複素アナログフィルタの基礎 - koba lab official … › lecture › mce2020_06.pdf1...
TRANSCRIPT
11
計測制御工学 第6回講義
複素アナログフィルタの基礎
小林春夫群馬大学大学院理工学府 電子情報部門
下記から講義使用 pdfファイルをダウンロードしてください。
出席・講義感想もここから入力してください。
https://kobaweb.ei.st.gunma-u.ac.jp/lecture/lecture.html
2020年6月1日(月)
/191
Kobayashi Lab.
Gunma University
欧州のアナログ集積回路研究教育事情
群馬大学 小林春夫
2020/5/11
Disclaimer:
15年前に訪問した時の情報もあり、一部は最新情報ではない
/192
欧州地図
/193
意外な欧州
● 日本は面積も人口も少ない?
欧州で面積&人口が日本より大の国なし
● 欧州は人口少ない、大都市少ない
アジアは多い
「アジアは米、欧州は小麦が主食のため」の説
米は何年も備蓄可、小麦は不可
● 「イギリス(ブリテン島)と欧州大陸は別」
の意識が双方にあり (?)
/194
米国と欧州は異なる
/195
欧州 2000年以上の西洋科学の伝統の底力
/196
オランダ
/197
ベルギー(大学)
/198
ベルギー(研究所)
/199
イタリア
/1910
イギリス、アイルランド
イギリス: 海外では UK を使う
England, Scotland, Wales, Northern Ireland の連合王国 United Kingdom
サッカー ワールドカップにも独立4チーム出場
/1911
ドイツ
/1912
スイス
/1913
北欧
/1914
ロシア
半導体メーカ技術者より
/1915
ポルトガル、イスラエル
/1916
スペイン バルセロナ市 雑景
サグラダファミリア(Sagrada Familia)
聖家族 贖罪教会
/1917
フランス
● グルノーブル(Grenoble): フランスのシリコンバレー
/1918
フランス パリ 雑景
パリ第6大学 セーヌ川 エッフェル塔 凱旋門
/1919
日本企業への売り込み
/1920
関係ファイル
https://kobaweb.ei.st.gunma-u.ac.jp/news/pdf/2019/oldrepo_NapoliUniv.pdf
https://kobaweb.ei.st.gunma-u.ac.jp/news/pdf/2019/oldrepo_Europe.pdf
https://kobaweb.ei.st.gunma-u.ac.jp/warehouse/20160722am9IMSTW.pdf
https://kobaweb.ei.st.gunma-u.ac.jp/warehouse/IMSTW20150703.pdf
Gunma University Kobayashi Lab
Complex Signal Processingin Analog / Mixed-Signal Circuits
Haruo Kobayashi
Minh Tri Tran, Koji Asami
Anna Kuwana, Hao San
Division of Electronics and InformaticsGunma University
IPS04 Analog and Power
3rd International Conference on Technology and Social Science
May 8, 2019Keynote Lecture 03
Full Body
Outline
● Motivation for Complex Signal Processing Research
● RC Polyphase Filter: Transfer Function
● RC Polyphase Filter: Flat Passband Gain Algorithm
● RC Polyphase Filter and Hilbert Filter
● Active Complex Bandpass Filters
● Conclusion
2/57
Outline
● Motivation for Complex Signal Processing Research
● RC Polyphase Filter: Transfer Function
● RC Polyphase Filter: Flat Passband Gain Algorithm
● RC Polyphase Filter and Hilbert Filter
● Active Complex Bandpass Filters
● Conclusion
3/57
Why My Research for Complex Signal Processing ?
About 15 years ago
at IEEE International Solid-State Circuits ConferenceSan Francisco, CAThe most prestigious conference in IC design
Katholieke Universiteit Leuven (KU Leuven), BelgiumWorld top research group in analog IC design
Some simple circuit with curious characteristics
presentation
However, I could not understand its principle
4/57
Basics of Complex Signal
2 real signals: I, Q
Vsignal = I + jQ Complex Signal
Vimage= I – jQ Image
I = [Vsignal + Vimage]/2
Q = [Vsignal – Vimage]/(2 j)
j・ j = -1
Basic complex signal processing blocks
5/57
Outline
● Motivation for Complex Signal Processing Research
● RC Polyphase Filter: Transfer Function
● RC Polyphase Filter: Flat Passband Gain Algorithm
● RC Polyphase Filter and Hilbert Filter
● Active Complex Bandpass Filters
● Conclusion
6/57
Research Goal of First Research
• To establish systematic design and analysis
methods of RC polyphase filters.
• As its first step,
to derive explicit transfer functions of
the 1st-, 2nd- and 3rd-order
RC polyphase filters.
7/57
Features of RC Polyphase Filter
• Its input and output are complex signal.
• Passive RC analog filter
• One of key components in wireless transceiver analog front-end
- I, Q signal generation
- Image rejection
• Its explicit transfer function
has not been derived yet.
8/57
First-order RC Polyphase Filter
R1C1
Iin+
Qin+
Iin-
Qin-
Iout+
Qout+
Iout-
Qout-
C1
C1
C1
R1
R1
R1
I: In-Phase, Q: Quadrature-Phase
Differential Complex Input: Vin = Iin + j Qin
Differential Complex Output: Vout = Iout + j Qout
9/57
I, Q Signal Generation
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
0.2 0.22 0.24 0.26 0.28 0.3 0.32 0.34 0.36 0.38 0.4
vo
lta
ge
[V
]
time [us]
Iout Qout
11
1
CRLO =
Polyphase
FilterQin = 0
Iin = cos (ωLO t)Iout = A cos (ωLO t+θ)
Qout = A sin (ωLO t+θ)
Single cosine Cosine, Sine signals
10/57
Cosine, Sine Signals in Receiver
IF Analog
Bandpass
Filter
cos(ω0 t)
I
Q
AD
Converter
- sin(ω0 t)
analog digital
AD
Converter
They are used for down conversion
11/57
Problem when ωLO≠1/R1C1
-3
-2
-1
0
1
2
3
0.2 0.22 0.24 0.26 0.28 0.3 0.32 0.34 0.36 0.38 0.4
vo
lta
ge
[V
]
time [us]
QoutIout
11
2
CRLO=
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
0.2 0.22 0.24 0.26 0.28 0.3 0.32 0.34 0.36 0.38 0.4
voltage [
V]
time [us]
Iout Qout
11
1
CRLO =
12/57
2nd-order RC Polyphase Filter
Iout Qout
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
0.2 0.22 0.24 0.26 0.28 0.3 0.32 0.34 0.36 0.38 0.4
voltage [
V]
time [us]
11
2
CRLO=
The problem of large
difference between
Iout, Qout amplitudes
can be alleviated
R1
R1
R1
R1
C1
C1
C1
C1 C2 C2
C2
C2
R2
R2
R2
R2
Iin+
Qin+
Iin-
Qin-
Iout+
Qout+
Iout-
Qout-
13/57
3rd-order RC Polyphase Filter
11
2
CRLO=
Iout Qout
The amplitude
difference problem
is further alleviated.
R1
R1
R1
R1
C1
C1
C1
C1C2
C2
C2
C2
R2
R2
R2
R2
R3C3
R3C3
R3C3
R3
C3
Iin+
Qin+
Iin-
Qin-
Iout+
Qout+
Iout-
Qout-
14/57
Pure I, Q Signal Generation
With3rd-order harmonics.
Without3rd-order harmonics.
15/57
Simulation of 3rd-order Harmonics Rejection
-8
-6
-4
-2
0
2
4
6
8
20 22 24 26 28 30 32 34 36 38 40
vo
lta
ge
[V
]
time [ns]
Iout Qout
)(sin)sin()(
)(cos)cos()(
3
3
tattQ
tattI
LOLOin
LOLOin
+=
+=
)sin()(
)cos()(
+=
+=
tAtQ
tAtI
LOout
LOout
-8
-6
-4
-2
0
2
4
6
8
0.2 0.22 0.24 0.26 0.28 0.3 0.32 0.34 0.36 0.38 0.4
voltage [
V]
time [us]
Iin Qin
11
13
CRLO=
16/57
Image Rejection Filter
Polyphase
Filter
Iin =
(A+B) cos(ωt)
Qin =
(A-B) sin(ωt)
Iout =
Acos(ωt)
Qout =
Asin(ωt)
tjtj BeAe −+tjAe
signal image
17/57
Approach (1)
Complex Input: Vin = Iin + j Qin
Complex Output: Vout = Iout + j Qout
Complex Signal Processing
There is NO physical complex signal.
It is only defined mathematically.
Complex Signal Processing is NOT Complex.
Gauss plane
18/57
Complex Transfer Function
• Complex Signal Theory
• Complex input
• Complex output
• Complex
Transfer Function
)(
)()(
jV
jVjG
in
out=
outoutout
ininin
QjIjV
QjIjV
+=
+=
)(
)(
19/57
Signals in RC Polyphase Filter
)()()(
)()()(
)()()(
)()()(
)()()(
)()()(
tjQtItV
tjQtItV
tQtQtQ
tItItI
tQtQtQ
tItItI
outoutout
ininin
outoutout
outoutout
ininin
ininin
+=
+=
−=
−=
−=
−=
−+
−+
−+
−+
Differential signal
Complex signal
R1
C1
Iin+
Qin+
Iin-
Qin-
Iout+
Qout+
Iout-
Qout-
C1
C1
R1
R1
R1C1
20/57
Transfer Function of RC Polyphase Filter
21
1
)(1
1)(
.1
1)(
RC
RCjG
RCj
RCjG
+
+=
+
+=
•Transfer Function
•Gain
RC
1−
Asymmetric
21/57
Explanation of I, Q Signal Generation by G1(jω)
𝑄𝑖𝑛 𝑡 ≡ 0, 𝐼𝑖𝑛 𝑡 = cos 𝜔𝑡
𝑉𝑖𝑛 𝑡 = 𝐼𝑖𝑛 𝑡 + 𝑗 𝑄𝑖𝑛 𝑡 = cos 𝜔𝑡 =1
2[𝑒𝑗𝜔𝑡 + 𝑒−𝑗𝜔𝑡]
𝐺1 𝑗𝜔𝜔=
1𝑅𝐶
= 0 , 𝐺1 𝑗𝜔𝜔=
1𝑅𝐶= 2 , ∠𝐺1 𝑗𝜔 = −
𝜋
4
𝑉𝑜𝑢𝑡 𝑡 =1
2[ 𝐺1 𝑗𝜔 𝑒𝑗 𝜔𝑡+∠𝐺1 𝑗𝜔 + 𝐺1 −𝑗𝜔 𝑒𝑗(−𝜔𝑡+∠𝐺1(−𝑗𝜔))]
=2
2cos 𝜔𝑡 −
𝜋
4+
𝑗 2
2sin(𝜔𝑡 −
𝜋
4)
Here𝑮𝟏 −𝒋𝝎 𝒆𝒋(−𝝎𝒕+∠𝑮𝟏(−𝒋𝝎))] = 𝟎
22/57
Outline
● Motivation for Complex Signal Processing Research
● RC Polyphase Filter: Transfer Function
● RC Polyphase Filter: Flat Passband Gain Algorithm
● RC Polyphase Filter and Hilbert Filter
● Active Complex Bandpass Filters
● Conclusion
23/57
Transfer Function of 2nd-order RC Polyphase Filter
Transfer Function
Derivation is very complicated, so we used “Mathematica.”
Gain |G2(jω)| characteristics
𝐺2 𝑗𝜔 =(1 + 𝜔𝑅1𝐶1)(1 + 𝜔𝑅2𝐶2)
1 − 𝜔2𝑅1𝐶1𝑅2𝐶2 + 𝑗𝜔(𝐶1𝑅1 + 𝐶2𝑅2 + 2𝑅1𝐶2)
24/57
Gain of 3rd-order RC Polyphase Filter
Gain:
Phase:
11
1
CR−
22
1
CR−
33
1
CR−
Gain characteristics
)(
)())(tan(
)()(
)()(
3
33
23
23
33
jD
jDjG
jDjD
jNjG
R
I
IR
−=
+=
25/57
Need for Flat Passband Gain Algorithm
Transfer Function
Gain |G2(jω)| characteristics
𝐺2 𝑗𝜔 =(1 + 𝜔𝑅1𝐶1)(1 + 𝜔𝑅2𝐶2)
1 − 𝜔2𝑅1𝐶1𝑅2𝐶2 + 𝑗𝜔(𝐶1𝑅1 + 𝐶2𝑅2 + 2𝑅1𝐶2)
We need flat passband gain
26/57
Four Design Parameters
4 parameters : 𝑅1, 𝑅2, 𝐶1, 𝐶2
𝜔1 =1
𝑅1𝐶1, 𝜔2 =
1
𝑅2𝐶2, 𝑋 =
1
𝑅2𝐶1, 𝑌 =
1
𝑅1𝐶2
4 constraints
27/57
Two Constraints from Filter Spec.
● 2 zeros : −𝜔1=− 1
𝑅1𝐶1, − 𝜔2=
−1
𝑅2𝐶2
are given from the filter specification.
28/57
Proposed Algorithm Uses Third Constraint
● We use the third constraint 𝑋 =1
𝑅2𝐶1
for passpand gain flattening.
● The fourth constraint is left for ease of IC realization.29/57
Nyquist Chart of G2(jω)
|G2(jω1)||G2(jω2)|
Gain characteristics |G2(jω)| Nyquist chart of G2(jω)=X(ω)+j Y(ω)
Y(ω
)X(ω)
|G2(jω1)|=|G2(jω2)|
But in general |G2(jω1)|=|G2(jω2)|=|G2(j√ω1ω2)|
|G2(j√ω1ω2)|
30/57
Our Idea for Flat Passband Gain Algorithm
If we make|G2(jω1)| =|G2(jω2)| =|G2(𝑗 𝜔1𝜔2)|,Passband gain becomes flat from ω1 to ω2.
Gain characteristics |G2(jω)| Nyquist chart of G2(jω)=X(ω)+j Y(ω)
|G2(jω1)||G2(jω2)|
|G2(j√ω1ω2)|
31/57
Outline
● Motivation for Complex Signal Processing Research
● RC Polyphase Filter: Transfer Function
● RC Polyphase Filter: Flat Passband Gain Algorithm
● RC Polyphase Filter and Hilbert Filter
● Active Complex Bandpass Filters
● Conclusion
32/57
Research Objective
Analyze RC polyphase filter
We found that relevance between
RC polyphase filter and Hilbert filter
RC Polyphase Filter
Analog
Complex (I, Q) input
Hilbert Filter
Digital
Real part (Vin) input
Vin
Iout+
Qout+
33/57
Hilbert Filter
◼ Characteristics・Hilbert transform・1 input and 2 outputs・It is often implemented in digital filter
Gain
Phase
34/57
Cosine, Sine Generation with Hilbert Filter
Gain
cos(ωt)+jsin(ωt)
cos(ωt)-jsin(ωt)
cos(ωt)+jsin(ωt)
2 cos(ωt)
cos(ωt)
sin(ωt)
xω-ω
ω component
-ω component
ω component
Hilbert filter
35/57
Hilbert Transform
Complex signal from real signal 𝑥 (𝑡)𝑥 𝑡 → 𝑥 𝑡 + 𝑗𝑦 𝑡
Hilbert transform
𝑦 𝑡 =1
𝜋න
−∞
∞𝑥(𝜏)
𝑡 − 𝜏𝑑𝜏 = 𝑥 𝑡 ∗
1
𝜋𝑡
Impulse response Fourier Transform
ℎ 𝑡 =1
𝜋𝑡𝑯 𝝎 = ቊ
−𝒋 (𝝎 ≥ 𝟎)𝒋 (𝝎 < 𝟎)
Frequency characteristic 𝐻 (ω)
𝒀 𝝎 = 𝑯 𝝎 𝑿(𝝎) = ቊ−𝒋𝑿 𝝎 (𝝎 ≥ 𝟎)
𝒋𝑿 𝝎 (𝝎 < 𝟎)
𝐹𝑜𝑢𝑟𝑖𝑒𝑟
David Hilbert1862-1943
Phase
36/57
余弦波の複素平面での解釈
37/57
「現世」と「あの世」
38/57
ヒルベルトフィルタで負周波数成分のカット
39/57
ヒルベルトフィルタで負周波数が見える
40/57
1st order RC Polyphase Filter:Analysis
R1
R1
R1
R1
C1
C1
C1
C1
Iin+
Qin+
Iin-
Qin-
Iout+
Qout+
Iout-
Qout-
Zero:
Pass bandStop band
: Transfer function𝐻1 𝑗𝜔 =1 + 𝜔𝑅1𝐶11 + 𝑗𝜔𝑅1𝐶1
ωk =1
𝑅𝑘𝐶𝑘41/57
1st order RC Polyphase Filter : Gain and Phase
H1re − H1im H1re + H1im
H1re = H1im
π
2phase lag
π
2phase lead
Gain Phase
𝐻1𝑟𝑒 𝑗𝜔 =𝐻1 𝑗𝜔 + 𝐻1
∗(−𝑗𝜔)
2=
1
1 + 𝑗𝜔𝑅1𝐶1
𝐻1𝑖𝑚 𝑗𝜔 =𝐻1 𝑗𝜔 − 𝐻1
∗(−𝑗𝜔)
2= −𝑗
𝜔𝑅1𝐶11 + 𝑗𝜔𝑅1𝐶1
𝐻1 𝑗𝜔 = 𝐻1𝑟𝑒 𝑗𝜔 + 𝑗𝐻1𝑖𝑚(𝑗𝜔)
ー H1reー H1imー H1
42/57
1st order case Analysis Results
Gain : Hilbert filter only at zero
Phase : Completely Hilbert filter
Gain Phase
Hilbert filter
RC Polyphase Filter
43/57
Results:2nd to 4th RC Polyphase Filter
Gain Phase
2nd
3rd
4th
ー H1reー H1imー H1
44/57
Analysis Results and Consideration
1st to 4th order RC Polyphase Filter Analysis results
Prove for general n-th order case(n = 1, 2, 3, 4, 5, ...)
Gain : Hilbert filter only at zero
Phase : Completely Hilbert filter
45/57
Order and Gain
-20 -10 0 10 20
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
相対角周波数 [rad/s] (RC=1/(2π))
|H|
-20 -10 0 10 20
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
相対角周波数 [rad/s] (RC=1/(2π))
|H|
-20 -10 0 10 20
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
相対角周波数 [rad/s] (RC=1/(2π))
|H|
-20 -10 0 10 20
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
相対角周波数 [rad/s] (RC=1/(2π))
|H|
1st 2nd 3rd 4th
The higher orders,Increase number of zeros|𝐻𝑟𝑒| and |𝐻𝑖𝑚| becomes close in wide range
Close to ideal Hilbert transform
46/57
Order and Phase
Phase characteristic is not
changed
There is always 90° phase
difference
1st 2nd 3rd 4th
Fulfill Hilbert transform in full range
-20 -10 0 10 20-π
π/2
0
π/2
π
相対角周波数 [rad/s] (RC=1/(2π))
∠H
-20 -10 0 10 20-π
π/2
0
π/2
π
相対角周波数 [rad/s] (RC=1/(2π))
∠H
-20 -10 0 10 20-π
π/2
0
π/2
π
相対角周波数 [rad/s] (RC=1/(2π))
∠H
-20 -10 0 10 20-π
π/2
0
π/2
π
相対角周波数 [rad/s] (RC=1/(2π))
∠H
47/57
Conclusion
RC polyphase filter is approximation of ideal Hilbert filter for complex input signal
-20 -10 0 10 20
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
相対角周波数 [rad/s] (RC=1/(2π))
|H|
RC Polyphase Filter Hilbert Filter
PhaseGain Gain Phase
-20 -10 0 10 20-π
π/2
0
π/2
π
相対角周波数 [rad/s] (RC=1/(2π))
∠H
48/57
Outline
● Motivation for Complex Signal Processing Research
● RC Polyphase Filter: Transfer Function
● RC Polyphase Filter: Flat Passband Gain Algorithm
● RC Polyphase Filter and Hilbert Filter
● Active Complex Bandpass Filters
● Conclusion
49/57
Gm : Transconductance
Input voltage: Vin
Output current : Iout Iout=gmVin
Vin
Ioutgm
I+ I-
Ib
Vin
+
-
dimension of gmR
1
Iout=I+-I-=gmVin
50/57
Complex Bandpass Gm-C Filter
Iin
Qin
C
C
gm
-
+
VIout
VQout
g0
g0
CsgCsgg
jgscg
jQinIin
jVV
m
m
QoutIout
02222
0
0
2+++
−+=
+
+
+
51/57
Gain of Complex Bandpass Gm-C Filter
ω0
C
gm
C
g
C
gm 0+C
g
C
gm 0−
0
1
2
1
g
0
1
g
Center Freq.C
gm
0g
gmQ =
Gain
52/57
Complex Bandpass Active RC Filter
Iout+
Qout-
Iin
Iout-
Qout+
Qin
+
-
+
-
g2
g2 g3
g3
g1
g1
C
C
2
1
2
3300
32
1
|)(|2
)()(
g
gjH
g
gQQ
C
g
Cgjg
gjH
===
+−+
−=
ゲイン値 中心周波数Center freq. Gain
53/57
Our Investigation Results
● Transfer functions of complex bandpass Gm-C and active RC filtersare the same.
● Both complex bandpass filters are NOTclose to Hilbert filter
Phase characteristics are far from Hilbert.
● RC polyphase filter is close to Hilbert filter.
54/57
Outline
● Motivation for Complex Signal Processing Research
● RC Polyphase Filter: Transfer Function
● RC Polyphase Filter: Flat Passband Gain Algorithm
● RC Polyphase Filter and Hilbert Filter
● Active Complex Bandpass Filters
● Conclusion
55/57
Conclusion
RC polyphase filter is simple, but very interesting
Even somewhat mysterious !
To understand its principle, we useits complex transfer function and
Hilbert transfer form.
These are useful for filter design as well as analysis
56/57
Final Statement
Complex Signal Processing is NOT Complex.
Quadrature Signals: Complex, But Not Complicated.
Our World is in Complex Domain.
Real Imaginary
57/57
付録: スイッチトキャパシタ回路
58/57
スイッチトキャパシタ回路の動作原理
59/57
スッチト・キャパシタ積分回路
60/57
H. San, Y. Jingu, H. Wada, H. Hagiwara,
A. Hayakawa, J. Kudoh2, K. Yahagi2,
T. Matsuura2, H. Nakane2, H. Kobayashi,
M. Hotta3, T. Tsukada2, K.Mashiko4, A.Wada5 1) Gunma University
2) Renesas Technology Corp.
3) Musashi Institute of Technology
4) STARC
5) Sanyo Electric Co., Ltd.
A Multibit Complex
Bandpass ∆ΣAD Modulator
with I,Q Dynamic Matching
and DWA Algorithm
2006/11/14 A-SSCC 2006 2
Motivation
Complex Bandpass Delta-Sigma
AD Modulator
Proposed Architecture
I, Q Dynamic Matching
Complex DWA Algorithm
Measured Results
Conclusion
Outline
2006/11/14 A-SSCC 2006 3
Motivation
Complex Bandpass Delta-Sigma
AD Modulator
Proposed Architecture
I, Q Dynamic Matching
Complex DWA Algorithm
Measured Results
Conclusion
Outline
2006/11/14 A-SSCC 2006 4
Low power ADC in low-IF receiver
targeted for bluetooth, wireless LAN.
Motivation
Complex bandpass delta-sigma
AD modulator
2006/11/14 A-SSCC 2006 5
fLOZero-IF
DC Frequency
Signaloffset
1/f noise
fLO
Signal
offset1/f noise
DC Frequency
Low-IFImage
fLO
Signal
DC Frequency
Low-IF
offset1/f noise
RF → Baseband
Zero-IF
⇒ No mage
Problem of DC offset, flicker noise
RF → Low-IF
No problem of DC offset, flicker noise.
Image as well as signal are
AD converted ⇒ Power is wasted
Image is not AD converted.
Direct conversion receiver
Low-IF receiver Conventional
Receiver Architecture Comparison
Quadrature-IF
2006/11/14 A-SSCC 2006 6
Motivation
Complex Bandpass Delta-Sigma
AD Modulator
Proposed Architecture
I,Q Dynamic Matching
Complex DWA Algorithm
Measured Results
Conclusion
Outline
2006/11/14 A-SSCC 2006 7
H(z)
Complex
Banpass Filter
ADCI
ADCQ
DACI
DACQ
+
+
Iin
Qin
Iout
Qout
Analog
Input
Digital
Output
Ei
Eq
-
-
)jEE(H1
1)jQI(
H1
H
jQI
qiinin
outout
Complex bandpass noise-shaping
Complex Bandpass Delta-Sigma Modulator
2006/11/14 A-SSCC 2006 8
Motivation
Complex Bandpass Delta-Sigma
AD Modulator
Proposed Architecture
I,Q Dynamic Matching
Complex DWA Algorithm
Measured Results
Conclusion
Outline
2006/11/14 A-SSCC 2006 9
Proposed Architecture
• New complex bandpass filter
• Multi-bit ADCs/DACs
• Complex DWA algorithm
2006/11/14 A-SSCC 2006 10
Proposed Structure
2006/11/14 A-SSCC 2006 11
Conventional complex filter
I &Q crossing paths
I,Q Dynamic Matching of Complex Filter
Proposed complex filter
Upper, lower separated paths
I,Q mismatch reduction.
Layout simplification.
2006/11/14 A-SSCC 2006 12
Operation of Proposed Complex Filter
Iout(n) = Iin(n-1) - Qout(n-1)
Qout(n) = Qin(n-1) + Iout(n-1)
2006/11/14 A-SSCC 2006 13
Complex BPDSM with Low-power
• 2nd order ---- low power
• 9-level ADCs/DACs
• Stability improvement
• Low quantization error
• Power reduction of amplifiers
I,Q mismatch
• Solved by dynamic matching
Nonlinearities of multibit DAC
• Solved by complex DWA
2006/11/14 A-SSCC 2006 14
Complex DWA (1)
jz
1)z(H1
jz)z(H2
Digital bandpass filter Analog band elimination filter
2006/11/14 A-SSCC 2006 15
Complex DWA (2)
2006/11/14 A-SSCC 2006 16
Motivation
Complex Bandpass Delta-Sigma
AD Modulator
Proposed Architecture
I,Q Dynamic Matching
Complex DWA Algorithm
Measured Results
Conclusion
Outline
2006/11/14 A-SSCC 2006 17
Chip Implementation
• 1P6M 0.18µm CMOS Process
• Core size 1.4 *1.3mm2.
2006/11/14 A-SSCC 2006 18
Measured Output Power Spectrum
2006/11/14 A-SSCC 2006 19
Effect of Complex DWA
2006/11/14 A-SSCC 2006 20
Summary of Modulator Performance
Technology 0.18-µm CMOS 1P6M
Supply voltage 2.8V
Sampling Frequency 20MHz
SNDR 64.5dB @ BW=78kHz
Power consumption 28.4mw
Active area 1.4mm*1.3mm
2006/11/14 A-SSCC 2006 21
Motivation
Complex Bandpass Delta-Sigma
AD Modulator
Proposed Architecture
I,Q Dynamic Matching
Complex DWA Algorithm
Measured Results
Conclusion
Outline
2006/11/14 A-SSCC 2006 22
• A 2nd-order multi-bit complex bandpass
delta-sigma modulator
Low power
• Complex filter with dynamic matching
– I,Q mismatch reduction
– Layout simplification
• Complex DWA
– Suppression of multibit DACs nonlinearities
• Chip measurements demonstrated these
Conclusion