy, as Analog Designer, work-
ing on shock and vibration sensing in hard disk motor controllers. In 2009 he moved
to Advanced Systems Technology (AST) division of STMicroelectronics (STM), as
Advanced Research Engineer on biomedical remote monitoring applications. He has
been resident engineer in Mayo Clinic (Rochester MN, USA) in May 2009, for testing
an innovative electronic patch (BodyGateWay) for remote monitoring of physio-
logical parameter, developed in a research project involving STM and Mayo Clinic.
Since 2011 he is also PhD student at The Biorobotics Institute of Scuola Superi-
ore SantAnna of Pisa in the framework of a collaboration between STM and the
school. He is also member of Complex Systems Community (CSC) of the University
of Siena. His main research interests are in the elds of wearable systems for remote
monitoring and biosignals acquisition and processing.
Marco Pessione was born in Carmagnola, Italy, in 1981 and received the Laurea
degree in Telecommunications Engineering (with honors) from the Politecnico di
Torino, Italy, in 2008. He discussed a thesis entitled Low-Complexity Adaptive
Reliable Electrocardiogram Feature Extraction, outcome of his activity in the AST
laboratory of STMicroelectronics, Agrate Brianza. He then joined the AST Remote
Monitoring group and contributed to the development of the BodyGuardian medi-
cal device as the main responsible in the design and implementation of algorithms
aimed at feature extraction from biological signals. His latest research interests focus
on gait analysis and movements disorders, through the use of attitude and head-
ing reference systems (AHRSs), and the characterization of sleep quality through
minimally invasive wearable devices.
Valeria Radicioni was born in Chiaravalle, Italy, in 1988 and received the Bach-
elor degree in Electronic Engineering from the Marche Polytechnic University in
Ancona, Italy, in 2012. She discussed a thesis entitled Design and realization of
a multiple purpose board for power, monitor and debug of microcontroller sys-
tems. She is now earning the Master degree in Telecommunication Engineering
from the Politecnico di Torino, Italy, with a thesis entitled Algorithm for cardiac
output estimation through wearable device for remote monitoring of physiolog-
ical parameters, outcome of her internship in the Advanced System Technology
laboratory of STMicroelectronics, Agrate Brianza, Italy.
Giovanni Baglione was born in Messina, Italy, in 1987. He received his bachelors
degree in biomedical engineering at university of Pisa in 2010 developing a thesis on
dielectric elastomers properties. In 2013 he received his Masters degree in biomed-
ical engineering at Pisa university developing a thesis on the bioimpedance signal
processing in wearable device in order to implement an algorithm to estimate blood
ow parameters through bioimpedance. The thesis was the outcome of his intern-
ship in STMicroelectrnics in the Advanced System Technology Group. At the moment
he is attending a second level master, in Smart Solution and Smart Communities,
at Scuola Superiore SantAnna in parterneship with Telecom Italia mobile, focused
 S. Rossi et al. / Sensors and Actuators A 232 (2015) 359367
to develop smart solution in biorobotics, telecomunication, managment, smart city
eld.
Dr. Monica Vatteroni is PostDoc at the Biorobotics Institute (BRI) since 2008. She
is responsible for the research and development of image sensors, vision systems
and sensorized platforms mainly focused on biomedical applications. She manages
a team for the design, production and testing of miniaturized vision systems, CMOS
imagers and system level platforms for biomedical parameters monitoring within
several European and Regional founded projects. Within the BRI Monica Vatteroni
has been involved in many research projects as operative responsible of technical
activities with an ofcial role in the technical management board in one case. She has
been scientic responsible for 3 regional founded projects. Her research interests
are in the eld of robotic sensitized platforms and imaging technologies, with a main
focus on vision systems and VLSI technology for CMOS camera design. She is author
and co-author of over 30 peer reviewed journal and conference publications and
holder of 6 patents. She is a regular invited reviewer for high impact journals and
conferences in the elds of biomedical engineering, sensors, electronics, imaging
and vision systems.
Prof. Paolo Dario is professor of Biomedical Robotics at SSSA and the Director of
The BioRobotics Institute, leading a large team of more than 100 researchers. Today,
Paolo Dario is the Coordinator of the PhD Program in BioRobotics at SSSA, currently
enrolling 92 PhD students. Paolo Dario is the Founding Coordinator of the Center of
Micro-BioRobotics@SSSA of the Italian Institute of Technology (IIT), and he is cur-
rently Senior Scientist of IIT. He has been and is the coordinator of more than 60
European projects, of more than 50 national projects, and of more than 50 indus-
trial projects. Paolo Dario is the editor of two books on the subject of robotics, and
the author of more than 500 scientic papers (250 on ISI journals). Paolo Darios main
eld of interest is biorobotics, a frontier research area that combines bio-inspiration
367
and bio-application. Paolo Dario has been and is working on theoretical models,
design methods, precision technologies (including micro, nano and bio) for fabri-
cation of novel components and integrated systems, machines and robots, and on
the application of such systems to different elds, including (micro) endoscopic and
endoluminal surgery, rehabilitation and assistance to disabled and elderly people,
and humanoid robotics. Paolo Dario has contributed to the foundation of such new
elds as micro and biomechatronics, and to the elds of biorobotics and neuro-
robotics, frontier research areas whose ultimate goals are to integrate biology and
neuroscience with robotics, and to educate the kernel of a new generation of engi-
neers, able to combine solid engineering capabilities, with curiosity and imagination,
and with the ability to incorporate in their engineering skills interdisciplinary and
transdisciplinary knowledge.
Luigi Della Torre was born in Monza, Italy in 1963 and started its activities in 1986
working in SGS (now STMicroelectronics) within the group that dealt Video, specif-
ically focusing on components for signal processing and audio/video interfaces. In
1994 he started to work on Set Top Box architectures, covering all aspects of the
product from the development to the production, in particular the denition of the
working platform for STB ITALTEL. In 1998 he became a member of the system
research team in STMicroelectronics, Advanced System Technology (AST), dealing
with multistandard decoder and digital receivers for analog TV. In 2000 he col-
laborated on the development of on chip architectures for a receiver for UMTS
Basestations and in 2002 he started to work on channel coding and source for video
streaming. In 2004 he took the responsibility of the group that deals with Con-
nectivity, for the development of communications solutions for High-Throughput
WLAN-based MIMO and UWB MBOA. Since 2008 he has been responsible of the AST
Remote Monitoring Group, leading the research activity about wearable systems for
remote monitoring of physiological parameters.