Professional Documents
Culture Documents
IMSE
Instituto de Microelectrnica de Sevilla
(IMSE-CNM-CSIC)
Avda. Reina Mercedes s/n,
41012-Sevilla, SPAIN
Phone: +34 95 505 6666/6669
FAX: +34 95 505 6686
email: angel@imse.cnm.es
Slide CoA 0
IMSE-
Design Group
Introduction
x
Analog Signal
x
ADC
clock @
TS
xD
fS
b 1, b 2, b N .
TS
xD
0 1
TS
Design of Embeddable Data Converters: Sigma-Delta Converters
Slide CoA 1
t
IMSE-
Design Group
Introduction
Mixed-Signal Systems
Common substrate
System-on-Boards
PCB
System-on-Chip
(ASICs)
Monolithic circuit
Multi-Chip Module
Analog inputs
Speech
Images
Radiation
Sensor signals
............................
Analog outputs
Analog Part
parameters
calibration
programming
coded signals
timing
Audio
Video
Actuators
............................
Digital outputs
Digital inputs
Digital Part
Design of Embeddable Data Converters: Sigma-Delta Converters
Slide CoA 2
IMSE-
Design Group
Introduction
Portable Systems
Slide CoA 3
IMSE-
Design Group
Introduction
Reduced Supply
Slide CoA 4
IMSE-
Design Group
Introduction
y
Quantization
x
TS
Slide CoA 5
IMSE-
Design Group
Introduction
22
20
18
16
14
12
10
8
6
4
102
103
104
105
106
107
108
109
(Low-Voltage)
(current mode)
Nyquist
Nyquist (Low-Voltage)
Slide CoA 6
IMSE-
Design Group
Introduction
10-2
10-3
10-4
10-5
10-6
10-7
10-8
10-9 2
10
103
104
105
106
107
108
109
(Low-Voltage)
(current mode)
Nyquist
Nyquist (Low-Voltage)
Power ( W )
FOM = ------------------------------------------------------------------------- 1012
2 resolution ( bit ) DOR(S/s)
[Good96]
Slide CoA 7
IMSE-
Design Group
Introduction
22
(LV)
(CM)
Nyquist
Nyquist (LV)
20
18
16
14
12
10
8
6
4
102
103
104
105
106
107
108
109
Slide CoA 8
IMSE-
Design Group
Basic Concepts
Uniform Quantization
y
Quantizer Resolution:
Quantizer
Resolution:
B
B =
= log
log22 (( #
# Levels
Levels ))
Separation
SeparationBetween
BetweenInput
InputLevels:
Levels:
X
X FS
FS=
--------XXLSB
=
--------------LSB
BB
2 2 1
Separation
SeparationBetween
BetweenOutput
OutputLevels:
Levels:
Y
Y
=
-------------- = --------------B
22 B 11
X LSB
Quantizer
QuantizerGain
Gain:
X FS
Y 2B
Slide CoA 9
IMSE-
Design Group
Basic Concepts
y
+
eq
eq
3-bit Quantizer
eq
Overloading of Quantizer
eq
x
Overload
Overload
eq
X FS
Slide CoA 10
IMSE-
Design Group
Basic Concepts
+
eq
Showed that
Errors Caused by Quantization
Produces Similar Effect as an
Independent Noise Source
When Many Quantization Levels
are Used
2
3
1
3
8n 2 2
2
4
n = 1n
= f fB
[Engel99]
Design of Embeddable Data Converters: Sigma-Delta Converters
Slide CoA 11
= x rms
IMSE-
Design Group
Basic Concepts
eq
Unfiltered
eq
Low-Pass Filtered@1Hz
e qLP
e qLP
Unfiltered
Design of Embeddable Data Converters: Sigma-Delta Converters
Low-Pass Filtered@1Hz
Slide CoA 12
IMSE-
Design Group
Basic Concepts
eq
Unfiltered
eq
Low-Pass Filtered@1Hz
e qLP
e qLP
Unfiltered
Design of Embeddable Data Converters: Sigma-Delta Converters
Low-Pass Filtered@1Hz
Slide CoA 13
IMSE-
Design Group
Basic Concepts
Input Signal:
Non-overloading
y = x + eq( x )
Random
Band-Limited Characteristic Function
p ( eq )
eq
t
2
( eq ) =
2
2
( e q p ( e q ) ) de q = -----12
2
eq
+
x
2
2
( e q ) = -----12
Slide CoA 14
IMSE-
Design Group
Basic Concepts
YD
X LSB
5
X LSB
4
3
Offset
INL4
2
1
0
X FS
X FS
Nonlinearity Errors:
Correct Offset and Gain Errors
Separation of Adjacent Transitions (DNL)
Deviations from Ideal Transitions (INL)
Slide CoA 15
IMSE-
Design Group
Basic Concepts
INL4
yD
yD
DNLk
Offset
Nonlinearity Errors:
Slide CoA 16
IMSE-
Design Group
Basic Concepts
Non-ideal
Quantizer
e qNI
G ADC
y
G DAC
E osDAC
E osADC
ADC Errors
DAC Errors
Ideal Quantizer
eq
y = x + e q + e ADC + e DAC
e ADC
e DAC
+
YD
Design of Embeddable Data Converters: Sigma-Delta Converters
Slide CoA 17
IMSE-
Design Group
Basic Concepts
,f [ 0, BW n ]
f S 2f S 3f S 4f S BW n
n = 9
n = 1
n = 2
n = 3
S *q [ f ( n f S ) ]
n =
n = 9
Sq
2
1
S q = ------ -----f S 12
fS 2
fS 2
Slide CoA 18
IMSE-
Design Group
Basic Concepts
fS
( n 1 )T SnT S ( n + 1 )T S
fB
fS
eq
S S
-, ----the band ------2 2
f S f B = 16
f S f B = 256
Slide CoA 19
IMSE-
Design Group
Basic Concepts
X
e qLP
Power
Estimation
e qLP
fB
eq
2
e qLP
meas
e qLP
ideal
= 120
W
= 212
W
f S fB = 256
Power
Estimation
Low-Pass Filtered@1Hz
e qLP
f S f B = 16
eq
e qLP
2
e qLP
meas
e qLP
ideal
= 55
W
f S f B = 1024
meas
= 13
W
e qLP
e qLP
Slide CoA 20
ideal
= 55
W
= 3
W
IMSE-
Design Group
Basic Concepts
e qLP
eq
f S = 16Hz
e qLP
2
e qLP
f S = 256Hz
= 11mW
meas
e qLP
= 11mW
ideal
Power
Estimation
e qLP
Power
Estimation
eq
Low-Pass Filtered@1Hz
e qLP
2
e qLP
meas
e qLP
ideal
= 640
W
f S = 1024Hz
meas
= 650
W
e qLP
e qLP
Slide CoA 21
ideal
= 160
W
= 164
W
IMSE-
Design Group
Basic Concepts
X
Sampling
dB
clock @ f S
fB
2
fS B
X FS f S
3
3
SNR max = 10log 10 --- ---------- --------- = 10log 10 --- --------- ( 2 1 )
2 2f B
2 2 2f B
dB
SNR dB
1
fB
fS
-------2f B
SNRmax
Signal Power X
------
Pq =
2 2f B
S q ( f ) df =
------ -------12 f S
fB
fB
SNR 0dB
X mim
DR
Slide CoA 22
X
X FS
dBv
IMSE-
Design Group
Basic Concepts
fB = f S1 2
dB
2
X FS
f S1
er P =
w
o
p
----=
-----------------------------------i se
q
o
2
n
12
f
B
d
n
12 ( 2 1 ) S2
in-ba
e power
is
o
n
.
t
n
a
qu
fB
X FS
r
e
w
P q = ------ = ----------------------------se p o
2
i
12
o
N
n
d
n
a
12
(
2
1
)
b
in2
f S2 2
fB
dB
f S2
B
1
N ------------------------------------------- log 2 1 + --- log ------ 2
2
2 f S1
6.02
Design of Embeddable Data Converters: Sigma-Delta Converters
Slide CoA 23
Basic Concepts
Measure SNR
Ideally the range of x ends at
XFS
In practice its ends at lower
value of x, namely Overload
Input, XOL
Filter
-25
-50
-80dB in-band
error power
-75
-100
-125
Overloading
SNRmax
-150
SNRpeak
SNR(dB)
SNR 0dB
X min
DR
N
X OL
ideal
-------------------------------------------
6.02
actual
---------------------------------------------
6.02
X FS
Slide CoA 24
IMSE-
Design Group
Basic Concepts
Amplitude (dBV)
-25
SFDR
-50
2nd har.
-75
3rd har.
-100
X 8
SNDR dB = 20log10 ------------------------------------------------------
noise + harmonics
-125
-150 0
(IM)
and
IP3:
IP 3 P x IM 3 2
dB
1e+06
0
-20
Magnitude (dB)
Intermodulation Products
Intercept Points (IP):
2e+05
IM3 =-43 dB
-40
IM3 =-54 dB
-60
-80
-100
0.22
0.23
0.24
0.25
0.26
0.27
0.28
Frequency/Sampling Frequency
Slide CoA 25
IMSE-
Design Group
Noise Shaping
fS
M = -------2f B
2
B
SNR max = 10log 10 3--- M ( 2 1 )
2
fB
f S2 2
f S1 2
f B
f S1 2
fB
dB
f S2 2
fB
Design of Embeddable Data Converters: Sigma-Delta Converters
f S1
f S2
Slide CoA 26
IMSE-
Design Group
Noise Shaping
10
(2
1 ) + 10 log
Quantizer
0.5V Uniform Random Signal with 1-bit Quantizer
eq
eq
M = 8
2
e qLP
10
(M)
Oversamplig
Low-Pass Filtered@1Hz
eq
M = 128
= 11mW
2
e qLP
= 650
W
12dB
M = 512
e qLP = 164
W
18dB
Slide CoA 27
IMSE-
Design Group
Noise Shaping
Basic Idea
e qHP ( n ) = e q ( n ) e q ( n 1 )
e qHP ( n ) = e q ( n ) + e q ( n 2 ) 2e q ( n 1 )
e qHP ( n ) = e q ( n ) + 3e q ( n 2 ) 3e q ( n 1 ) e q ( n 3 )
eq
N TF ( z ) =
eq
L = 0
e qHP
Low-Pass Filtered@1Hz
eq
L = 1
1 L
(1 z )
L = 2
= 65
W
f S = 16Hz e qHP
meas
Slide CoA 28
IMSE-
Design Group
Noise Shaping
1 L
E qHP ( z ) = N TF ( z ) E q ( z ) = ( 1 z ) E q ( z )
z = e
f
j2
----fS
= e
1 f
j
----- ----M fB
max = ---M
Real ( z )
N TF ( )
2
2L
Pq =
S
( f ) df ------ ------------------------------------------ f B qHP
12 ( 2L + 1 )M 2L + 1
Slide CoA 29
IMSE-
Design Group
Noise Shaping
SNR
max
= 1.76 + 20 log
2L + 1
2 B 1 + 10 log ( M ) + 20 L log ( M ) + 20 log -------------------
10
10
10
10
L
Quantizer
Plain Oversamplig
Noise Shaping
Plain Oversampling
Oversampling
Noise-Shaping
plus
SNR max
Add 3.01dB
per octave of M
M = 128
M = 64
M = 32
Add 3.01+6.02dB
per octave of M for L = 1
Offset = - 5.17dB
M = 16
M = 8
Add 3.01+12.04dB
per octave of M for L = 2
Offset = - 12.89dB
M = 4
Add 3.01+18.06dB
per octave of M for L = 3
Offset = - 21.37dB
Design of Embeddable Data Converters: Sigma-Delta Converters
L
Slide CoA 30
IMSE-
Design Group
Noise Shaping
eq
H(z)
[Cuttler, 1954]
YD
fB
DAC
y
Y ( z) = Eq ( z ) + U ( z ) = X ( z ) + [ 1 H ( z ) ] E q ( z )
S TF ( z ) = 1
N TF ( z ) = 1 H ( z )
Slide CoA 31
IMSE-
Design Group
Noise Shaping
T( z)
e DAC
e ADC
+
u
YD
YD
Linear
Modeling
Quantizer
F( z)
DAC
y
of
G T( z)
1
G T( z ) F( z) E
Y ( z ) = ----------------------------------------------X
( z ) + ---------------------------------------------( E q ( z ) + E ADC ( z ) ) ---------------------------------------------(z)
1 + G T (z ) F(z )
1 + G T (z ) F (z )
1 + G T ( z ) F ( z ) DAC
Delta Modulators
Interpolative Modulators
Sigma-Delta Modulators
G T ( z ) 1 in baseband
F ( z ) to unity
Slide CoA 32
IMSE-
Design Group
Noise Shaping
fS
YD
T(z)
F(z)
DAC
1 N TF ( z )
F ( z ) = --------------------------S TF ( z )
BandPass M
LowPass M
fS 2
f B
fB
Low-Pass to Band-Pass
Transformation, Among
Others
fS 2
fS 2
2f B
Slide CoA 33
fS 2
IMSE-
Design Group
Architectures
Antialiasing filter
Modulator
Rough quantization
x sh
fS 2
Decimator
xa
H(z )
S/H
y
n
fN 2
Noise-shaping
Oversampling
D/A
yf( n )
fB fS 2
fS
yN ( n )
...
n
spurious
Digital processing
y(n)
...
X ( f ) antialiasing filter
Downsampling
Digital filter
x a ( t ), x sh ( t )
yN
yf
...
n
...
n
quantization error
with noise-shaping
Y(f)
Yf( f )
YN( f )
digital filter
X sh ( f )
...
fB fS 2
fS
f B fS 2
fS
fB fS 2
fS
Slide CoA 34
fB f N
fS
IMSE-
Design Group
Architectures
fS
DAC
g
y = ------ x
g
Exemplary Waveforms
Zero Input
Positive Input
0
Negative Input
+ + + +
+ + + +
+
+ + + +
+
Caution!!! : Periodicity is a non-desirable feature happening for some inputs and some architectures
Design of Embeddable Data Converters: Sigma-Delta Converters
Slide CoA 35
IMSE-
Design Group
Architectures
YD
fS
DAC
Slide CoA 36
IMSE-
Design Group
Architectures
eq
clock @
fS
u
DAC
y
Ggz
1z
Y ( z ) = X ( z ) ------------------------------------------ + E q ( z ) -----------------------------------------1
, 1
1
, 1
1 z + Gg z
1 z + Gg z
,
For Gg = 1 ,
g-z 1
S TF = ---,
g
YD
Pq = 2
min
2 2
1
---------------------------------------------------------- --------- d
j
j
24
1 + G T(e ) F(e )
N TF = 1 z
2
2 2
1
Ps +
---------------------------------------------------------- --------- d = -----
4
j
j
24
1 + G T ( e ) F ( e )
for P s = 0 .
Slide CoA 37
IMSE-
Design Group
Architectures
tio
ns
N TF ( z ) = ( 1 z 1 )
F ( z ) = 2z 1
clock @ f S
Af
te
rs
om
e
S TF ( z ) = z 2
ma
G T (z )
1
Y ( z ) = ----------------------------------------------X
( z ) + ----------------------------------------------E
(z)
1 + G T(z) F(z)
1 + G T( z ) F( z) q
nip
ula
z
G T ( z ) = -----------------------1 2
(1 z )
YD
y
DAC
G = 1
Slide CoA 38
IMSE-
Design Group
Architectures
Stability Problems
u(n)
u(n)
sine-input@X = X FS 2
sine-input@X = 5X FS 4
Low-frequency,
large-amplitude oscillations are observed
sine-input@X = X FS
Design of Embeddable Data Converters: Sigma-Delta Converters
sine-input@X = 2X FS
Slide CoA 39
IMSE-
Design Group
Architectures
g2
g1
x
g 1'
g 2'
YD
y
DAC
usually g 1 = g 1'
There are two degrees of freedom to control the output swing of the integrators
Overloading typically occurs at
X OL 0.8 X FS
Slide CoA 40
IMSE-
Design Group
Architectures
clock @
fS
P+ P y = ------------------E = x
P + + P- r
y
DAC
y ( n ) = E r sgn ( u ( n ) )
e q ( 1 ) = 0
x=0
x = 1/3
x = 1/2
u(n)
y(n)
eq(n)
u(n)
y(n)
eq(n)
u(n)
y(n)
eq(n)
1/3
2/3
1/2
1/2
-1
-1
-1/3
-1
-2/3
-1/2
-1
-1/2
1/3
2/3
1/2
1/2
3
4
Slide CoA 41
IMSE-
Design Group
If the input x
In-band quantization error
power (dB)
is an irrational number such that x 1 , the output is not periodic. However its
spectrum is discrete.
Some Properties of Pattern Noise
-40
-45
-50
-55
-60
-65
2 2
------ ----------The average noise is given by P Q =
-70
12 3M 3
Slide CoA 42
IMSE-
Design Group
Architectures
2nd Order M
In-band quantization noise power
(dB)
1st Order M
-40
-45
-50
-55
-60
-65
-70
-30
-40
-50
-60
-70
-80
-90
Slide CoA 43
IMSE-
Design Group
Architectures
g1
g2
g 1'
g 2'
+E r
D/A
Er
YD
SNRpeak
Overloading
X OL 0.8 X FS
for 2nd-Order SDM
SNR0dB
X min
DR
X OL X FS
Weight
[Bose,88]
[Yin,94]
[Marques, 97]
[Medeiro, 98]
g 1 , g 1'
0.5
0.25
1/3
0.25
g2
0.5
0.5
0.6
g 2'
0.5
0.25
0.4
0.5
Total int. OS / Er
3.5
2.4
# unitary caps.
10
12
Slide CoA 44
IMSE-
Design Group
High-Order Single-Loop Ms
Architectures
S TF ( z ) = z L
N TF ( z ) = ( 1
T(z )
2
2L
P q ------ ------------------------------------------12 ( 2L + 1 )M 2L + 1
YD
F( z)
N TF = 2
DAC
Tradeoff Stability-SNR
Analytical techniques are not available
for evaluation of the optimum noise
transfer function.
clock @
L
z1 )
fS
1 z1 )
N TF ( z ) = (-----------------------D( z)
Maximum Input
Rule of Thumb
j
N TF ( e ) < 1.5 or 2
Relationship Between
N TF
Slide CoA 45
1 S q ( ) -----------------2
D( )
IMSE-
Design Group
Architectures
z
---------------1
1z
g1
z
---------------1
1z
g2
g 1'
fS
z
---------------1
1z
g3
YD
g L'
g 2'
DAC
1 S TF ( z ) = ----------D( z)
( 1 z 1 ) N TF ( z ) = -----------------------D(z )
Butterworth approximation
Assuming g k = g k' , and 1-bit quantizer, g k' G = 2 for good fitting to the simulated psd
Increasing
k = 1, L
k = 1, L
Difficult to stabilize
Significant SNR lost
Large power consumption
FOM@Power
8KHz
100KHz
12
14
5@0.34mW(2V)
195@160mW(5V)
Third-Order
Third-Order
Slide CoA 46
[Au, 97]
[OptEynde, 91]
IMSE-
Design Group
Architectures
g1
clock @
g2
g 1'
fS
g1
g 2'
YD
g1
g 1'
g 1'
DAC
L2
( 1 z1 )
[ 1 ( 2 )z 1 + z 2 ]
N TF ( z ) = -------------------------------------------------------------------------------------------D(z)
Zeros at DC
Other zeros on the unity circle
Also feedforward inputs to integrators
clock @
g2
g2
g2
g2
fS
YD
DAC
Slide CoA 47
IMSE-
Design Group
Architectures
..
.
B2
B1
AL
..
.
eq
A2
+
YD
A1
A0
D/A
N TF
( z 1 )L Bi ( z 1 )L i
Ai ( z 1 )N i
i=1
i=0
= --------------------------------------------------------------------------------------------------------------------------- S TF = --------------------------------------------------------------------------------------------------------------------------L
L
L
L
z ( z 1 )L Bi ( z 1 )L i + Ai ( z 1 )L i
z ( z 1 ) L B i ( z 1 )L i + Ai ( z 1 )L i
i=1
i=0
i=1
Slide CoA 48
i=0
IMSE-
Design Group
Architectures
Pros of MultiBit
X FS 1
---------- ----12 f S
1-bit
2
X FS
Pq =
------ = --------12
12
B-bits
X FS
P q = ------ = ----------------------------2
12
B
12 ( 2 1 )
X FS
1
-------------------------- ----B
12 ( 2 1 ) fS
Counters of MultiBit
DAC errors appear at the modulator input
Produces extra noise and distortion
e ADC
eq
x
G (z )
u
G
YD
e DAC
2
2 2
e DAC 1--- ( INL
)
LSB
2
Design of Embeddable Data Converters: Sigma-Delta Converters
Slide CoA 49
IMSE-
Design Group
Architectures
2 Unity Elements
Parallel Bank of
Comparators
Unit Elements
Thermometer
Unit Elements
type
decoder
Digital
Input
Unit Elements
Unit Elements
ADC
DAC
Unit Elements
Equal-valued resistors, capacitors or transistor current sources
Mismatching of elements determines the DAC nonlinearity
U
y
1
------- = -------------- ----------e-
y
U
B
2 2 e
U
----------e- unit element error
Ue
Others
Slide CoA 50
IMSE-
Design Group
Architectures
2 Unity Elements
Unit Elements
Thermometer
type
Shuffler
Unit Elements
decoder
Digital
Input
Unit Elements
Unit Elements
For each input code, select different unit elements at different times
Random Selection:
Convert the mismatch error into unshaped white noise
Conceptually simple implementation
Algorithmic Selection:
Reduce the in-band power of mismatch error
Different Algorithms: CLA, ILA, DWA, . . .
Noise-Shaping DEM:
Second-order LowPass or BandPass shaping of mismatch error
Design of Embeddable Data Converters: Sigma-Delta Converters
Slide CoA 51
IMSE-
Design Group
Architectures
B1
Slide CoA 52
IMSE-
Design Group
Architectures
output
Multibit
Digital
Correction
(RAM)
M
M
Decimation
Filter
N bit
DAC
RAM
Data in
G(z)
Address
Up/Down
Counter
N-bit counter
Design of Embeddable Data Converters: Sigma-Delta Converters
Slide CoA 53
IMSE-
Design Group
Architectures
G1 T1 ( z )
Y ( z ) = --------------------------------------------- X( z) +
1 + G 1 T 1 ( z )F 1 ( z )
fS
Y 1D
T1( z )
y1
F1( z )
H ( z )G 2 T 2 ( z )
---------------------------------------------E (z ) +
1 + G 2 T 2 ( z )F 2 ( z ) q1
H(z )
---------------------------------------------- E q2 ( z )
1 + G 2 T 2 ( z )F 2 ( z )
DAC
clock @
G 1 T1 ( z )
S TF ( z ) = --------------------------------------------1 + G 1 T 1 ( z )F 1 ( z )
fS
Y 2D
T2( z )
H ( z )G 2 T 2 ( z )
1
--------------------------------------------- = --------------------------------------------1 + G 1 T 1 ( z )F 1 ( z )
1 + G 2 T 2 ( z )F 2 ( z )
Then:
e q1
YD
1
+ ---------------------------------------------E (z ) +
1 + G 1 T 1 ( z )F 1 ( z ) q1
H(z)
N TF ( z ) = ---------------------------------------------1 + G 2 T 2 ( z )F 2 ( z )
For First Order,
H(z ) = 1 z
F2 ( z )
y2
DAC
N TF ( z ) = ( 1 z 1 )
by adding a delay to
Slide CoA 54
Y 1D
IMSE-
Design Group
Architectures
Dual Quantization
clock @
G(z)
fS
YD
1-bit
ADC
1b
H1( z )
YD
1-bit
DAC
clock @
fS
YD
B-bit
ADC
YD
YD
Bb
1b
Bb
H2( z )
= S TF X + N TF E q1
Y D = S TF ( H 1 + H 2 ) x + ( H 1 N TF + H 2 N TF H 2 ) E q1 + H 2 E qB
= U + E qB = Y D
E q1 + E qB
1b
H 1 = 1 N TF
[Leslie, 92]
H2 = N TF
Y D = S TF x + N TF E qB
Design of Embeddable Data Converters: Sigma-Delta Converters
Slide CoA 55
IMSE-
Design Group
Architectures
L
x
Modulator Topology:
q1
1-bit
1D
Eq2
2
L
1-bit
2D
X3
. .
.
E qN
B-bit Y N D
Q
LN
CANCELLATION LOGIC
Cascade Topologies
Li
YD
-------------B
2 1
L = L 1 + L2 + + LN
( L LN )
1 L
1
X ( z ) + d ( 1 z ) E qN ( z ) d ( 1 z )
ED ( z )
Signal
Lth-order shaping in
last-stage quantization error
Design of Embeddable Data Converters: Sigma-Delta Converters
(L-LN)th-order shaping in
last-stage DAC error
Slide CoA 56
IMSE-
Design Group
Architectures
g1
g '
1
Eq1
g2
g '
2
Y1D
H1(z)
Analog
Digital/Analog
Digital
g 1' = g 1
g3'
d 0 = 1 -------------------g1 g2 g3
H1 ( z ) = z 1
g 2' = 2g 1'g 2
g 3''
d 1 = -------------------g1 g2 g3
H2 ( z ) = ( 1 z 1 ) 2
g 4' = g 3''g 4
d2 = 0
H 3 ( z ) = z 1
g 4''
d 3 = --------------------------g1 g2 g3 g4
H4 ( z ) = ( 1 z 1 ) 3
DAC
Eq2
g3
g '
3
d0
Y2D
g 3''
d1
H2(z)
DAC
g '
4
g ''
4
H3(z)
Eq3
g4
ED
d2
B-bit
ADC
B-bit
DAC
Y3D
B
d3
H4(z)
+
YD
Cancellation Logic
Y ( z ) = z 4 X ( z ) + d 3 ( 1 z 1 ) 4 E q3 ( z ) d 3 ( 1 z 1 ) 3 E D ( z )
1 2
q = ------ -----------------
12 B
2 1
2
2
1 2 INL 2
D = --- ----------
2
100
8
6
2 2
2
P q = d 3 q ----------- + D -----------
9
7
9M
7M
Slide CoA 57
IMSE-
Design Group
Architectures
x
g1
g '
1
g2
g '
2
Cancellation Constraints
Y1
H1(z)
Analog
Digital/Analog
Digital
g 1' = g 1
g 3'
d 0 = 1 -------------------g1 g2 g3
H1 ( z ) = z 1
g 2' = 2g 1'g 2
g 3''
d 1 = -------------------g1 g2 g 3
H2 ( z ) = ( 1 z 1 ) 2
g 4' = g 3''g 4
d2 = 0
H 3 ( z ) = z 1
g 5' = g 4''g 5
g 4''
d 3 = --------------------------g1 g2 g 3 g 4
H4 ( z ) = ( 1 z 1 ) 3
d4 = 0
H5 ( z ) = z 1
DAC
g3
g '
3
g ''
Eq2
d0
Y2
d1
+ H2(z) +
DAC
g4
g '
4
g ''
Eq3
d2
Y3
DAC
g5
g '
5
g ''
5
+ H4(z)
g 5''
d 5 = ---------------------------------- H6 ( z ) = ( 1 z 1 ) 4
g 1 g2 g3 g 4 g 5
H5(z)
Eq4
B-bit
ADC
ED
B-bit
DAC
d3
H3(z)
d4
Y4
B
d5 + H6(z)
Cancellation Logic
Y ( z ) = z 5 X ( z ) + d 5 ( 1 z 1 ) 5 E q4 ( z ) d 5 ( 1 z 1 ) 4 E D ( z )
2
1 2
q = ------ -----------------
12 B
2 1
2
2
1 2
D = --- INL
--------2 100
8
8
2 2
2
P q = d 5 q ----------------- + D -----------
9
11M 11
9M
Slide CoA 58
IMSE-
Design Group
Architectures
2
2 1 mb
8
6
2 2
2
= d 3 q ----------- + D -----------
9
7
9M
7M
2 2
2
= d 5 q ---------------- + D ----------
11
9
10
2 1 mb
11M
9M
2-13mb M
g1
0.25
g3
g4
d0
-1
g1 0.25 g3
g1 0.25 g3
g1
0.25
g3
0.5
g4
d1
g2
g2
g3
0.5
g4
d2
g2 0.5
g2
0.5
d3
g3
g4
g5
d0
-1
g4 0.5 g5
d1
g4 0.5 g5
d2
d3
d4
d5
Slide CoA 59
0.5
IMSE-
Design Group
Architectures
V in +
16
5b
DAC
5b
DAC
5b
DAC
3b
3b
DAC DAC
3b
3b
DAC DAC
5b
DAC
4b
DAC
SHUFFLE
M out
8 LSBs
LSB
Differentiator
C out
[Brook, 97]
Design of Embeddable Data Converters: Sigma-Delta Converters
Slide CoA 60
IMSE-
Design Group
Architectures
High-Order, Single-Loop
BW
Power
Type
[Kasha 98]
400Hz
20bit
16mW (5V)
[Rito, 91]
24kHz
16
200mW (5V)
[Welland, 89]
24kHz
16
450mW (10V)
[Adams, 91]
24kHz
18
1100 mW (10V)
[Sau, 95]
24kHz
15
1mW (2.6V)
FB + FF input + 1 resonator
Slide CoA 61
IMSE-
Design Group
Architectures
Multi-Bit
BW
N, bits
Power
[Nys, 96]
400Hz
19
2.7mW (5V)
DWA
2nd / 3b
[Yasuda, 98]
100kHz
12.8
14.8 mW (2.7V)
TNSDEM
3rd / 3b
[Sarhang, 93]
20.5kHz
16
Digital Correction
2nd / 4b
[Baird, 96]
500kHz
14
58 mW (5V)
Digital Calibration
4th / 4b
2nd / 3b
Order /
Number of bit
[Fattaruso, 93]
24kHz
15
100mW (5V)
Self-calibration &
random dynamic
matching
[Chen, 95]
20kHz
16
67.5mW (5V)
ILA
2nd / 3b
[Fongleman, 00]
24kHz
16
66mW (5V)
2nd / 33 levels
[Vink, 98]
5MHz
1MHz
6.8
10
60mW (5V)
No Calibration
2nd / 5b
[Fuji, 00]
2.5MHz
15
270mW (5V)
Bi-directional DWA
2-1-1 / 4b
[Geert, 00]
2.5MHz
16
295mW (5V)
Speed-optimized
DWA
3rd / 4b
Slide CoA 62
IMSE-
Design Group
Architectures
Cascade
N, bits
DOR (MS/s)
Process / Supply
Power (mW)
Topology
[Geerts00]
15.8
2.5
0.65m DP (5V)
295
3rd-order single-loop M
[Fujimori00]
15.0
2.5
0.5m DP (5V)
105
[Paul99]
12.0
18.0
1.2m DP (5V)
324
pipeline / oversampling
[Paul99]
10.7
30.0
0.6m DP (5V)
230
pipeline / oversampling
[Geerts99]
15.0
2.2
0.5m DP (3.3V)
200
2-1-1 cascade M
[Medeiro99]
13.0
2.2
55
[Feldman98]
13.0
1.4
0.7m DP (3.3V)
81
2-2-2 cascade M
[Marques98]
14.8
2.0
1m DP (5V)
230
2-1-1 cascade M
[Brooks97]
14.5
2.5
0.6m DP (5V)
550*
pipeline / oversampling
Slide CoA 63
IMSE-
Design Group