EE247
Lecture 22
Pipelined ADCs (continued)
Effect subADC, gain stage, subDAC nonidealities on overall ADC
performance (continued)
Correction for interstage gain nonlinearity
Implementation
Combining the bits
Practical circuits
Stage scaling
Stage implementation
Circuits
Noise budgeting
How many bits per stage?
Algorithmic ADCs utilizing pipeline structure
Advanced background calibration techniques
Time Interleaved Converters
ADC figures of merit
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 2
Pipeline ADC
Block Diagram
Idea: Cascade several low resolution stages to obtain high overall resolution
(e.g. 10bit ADC can be built with series of 10 ADCs each 1bit only!)
Each stage performs coarse A/D conversion and computes its quantization
error, or "residue
Stage 1
B
1
Bits
Digital Output (B
1
+B
2
+..B
k
) Bits
V
in
MSB....... ...LSB
V
res1 V
res2 Stage 2
B
2
Bits
Stage k
B
k
Bits
+

DAC ADC
Align and Combine Data
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 3
Summary So Far
Pipelined A/D Converters
V
in
ADC
2
B1eff 2
B2
2
B3
B
2
bits B
1
bits
Cascade of low resolution stages
By adding interstage gain= 2
Beff
No need to scale down Vref for stages down the pipe
Reduced accuracy requirement for stages coming after stage 1
Addition of Track & Hold function to interstagegain
Stages can operate concurrently
Throughput increased to as high as one sample per clock cycle
Latency function of number of stages & conversionperstage
Correction for circuit nonidealities
Builtin redundancy compensate for subADC inaccuracies such as
comparator offset (interstage gain: G=2
Bneff
, B
neff
< B
n
)
Error associated with gain stage and subDAC calibrated out
B
3
bits
2
B2eff
2
B3eff
V
ref
V
ref
V
ref
V
ref
T/H+Gain
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 4
Pipelined ADC
Error Correction/Calibration Summary
V
IN1
V
RES1
DAC
D
1

a
3
V
3
+
2
2
ADC
+
V
OS
+
+
c
ADC
+
c
DAC
c
gain
Error Correction/Calibration
c
ADC
, V
os
Redundancy either same stage or next stage
c
gain
Digital adjustment
c
DAC
Either sufficient component matching or digital
calibration
Interstage amplifier nonlinearity ?
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 5
Interstage Gain Nonlinearity
Ref: B. Murmann and B. E. Boser, "A 12b, 75MS/s Pipelined ADC using OpenLoop Residue
Amplification," ISSCC Dig. Techn. Papers, pp. 328329, 2003
Invert gain stage nonlinear polynomial
Express error as function of V
RES1
Push error compensator into digital domain through backend ADC
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 6
Interstage Gain Nonlinearity
Ref: B. Murmann and B. E. Boser, "A 12b, 75MS/s Pipelined ADC using OpenLoop Residue
Amplification," ISSCC Dig. Techn. Papers, pp. 328329, 2003
V
RES1
a
3
V
X
3
+ Backend
D
B
D
B,corr
(...)
+

... D 12p D 3p D p ) p , (D
7
B
3
2
5
B
2
2
3
B 2 2 B
+ + = c
V
X
c
(D
B
, p
2
)
3 3
3
2
c
gain
)
(2
a
p
+
=
2
3
c
gain
Premeasured & stored in table lookup form
p
2
continuously estimated & updated (account for temp. & other variations)
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 7
Interstage Gain Nonlinearity Compensation
Proof of Concept Evaluation Prototype
Ref: B. Murmann and B. E. Boser, "A 12b, 75MS/s Pipelined ADC using OpenLoop Residue
Amplification," ISSCC Dig. Techn. Papers, pp. 328329, 2003
Reused 14bit ADC in 0.35m from Analog Devices [Kelly, ISSCC 2001]
Modified only 1
st
stage with 3b
eff
openloop amplifier built with simple diffpair +
resistive load instead of the conventional feedback around highgain amp
Conventional 9b
eff
backend, 2bit redundancy in 1
st
stage
Realtime postprocessor offchip (FPGA)
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 8
Measurement Results
12bit ADC w Extra 2bits for Calibration
0 1000 2000 3000 4000
1
0.5
0
0.5
1
(b) with calibration
Code
0 1000 2000 3000 4000
10
0
10
(a) without calibration
Code
I
N
L
[
L
S
B
]
RNG=0
RNG=1
0 1000 2000 3000 4000
10
0
10
(b) with calibration
Code
I
N
L
[
L
S
B
]
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 9
Combining the Bits
Example: Three 2bit stages, no redundancy
Stage 2
V
in Stage 3 Stage 1
D
out
+ +
2 2 2
6
B
1
=2
B
1eff
=2
B
2
=2
B
2eff
=2
B
3
=2
1/2
2
1/2
2
D
1
D
2
D
3
3 2 1
3
2 1
2
1
1
16
1
4
1
2 2
1
2
1
D D D D
D D D D
out
eff B eff B eff B
out
+ + =
+ + =
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 10
Combining the Bits
Only bit shifts
No arithmetic
circuits needed
D
1
XX
D
2
XX
D
3
XX

D
out
DDDDDD
Stage 2
V
in Stage 3 Stage 1
D
out[5:0]
B
1
=2
B
1eff
=2
B
2
=2
B
2eff
=2
B
3
=2
MSB
LSB
D
1
D
2
D
3
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 11
Combining the Bits
Including Redundancy
Example: Three 2bit stages, incorporating 1 bit
redundancy in stages 1 and 2
Stage 2
V
in Stage 3 Stage 1
B
1
=3
B
1eff
=2
B
2
=3
B
2eff
=2
B
3
=2
8 Wires
6 Wires
???
D
out[5:0]
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 12
Combining the Bits
Bits overlap
Need adders
D
1
XXX
D
2
XXX
D
3
XX

D
out
DDDDDD
Stage 2
V
in Stage 3 Stage 1
D
out[5:0]
B
1
=3
B
1eff
=2
B
2
=3
B
2eff
=2
B
3
=2
HADD HADD FADD HADD HADD
D
1
D
2
D
3
3 2 1
3
2 1
2
1
1
16
1
4
1
2 2
1
2
1
D D D D
D D D D
out
eff B eff B eff B
out
+ + =
+ + =
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 13
Combining the Bits
Example
D
1
001
D
2
111
D
3
10

D
out
011000
Stage 2
V
in Stage 3 Stage 1
D
out[5:0]
B
1
=3
B
1eff
=2
B
2
=3
B
2eff
=2
B
3
=2
HADD HADD FADD HADD HADD
D
1
D
2
D
3
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 14
Pipelined ADC
Stage Implementation
Each stage needs T/H hold function
Track phase: Acquire input/residue from previous stage
Hold phase: subADC decision, compute residue
Stage 1 Stage 2 V
in
Stage n
acquire
convert
convert
acquire
...
...

1
...
CLK
+

DAC ADC
G
V
res
T/H

2

1

1

2
V
in
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 15
Stage Implementation
Usually no dedicated T/H amplifier in each stage
(Except first stage in some cases why?)
T/H implicitely contained in stage building blocks
V
in
+

DAC ADC
G
V
res
T/H T/H
T/H
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 16
Stage Implementation
DACsubtractgain function can be lumped into
a single switched capacitor circuit
"MDAC"
V
in
+

DAC ADC
G
V
res
T/H
T/H
MDAC
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 17
1.5Bit Stage Implementation Example
Ref: A. Abo, "Design for Reliability of Low voltage, Switchedcapacitor Circuits," UCB PhD Thesis,
1999
D1,D0
V
DAC
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 18
1.5Bit Stage Implementation
Acquisition Cycle
Ref: A. Abo, "Design for Reliability of Low voltage, Switchedcapacitor Circuits," UCB PhD Thesis,
1999
D1,D0
V
DAC
Vc
f
=Vc
s
=V
i
Q
Cs
=C
s
xV
i
Q
Cf
=C
f
xV
i
u
1
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 19
1.5Bit Stage Implementation
Conversion Cycle
Ref: A. Abo, "Design for Reliability of Low voltage, Switchedcapacitor Circuits," UCB PhD Thesis,
1999
D1,D0
V
DAC
u
2
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 20
1.5Bit Stage Implementation
Conversion Cycle
D1,D0 V
DAC
V
i
>V
R/
4 1 1 V
R
V
R
/4<V
i
<V
R
/4 0 1 0
V
i
<V
R
/4 0 0 +V
R
V
DAC
V
O

+
2 1 1
1
2
DAC s o f
Cf DAC s
Total
Cf Cf Cs DAC s
o f i f i s DAC s
s s
o i DAC
f f
o i DAC
V C V C
Q V C
Q Q Q V C
V C V C V C V C
C C
V V V
C C
s f
V V V
C C
  
=
=
= +
= +
= +
=
 


\ .
=
C
s
C
f
V
DAC
induces equal currents in C
s
and C
f
:
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 21
1.5 Bit Stage Implementation Example
Ref: A. Abo, "Design for Reliability of Low voltage, Switchedcapacitor Circuits," UCB PhD Thesis,
1999
Note:
Interstage gain set by C ratios
Accuracy better than 0.1%
Up to 10bit level no need for
gain calibration
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 22
1.5Bit Stage Implementation
Timing of Stages
V
DAC
V
DAC
Conversion Acquisition
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 23
Pipelined ADC Stage Power Dissipation & Noise
Typically pipeline ADC noise dominated by inter stage gain blocks
SubADC comparator noise translates into comparator threshold
uncertainty and is compensated for by redundancy
V
in Stage 1 Stage 3 Stage 2
V
n2
V
n3
V
n1
G1 G2 G3
V
in
2 2
i n 2 n2 n3
noi se n1
2 2 2
V V
V V . . .
G1 G1 G2
= + + +
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 24
Pipelined ADC Stage Scaling
Example: Pipeline using 1bit
eff
stages
Total input referred noise power:
C
1
/2
C
1
Gm
C
2
/2
C
2
Gm
C
3
/2
C
3
Gm V
in
V
n2
V
n3
V
n1
G1=2 G2=2 G3=2
V
in
t ot
2 2 2
1 2 3
t ot
1 2 3
1 1 1
N kT . . .
C G1 C G1 G2 C
1 1 1
N kT . . .
C 4C 16C
(
+ + +
(
(
(
+ + +
(
(
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 25
C
1
/2
C
1
Gm
C
2
/2
C
2
Gm
C
3
/2
C
3
Gm V
in
If all caps made the same size, backend stages contribute very
little noise
Wasteful powerwise, because:
Power ~ Gm
Speed ~ Gm/C
Fixed speed Gm/C filxed Power ~ C
(
+ + + ...
16
1
4
1 1
3 2 1
C C C
kT N
tot
Pipelined ADC Stage Scaling
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 26
How about scaling caps down by G
2
=2
2
=4x per stage?
Same amount of noise from every stage
All stages contribute significant noise
To keep overall noise the same noise/stage must be
reduced
Power ~ Gm ~ C goes up!
C
1
/2
C
1
Gm
C
2
/2
C
2
Gm
C
3
/2
C
3
Gm V
in
(
+ + + ...
16
1
4
1 1
3 2 1
C C C
kT N
tot
Pipelined ADC Stage Scaling
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 27
Stage Scaling
Example: 2bit
eff
/stage
Optimum capacitior scaling lies approximately midway between
these two extremes
Ref: D. W. Cline, P.R. Gray "A power optimized 13b 5MSamples/s pipelined analogtodigital
converter in 1.2um CMOS," JSSC 3/1996
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 28
Pipeline ADC
Stage Scaling
Power minimum is "shallow
Near optimum solution in practice: Scale capacitors
by stage gain
E.g. for effective stage resolution of 1bit (Gain=2):
C/2
C
Gm
C/4
C/2
Gm
C/8
C/4
Gm V
in
1 1 1
...
2 4
tot
N kT
C C C
(
(
+ + +
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 29
Stage Scaling Example
Ref: D. W. Cline, P.R Gray "A power optimized 13b 5 MSamples/s pipelined analogtodigital
converter in 1.2um CMOS," JSSC 3/1996
Note:
Resolution
per stage:
2bits
G=4
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 30
How Many Bits Per Stage?
Many possible architectures
E.g. B
1eff
=3, B
2eff
=1, ...
vs. B
1eff
=1, B
2eff
=1, B
3eff
=1, ...
Complex optimization problem, fortunately optimum tends to be
shallow...
Qualitative answer:
Maximum speed for given technology
Use small resolutionperstage (large feedback factor)
Maximum power efficiency for fixed, "low" speed
Try higher resolution stages
Can help alleviate matching & noise requirements in stages
following the 1
st
stage
Ref: Singer VLSI 96, Yang, JSSC 12/01 (14bit ADC w/o calibration)
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 31
14 & 12Bit StateoftheArt Implementations
Reference
Yang
(JSSC 12/2001)
0.35/3V
Loloee
(ESSIRC 2002)
0.18/3V
Bits
14 12
Architecture
3111111113 11111111112
SNR/SFDR
~73dB/88dB ~66dB/75dB
Speed
75MS/s 80MS/s
Power
340mW 260mW
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 32
10 & 8Bit StateoftheArt Implementations
Reference
Yoshioko et al
(ISSCC 2005)
0.18/1.8V
Kim et al
(ISSCC 2005)
0.18/1.8V
Bits
10 8
Architecture
1.5bit/stage 2.8 2.8  4
SNR/SFDR
~55dB/66dB ~48dB/56dB
Speed
125MS/s 200MS/s
Power
40mW 30mW
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 33
Algorithmic ADC
Essentially same as pipeline, but a single stage is reused for all partial conversions
For overall B
overall
bits need B
overall
/B
stage
clock cycles per conversion
Small area, slow
Trades conversion time for area
T/H
subADC
(1.6 Bit)
Digital Output
V
IN
Residue
DAC
Shift Register
& Correction Logic
start of conversion
2
B
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 34
Least Mean Square Adaptive Digital Background
Calibration of Pipelined AnalogtoDigital Converters
Ref: Y. Chiu, et al, Least Mean Square Adaptive Digital Background Calibration of Pipelined
AnalogtoDigital Converters, IEEE TRANS. CAS, VOL. 51, NO. 1, JANUARY 2004
Slow, but accurate ADC operates in parallel with pipelined (main) ADC
Slow ADC samples input signal at a lower sampling rate (f
s
/n)
Difference between corresponding samples for two ADCs (e) used to correct
fast ADC digital output via an adaptive digital filter (ADF) based on
minimizing the LeastMeanSquared error
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 35
Example: "A 12bit 20MS/s pipelined analogtodigital
converter with nested digital background calibration"
Ref: X. Wang, P. J. Hurst, S. H. Lewis, " A 12bit 20Msample/s pipelined analogtodigital converter
with nested digital background calibration, IEEE JSSC, vol. 39, pp. 1799  1808, Nov. 2004
Pipelined ADC operates at 20Ms/s @ has 1.5bit/stage
Slow ADC Algorithmic type operating at 20Ms/32=625ks/s
Digital correction accounts for bit redundancy
Digital error estimator minimizes the meansquarederror
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 36
Algorithmic ADC Used for Calibration of
Pipelined ADC (continued from previous page)
Ref: X. Wang, P. J. Hurst, S. H. Lewis, " A 12bit 20MS/s pipelined analogtodigital converter
with nested digital background calibration, IEEE JSSC, vol. 39, pp. 1799  1808, Nov. 2004
Uses replica of pipelined ADC stage
Requires extra SHA in front to hold residue
Undergoes a calibration cycle periodically prior to being used to calibrate
pipelined ADC
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 37
12bit 20MS/s Pipelined ADC with Digital
Background Calibration
Ref: X. Wang, P. J. Hurst, S. H.
Lewis, " A 12bit 20MS/s
pipelined analogtodigital
converter with nested digital
background calibration,
IEEE JSSC, vol. 39, pp.
1799  1808, Nov. 2004
Sampling capacitors scaled (1B
eff
/stage):
Input SHA: 6pF
Pipelined ADC: 2pF,0.9,0.4,0.2, 0.1,0.1
Algorithmic ADC: 0.2pF
Chip area: 13.2mm
2
Does not include digital
calibration circuitry estimated
~1.7mm
2
Area of Algorithmic ADC <20%
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 38
Ref: X. Wang, P. J. Hurst, S. H. Lewis, " A 12bit 20MS/s pipelined analogtodigital converter
with nested digital background calibration, IEEE JSSC, vol. 39, pp. 1799  1808, Nov. 2004
Without
Calibration
INL<4.2LSB
With
Calibration
INL<0.5LSB
Measurement Results
12bit 20MS/s Pipelined ADC with Digital Background Calibration
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 39
Ref: X. Wang, P. J. Hurst, S. H. Lewis, " A 12bit 20MS/s pipelined analogtodigital converter
with nested digital background calibration, IEEE JSSC, vol. 39, pp. 1799  1808, Nov. 2004
Measurement Results
12bit 20MS/s Pipelined ADC with Digital Background Calibration
Nyquist
rate
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 40
Ref: X. Wang, P. J. Hurst, S. H. Lewis, " A 12bit 20MS/s pipelined analogtodigital converter
with nested digital background calibration, IEEE JSSC, vol. 39, pp. 1799  1808, Nov. 2004
Measurement Results
12bit 20MS/s Pipelined ADC with Digital Background Calibration
Does not include
digital calibration
circuitry estimated
~1.7mm
2
Alg. ADC SNDR
dominated by noise
EECS 247 Lecture 22: Data Converters Nyquist Rate ADCs 2010 Page 41
ADC Architectures
Slope type converters
Successive approximation
Flash
Interpolating & Folding
Residue type ADCs
Twostep Flash
Pipelined ADCs
Timeinterleaved / parallel converter
Oversampled ADCs
EECS 247 Lecture 22 Data Converters Nyquist Rate ADCs
2010 Page 42
Time Interleaved Converters
Example:
4 ADCs operating in parallel at
sampling frequency f
s
Each ADC converts on one of
the 4 possible clock phases
Overall sampling frequency= 4f
s
Note T/H has to operate at 4f
s
!
Extremely fast:
Typically, limited by speed of T/H
Accuracy limited by mismatch
among individual ADCs (timing,
offset, gain, )
T/H
4f
s
ADC
f
s
ADC
ADC
ADC
O
u
t
p
u
t
C
o
m
b
i
n
e
r
V
IN
D
i
g
i
t
a
l
O
u
t
p
u
t
f
s
+T
s
/4
f
s
+2T
s
/4
f
s
+3T
s
/4
EECS 247 Lecture 22 Data Converters Nyquist Rate ADCs 2010 Page 43
Time Interleaved Converters
Timing
Note: Effective sampling rate4xf
s
1/4T
s
T
s
=1/f
s
1/4T
s
1/4T
s
1/4T
s
Input
signal
sampled
EECS 247 Lecture 22 Data Converters Nyquist Rate ADCs
2010 Page 44
ADC Figures of Merit
Objective: Establish measure/s to compare
performance of various ADCs
Can use FOM to combine several
performance metrics to get one single
number
What are reasonable FOM for ADCs?
EECS 247 Lecture 22 Data Converters Nyquist Rate ADC 2010 Page 45
ADC Figures of Merit
This FOM suggests that adding an extra bit
to an ADC is just as hard as doubling its
bandwidth
Is this a good assumption?
ENOB
s
f FOM 2
1
=
Ref: R. H. Walden, "Analogtodigital converter survey and analysis," IEEE Journal
on Selected Areas in Communications, April 1999
EECS 247 Lecture 22 Data Converters Nyquist Rate ADCs 2010 Page 46
Survey Data
1bit/Octave
Ref: R. H. Walden, "Analogtodigital converter survey and analysis," IEEE Journal on Selected
Areas in Communications, April 1999
EECS 247 Lecture 22 Data Converters Nyquist Rate ADCs 2010 Page 47
ADC Figures of Merit
Sometimes inverse of this metric is used
In typical circuits power ~ speed, FOM
2
captures this tradeoff correctly
How about power vs. ENOB?
One more bit 2x in power?
Ref: R. H. Walden, "Analogtodigital converter survey and analysis," IEEE Journal on Selected
Areas in Communications, April 1999
] / [
2
2
conv J
f
Power
FOM
ENOB
s
=
EECS 247 Lecture 22 Data Converters Nyquist Rate ADCs 2010 Page 48
ADC Figures of Merit
One more bit means...
6dB SNR, 4x less noise power, 4x larger C
Power ~ Gm ~ C increases 4x
Even worse: Flash ADC
Extra bit means 2x number of comparators
Each of them needs double precision
Transistor area 4x, Current 4x to keep same
current density
Net result: Power increases 8x
EECS 247 Lecture 22 Data Converters Nyquist Rate ADCs 2010 Page 49
ADC Figures of Merit
FOM
2
seems not entirely appropriate, but
somehow still standard in literature, papers
"Tends to work" because:
Not all power in an ADC is "noise limited
E.g. Digital power, biasing circuits, etc.
Better use FOM
2
to compare ADCs with
same resolution!
EECS 247 Lecture 22 Data Converters Nyquist Rate ADCs 2010 Page 50
ADC Figures of Merit
Speed
Power
FOM =
3
Compare only power of ADCs with
approximately same ENOB
Useful numbers:
10b (~9 ENOB) ADCs: 1 mW/MSample/sec
Note the ISSCC 05 example: 0.33mW/MS/sec!
12b (~11 ENOB) ADCs: 4 mW/MSample/sec
EECS 247 Lecture 22 Data Converters Nyquist Rate ADCs 2010 Page 51
10Bit ADC Power/Speed
Yoshioko ISSCC 05
EECS 247 Lecture 22 Data Converters Nyquist Rate ADCs 2010 Page 52
12Bit ADC Power/Speed
Loloee
(ESSIRC 2002)
EECS 247 Lecture 22 Pipelined ADCs and More 2010 Page 53
1.0E+09
1.0E+10
1.0E+11
1.0E+12
1985 1990 1995 2000 2005
B
a
n
d
w
i
d
t
h
x
R
e
s
o
l
u
t
i
o
n
[
H
z

L
S
B
]
Performance Trend
~ 2x/5 years
EECS 247 Lecture 22: Data Converters Nyquist Rate ADCs 2010 Page 54
ADC Architectures
Slope type converters
Successive approximation
Flash
Interpolating & Folding
Residue type ADCs
Twostep Flash
Pipelined ADCs
Timeinterleaved / parallel converter
Oversampled ADCs
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 55
AnalogtoDigital Converters
Two categories:
Nyquist rate ADCs f
sig
max
~ 0.5xf
sampling
Maximum achievable signal bandwidth higher compared
to oversampled type
Resolution limited to max. ~14 to 16bits
Oversampled ADCs f
sig
max
<< 0.5xf
sampling
Maximum possible signal bandwidth lower compared to
nyquist rate ADCs
Maximum achievable resolution high (18 to 20bits!)
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 56
The Case for Oversampling
Nyquist sampling:
Oversampling:
Nyquist rate f
N
~2B
Oversampling rate M = f
s
/f
N
>> 1
f
s
>2B +o
Freq
B
Signal
narrow
transition
Sampler AAFilter
Nyquist
ADC
DSP
=M
Freq
B
Signal
wide
transition
Sampler AAFilter
Oversampled
ADC
DSP
f
s
f
s
f
N
f
s
>> f
N
??
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 57
Nyquist v.s. Oversampled Converters
Antialiasing Requirements
X(f)
frequency
frequency
frequency
f
B
f
B
2f
s
f
B
f
s
Input Signal
Nyquist Sampling
Oversampling
f
S
~2f
B
f
S
>> 2f
B
f
s
Antialiasing Filter
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 58
Oversampling Benefits
Almost no stringent requirements imposed on
analog building blocks
Takes advantage of the availability of low cost,
low power digital filtering
Relaxed transition band requirements for
analog antialiasing filters
Reduced baseband quantization noise power
Allows trading speed for resolution
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 59
ADC Converters
Baseband Noise
For a quantizer with quantization step size A and sampling rate f
s
:
Quantization noise power distributed uniformly across Nyquist
bandwidth ( f
s
/2)
Power spectral density:
Noise is distributed over the Nyquist band f
s
/2 to f
s
/2
2 2
e
s s
e 1
N ( f )
f 12 f
  A
= =

\ .
f
B
f
s
/2 f
s
/2 f
B
N
e
(f)
N
B
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 60
Oversampled Converters
Baseband Noise
B B
B B
f f 2
B e
f f s
2
B
s
B s
2
B0
B B0
B B0
s
s
B
1
S N ( f )df df
12 f
2 f
12 f
where f or f f / 2
S
12
2 f S
S S
f M
f
where M oversampl i ng rat i o
2 f
  A
= =

\ .
 
A
=


\ .
=
A
=
 
= =


\ .
= =
} }
f
B
f
s
/2 f
s
/2 f
B
N
e
(f)
N
B
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 61
Oversampled Converters
Baseband Noise
2X increase in M
3dB reduction in S
B
bit increase in resolution/octave oversampling
B B0
B B0
s
s
B
2 f S
S S
f M
f
where M oversampl i ng rat i o
2 f
 
= =


\ .
= =
To further increase the improvement in resolution:
Embed quantizer in a feedback loop (patented by Cutler
in 1960s!)
Noise shaping (sigma delta modulation)
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 62
PulseCount Modulation
V
in
=2/8
Nyquist
ADC
t/Ts
0 1 2
Oversampled
ADC, M = 8
t/Ts
0 1 2
V
in
=2/8
Mean of pulsecount signal approximates analog input!
010
010
2/8
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 63
PulseCount Output Spectrum
f
Magnitude
Signal band of interest: low frequencies, f < B << f
s
Quantization error: high frequency, B f
s
/ 2
Separate with digital lowpass filter!
2/8
f
s
/4
Digital
filter
f
s
/2
B
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 64
Oversampled ADC
Predictive Coding
Quantize the difference signal rather than the signal itself
Smaller input to ADC Buy dynamic range
Only works if combined with oversampling
1Bit digital output
Digital filter computes average Nbit output
+
_
v
IN
D
OUT
Predictor
ADC
Digital
Filter
Nbit
1bit
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 65
Oversampled ADC
Decimator:
Digital (lowpass) filter
Removes quantization noise for f > B
Provides antialias filtering for DSP
Narrow transition band, highorder (digital filters with high order
consume significantly smaller power & area compared to analog filters)
1Bit input, NBit output (essentially computes average)
f
s
= Mf
N
Freq
B
Signal
wide
transition
Sampler
Analog
AAFilter
E.g.
PulseCount
Modulator
Decimator
narrow
transition
f
s1
= M f
N
DSP
Modulator Digital
AAFilter
f
s2
= f
N
+ o
1Bit Digital NBit
Digital
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 66
Modulator
or Analog Front End (AFE)
Objectives:
Convert analog input to 1Bit pulse density stream
Move quantization error to high frequencies f >>B
Operates at high frequency f
s
>> f
N
M = 8 256 (typical).1024
Since modulator operated at high frequencies
need to keep analog circuitry simple
EA = AE Modulator
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 67
Sigma Delta Modulators
Analog 1Bit EA modulators convert a continuous time
analog input v
IN
into a 1Bit sequence D
OUT
H(z)
+
_
V
IN
D
OUT
Loop filter 1b Quantizer (comparator)
f
s
DAC
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 68
SigmaDelta Modulators
The loop filter H can be either switchedcapacitor or continuous time
Switchedcapacitor filters are easier to implement + frequency
characteristics scale with clock rate
Continuous time filters provide antialiasing protection
H(z)
+
_
V
IN
D
OUT
f
s
DAC
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 69
Oversampling A/D Conversion
Analog frontend oversampled noiseshaping modulator
Converts original signal to a 1bit digital output at the high rate of
(2BXM)
Digital backend digital filter (decimator)
Removes outofband quantization noise
Provides antialiasing to allow resampling @ lower sampling rate
1bit
@ f
s
Nbit
@ f
s
/M
Input Signal Bandwidth
B=f
s
/2M
Decimation
Filter
Oversampling
Modulator
(AFE)
f
s
f
s
= sampling rate
M= oversampling ratio
f
s
/M
EECS 247 Lecture 22 Oversampled ADCs 2010 Page 70
1
st
Order EA Modulator
1
st
order modulator, simplest loop filter an integrator
+
_
V
IN
D
OUT
}
H(z) =
z
1
1 z
1
DAC
Note: Nonlinear system with memory difficult to
analyze