Technical Note
Two-electrode low supply voltage
electrocardiogram signal amplifier
D. Dobrev
Centre of Biomedical Engineering, Bulgarian Academy of Sciences, Sofia, Bulgaria
Abstract—Portable biomedical instrumentation has become an important part of
diagnostic and treatment instrumentation, including telemedicine applications. Lowvoltage and low-power design tendencies prevail. Modern battery cell voltages in the
range of 3–3.6 V require appropriate circuit solutions. A two-electrode biopotential
amplifier design is presented, with a high common-mode rejection ratio (CMRR),
high input voltage tolerance and standard first-order high-pass characteristic. Most of
these features are due to a high-gain first stage design. The circuit makes use of
passive components of popular values and tolerances. Powered by a single 3 V
source, the amplifier tolerates 1 V common mode voltage, 50 mA common mode
current and 2 V input DC voltage, and its worst-case CMRR is 60 dB. The amplifier is
intended for use in various applications, such as Holter-type monitors, defibrillators,
ECG monitors, biotelemetry devices etc.
Keywords—ECG amplifier, Biopotential amplifier, Low supply voltage amplifier,
AC coupled amplifier
Med. Biol. Eng. Comput., 2004, 42, 272–276
1 Introduction
A MODERN tendency in patient diagnosis and treatment involves
the use of personalised portable biomedical instrumentation. In
addition to well-known Holter-type ambulatory recorders of
electrocardiogram (ECG) and blood pressure signals, various
telemedicine applications require instruments of improved
design, compatible with modern microcomputers and microcontrollers.
Low voltage and low power are among the most important
requirements for such instrumentation. Present-day rechargeable
or non-rechargeable 3.6 V or 3 V battery voltages need adequate
biopotential amplifiers. High performance should be obtained in
spite of the low supply voltage limitation, especially concerning
electrode polarisation voltage and common-mode input voltage
tolerance.
The most widely used circuits for biosignal amplifiers are
based on the three-operational-amplifier configuration, or instrumentation amplifier, followed by an additional AC-coupled
stage (NEUMANN, 1998).
Usually, the ‘classical’ amplifier gain is split between the
instrumentation amplifier and the stage after the high-pass
decoupling filter. The first stage gain is set to low values,
because of the electrode polarisation potentials. Their voltage
difference can reach up to about 200 mV, depending on various
factors (electrode metal, conductive gel, patient skin etc.), and
appears as an input signal DC component (NEUMAN, 1995).
Correspondence should be addressed to Dr Dobromir Dobrev;
email: ikdas@argo.bas.bg
Paper received 7 April 2003 and in final form 11 November 2003
MBEC online number: 20043858
# IFMBE: 2004
272
The main performance characteristics of ECG amplifiers can
be summarised as follows:
frequency band at 3 dB from 0.05 to 100 Hz, with first-order
high-pass filter
tolerance of DC input voltage (of level depending on the
type of electrode) without input stage saturation
overall gain in the range 200–1000 (46–60 dB), with a
maximum input signal of about 5 mV without output
stage saturation
differential input impedance >5 MO in the entire frequency
band
common mode rejection ratio (CMRR) >60 dB
for a two-electrode amplifier, the inputs should tolerate at
least 3 mA common mode current per input, without saturation of the input stage.
The last requirement corresponds to interference level, which
commonly occurs in a hospital room environment, according to
our previous experience (DOBREV and DASKALOV, 2002;
DOBREV, 2002). Even with a battery-supplied amplifier, input
common mode currents can often reach 1.5 mA per input.
The idea of setting high gain in the first amplifier stage is well
known. It allows a high CMRR to be obtained easily. The
simplest solution is to add a capacitor in series with the gainsetting resistor of the differential amplifier (MCCLELLAN, 1981;
PALLAS-ARENY and WEBSTER, 1993), but its value can be
inconveniently high, depending on the high-pass cutoff
frequency. A version of this circuit, having the same disadvantage, was patented by CHEE (2002). In addition, as the first stage
is a differential follower, any DC input voltage is amplified by
the second stage. An old solution, using differential high-pass
filters at the inputs, has been reconsidered by BURKE and
GLEESON (2000). The circuit needs a reference electrode,
otherwise the input stage would be saturated even by very
Medical & Biological Engineering & Computing 2004, Vol. 42
small common mode currents. Bootstrapped input stages also
suffer from saturation by relatively low input voltages (THAKOR
and WEBSTER, 1980).
SPINELLI and MAYOSKY (2000) proposed the use of optocouplers in photovoltaic mode and an integrator, included in a
negative feedback loop, for input DC voltage compensation and
high-pass filtering. The optocoupler transfer characteristics are
non-linear, and there is a wide variation between specimens of
the current-to-current transfer ratio (about twice). This leads to
low accuracy of the high-pass cutoff frequency.
In a similar design, JORGOVANOVIC et al. (2001) used
differential-to-differential amplifiers instead of optocouplers.
The circuit is unacceptable for low-power systems, as these
types of amplifier, designed for high-frequency operation,
consume large amounts of current (20 mA or more).
These and other inconveniences of existing solutions stimulated us to try and develop a low-voltage, low-power, twoelectrode amplifier, satisfying the above requirements.
Ad ¼ 1 þ
2 Amplifier circuit concept
The simplified amplifier circuit is shown in Fig. 1a. The
general principle is that the input signal is buffered (two buffers
marked ‘1’) and AC decoupled by the capacitor C and the
resistors R3. The second stage consists of two differential
amplifiers Ad. Each of them amplifies half of the differential
input signal. By summing, the output signal is obtained as
Vout ¼ Ad (Va Vb þ Vc Vd )
s2R3 C
(V VinN )
¼ Ad 1 þ s2R3 C inP
Here a, b and c, d are the two differential amplifier inputs, and Ad
is the gain. The high-pass cutoff frequency is defined by the time
constant 2R3C.
The detailed circuit is shown in Fig. 1b. The input stage
consists of operational amplifiers A1, A2, A3 and A4. A1 and A2
are the main gain stages, and A3 and A4 are unity gain buffers. As
the non-inverting input voltages of A3 and A4 are equal to their
respective output voltages, resistors R2 and R3 are virtually in
parallel. Therefore the ratio of the currents in R2 and R3 is
IR2=IR3 ¼ R3=R2. The current in R1 is the sum of the currents in R2
and R3: IR1 ¼ IR2 þ IR3 ¼ (1 þ R3=R2)IR3.
As mentioned above, the resistors R3 and C form a first-order
low-pass filter, and the AC component on C decreases with 6 dB
Oct1 and becomes practically zero for the operating frequency
band. The A1 and A2 amplifiers take one-half of the differential
input AC signal each. The input DC component is filtered by C
and appears at the A3 and A4 outputs.
The second stage is a unity gain four input adder=subtractor
stage. It implements (1), where Ad is as follows:
(1)
R1
R2 kR3
with R3 4R2 , Ad ¼ 1 þ
R1
R2
Another solution for the second stage could be by two differential channel analogue-to-digital converters (ADCs), producing a digitised Vout, ready for microcomputer processing. When
5 V supply voltage is available, it is possible to obtain Vout by
two difference amplifiers in a microchip, such as INA2134, for
example.
The first stage has unity common mode voltage gain. The
second stage has unity differential mode voltage gain. The
minimum CMRR can be calculated as
Ad14 Ad5
A
1
¼ d
Acm14 Acm5
1 4d=(1 þ R4 =2R4 )
1:5
(2)
¼ Ad 4d
where d is the tolerance of the R4 resistors used.
If Ad ¼ 1000 and d ¼ 1%, the theoretically computed
minimum CMRR (assuming ideal operational amplifiers) is
91.5 dB, taking opposite signs for the resistor tolerances. With
Ad ¼ 200, CMRR becomes 77.5 dB. Taking into account real
operational amplifiers (with CMRRmin ¼ 75 dB) and with
Ad ¼ 200, the real minimum CMRR is 60.3 dB.
A very important parameter is the operational amplifiers’
input offset voltage, especially concerning A3 and A4. The A1
and A2 offsets do not contribute to error, as they are added to the
input signal DC component, which is cancelled by the capacitor C.
The maximum output voltage error due to the operational
amplifiers’ input offset voltage is
R1
Voomax ¼ (VioA3 max þ VioA4 max ) 1 þ
R2
CMRR ¼
þ 3VioA5 max 2Ad VioA3;4 max
Here Viomax are the maximum offset voltages of the corresponding operational amplifiers.
When selecting operational amplifiers, the following should
be respected: A3, A4 and A5 to be low input offset voltage and
high CMRR types; A1 and A2 to be of high open-loop gain, high
CMRR and high gain-bandwidth product.
3 Body–amplifier interface
Fig. 1
Basic amplifier circuit concept. (a) Simplified and (b) detailed
circuits
Medical & Biological Engineering & Computing 2004, Vol. 42
As mentioned above, for a two-electrode amplifier, the inputs
should tolerate input common mode currents of at least 3 mA per
input. If the supply voltage is only 3 V, this cannot be done using
passive components, resistors and capacitors. The only solution
273
Fig. 3 Practical amplifier circuit
consideration. With R7 C4 ¼ (R1 k R2 k R3)C2 R2C2, the highfrequency zero in the amplifier transfer function is cancelled
Ad (s) ¼
Fig. 2
Amplifier with bidirectional current sources connected to
inputs
is common mode input impedance reduction by voltagecontrolled current sources (DOBREV and DASKALOV, 2002)
using negative shunt-shunt feedback.
Such a circuit is shown in Fig. 2. It makes use of the first stage
circuit shown above. In addition, two bidirectional current
sources are connected to the amplifier inputs. Thus frequencydependent differential and frequency-independent common
mode input impedances are obtained. If the current source
transconductance is gm, it can be seen (Fig. 2) that
Zcm ¼
1
2gm
Zd ¼
2
(1 þ s2R3 C)
gm
(3)
Here Zcm and Zd are the common-mode and differential input
impedances, respectively. Zcm has only resistive character,
whereas Zd has resistive (2=gm) and inductive (4R3C=gm)
components.
Controlling the amplifier differential input impedance yields
an advantage: the polarisation potentials’ effect is automatically balanced. Owing to the low resistive value of the
differential input impedance, they recharge and tend to equalise
each other.
Vout
1
sC3 2R3
¼
VInP VInN 1 þ sC4 R7 1 þ sC3 2R3
R1
1 þ sC2 (R1 kR2 kR3 )
1þ
1 þ sC2 R1
R2 kR3
(4)
Inserting C5 capacitors ensures the circuit stability. The input
impedances are implemented by two bidirectional modified
Howland voltage controlled current sources (VCCSs), described
in DOBREV and DASKALOV (2002). The VCCS transconductance can be chosen in the range of 1=20–1=100 kO. Thus a high
VCCS output minimum resistance is ensured for a given resistor
tolerance and signal frequency band. The corresponding input
impedance (3) differential and common mode resistive components including the input RF filters are
1
¼ 2(R5 kR6 þ R7 ) 84 kO
Rd ¼ 2
gm þ R7
(1=gm þ R7 ) (R5 kR6 þ R7 )
¼
21 kO
Rcm ¼
2
2
In the signal frequency band, Zd also has an inductive component (3)
Ld ¼
4R3 C
¼ 461:6 MO61 mF6(R5 kR6 ) 205 kH
gm
The simulated Ad, differential Zd and common mode Zcm input
impedances for this amplifier (circuit of Fig. 3) are shown in
Fig. 4.
The frequency band is 0.05–100 Hz, as is usual for ECG
amplifiers. The circuit tolerates up to 50 mA common mode
4 Practical amplifier circuit
The two-electrode amplifier design was implemented in a
practical circuit shown in Fig. 3. It is powered by a single 3 V
supply voltage. Several operational amplifiers types can be used,
e.g. MCP607, OPA2336 or similar. Because of the input
common mode voltage range, the signal ground is set to onethird of the supply voltage (U4B). The diodes D1–D4 prevent
latch up of the circuit. The inputs are RF noise-protected by the
RC networks R7, C4. Its value was derived from the following
274
Fig. 4
Simulated gain, differential Zd and common mode Zcm input
impedances of practical amplifier circuit. (u u) Ad, dB; (s s) Zd
O; (, ,) Zcm, O
Medical & Biological Engineering & Computing 2004, Vol. 42
currents and up to about 2 V DC differential signal. The current
consumption is 150 mA (0.45 mW) at 3 V supply voltage.
5 Multichannel ground free circuit
Multichannel amplifiers can be built according to the design
described above, as shown in Fig. 5. One of the electrodes (REF)
is buffered and is common for all channels. The same electrode is
connected to a current source transconductance reciprocal
resistor to signal ground (or half of the ADC reference). The
remaining channels have VCCSs connected to their inputs. Each
VCCS is driven by the difference between signal ground (or half
of the ADC reference) and a filtered DC component input
voltage. Each channel amplifies the voltage difference
between its input, referenced to the common electrode REF.
Thus a pseudo-differential multichannel system is achieved.
The output signal can be directly obtained by ADC
or referenced to the circuit common point by differential
amplifiers. The amplified differential voltage between input 1
and REF is
Fig. 6 Lead I electrocardiogram of volunteer taken simultaneously
by (a) commercial electrocardiograph and (b) the amplifier of
Fig. 4
V (1a ) V (1b ) V (REF)
The REF electrode common for all channels is usually connected
to the left leg. Setting the current source transconductance
depends on the number of channels, and it can be selected
slightly lower than for the single-channel amplifier, for example
in the range of 1=500–1=100 kO.
6 Results and conclusions
A sample recording of an ECG signal acquired using a
commercial electrocardiograph* and the proposed ECG amplifier is shown in Fig. 6. This type of three-channel electrocardiograph was selected owing to its abilities to record one lead I ECG
synchronously with two ‘experimental inputs’, where external
units can be connected. The trace in Fig. 6a was obtained by the
electrocardiograph own amplifier (lead I) and, in Fig. 6b, the
signal from the proposed amplifier output is displayed. Standard
stick-on disposable ECG electrodes were used, two for the ECG
channel and two for the tested amplifier, at 5 cm distance from
each other on the arms, plus a third one for the ECG unit, which
required a reference electrode. The two signals were identical,
except for a small difference in channel sensitivities. Lowamplitude electromyogram signals can be observed in both
traces.
The measured CMRR was 60 dB, using 1 Vpp 50 Hz common
mode voltage. The measurements were extended for the
frequency range of 3–129 Hz, yielding the same value. In
addition, this value includes common mode input voltage and
input current simultaneously, owing to the low common mode
input impedance (21 kO). The common mode input current was
48 mApp.
Eliminating the two current sources at the amplifier inputs
produced CMRR ¼ 66 dB. Obviously, the price for the common
mode input impedance reduction (which prevents saturation by a
high level of common mode noise) was the loss of 6 dB CMRR,
mainly owing to non-ideal resistor matching in the current
sources.
The following advantages of this circuit should be pointed out:
(i)
the overall gain is ensured by the first stage; thus a high
CMRR is obtained without the use of high-precision
resistors in the second stage
(ii) additional input buffers are avoided by connecting the low
frequency determining RC network to the inverting inputs
of op amp pair which amplifies the input signal
(iii) implementing different common mode and differential
mode input impedances achieves two goals:
–
Fig. 5 Ground-free multichannel amplifier principle
improved tolerance to input common mode currents, thus avoiding saturation even with low supply
voltage;
– low resistive differential impedance component,
helping to minimize and equalize electrode polarisation potentials difference
(iv) low supply current and power consumption: 150 mA,
0.45 mW
(v) acceptable input common mode currents (<50 mA) and
input DC differential voltage (2 V)
Acknowledgments—The author thanks Professor I. K. Daskalov
*EK53R, Hellige
Medical & Biological Engineering & Computing 2004, Vol. 42
for the useful discussions and help.
275
References
BURKE, M. J., and GLEESON, D. T. (2000): ‘A micropower dryelectrode ECG preamplifier’, IEEE Trans. Biomed. Eng., 47,
pp. 155–162
CHEE, J. (2002): ‘Low-frequency high gain amplifier with high DCoffset voltage tolerance’. US patent, US6396343 B2
DOBREV, D. P., and DASKALOV, I. K., (2002): ‘Two-electrode biopotential amplifier with current-driven inputs’, Med. Biol. Eng. Comp.,
40, pp. 122–127
DOBREV, D. P. (2002): ‘Two-electrode biopotential amplifier’, Med.
Biol. Eng. Comp., 40, pp. 546–549
JORGOVANOVIC, N., PETROVIC, R., DOSEN, S., and POPOVIC, D.
(2001): ‘A novel AC amplifier for electrophysiology: active DC
suppression with differential to differential amplifier in the feedback
loop’. 23rd Ann. Internat. Conf. IEEE EMBS, CD-ROM, paper
441.pdf
MCCLELLAN, A. D. (1981): ‘Extracellular amplifier with bootstrapped
input stage results in high common-mode rejection’, Med. Biol.
Eng. Comp., 19, pp. 657–658
NEUMAN, M. R. (1998): ‘Biopotential amplifiers’ in WEBSTER, J. G.
(Ed.): ‘Medical instrumentation: applications and design’, 3rd edn
(John Wiley & Sons, New York, 1998), pp. 262–264
NEUMAN, M. R. (1995): ‘Biopotential electrodes’, in BRONZINO, J. D.
(Ed): ‘Biomedical engineering handbook’ (CRC Press, Boca Raton,
1995), pp. 745–757
276
PALLAS-ARENY, R., and WEBSTER, J. G. (1993): ‘AC instrumentation
amplifier for bioimpedance measurements’, IEEE Trans. Biomed.
Eng., 40, pp. 830–833
SPINELLI, E. M., and MAYOSKY, M. A. (2000): ‘AC coupled three opamp amplifier with active DC suppression’, IEEE Trans. Biomed.
Eng., 47, pp. 1616–1619
THAKOR, N., and WEBSTER, J. (1980): ‘Ground-free ECG recording
with two electrodes’, IEEE Trans. Biomed. Eng., 27, pp. 699–704
Author’s biography
DOBROMIR DOBREV obtained his MSc in electronic engineering from
the Technical University of Sofia, in 1994. He specialized in medical
electronics, with a diploma thesis on filtering and amplification of
biosignals. He has worked in the Institute of Medical Engineering of
the Medical Academy as a Research Assistant and, since 1997, has
been with the Centre of Biomedical Engineering of the Bulgarian
Academy of Sciences. His PhD is in the field of neonatal monitoring.
The study of analogue circuits, including the design, simulation and
integration of biosignal amplifiers and filters, electrical impedance
measurement circuits and transient processes in amplifiers, are among
his present research interests.
Medical & Biological Engineering & Computing 2004, Vol. 42