Professional Documents
Culture Documents
JSSC 2013 ChopperNoise
JSSC 2013 ChopperNoise
Abstract—This paper presents a theoretical analysis and mea- the order of several tens of fA/sqrt(Hz). This is roughly a hun-
surements of the current noise of several chopper instrumentation dred times higher than the current noise of conventional CMOS-
amplifiers, which demonstrate that the charge injection and clock or JFET-input amplifiers [10]. Some commercially available
feed-through associated with the MOSFETs of the input chopper
give rise to significant input current and current noise. In combi- chopper amplifiers exhibit even higher current noise densities
nation with high source impedances, this “chopper noise” is con- ( 100 fA/sqrt(Hz)) [12]–[14]. When used with high-impedance
verted to voltage noise, which may then be a significant contributor sensors such as dry bio-potential electrodes, photodiodes and
to the amplifier’s total input-referred voltage noise. Chopper noise piezoelectric sensors, this current noise will be converted into
has a white power spectral density, whose magnitude is roughly voltage noise, which will then add to, and may even dominate,
proportional to the chopping frequency. Design guidelines are pro-
posed to reduce chopper noise, as well as the use of a clock-boot- the amplifier’s total input-referred voltage noise.
strapped chopper, which generates significantly less noise than a This paper presents the results of noise measurements on sev-
traditional chopper. eral chopper amplifiers [15] and proposes a theory that explains
Index Terms—Charge injection, chopper amplifier, chopping,
the behavior of the measured current noise, which we will refer
current noise, high-impedance sensors, noise. to as chopper noise in the remaining of this paper. The measure-
ments reveal that chopper noise has a white power spectral den-
sity (PSD) and is caused by the clock feed-through and charge
I. INTRODUCTION injection of the MOSFETs of the input chopper. To first order,
both the resulting input current and the PSD of the associated
Fig. 1. Excess voltage noise due to the source impedance and input current
noise of a chopper amplifier.
Fig. 3. Periodic charge injection and associated transient current of an input
chopper switch.
This periodic charge injection and clock feed-through at the The clock driver circuit is another possible source of noise.
chopping frequency causes transient current spikes (Fig. 3), As shown in Fig. 4, it can be modeled as a resistance in
whose average value is given by series with the gate-source capacitance . Since this resistor
(and any other series resistance in the gate charging circuit) will
(3) generate thermal noise, the channel charge will fluctuate and
so a certain noise charge will be injected into the surrounding
XU et al.: MEASUREMENT AND ANALYSIS OF CURRENT NOISE IN CHOPPER AMPLIFIERS 1577
TABLE I
PARAMETERS OF A TYPICAL MOSFET SWITCH AND THE CALCULATED
CURRENT NOISE VOLTAGE DENSITY CONTRIBUTION OF A CMOS CHOPPER
CONSISTING OF EIGHT MOSFETS ( C OR 298 K)
E. Summary
Fig. 5. Input parasitic SC resistance of a chopper amplifier.
The total chopper noise PSD is obtained by sum-
ming the contributions of all the above-mentioned current noise
circuitry every time the MOSFET is turned off. The rms value sources. Table I shows the parameters of the MOSFET switches
of this noise charge can be expressed as (in an ON Semiconductor 0.5 m CMOS process) used in the
input chopper of a CMOS chopper instrumentation amplifier
(6)
[4]. Also shown is the calculated contribution of each noise
source assuming that eight of these transistors (four NMOS
and four PMOS) are used to realize the four complementary
As before, this periodically injected noise charge will give
switches of its input chopper. The results show that the total
rise to an average rms noise current of
current noise is dominated by the contribution of charge injec-
(7) tion, and so from (5), the chopper noise PSD should be linearly
proportional to the chopping frequency, which is in line with the
Assuming that this impulsive noise is approximately white measurements reported in [9].
and distributed over the fundamental interval between 0 and
, then its PSD is given by III. INPUT CURRENT NOISE MEASUREMENTS ON
CHOPPER AMPLIFIERS
(8) In this section, noise measurements on different types of
chopper instrumentation amplifiers are presented.
This PSD is also a linear function of the chopping frequency.
A. Conventional Chopper Modulated Instrumentation
D. Parasitic Switched-Capacitor (SC) Resistance Amplifier
Due to the action of the input chopper, the amplifier’s input Fig. 6 shows the setup used to measure the current noise of
parasitic capacitances will be charged and discharged by a chopper instrumentation amplifier (IA) [4]. Its input chopper
the input voltage and give rise to a net DC current [6], [7]. As consists of 4 complementary CMOS switches, whose character-
shown in Fig. 5, this effect can be modeled by a switched-ca- istics are shown in Table I. The IA is configured with a voltage
pacitor resistance at the amplifier’s input [22], [23]. This gain of 800 and a bandwidth of 200 Hz. Since it was intended
resistance generates current noise in the same manner as a phys- for biomedical applications, an internal DC-servo loop ensures
ical resistor [24]: that the amplifier has a high-pass characteristic with a corner
frequency around 0.5 Hz.
with (9) A low-noise input bias voltage, , is generated from a 3.3-V
battery. Two large series resistors ensure that the
amplifier’s current noise is the dominant contributor to its total
1578 IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 48, NO. 7, JULY 2013
Fig. 6. Current noise measurement of a chopper instrumentation amplifier. Fig. 8. Measured input-referred noise of the chopper amplifier.
input-referred noise, i.e., that (10) and (11) are satisfied [25] as The measured input-referred current noise PSD at various
follows: chopping frequencies is shown in Fig. 9 and Fig. 10. As pre-
dicted by (5), the PSD of this chopper noise is linearly propor-
(10) tional to . Fig. 11 shows that, the measured input current
noise density is independent of the value of source resistance,
(11)
as expected.
Fig. 12 shows the measured SC input impedance of the
The input current noise PSD can then be determined from
chopper amplifier at different chopping frequencies. The
(12), where is the measured output noise voltage, is
smallest input impedance is about 250 M , which corresponds
the amplifier’s voltage gain, and the amplifier’s voltage noise
to the highest chopping frequency of 16 kHz. Hence, the
and source resistance are known. Note that the thermal
maximum current noise density associated with this SC input
noise of the choppers’ on-resistance is included in the measured
impedance is only 8 fA/sqrt(Hz), which is negligible compared
:
to the measured total current noise density of 158 fA/sqrt(Hz)
at this chopping frequency.
Fig. 14. Measured and simulated input referred noise of inverting CCOPA
without chopping.
Fig. 11. Measured input noise current density with different source 0.18- m process and its input chopper consists of four NMOS
resistors kHz .
devices .
Due to the presence of C in the feedback path, the excess
sistor and a capacitor C in the feedback path to define its voltage noise PSD exhibits a spectrum (with a pole at
voltage gain and establish a high-pass corner at about 0.5 Hz. As ), which is given by
a result, the amplifier’s virtual ground is a high-impedance node,
which converts chopper noise into significant amounts of ex-
(13)
cess voltage noise. The amplifier was implemented in a TSMC
1580 IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 48, NO. 7, JULY 2013
Fig. 15. Measured and simulated input referred voltage noise of inverting CCOPA at variable chopping frequencies.
Fig. 16. Non-inverting capacitively coupled chopper amplifier (CCOPA). Fig. 17. Measured and simulated input referred voltage noise of non-inverting
CCOPA without chopping.
With chopping disabled, the current noise is quite low, and so
any noise is buried in noise. This has been verified by current noise densities ranged from 7.5 fA/sqrt(Hz) at 500 Hz
periodic noise simulations and measurements (Fig. 14). As can to 21 fA/sqrt(Hz) at 5 kHz, and in line with (5), are roughly
be seen, the simulation results match the measurement results proportional to the square root of the chopping frequency.
well. A similar effect occurs in the non-inverting opamp shown
With chopping enabled, however, the ensuing chopper noise in Fig. 16, which was also implemented in a TSMC 0.18- m
results in excess voltage noise, which dominates the process [27]. The amplifier employs a CMOS chopper with
amplifier’s noise performance (Fig. 15). In order to simulate equally sized PMOS and NMOS devices .
the effect of chopper noise, a current noise source at the With chopping disabled, the amplifier’s noise is dom-
high-impedance chopping node (Fig. 13) was added. As shown inant and the measured noise is in good agreement with
in Fig. 15, its magnitude was then adjusted to fit the measure- simulations (Fig. 17). With chopping enabled, noise
ments obtained at different chopping frequencies. The resulting becomes dominant since the amplifier’s inverting input is a
XU et al.: MEASUREMENT AND ANALYSIS OF CURRENT NOISE IN CHOPPER AMPLIFIERS 1581
Fig. 18. Measured and simulated input referred voltage noise of non-inverting CCOPA at variable chopping frequencies.
Fig. 20. Four types of input chopper switches: (a) NMOS, (b) NMOS with dummy switches, (c) CMOS, (d) NMOS with bootstrapped clock drivers.
Fig. 21. Current noise PSD comparison of NMOS chopper amplifiers at var- Fig. 22. Current noise PSD comparison of four chopper amplifiers
ious chopping frequencies. kHz .
is related to the charge injection, and hence the shot noise, As shown in Fig. 23, all the alternative chopper architectures
associated with the individual chopper switches. As such it do reduce the amplifier’s DC input current to various degrees.
cannot be canceled by using dummy or complementary MOS- Apparently, the charge injection of the main NMOS switches
FETs. In fact, the use of additional MOSFETs only increases can be significantly reduced by a low-swing clock driver, and ef-
the total amount of charge injection and hence the total amount fectively canceled by the use of simultaneously clocked PMOS
of current noise. However, the bootstrapped NMOS chopper is or dummy switches, thus leading to lower input currents. As ex-
driven by a low-swing clock, which reduces its charge injection pected from (3), the input current of all four chopper amplifiers
and thus leads to less current noise. increases monotonically with .
XU et al.: MEASUREMENT AND ANALYSIS OF CURRENT NOISE IN CHOPPER AMPLIFIERS 1583
[27] S. Mitra, J. Xu, A. Matsumoto, K. A. A. Makinwa, C. Van Hoof, and R. Chris Van Hoof (M’87) received the Ph.D. degree in
F. Yazicioglu, “A 700 8-channel EEG/contact-impedance acquisi- electrical engineering from the University of Leuven,
tion system for dry-electrodes,” in Proc. Symp. VLSI Circuits (VLSIC), Leuven, Belgium, in 1992.
Jun. 2012, pp. 68–69. Since 1992, he has been with IMEC, Leuven. He
[28] Q. Fan, J. H. Huijsing, and K. A. A. Makinwa, “A 78 30 V is currently the Director of Heterogeneous Integrated
input common-mode range and 160 dB CMRR capacitively-coupled Microsystems and the Director of the HUMAN++
chopper instrumentation amplifier with 5 offset for high-side cur- healthcare program at IMEC in Leuven and at the
rent-sensing applications,” in IEEE ISSCC Dig. Tech. Papers, Feb. Holst Centre, a collaboration between imec and TNO
2012. in Eindhoven, The Netherlands.
[29] Y. Kusuda, “A 5.9 nV Hz chopper operational amplifier with 0.78 His work focuses on wearable healthcare and its
V maximum offset and 28.3 nV C offset drift,” in IEEE ISSCC Dig. enabling ultra low-power design and technology,
Tech. Papers, 2011, pp. 242–244. smart sensors and imagers, and heterogeneous integration. He has a track
record of 20 years of initiating, executing, and leading R&D contracts with
industry, and among others, he delivered space-qualified flight hardware to
two cornerstone European Space Agency missions. He has published over 500
papers and given 40 invited talks or keynotes.
Jiawei Xu received the M.Sc. degree in micro-
electronics from Delft University of Technology,
Delft, The Netherlands, in 2006. He is currently
pursuing the Ph.D. degree at the Delft University of Refet Firat Yazicioglu received the Ph.D. degree in
Technology. electronics engineering from Katholieke Universiteit
After the M.Sc. degree, he started as a Researcher Leuven, Leuven, Belgium, in 2008 in collaboration
on analog IC design at the imec/Holst Centre in with imec, Belgium.
Eindhoven, The Netherlands, where he worked on He is currently working at imec, Belgium as R&D
low-power readout circuits for smart sensors. He Team Leader, where he is leading the “Biomedical
is currently working on low-power biopotential Integrated Circuits” team focusing on analog and
readout circuits. mixed signal integrated circuit design for biomedical
applications. During his research, he has (co)au-
thored over 50 publications, three book chapters,
and a book on ultra-low-power circuit and system
Qinwen Fan (S’10) was born in Inner Mongolia, design for biomedical applications, and authored several patents in this field.
China. She received the B.Sc. degree in electronics He has developed several generations of integrated circuits for wearable and
science and technology from Nankai University, implantable healthcare applications.
Tianjin, China, in 2006 and the M.Sc. degree in mi- Dr. Yazicioglu is the corecepient of Best Paper Award in Biomedical Circuits
croelectronics from Delft University of Technology and System Conference 2011, Smart Systems Integration Conference, 2008, and
(cum laude), Delft, Netherlands, in 2008. She is Sensors & Transducers Journal 2008. He serves in the technical program com-
currently pursuing the Ph.D. degree at TU Delft. mittees of European Solid-State Circuits Conference (ESSCIRC) and Interna-
From August 2007 to August 2008, she was an in- tional Solid-State Circuit Conference (ISSCC). He is also co-chair of Biomed-
tern at NXP Research Laboratories, Eindhoven, The ical Circuits and Systems Conference (BioCAS) 2013 in Rotterdam.
Netherlands, where she worked on a precision instru-
mentation amplifier for biomedical purposes. In Oc-
tober 2012, she joined Maxim integrated in Delft, The Netherlands. Her research
interests include precision analog amplifiers, biomedical interface circuits, and Kofi A. A. Makinwa (M’97–SM’05–F’11) received
mixed-signal integrated circuits. the B.Sc. (First Class Honsors) and M.Sc. degrees
from Obafemi Awolowo University, Ile-Ife, Nigeria,
in 1985 and 1988, respectively, the M.E.E. degree
(cum laude) from the Philips International Institute,
Johan H. Huijsing (SM’81–F’97–LF’10) was born Eindhoven, The Netherlands, and the Ph.D. degree
on May 21, 1938. He received the M.Sc. degree in from Delft University of Technology, Delft, The
electrical engineering and the Ph.D. degree from the Netherlands, in 1989 and 2004, respectively.
Delft University of Technology, Delft, The Nether- He is currently an Antoni van Leeuwenhoek
lands, in 1969 and 1981, respectively. Professor with the Faculty of Electrical Engineering,
He has been an Assistant and Associate Professor Computer Science and Mathematics, Delft Uni-
in Electronic Instrumentation at the Faculty of Elec- versity of Technology, which he joined in 1999. From 1989 to 1999, he was
trical Engineering, Delft University of Technology, a Research Scientist with Philips Research Laboratories, Eindhoven, where
since 1969. He became a Full Professor in the chair he worked on interactive displays and digital recording systems. His current
of Electronic Instrumentation in 1990, and Professor research interests include the design of precision analog circuitry, sigma-delta
Emeritus since 2003. From 1982 through 1983, he modulators, smart sensors, and sensor interfaces. This has led to four books, 18
was a Senior Scientist at Philips Research Labs., Sunnyvale, CA, USA. From patents and over 170 technical papers.
1983 until 2005, he was a Consultant for Philips Semiconductors, Sunnyvale, Prof. Makinwa is on the Program Committees of the European Solid-State
CA, USA, and since 1998 also a consultant for Maxim, Sunnyvale, CA, USA. Circuits Conference (ESSCIRC) and the Advances in Analog Circuit Design
The research work of Johan Huijsing is focused on operational amplifiers, (AACD) Workshop. From 2006 to 2012, he was on the Program Committee of
analog-to-digital converters and integrated smart sensors. He has supervised the International Solid-State Circuits Conference (ISSCC). He has been a Guest
30 Ph.D. students. He is author or coauthor of more than 300 scientific papers, Editor of the IEEE JOURNAL OF SOLID-STATE CIRCUITS (JSSC) and a Distin-
40 U.S. patents and 15 books. In 1992 he initiated the international Workshop guished Lecturer of the IEEE Solid-State Circuits Society (2008 to 2011). For
on Advances in Analog Circuit Design. He co-organized it yearly until 2003. his doctoral research, he was awarded the 2005 Simon Stevin Gezel Award from
Prof. Huijsing has been a member of the programme committee of the Euro- the Dutch Technology Foundation. He is also a corecipient of several best paper
pean Solid-State Circuits Conference from 1992 until 2002. He was chairman awards: from the JSSC, ISSCC, ESSCIRC and Transducers, among others. He
of the Dutch STW Platform on Sensor Technology and of the biannual national is an alumnus of the Young Academy of the Royal Netherlands Academy of
Workshop on Sensor Technology from 1991 until 2002. He was awarded the Arts and Sciences and an Elected Member of the IEEE Solid-State Circuits So-
title of Simon Stevin Meester by the Dutch Technology Foundation. ciety AdCom, the society’s governing board.