You are on page 1of 8

Microelectronics Journal 92 (2019) 104606

Contents lists available at ScienceDirect

Microelectronics Journal
journal homepage: www.elsevier.com/locate/mejo

A 0.25-V fifth-order Butterworth low-pass filter based on fully differential


difference transconductance amplifier architecture
Paulo M. Pinto, Luís H.C. Ferreira ∗ , Gustavo D. Colletta, Rodrigo A.S. Braga
Institute of Systems Engineering and Information Technology, Federal University of Itajubá, Itajubá, 37500-903, Brazil

A R T I C L E I N F O A B S T R A C T

Keywords: This paper presents a fifth-order Butterworth low-pass filter based on the fully differential difference transcon-
Fifth-order Butterworth filter ductance amplifier (FDDTA) building blocks. At first, the FDDTA has been stated and its operation has been
Weak inversion
evaluated. Then, a practical implementation using two fully differential inverter-based operational transconduc-
Analog filtering
tance amplifiers (OTA) was investigated. This particular FDDTA implementation relies on two main features:
Fully differential difference transconductance
amplifier
the intrinsically matched transistors that assure similar transconductances and output conductances for both
Ultra-low-voltage and ultra-low-power circuits inverter-based OTA instances; and the inverter-based approach without internal nodes that reduces circuit com-
and applications plexity and power consumption since it requires no supplementary external calibration circuit such as tail current
or bias voltage sources. Finally, the filter architecture, which consists of one inverter-based OTA input stage and
five FDDTAs in a cascade connection has been evaluated, showing that it presents the expected fifth-order trans-
fer function according to the Butterworth theory. The prototypes, implemented in a 130 nm CMOS process,
operate in weak inversion supplied with 0.25 V and consumes 603 nW. Furthermore, they feature a DR of 57 dB
in a 100 Hz bandwidth and a maximum THD of 54 dB, therefore, accomplishing specifications that suit for
low-frequency applications.

1. Introduction output stage. Since they are intended to build Gm -C filters, a high
linear and low transconductance, and a high output impedance are the
Nowadays, practical Butterworth filter implementations target the main requirements. Moreover, some FDDTAs implementations using
analog front-ends (AFE) for wearable biomedical applications, like ECG uniform transistors operating in strong inversion have been reported
monitors, where they are used as anti-aliasing filters [1–3]. Modern in Ref. [11], while a floating gate (FG) transistor implementation
AFEs are mobile devices that communicate wirelessly to a smartphone operating with a reduced supply voltage was reported in Ref. [12],
or cloud services, and therefore they require low power circuit solutions and a quasi-floating gate (QFG) transistor implementation has been
[4–7]. reported in Ref. [13].
Regarding the system on chip (SoC) solutions, the main strategy Although we can see significant improvement in the FDDTA archi-
to save power relies on reducing the number of active elements in tectures over the years, they still require controller circuits to stabi-
the Gm -C Butterworth filter implementations. In general, the conven- lize or tune the transconductance, and therefore they consume more
tional transconductor element based on the operational transconduc- power.
tance amplifier (OTA) has been replaced by the fully differential differ- Regarding this scenario, we propose a low complexity circuit imple-
ence transconductance amplifiers (FDDTA) [8–10]. Moreover [9,10], mentation of a fifth-order anti-aliasing Butterworth filter that also relies
operate in weak inversion to reduce power consumption, but they on a particular implementation of the FDDTA. Since our filter does
require a triple well CMOS process and controller circuits to stabilize not require tuning, the FDDTA implementation uses two intrinsically
and tune the transconductance of the transconductor building block. matched inverter-based transconductors without external circuit for
The FDDTA is a six-terminal device that comprises two differential common mode control or tuning. Therefore the whole filter benefits
voltage input ports which converts them into a differential current of power savings.

∗ Corresponding author.
E-mail addresses: pmarcos@unifei.edu.br (P.M. Pinto), luis@unifei.edu.br (L.H.C. Ferreira), gustavo.colletta@unifei.edu.br (G.D. Colletta), rodrigobraga@unifei.
edu.br (R.A.S. Braga).

https://doi.org/10.1016/j.mejo.2019.104606
Received 6 June 2019; Received in revised form 4 August 2019; Accepted 22 August 2019
Available online 29 August 2019
0026-2692/© 2019 Elsevier Ltd. All rights reserved.
P.M. Pinto et al. Microelectronics Journal 92 (2019) 104606

The remainder of this manuscript is organized as follows: Section 2 (0.4 μm/0.6 μm), enabling threshold voltages of 230 mV and 190 mV
states the FDDTA building block operation and implementation. Section respectively.
3 provides the proposed fifth-order Butterworth low-pass filter archi- Furthermore, its intrinsically matched feature assures the same
tecture. The circuit characterization and discussions are presented in transconductance and output impedance to both OTA instances; hence
Section 4. Section 5 concludes our contributions. simplifying the FDDTA overall design. For your convenience, we
reprinted the inverter-based OTA in Fig. 1d and the performance bench-
2. The fully differential difference transconductance amplifier marks in Table 1.
The inverter-based OTA features a high-linear and low transconduc-
This section describes the operation and implementation of the tance of 2.46 μS, which is mandatory for Gm -C applications. In addition,
FDDTA that we use as a building block to accomplish the fifth-order the power consumption of 55 nW is appropriate for low-power applica-
Butterworth filter. Furthermore, it explains the reasons that make the tions.
inverter-based transconductor reported in Ref. [14] a good candidate Since the inverter-based OTA presents no internal nodes, the invert-
to implement the FDDTA. ers Inv3, Inv4, Inv5, and Inv6 of both OTA instances are connected in
The FDDTA, illustrated in Fig. 1a, is a six-terminal device that parallel when we short circuit the outputs of the two OTA instances.
comprises two differential voltage input ports, (Vpp − Vpn ) and Fig. 1e shows that the parallel connected inverters, in fact, act as single
(Vnp − Vnn ) which converts the two differential voltages into a differ- inverters with doubled aspect ratios, and therefore simplify the overall
ential current output stage (Iop − Ion ). Operating in the linear range, circuit.
the output is In summary, the proposed FDDTA architecture adds two input
inverters to the OTA architecture and doubles the aspect ratios of the
Iop − Ion = Gm [(Vpp − Vpn ) − (Vnp − Vnn )] , (1) inverters Inv3, Inv4, Inv5, and Inv6, still keeping the low-complexity
approach to control the common mode voltage. Moreover, since the
where Gm states the FDDTA transconductance.
proposed FDDTA architecture requires no additional voltage drops it
In this sense, we propose to implement the FDDTA architecture,
still operates supplied with 0.25 V.
illustrated in Fig. 1b, in terms of two fully differential OTA instances
In fact, a similar transconductor has been reported in Refs. [9,10],
operating in weak inversion with short circuit outputs.
where simulations validated its use in applications of radio-frequency
In addition, we can apply two small differential signals (vid1 , vid2 ) to
communications such as Zigbee and Bluetooth. It is constructed in a
the FDDTA inputs according to
triple well CMOS process that enables to stabilize and tune the transcon-
vid1 vid1 ductance through the bulk of the transistors. Although it operates in
Vpp = VCM + , Vpn = VCM − ; (2a)
2 2 weak inversion it still requires a controller circuit and therefore con-
and sumes additional power.
vid2 vid2 Since low-frequency anti-aliasing filters, in general, do not require
Vnp = VCM + , Vnn = VCM − , (2b) tuning, our approach relies on a digital CMOS process with transistors
2 2
that features reduced threshold voltages and smaller channel lengths.
where VCM is the common-mode voltage. Thereby, our transconductor can operate in weak inversion at a reduced
Moreover, the small-signal AC model, illustrated in Fig. 1c, shows supply voltage of 0.25 V. In addition, the distributed layout technique
that the differential output voltage is assures a stable transconductance, eliminating the controller circuit,
Gm1 vid1 (s) − Gm2 vid2 (s) and therefore saving power.
vod (s) = , (3)
(CP1 + CP2 )s + Go1 + Go2
3. Proposed fifth-order Butterworth filter
where Gm stands for the OTA transconductance while CP and Go stand
for the OTA output (parasitic) capacitance and output conductance, This section details our proposed fifth-order Butterworth filter archi-
respectively, per each single-ended output branch. Assuming two iden- tecture built with the FDDTA blocks. Furthermore, we carried a circuit
tical matched OTAs, Gm1 = Gm2 = Gm , CP1 = CP2 = CP and Go1 = Go2 = analysis showing that we can, in fact, accomplish a fifth-order Butter-
Go , leads to the output voltage, worth filter by setting the appropriate values to some parameters.
vod (s) Gm Feasible implementation of a low-pass filter may rely on the Butter-
= , (4)
worth theory [15], that shows an m-order low-pass filter design accord-
vid1 (s) − vid2 (s) 2CP s + 2Go
ing to
showing that the FDDTA features the same transconductance of a sin-
T0
gle OTA and an increased output admittance. For low frequencies, the T (s ) = ( ) , (6)
differential output current, iod = iop − ion , can be given by Bm 𝜔s
c

iod = Gm (vid1 − vid2 ), (5) where 𝜔c is the cutoff frequency and T0 is the DC gain. The polynomials
are normalized by setting 𝜔c = 1, therefore presenting the form of
showing that the differential current in terms of the original input sig-
nals, Vpp , Vpn , Vnp , and Vnn leads to (1); hence testifying the FDDTA ∏[ (
2k + m − 1
) ]
Bm (s) = (s + 1) s2 − 2s cos 𝜋 + 1 , m = odd. (7)
operation. 2m
k
Although we can pick any fully differential transconductor to imple-
ment the FDDTA block illustrated in Fig. 1b, our choice relies on the where, for a normalized fifth-order Butterworth polynomial, the coeffi-
inverter-based OTA that we previously reported in Ref. [14], since it cients are
operates in weak inversion with a power supply of 0.25 V and requires
B5 (s) = (s + 1)(s2 + 0.6180s + 1)(s2 + 1.6180s + 1) , (8)
no supplementary external calibration circuits, and therefore saves
power. In addition, it uses the design technique of distributed layout showing that there is one pole located at 0◦ and that the others are
transistors that matches all the base inverters, enabling an efficient apart by multiples of 36◦ [15].
(low-complexity) approach to improve the common-mode voltage con- The conventional fifth-order Butterworth low-pass Gm -C filter imple-
trol. mentation, depicted in Fig. 2a, shows that it requires eleven transcon-
All unity pMOS and nMOS transistors within the distributed lay- ductors properly connected. Our approach intends to reduce the num-
out have their aspect ratio (W∕L) equal to (2.0 μm/2.0 μm) and ber of circuit elements by replacing the highlighted transconductors

2
P.M. Pinto et al. Microelectronics Journal 92 (2019) 104606

Fig. 1. The FDDTA block: symbol and implementation.

by the FDDTA building block, since the differential outputs of those and OTA2b, OTA3a and OTA3b, OTA4a and OTA4b, and OTA5a and
transconductors are connected in short circuit, in the same fashion of OTA5b are replaced by five FDDTA instances. Thereby, the proposed
the FDDTA block. Therefore, the instances OTA1a and OTA1b, OTA2a filter architecture, depicted in Fig. 2b, comprises one inverter-based

3
P.M. Pinto et al. Microelectronics Journal 92 (2019) 104606

Table 1 And finally, assuming C5 ≫ CP , the fifth FDDTA (FDDTA5) transfer


Summary of inverter-based OTA performance [14]. function is
Parameter Value vop (s) − von (s) Gm
= . (13)
Technology 130-nm [von (s) − von4 (s)] − [vop (s) − vop4 (s)] 2sC5 + 2Go
Power supply 0.25-V
Transconductance 2.46-μS Manipulating (9), (10), (11), (12) and (13), give us the transfer func-
Output conductance 56.73-nS tion of the complete architecture according to
Open loop gain 24.92-dB
Output (parasitic) capacitance 13.4-pF
√ vod (s) G5m
T (s ) = = 5 , (14)
White noise floor 139-nVrms ∕ Hz vid (s) 𝛼 s + 𝛽 s + 𝛾 s3 + 𝛿 s2 + 𝜖 s + 𝜃
4
Linear input range HD3 ≤ 0.1% 18.8-mVpk @10-Hz
Offset voltage ≤0.78-mV where vid (s) = vip (s) − vin (s) and vod (s) = vop (s) − von (s). Table 2
Power consumption 54.75-nW summarizes the coefficients values. Note that, all coefficient values
(𝛼, 𝛽, 𝛾, 𝛿 and 𝜖 ) rely on the constant values C1 , C2 , C3 , C4 , C5 , Go ,
and Gm . It means that we can accomplish the fifth-order Butter-
worth filter stated in (6) by choosing appropriate values for the
OTA input stage (OTA1 ) and five FDDTAs stages (FDDTA1 -FDDTA5 ) capacitance, conductance, and transconductance parameters. In addi-
associated in a cascaded connection. tion, the open loop gain is -6dB when Gm > Go , because T (0) =
Assuming C1 ≫ CP , the transfer function of the first FDDTA stage G5m ∕𝜃 ≈ 1∕2. Moreover, Fig. 3 shows the location of the poles for
(FDDTA1 ), including the OTA input stage (OTA1 ), is the transfer function with a Butterworth response, where each pole
lies along a semicircle in the left half-plane and are spaced at
vop1 (s) − von1 (s) Gm equal angular distances, as required by (8) to be a Butterworth
= .
[vip (s) − vin (s)] + [von2 (s) − vop1 (s)] − [vop2 (s) − von1 (s)] 2sC1 + 3Go filter.

(9)
4. Measurements and results
Assuming C2 ≫ CP , the transfer function of the second FDDTA
stage (FDDTA2 ) is Fig. 4 shows the micrograph of the fabricated circuit obtained
with Cascade Microtech MPS150 Probe Station where the layout
vop2 (s) − von2 (s) Gm is overlaid for better visualization. Our current layout for the
= . (10)
[von3 (s) − von1 (s)] − [vop3 (s) − vop1 (s)] 2sC2 + 2Go proposed fifth-order Butterworth filter architecture, actually, reuses
the layouts of the OTA circuit that we have previously published
Assuming C3 ≫ CP , the transfer function for the third FDDTA in Ref. [14].
(FDDTA3) is We have interconnected the OTAs to build the FDDTAs build-
vop3 (s) − von3 (s) ing blocks, then we made a cascade connection of the FDDTAs,
Gm
= . (11) leading to the proposed filter architecture. For measurement conve-
[von4 (s) − von2 (s)] − [vop4 (s) − vop2 (s)] 2sC3 + 2Go
nience, we used external capacitors within the commercial range.
Assuming C4 ≫ CP , the fourth FDDTA (FDDTA4) features the The whole architecture has been simulated in the Spectre simulator
transfer function according to, using BSIM4 models and implemented using the GF 130-nm CMOS
process.
vop4 (s) − von4 (s) Gm Fig. 5a shows the measured frequency response of the fifth-order
= . (12)
[von (s) − von3 (s)] − [vop (s) − vop3 (s)] 2sC4 + 2Go Butterworth filter taken with the Keysight 35670A Dynamic Signal Ana-

Fig. 2. Fifth-order Butterworth low-pass filter implementations.

4
P.M. Pinto et al. Microelectronics Journal 92 (2019) 104606

Table 2
Coefficients values of the filter transfer function.
Coefficient Value
𝛼 32C1 C2 C3 C4 C5
𝛽 16Gm [C1 C2 C3 C4 + C2 C3 C4 C5 ] + 16Go [2C1 C2 C3 C4 + 2C1 C2 C3 C5 + 2C1 C2 C4 C5 + 2C1 C3 C4 C5 + 3C2 C3 C4 C5 ]
𝛾 8G2m [C1 C2 C3 + C1 C2 C5 + C1 C4 C5 + C2 C3 C4 + C3 C4 C5 ] + 16G2o [2C1 C2 C3 + 2C1 C2 C4 + 2C1 C2 C5 + 2C1 C3 C4 + 2C1 C3 C5 + 2C1 C4 C5 +
3C2 C3 C4 + 3C2 C3 C5 + 3C2 C4 C5 + 3C3 C4 C5 ] + 8Go Gm [2C1 C2 C3 + 2C1 C2 C4 + 2C1 C3 C4 + 5C2 C3 C4 + 2C3 C4 C5 + 2C2 C3 C5 + 2C2 C4 C5 ]
𝛿 4G3m [C1 C2 + C2 C3 + C2 C5 + C3 C4 + C1 C4 + C4 C5 ] + 4Go G2m [4C1 C2 + 2C1 C3 + 4C1 C5 + 5C2 C3 + 3C2 C5 + 2C2 C4 + 2C1 C4 + 5C4 C5 +
4C3 C4 + 2C3 C5 ] + 8G2o Gm [2C1 C2 + 2C1 C3 + 5C2 C3 + 2C1 C4 + 5C2 C4 + 2C2 C5 + 5C3 C4 + 2C3 C5 + 2C4 C5 ] + 16G3o [2C1 C2 + 2C1 C3 +
2C1 C4 + 2C1 C5 + 3C2 C3 + 3C2 C4 + 3C2 C5 + 3C3 C4 + 3C3 C5 + 3C4 C5 ]
𝜖 2G4m [C1 + C2 + C3 + C4 + C5 ] + 2Go G3m [4C1 + 7C2 + 4C3 + 7C4 + 4C5 ] + 4G2o G2m [6C1 + 8C2 + 7C3 + 7C4 + 8C5 ] + 8G3o Gm [2C1 + 5C2 +
5C3 + 5C4 + 2C5 ] + 16G4o [2C1 + 3C2 + 3C3 + 3C4 + 3C5 ]
𝜃 48G5o + 40G4o Gm + 52G3o G2m + 28G2o G3m + 11Go G4m + 2G5m

Fig. 3. The pole plot of the filter transfer function.

Fig. 4. Circuit micrograph overlayed with layout.


Fig. 5. Fifth-order Butterworth measurements.

lyzer. In addition, it also shows the ideal simulation (dashed line) for
reference. Using C1 = C5 = 1 nF, C2 = C4 = 2.7 nF, and C3 = 3.3 nF fore, our proposed circuit, in fact, performs a fifth-order Butterworth
we can set the cut-off frequency to 100-Hz which is compliant to filter.
biomedical applications. Although we have ranged the differential input Some minor divergences around 100 Hz in both magnitude and
with the amplitudes of 14.2 mVpk , 28.4 mVpk , and 42.6 mVpk , the cir- phase are due to the tolerance of 20% of the external capacitors. Above
cuit keeps the low-pass characteristic with a 100 dB/dec roll-off. There- 350 Hz the signal amplitude is attenuated outside the instrument mea-

5
P.M. Pinto et al. Microelectronics Journal 92 (2019) 104606

35670 A Dynamic Signal Analyzer is 250 μVpk , the measurements above


350 Hz are not reliable.
We also have evaluated the dynamic performance of the pro-
posed Butterworth filter by carrying an FFT analysis using the 35670A
Dynamic Signal Analyzer. Fig. 5b shows the measured output spec-
trum with 31.25 mHz resolution where a 125 mV common mode
voltage was applied to both inputs with three differential sinusoidal
signals of 10 Hz and amplitudes of ±14.2 mVpk , ±28.4 mVpk , and
±42.6 mVpk at each time. Increasing the amplitude of the input volt-
age enables us to investigate the second and third harmonic distor-
tions. In both scenarios the THD, closer to HD3 , reveals a controlled
HD2 .
Since we use an intrinsically matched FDDTA building block, we
expect the fifth-order Butterworth filter presents the same feature.
Therefore, we conducted some Monte Carlo simulation to validate this
feature by evaluating the THD and HD2 distribution due to mismatch
variation. Fig. 6a shows the 1000 runs Monte Carlo simulation for
THD regarding the three input voltages of 14.2 mVpk , 28.4 mVpk , and
42.6 mVpk . In addition, process, power supply, and temperature are
in the typical range. Increasing the input voltage leads to increases
in the THD as expected. Furthermore, the THD spreads over a 6 dB
range within 99.74% of confidence regarding the worst spread scenario
(14.2 mVpk ).
Fig. 6b shows the 1000 runs Monte Carlo simulation for HD2 regard-
ing the same three input voltages and the same process, power sup-
ply, and temperature conditions. Increasing the input voltage leads
to a minor increase in the HD2 ; hence, showing that HD2 is under
control. In addition, the HD2 features practically the same distri-
bution for the three input voltages, highlighting that it is under
control.
In summary, the Monte Carlo simulations validate the intrinsically
matched feature of the fifth-order Butterworth filter, once the HD2 is
low and under control.
The measured power spectral density of input-referred noise is illus-
trated in Fig. 7a. Integrating the in-band noise from 62.5 mHz to
100 Hz provides an input-referred noise voltage value of 4.7 μVrms .
Fig. 7b depicts the relationship between the differential input range
and the signal-to-noise and distortion ratio (SINAD) of the filter. The
frequency of the input signal is fixed at 10 Hz. The dynamic range
is 57 dB under a 0.25-V power supply. Table 3 shows a performance
Fig. 6. Fifth-order Butterworth Monte Carlo analysis. comparison between this work and other low-voltage Butterworth
filter where our proposed architecture features one of the smallest
surement range, and therefore some measurement fails occurs. For THDs and high DR within the hundreds of nW range, operating under
instance, at 350 Hz the attenuation is – 60 dB, which represents 14 0.25 V. Moreover, the proposed filter presents the lowest input referred
μVpk for a given input of 14 mVpk . Since the resolution of the Keysight noise.

Fig. 7. Fifth-order Butterworth noise and signal analysis.

6
P.M. Pinto et al. Microelectronics Journal 92 (2019) 104606

5. Conclusion

TBioCAS 2009 [21]

Fully Differential
This paper presents a fifth-order Butterworth low-pass filter archi-

Butterworth

300-μVrms
0.13 mm2
tecture which relies on the FDDTA building block. Since the FDDTA

90.6-nW
453-nW

48.6-dB
50.0-dB
180-nm

250-Hz
1.0-V
building block is intrinsically matched, the proposed Butterworth fil-

OTA
ter requires no supplementary external calibration circuit, such as

5
tail current or bias voltage sources, and therefore reduces circuit
TBioCAS 2013 [20]

complexity.
The prototypes, implemented in a 130 nm CMOS process, operate
Butterworth

Differential
in weak inversion supplied with 0.25 V and consumes 603 nW. Fur-
0.11 mm2

31-μVrms
3.75-nW

55.2-dB
66.7-dB
350-nm

100-Hz
15-nW
thermore, they feature a DR of 57 dB in a 100 Hz bandwidth and a
3.0-V
SSF

maximum THD of 54 dB; hence accomplishing specifications that suits


4

for low-frequency applications.


Springer 2013a [10]

Fully Differential

43.84–43.67-dB
Acknowledgment

MODI-OTA: multiple-output differential-input – operational transconductance amplifier. DDA: differential difference amplifier. SSF: subthreshold-source-follower.
69–90-μVrms
Butterworth

2.77-mW

1–2-MHz

This work was supported in part by the Brazilian National Council


230-μW
130-nm

FDDTA

0.5-V

for Scientific and Technological Development (PQ 303090/2018-9 and


12

GD 140929/2017-7) and FAPEMIG. The authors would like to thank


MOSIS for the chip fabrication.
TCAS II 2017 [19]

Fully-Differential
Butterworth

0.21 mm2

1.35-mW
8.07-mW

65-MHz
180-nm

Appendix A. Supplementary data


1.8-V
DDA
6



Supplementary data to this article can be found online at https://


doi.org/10.1016/j.mejo.2019.104606.
0.09–1.96-MHz
MEJ 2018 [18]

1.6–2.1-mW
Butterworth

Differential

0.35 mm2

References
400-μW
180-nm

1.8-V
OTA

[1] K. Lasanen, J. Kostamovaara, A 1-V analog CMOS front-end for detecting QRS
4



complexes in a cardiac signal, IEEE Trans. Circuits Syst. I: Reg. Pap. 52 (12)
TCAS II 2018 [2]

Fully Differential

(2005) 2584–2594.
[2] C.-Y. Sun, S.-Y. Lee, A fifth-order Butterworth OTA-C LPF with multiple-output
Butterworth

MODI-OTA

differential-input OTA for ECG applications, IEEE Trans. Circuits Syst. II: Express
0.12 mm2

Briefs (2018) 421–425.


350-nW

49.9-dB
49.8-dB
180-nm

70-nW
50-Hz

[3] A. Rao, Y.-C. Teng, C. Schaef, E.K. Murphy, S. Arshad, R.J. Halter, K. Odame, An
1.0-V

analog front end ASIC for cardiac electrical impedance tomography, IEEE Trans.
5

Biomed. Circuits Syst. 12 (4) (2018) 729–738.


[4] T.-H. Tsai, J.-H. Hong, L.-H. Wang, S.-Y. Lee, Low-power analog integrated circuits
TCAS I 2018 [17]

Fully Differential

for wireless ECG acquisition systems, IEEE Trans. Inf. Technol. Biomed. 16 (5)
(2012) 907–917.
80.5-μVrms

[5] D. Gangopadhyay, E.G. Allstot, A.M. Dixon, K. Natarajan, S. Gupta, D.J. Allstot,
0.11 mm2
Performance comparison between proposed filter and other low-voltage low-pass filters.

4.26-nW
1.06-nW

48.2-dB
50.0-dB
350-nm

Compressed sensing analog front-end for bio-sensor applications, IEEE J. Solid


100-Hz
Biquad

0.9-V

State Circuits 49 (2) (2014) 426–438.


OTA

[6] Z. Zhu, W. Bai, A 0.5-V 1.3-μW analog front-end CMOS circuit, IEEE Trans.
4

Circuits Syst. II: Express Briefs 63 (6) (2016) 523–527.


[7] H. Bhamra, J. Lynch, M. Ward, P. Irazoqui, A noise-power-area optimized
Fully Differential
VLSI 2018 [16]

biosensing front end for wireless body sensor nodes and medical implantable
Butterworth

46.27-μVrms

devices, IEEE Trans. Very Large Scale Integr. (VLSI) Syst. 25 (10) (2017)
0.168 mm2

0.225-nW

2917–2928.
47.0-dB
60.0-dB
350-nm

0.9-nW

101-Hz

[8] S.-Y. Lee, C.-P. Wang, Y.-S. Chu, Low-voltage OTA–C filter with an area-and
0.6-V
OTA

power-efficient OTA for biosignal sensor applications, IEEE Trans. Biomed. Circuits
4

Syst. 13 (1) (2019) 56–67.


[9] R. Arya, G. Souliotis, S. Vlassis, C. Psychalinos, 0.5V 3rd -order tunable g m -C filter,
TBioCAS 2019 [8]

Fully Differential

Radioengineering 22 (1) (2013) 174–178.


[10] R. Arya, G. Souliotis, S. Vlassis, C. Psychalinos, A 0.5V tunable complex filter for
Butterworth

MOFD-OTA

Bluetooth and Zigbee using OTAs, Analog Integr. Circuits Signal Process. (2014)
134-μVrms
0.24 mm2

73–81.
61.2-dB
180-nm

250-Hz
8.2-nW
41-nW

[11] S.A. Mahmoud, A.M. Soliman, New CMOS fully differential difference
1.0-V

transconductors and application to fully differential filters suitable for VLSI,


5

Microelectron. J. 30 (2) (1999) 169–192.


[12] M. Kumngern, F. Khateb, Fully differential difference transconductance amplifier
Fully Differential

using FG-MOS transistors, in: 2015 International Symposium on Intelligent Signal


Processing and Communication Systems (ISPACS), 2015, pp. 337–341.
Butterworth
This work

0.67 mm2

120.6-nW

[13] F. Khateb, M. Kumngern, T. Kulej, V. Kledrowetz, Low-voltage fully differential


4.7-μVrms
603-nW

57.0-dB
54.0-dB
130-nm

100-Hz
FDDTA

difference transconductance amplifier, IET Circuits, Devices Syst. 12 (1) (2018)


0.25-V

73–81.
5

[14] R.A. Braga, L.H. Ferreira, G.D. Coletta, O.O. Dutra, A 0.25-V calibration-less
Simulated values.

inverter-based OTA for low-frequency Gm-C applications, Microelectron. J. 83


Power consumption

(2019) 62–72.
CMOS technology

[15] M.E. Van Valkenburg, Analog Filter Design, Holt, Rinehart, and Winston, 1982.
Power per pole
Filter order (n)

Active deviceb

Power supply

[16] C. Sawigun, S. Thanapitak, A 0.9-nW, 101-Hz, and 46.3-μ V rms IRN low-pass filter
THD or HD3
Architecture

Bandwidth

for ECG acquisition using FVF biquads, IEEE Trans. Very Large Scale Integr. (VLSI)
Filter type
Parameter

Syst. 26 (11) (2018) 2290–2298.


Table 3

[17] S. Thanapitak, C. Sawigun, A subthreshold buffer-based biquadratic cell and its


Area

IRN
DR

application to biopotential filter design, IEEE Trans. Circuits Syst. I: Reg. Pap. 65
b
a

(9) (2018) 2774–2783.

7
P.M. Pinto et al. Microelectronics Journal 92 (2019) 104606

[18] J. Wu, Z. Xie, T. Yu, C. Chen, A wide tuning range Gm-C complex filter with [20] T.-T. Zhang, P.-I. Mak, M.-I. Vai, P.-U. Mak, M.-K. Law, S.-H. Pun, F. Wan, R.P.
master-slave automatic frequency tuning based switched-capacitor, Microelectron. Martins, 15-nW biopotential LPFs in 0.35-μm CMOS using
J. 81 (2018) 200–207. subthreshold-source-follower biquads with and without gain compensation, IEEE
[19] J.S. Mincey, C. Briseno-Vidrios, J. Silva-Martinez, C.T. Rodenbeck, Low-power Trans. Biomed. Circuits Syst. 7 (5) (2013) 690–702.
Gm-C filter employing current-reuse differential difference amplifiers, IEEE Trans. [21] S.-Y. Lee, C.-J. Cheng, Systematic design and modeling of a OTA-C filter for
Circuits Syst. II: Express Briefs 64 (6) (2017) 635–639. portable ECG detection, IEEE Trans. Biomed. Circuits Syst. 3 (1) (2009) 53–64.

You might also like