noise improvement for different bp filter...
TRANSCRIPT
Noise improvement for different BP filter designs
Z. Šverko*, N. Stojković*, S. Vlahinić* and I. Markovinović* * University of Rijeka – Faculty of Engineering/Department of automatics and electronics, Rijeka, Croatia
[email protected], [email protected], [email protected], [email protected]
Abstract - In this paper an analysis of noise through four
different approaches to filter design is performed. The
voltage noise spectral density and root mean square (RMS)
value of noise voltage for the simply designed (SD) active
RC, noise optimized (NO) active RC, operational
transconductance amplifiers (OTA-C) and switched
capacitor (SC) filter design are calculated for fourth-order
Chebyshev band-pass filter, with central frequency fc=4kHz,
bandwidth 1kHz and pass band ripple αmax=-0.1dB. For the
achievement realization of all four designs a cascade of two
second order sections are used. The analyses is performed
using Matlab and SPICE (LTspice) programming tools.
Keywords - band-pass filter; noise; SC filter
I. INTRODUCTION
There are different ways to achieve the desired transfer function of the filter. The most commonly used realization is with the operational amplifiers (OA) which gives a stable (solid) voltage point output and also low impedance output. On the other hand, if high impedance output is desired, then realization with general transconductance amplifier (OTA-C) sections will be used. Such design gives us a constant current source on output.
Likewise, one of the realization of the same transfer function can be with switched capacitors (SC) where resistors are replaced with switched capacitors [1, 2]. This realization is good for design in monolithic integrated technology, where the filter is completely implemented in chip using thick and thin film technology.
Except satisfaction of elementary specifications on amplitude and phase filter characteristics, there exist additional parameters of filter quality. One of them is noise. The goal is to reduce noise as much as possible, because with superposition of noise on observed signal the information that the signal transmits can be covered. Through the sections of the paper will be demonstrating how to design filters on above suggested approaches with the desire to reduce noise in filter.
II. REALIZATION
The normed transfer functions of the observed Chebyshev band-pass filter are given by:
)(2)(1)( sHsHsH a
7066.02456.02
4525.0)(1
ss
ssH b
415.13475.02
4525.0)(2
ss
ssH c
15931.02207.235931.04
22048.0)(
ssss
ssH d
The transfer functions of fourth-order Chebyshev band-pass filter sections denormed on a frequency of 4kHz are given by:
)(2)(1)( sHsHsH a
8105469.73101764.42
3103542.5)(1
ss
ssH b
8102869.53104956.32
3103969.6)(2
ss
ssH c
171099.312108462.4
29102979.13310672.74
27104251.3)(
s
sss
ssH d
A. Continuous-time active-RC filters
In Fig. 1 the Single Amplifier Biquad (SAB) topology of the band-pass filter is given.
Realization is performed as a cascade of two second order sections. The circuit transfer function is given as:
2,1 )()( i si
HsH a
R111
R121
R21
R31
R41
C21
C11
-
+
R112
R122
R22
R32
R42
C22
C12
-
+
Uin
Uout
Figure 1. Fourth-order OA band-pass filter
126 MIPRO 2019/MEET
iCiC
iGiG
iGiC
iGiG
iC
iG
iC
iGss
s
iC
iG
iG
iG
siH
21
21
31
41
2
2
1
22
1
11
3
41
)(
b
Using given filter parameters, element values can be calculated by comparing equations (2a) to (2c) with (3a) and (3b) with assumptions C1i=C2i and R1i=R2i. Values taken over from [3] are given in Table 1.
Fig. 2 shows the frequency responses of the filter. The simulated values of the frequency response are given by simulating circuit shown in Fig.1 using element values from Table 1. Unlike the former, frequency response of calculated values is given by the filter transfer function (2a) to (2c). In the zoomed inset of Fig. 2 there can be seen a minimal difference between these two characteristics (simulated-dashed, calculated-full).
B. OTA-C filter
Realization is performed as a cascade of two second order general transconductance amplifier sections. This
circuit is shown in Fig. 3. The circuit transfer function is given as:
2,1 )()( i si
HsH a
iCiC
imgimg
iC
imgss
s
iC
img
siH
21
21
2
32
2
4
)( b
Comparing the equations (4a, 4b) with (2a-2c) element values of OTA-C filter were calculated and shown in Table 2. Frequency responses are shown in Fig. 4.
4000-100
-50
0
20
Ma
gn
itu
de
[d
B]
Frequency [Hz]
400 4000 40000-200
-100
0
100
200
Ph
ase
[d
eg
]
Frequency [Hz]
4000
-0.1
0
Frequency [Hz]
Figure 2. Frequency response of the Fourth-order continuous-time
active-RC (SD) band-pass filter
TABLE II. ELEMENT VALUES OF OTA-C FILTER
1st section 2nd section
gm1i 0.27472 mS 0.22993 mS
gm2i 0.27472 mS 0.22993 mS
gm3i 0.041764 mS 0.034956 mS
gm4i 0.053542 mS 0.063969 mS
C1i 100 nF 100 nF
C2i 100 nF 100 nF
-
+
+
-
-
+
+
-
C11
C21
gm11
gm21
gm31
gm41Uin
-
+
+
-
-
+
+
-
C12
C22
gm12
gm22
gm32
gm42
Uout
Figure 3. Fourth-order OTA-C band-pass filter
4000-100
-50
0
20
Ma
gn
itu
de
[d
B]
Frequency [Hz]
400 4000 40000-200
-100
0
100
200
Ph
ase
[d
eg
]
Frequency [Hz]
4000
-0.1
0
Frequency [Hz]
Figure 4. Frequency response of the Fourth-order OTA-C BP filter
TABLE I. ELEMENT VALUES OF SIMPLY DESIGNED (SD) AND
NOISE OPTIMIZED (NO) CONTINUOUS ACTIVE-RC FILTERS [3]
SD NO
1st section 2nd section 1st section 2nd section
R11i 53.1915 kΩ 44.5208 kΩ 21.3238 kΩ 21.3238 kΩ
R12i 3.9075 kΩ 4.81995 kΩ 1.5299 kΩ 1.8538 kΩ
R2i 3.6401 kΩ 4.3491 kΩ 9.2824 kΩ 11.0902 kΩ
R3i 7.3529 kΩ 7.35294 kΩ 0.9866 kΩ 0.9866 kΩ
R4i 3.9789 kΩ 3.9789 kΩ 3.9789 kΩ 3.9789 kΩ
C1i 10 nF 10 nF 10 nF 10 nF
C2i 10 nF 10 nF 10 nF 10 nF
ki 1 1 2.55 2.55
MIPRO 2019/MEET 127
C. Discrete time SC filter
The discrete time SC filter based on Sallen-Key (SAK) topology is shown in Fig. 5.
The forward Euler transformation (s=>(z-1)/Ts) was applied for calculation of the discrete time transfer function. Where Ts (fs=4900fc) is a sampling time and which has been calculated as Ts=1/fs .
)(2)(1)( zHzHzH a
9997.09997.12
)1(4108023.5)(1
zz
zzH b
9996.09996.12
)1(4108023.5)(2
zz
zzH c
9992.09977.3
29977.539992.34
1227103667.3
)(
z
zzz
zz
zH d
The transfer function of discrete time SC filter based on Sallen-Key topology is given by:
2,143
22
)1(1)(
ii
azi
azi
a
zi
a
iGzH a
i
Ci
Ci
G5
/4
1 b
iR
Ci
Ci
Ci
a1211
c
RfiC
iRC
iC
RfiC
iRC
iC
RfiC
iC
iC
iRC
iRC
iC
iRC
iRC
iC
iRC
iC
iC
iRC
iC
iRC
iC
iC
iC
iC
ia
2221
21212
211221
2
2
11212
2
12
d
RfiC
iC
iC
iG
RfiC
iRC
iC
RfiC
iC
iC
iRC
iRC
iC
iRC
iC
iC
iRC
iC
iRC
iC
iC
iC
iC
ia
21
2221
212221
2
2
11212
2
12
3
e
Rf
Ci
Ci
Ci
Gi
Ci
Ci
a212
2
14 f
Comparing the equations (5a-5c) with (6a-6f) element values were calculated and presented in Table 3. From Fig. 6 and Fig 2. it can be seen that the Z-transformation hasn't got undesired effects on filter specifications.
TABLE III. ELEMENT VALUES OF SC FILTER
1st section 2nd section
CR1i 37.9 pF 53.7 pF
CR2i 2.82 pF 2.52 pF
C1i 120 nF 120 nF
C2i 3.3 nF 3.3 nF
CRfi 125.5 pF 313.5 pF
C4i 2.15 nF 1.53 nF
C5i 2.55 nF 5.1 nF
CR11 C11
C21
CR21
+
-
CRf1
CR51
CR41
CR12 C12
C22
CR22
+
-
CRf2
CR52
CR42
UoutUin
Figure 5. Fourth-order SC band-pass filter (SAK topology)
4000-100
-50
0
20
Ma
gn
itu
de
[d
B]
Frequency [Hz]
400 4000 40000-200
-100
0
100
200
Ph
ase
[d
eg
]
Frequency [Hz]
4000
-0.1
0
Ma
gn
itu
de
[d
B]
Frequency [Hz]
Figure 6. Frequency response of the Fourth-order SC band-pass filter
128 MIPRO 2019/MEET
III. NOISE
A. Noise in filter with OA
Fig 7. shows an equivalent circuit of the active-RC
filter for the calculation of the voltage noise spectral
density. Parallel to the resistors have been added current
noise sources: i/RkTniI 4 , whilst to the voltage
noise sources En = 2.5 nV/√Hz for LT1007 have been
added at the input of the operational amplifiers. The input
current noise source of operational amplifiers were
neglected, because its values are much smaller than
voltage noise sources, for instance In=0.4 pA/√Hz .
The voltage noise spectral density is calculated by the given equation [4]:
n
lnEj
lVT
knI
m
jkI
Tn
V2
),
(
2
)(,
2)
,(
2
)(,
)(2
In this equation TV,l and TI,k represent the transfer functions of the voltage noise sources En,i and the current noise sources In,i, respectively. Fig. 8 (a) shows the total voltage noise spectral density while Fig. 8 (b, c) shows noise components for both sections. The differences between simulated and calculated values can be considered as a result of using theoretical values based on thermal noise sources in calculations and real elements noise models using SPICE. The RMS value has been calculated by the equation:
2
1
)(2)2
(
dn
Vefn
E
In the desire to reduce voltage noise spectral density of filter we used the optimization algorithm given in [3]. All element values are taken from [3], recalculated and tested in Matlab and SPICE in order to give us smaller noise results. The total voltage noise spectral density of NO filter design is shown in Fig. 9 (a). Noise components for both sections are given through the Fig. 9 (b, c).
Comparing the noise RMS values of filters, we can conclude that optimized filter gave smaller noise RMS value.
B. Noise in OTA-C filter
The equivalent circuit for the calculation of the
voltage noise spectral density of the observed OTA-C
filter is shown in Fig. 10.
Voltage noise spectral density (7) is calculated. The
total voltage noise spectral density components are given
in Fig. 11 (a). The differences between simulated and
calculated values can be considered as a result of using
theoretical values based on thermal noise sources in
calculations using MATLAB and real elements noise
models using SPICE.
(a)
400 4000 400000
100
200
300
Frequency [Hz]
Nois
e s
pectr
al d
ensity
[nV
/Hz1
/2]
In11,1
In12,1
In2,1
In3,1
In4,1
En1,1
(b)
400 4000 400000
100
200
300
400
Frequency [Hz]
Nois
e s
pectr
al d
ensity
[nV
/Hz1
/2]
In11,2
In12,2
In2,2
In3,2
In4,2
En1,2
(c)
Figure 8. Noise voltage spectral density of Fourth-order SD band-pass
filter and its components. (a) Total noise voltage spectral density: simulated (dashed) and calculated (full). Noise RMS value is
En=54.556µV. (b) Section 1 noise voltage spectral density components.
Noise RMS value is En=34.805 µV. (c) Section 2 noise voltage spectral density components. Noise RMS value is En=42.011 µV.
R111
R121
R21
R31
R41
C21
C11
-
+
R112
R122
R22
R32
R42
C22
C12
-
+
Uin
Uout
+
+
In111
In21
In121
In31
In41
In42
In112
In22
In122
In32
En11
En12
Figure 7. Equivalent electrical noise circuit of the Fourth-order SD and
NO band-pass filter
MIPRO 2019/MEET 129
Fig. 11 (b, c) shows noise components for both
sections.
LT1228 was used as a transconductance amplifier in
this design. The input voltage noise source of this
-
+
+
-
-
+
+
-
C11
C21
gm11
gm21
gm31
gm41
-
+
+
-
-
+
+
-
C12
C22
gm12
gm22
gm32
gm42Uin
Uout
+
En11 +
En21
+
En41
+
En31
+
En12 +
En22
+
En42
+
En32
Figure 10. Equivalent electrical noise circuit of the Fourth-order OTA-C band-pass filter
(a)
400 4000 400000
40
80
120
Frequency [Hz]
Nois
e s
pectr
al d
ensity
[nV
/Hz1
/2]
In11,1
In12,1
In2,1
In3,1
In4,1
En1,1
(b)
400 4000 400000
50
100
150
Frequency [Hz]
Nois
e s
pectr
al d
ensity
[nV
/Hz1
/2]
In11,2
In12,2
In2,2
In3,2
In4,2
En1,2
(c)
Figure 9. Noise voltage spectral density of Fourth-order NO band-pass
filter and its components. (a) Total noise voltage spectral density:
simulated (dashed) and calculated (full). Noise RMS value is
En=20.412µV. (b) Section 1 noise voltage spectral density components. Noise RMS value is En=12.422 µV. (c) Section 2 noise voltage spectral
density components. Noise RMS value is En=16.197 µV.
(a)
400 4000 400000
10
20
Frequency [Hz]
Nois
e s
pectr
al d
ensity
[nV
/Hz1
/2]
En1,1
En2,1
En3,1
En4,1
(b)
400 4000 400000
10
20
Frequency [Hz]
Nois
e s
pectr
al d
ensity
[nV
/Hz1
/2]
En1,2
En2,2
En3,2
En4,2
(c)
Figure 11. Noise voltage spectral density of Fourth-order OTA-C band-
pass filter and its components. (a) Total noise voltage spectral density: simulated (dashed) and calculated (full). Noise RMS value is
En=2.4384µV. (b) Section 1 noise voltage spectral density components.
Noise RMS value is En=1.5715 µV. (c) Section 2 noise voltage spectral density components. Noise RMS value is En=1.8645 µV.
130 MIPRO 2019/MEET
transconductance amplifier is En=6 nV/√Hz and the
input current noise source is In=1.4 pA/√Hz. The current
noise source is much smaller and because of that it was
neglected in calculation procedure.
C. Noise in SC filter
The fourth-order SC band-pass filter is shown in Fig.
5. The equivalent circuit for the calculation of voltage
noise spectral density is based on [5]. Fig. 12 shows total
voltage noise spectral density and its components.
D. Comparison
If the RMS voltage noise value is measured between
0.1fc to 10fc as the indicator of noise reduction,
comparison of noise in all presented designs is given in
Table 4. Fig. 13 shows voltage noise spectral density for
all filter designs which are discussed in this paper.
IV. CONCLUSION
In this paper, noise improvement using different filter
designs is analyzed. Analyses of four different designs
are done. As noise measure RMS voltage noise value is
calculated between 400 Hz and 40 kHz. Noise analysis is
done for SD and for NO design filters based on
operational amplifiers. Noise improvement for NO is 2.5
times. Noise analyses for OTA-C and SC designs are also
done. They gave improvement of more than 20 times for
OTA-C design in comparison with SD and 10 times for
SC design in comparison with SD. In the end, considering
RMS voltage noise as a numeric indicator, it is shown
that the OTA-C filter design gives the best noise
reduction.
REFERENCES
[1] H. A. de Andrade Serra, “Design of Switched-Capacitor Filters using Low Gain Amplifiers”, PhD thesis, Faculdade de Ciencias e Technologia, Universidade Nova de Lisboa, 2012.
[2] I. Volarić, N. Stojković and S. Vlahinić, "Noise improvement using SC filters," 2015 38th International Convention on Information and Communication Technology, Electronics and Microelectronics (MIPRO), Opatija, 2015, pp. 121-126.
[3] N. Stojković, E. Kamenar, M. Šverko, “Optimized Second- and Fourth- Order LP and BP Filters”, Automatika, no 52, vol. 2, pp. 152-168, 2011.
[4] W. K. Chen, The Circuits and Filters Handbook, 2nd ed. CRC Press, 2003.
[5] J. H. Fisher, “Noise sources and calculation techniques for switched capacitor filters”, IEEE Journal of Solid-State Circuits, vol. 17., pp. 742-752, August 1982.
TABLE IV. NOISE RMS VALUE COMPARISON
SD NO OTA-C SC
En 54.556 µV 20.412 µV 2.4384 µV 5.3954 µV
(a)
(b)
(c)
Figure 12. Noise voltage spectral density of Fourth-order discrete time
SC band-pass filter and its components. (a) Total noise voltage spectral density: simulated (dashed) and calculated (full). Noise RMS value is
En=5.3954 µV. (b) Section 1 noise voltage spectral density components.
Noise RMS value is En=3.2451 µV. (c) Section 2 noise voltage spectral density components. Noise RMS value is En=4.3022 µV.
Figure 13. Noise voltage spectral density for: SD (dashed), NO
(dashed-dot), OTA-C (full), SC (dot-dot)
MIPRO 2019/MEET 131