A Study of Successive Approximation

Registers and Implementation of an UltraLow Power 10-bit SAR ADC in 65nm CMOS
Technology
Masters thesis performed in
Electronic Devices
by
Raheleh Hedayati

Reg nr: LiTH-ISY-EX--11/4512--SE

September 2011

ii

A Study of Successive Approximation

Registers and Implementation of an UltraLow Power 10-bit SAR ADC in 65nm CMOS
Technology
Masters thesis

Performed in Electronic Devices,

Dept. of Electrical Engineering
Linkping Institute of Technology

By Raheleh Hedayati
Reg nr: LiTH-ISY-EX--11/4512--SE

Supervisor: Professor Atila Alvandpour

Linkping University
Examiner: Professor Atila Alvandpour
Linkping University
Linkping, September 2011

iii

iv

Presentation Date
2011-09-XX

Department and Division

Publishing Date (Electronic version)

Department of Electrical Engineering

Electronic Devices

Language

Type of Publication

English
Other (specify below)

Licentiate thesis
Degree thesis
Thesis C-level
Thesis D-level
Report
Other (specify below)

Number of Pages

ISBN (Licentiate thesis)

ISRN: LiTH-ISY-EX--11/4512--SE
Title of series (Licentiate thesis)
Series number/ISSN (Licentiate thesis)

URL, Electronic Version

http://www.ep.liu.se
Publication Title

A Study of Successive Approximation Registers and Implementation of An Ultra-Low Power 10-bit

SAR ADC in 65nm CMOS Technology
Author(s)

Raheleh Hedayati
Abstract

In recent years, there has been a growing need for Successive Approximation Register
(SAR) Analog-to-Digital Converter in medical application such as pacemaker. The
demand for long battery life-time in these applications poses the requirement for
designing ultra-low power SAR ADCs.
This thesis work initially investigates and compares different structures of SAR control
logics including the conventional structures and the delay line based controller.
Additionally, it focuses on selection of suitable dynamic comparator architecture.
Based on this analysis, dynamic two-stage comparator is selected due to its energy
efficiency and capability of working in low supply voltages. Eventually, based on these
studies an ultra-low power 10-bit SAR ADC in 65 nm technology is designed.
Simulation results predict that the ADC consumes 12.4nW and achieves an energy
efficiency of 14.7fJ/conversion at supply voltage of 1V and sampling frequency of
1kS/s. It has a signal-to-noise-and-distortion (SINAD) ratio of 60.29dB and effectivenumber-of-bits (ENOB) of 9.72 bits. The ADC is functional down to supply voltage of
0.5V with proper performance and minimal power consumption of 6.28nW.

Keywords

SAR ADC, SAR Logic, Dynamic Comparator, Low Power

v

vi

Abstract
In recent years, there has been a growing need for Successive
Approximation Register (SAR) Analog-to-Digital Converter in
medical application such as pacemaker. The demand for long
battery life-time in these applications poses the requirement for
designing ultra-low power SAR ADCs.
This thesis work initially investigates and compares different
structures of SAR control logics including the conventional
structures and the delay line based controller. Additionally, it
focuses on selection of suitable dynamic comparator architecture.
Based on this analysis, dynamic two-stage comparator is selected
due to its energy efficiency and capability of working in low
supply voltages. Eventually, based on these studies an ultra-low
power 10-bit SAR ADC in 65 nm technology is designed.
Simulation results predict that the ADC consumes 12.4nW and
achieves an energy efficiency of 14.7fJ/conversion at supply
voltage of 1V and sampling frequency of 1kS/s. It has a signal-tonoise-and-distortion (SINAD) ratio of 60.29dB and effectivenumber-of-bits (ENOB) of 9.72 bits. The ADC is functional down
to supply voltage of 0.5V with proper performance and minimal
power consumption of 6.28nW.

Keywords: SAR ADC, SAR Logic, Dynamic Comparator, Low

Power

vii

viii

Acknowledgment
This thesis would not have been possible without the supervision of
Professor Atila Alvandpour with whom I have had great discussions. It
is also my pleasure to thank Dai, Ameya, and all other PhD students in
Electronic Devices for their helpful suggestions.
I am grateful to my friends, Iman and Farrokh for their support during
my research life. Furthermore, I would like to thank Hoda, Mahboobeh,
Sepehr, Golnaz, Asieh, Leila, Maryam, Farid, Milad, Naeim, Peyman,
and Mahira for the happy moments we shared during my master studies
in Linkping.
I owe my sincere gratitude to my parents who always encouraged me to
pursue my education. I will be forever thankful to my mom for her love
and prayers which I will never forget. Last but not least, I
wholeheartedly thank to my husband, Hojat, for his love and patience.

ix

Abbreviation
ADC
BWC
DAC
DNL
DFF
DL
DR
ENOB
FOM
FFT
HD
INL
LSB
MSB
SAR
SINAD
SFDR
SNR
SHC
TWC

Analog to Digital Converter

Binary-Weighted Capacitor
Digital to Analog Converter
Differential Non Linearity
Delay type Flip Flop
Delay Line
Dynamic Range
Effective Number of Bits
Figure of Merit
Fast Fourier Transform
Harmonic Distortion
Integral Non Linearity
Least Significant Bit
Most Significant Bit
Successive Approximation Register
Signal to Noise and Distortion Ratio
Spurious-Free Dynamic Range
Signal to Noise Ratio
Sample and Hold Circuit
Two-Stage Weighted Capacitor

Contents
Abstract..................................................................................................................... vii
Abbreviation ............................................................................................................... x
Chapter 1. Introduction............................................................................................. 1
1.1

Background .................................................................................................. 1

1.2

Motivation .................................................................................................... 2

1.3

Thesis outline ............................................................................................... 2

Chapter 2. ADC Principles and Performance Metrics ........................................... 4

2.1 Analog to Digital Conversion fundamentals ...................................................... 4
2.1.1 Resolution .................................................................................................. 4
2.1.2 Quantization Error ..................................................................................... 4
2.2 Static Performance ............................................................................................. 5
2.2.1 Offset and Full Scale Error ......................................................................... 5
2.2.2 Differential Nonlinearity ............................................................................ 6
2.2.3 Integral Nonlinearity .................................................................................. 6
2.2.4 Missing Code ............................................................................................. 6
2.3 Dynamic Performance ...................................................................................... 7
2.3.1 Signal-to-Noise Ratio................................................................................. 7
2.3.2 Spurious-Free Dynamic Range .................................................................. 7
2.3.3 Signal-to-Noise-and-Distortion Ratio ........................................................ 7
2.3.4 Effective Number of Bits ........................................................................... 8
Chapter 3. Successive Approximation Register ADC............................................. 9
3.1 Successive Approximation Algorithm .............................................................. 9
3.2 Two Different Architectures of SAR ADC..................................................... 10
3.2.1 Separate DAC and Sample and Hold Circuit ........................................... 10
3.2.2 Charge Redistribution Architecture ......................................................... 11
3.3 Sub-blocks of SAR ADC ................................................................................ 12
3.3.1 Sample and Hold ...................................................................................... 12
3.3.2 Successive Approximation Register ........................................................ 12
3.3.3 Comparator .............................................................................................. 12
3.3.4 Digital to Analog Converter..................................................................... 13
Chapter 4. Successive Approximation Register (SAR) Control Logic ................ 16
xi

4.1 SAR Logic Operation ..................................................................................... 16

4.2 SAR Logic Type1 ........................................................................................... 17
4.3 SAR Logic Type2 .............................................................................................. 22
4.3.1 Non-redundant SAR with counter........................................................... 26
4.4 Delay Line Based SAR Control Logic ........................................................... 32
4.4.1 Inverter Based Delay Line ...................................................................... 32
4.4.2 Current-Starved Inverter Delay Line ...................................................... 34
4.4.3 Capacitive-Controlled Delay Element .................................................... 35
4.5 Comparison .................................................................................................... 36
Chapter 5. Comparator Design .............................................................................. 38
5.1

Comparator ............................................................................................... 38

5.2

Performance Metrics of Comparators ........................................................ 39

5.2.1 Resolution ............................................................................................... 39

5.2.2

Propagation Delay .............................................................................. 39

5.2.3

Comparison Rate ................................................................................ 40

5.2.4

Input-Referred Offset ......................................................................... 40

5.2.5

Kickback Noise .................................................................................. 41

5.2.6

Metastability ..................................................................................... 41
Comparator Architectures ......................................................................... 42

5.3
5.3.1

Open-Loop Comparator ..................................................................... 42

5.3.2

Latched Comparator Following Pre-amplifier ................................... 43

5.3.3 Dynamic Latched Comparator ............................................................... 45

Chapter 6. Implementation of Dynamic Latched Comparators .......................... 47
6.1

Simple Architecture of Dynamic Latched Comparator ............................. 47

6.1.1 Simulation Results of N-type Latched Comparator ................................ 49

6.2

Energy Efficient Dynamic Two-Stage Latched Comparator ................... 52

6.2.1 Simulation Results of Two-Stage Latched Comparator.......................... 54

6.3

Modified Two-Stage Latched Comparator ................................................ 56

6.3.1 Simulation Results of Modified Two-Stage Latched Comparator ........ 58

Chapter 7. Implementation of SAR ADC and Performance Evaluation ............ 62
7.1 Implementation of SAR ADC ......................................................................... 62
7.1.1 Design of 10-bit D/A Converter.............................................................. 62
xii

7.2 Performance Evaluation .................................................................................. 66

7.2.1 Power Consumption measurement........................................................... 66
7.2.2 Dynamic Performance Evaluation ........................................................... 68
Chapter 8. Summary ............................................................................................... 71
References ................................................................................................................. 72

xiii

Chapter 1

Introduction
Pacemaker is an example of implantable devices for medical application. In
this chapter, a concise background about pacemaker and its building blocks
are provided. Then, the motivation and organization of this thesis work are
presented.

1.1 Background
Pacemakers directly control the pattern and speed of the heartbeat. When the
heart stops beating or it beats too slowly, pacemaker provides weak electrical
signals with approximately 70 beats per minute to correct the timing of the
heart beat [1].
This medical device contains a battery, a generator and pacing leads. The
leads connect the pacemaker to the heart and stimulate the heart with the
pulses generated in pacemaker. Battery and generator are inside a titanium
container which is placed inside the body.

Figure 1.1. Pacemaker functional blocks borrowed from [2]

Figure 1.1 shows the block diagram of a pacemaker. The main blocks fall into
four parts [2]:
1) At the input, there are sensing system, amplifier, filter, and analog to
digital converter.
2) The digital output of the ADC is fed to the logic block which consists
of a programmable logic, timing control system and therapy
algorithms.
3) Current and voltage reference generator and battery power
management
4) At the output of the pacemaker, high voltage pulse generator and
multiplier exist.

1.2 Motivation
The life time of the artificial pacemakers should last up to 10 years which
mandate low power consumption per operation [2]. The analog to digital
converter is the crucial part of an implantable pacemaker since it consumes a
large amount of power as the interface between sensed analog signal and
digital signal processor block. Therefore, decreasing the power consumption
of the ADC is a major concern.
Low power ADCs with moderate resolution and low sampling frequency is
suited for biomedical application. These specifications make SAR ADC the
suitable choice. It consumes low power due to its simple structure. Moreover,
SAR ADC is scalable with the technology scaling since most parts of the
architecture apart from the comparator are digital.
In this thesis, different structures of SAR control logics and dynamic latched
comparators are studied; then, a 10-bit SAR ADC is designed and
implemented in 65nm CMOS technology. The main target is to design an
ultra-low power 10-bit SAR ADC operating at fs= 1kS/s.

1.3 Thesis outline

In this thesis, Chapter 2 reviews the fundamentals of A/D conversion and
performance metrics of A/D converter.
Chapter 3 presents the different architectures of SAR ADC and introduces
sub-modules of the SAR ADC.

In chapter 4, various structures of SAR control logic are designed. In the

following delay line based SAR controllers are investigated and a comparison
in terms of power efficiency is presented.
Chapter 5 reviews the performance metric and types of comparator.
In chapter 6, three different dynamic latch comparators are implemented and a
comparison is presented.
Chapter 7 deals with the design of the DAC and implementation of the SAR
ADC. Additionally presents the performance evaluation of the designed ADC.
In chapter 8, a concise summary of the thesis work is presented.

Chapter 2

ADC Principles and Performance

Metrics
Analog to digital converters are the interface between the analog input signal
and digital signal processing block. In this chapter, analog to digital
conversion principles are briefly reviewed. In the following, the static and
dynamic performance metrics are presented.

2.1 Analog to Digital Conversion fundamentals

2.1.1 Resolution
The number of bits at the output of ADC is called resolution. VLSB is the
smallest step that can be detected which is equal to

for an N bit ADC.

However in reality the effective resolution is lower than N bits due to different
error sources [3].

2.1.2 Quantization Error

In an A/D converter, when the analog input data is quantized to a finite
number of steps, quantization error occurs. The maximum quantization error
is

. This phenomenon even takes place in an ideal ADC. Assuming that

the quantization error is uncorrelated, it can be modeled as white noise.

Figure 2.1. Quantization error of an ideal ADC, borrowed from [4]

2.2 Static Performance

2.2.1 Offset and Full Scale Error
Offset error is the deviation of code transition voltage at first step from the
ideal one which is

. Full scale error is the deviation of the last code

transition voltage from the ideal one. These errors are shown in Figure 2.2.

Figure 2.2. Offset and full scale error, borrowed from [5]

2.2.2 Differential Nonlinearity

Deviation of the code transition width from the ideal one (1 LSB) is called
differential nonlinearity (DNL). For narrow code width, DNL is negative
while for the wide one DNL is positive. In an ideal ADC the code width is
always one, thus, DNL is zero.

2.2.3 Integral Nonlinearity

Integral nonlinearity (INL) is the difference between the code centers from the
ideal line. INL can also be specified as the sum of DNLs [5]. Additionally,
INL can be defined as the distance of the code centers with the best fit line.
Figure 2.4 depicts the maximum INL which is measured with the ideal one.

Figure 2.4. DNL and maximum INL, borrowed from [4]

2.2.4 Missing Code

When a digital code at the ADC output is not produced for the corresponding
input voltage, there is a missing code. An example of missing code is
illustrated in Figure 2.4. Whenever there is a missing code DNL is -1.

2.3 Dynamic Performance

2.3.1 Signal-to-Noise Ratio
Signal to Noise ratio (SNR) is the ratio of the input signal power over the total
noise power.
(2.2)
In an ideal ADC the only noise source is the quantization error.
(2.3)

(2.4)

By inserting
as below:

and

into Equation 2.2 SNR of an ideal ADC is expressed

= 6.02N + 1.76

(2.5)

2.3.2 Spurious-Free Dynamic Range

Spurious-Free Dynamic Range (SFDR) is defined as the ratio of the input
signal to the largest peak of spur or harmonic distortion tone. SFDR is
presented mathematically in Equation 2.6.
(2.6)

2.3.3 Signal-to-Noise-and-Distortion Ratio

Signal-to-Noise-and-Distortion Ratio (SINAD) is the value of the input signal
over the sum of the total noise and harmonic components. Equation 2.7
indicates the SINAD mathematically.
(2.7)

2.3.4 Effective Number of Bits

Effective Number of Bits (ENOB) is obtained from SINAD. ENOB is
commonly used instead of SINAD, since it presents SINAD in the number of
bits. ENOB is achieved using Equation 2.5.
(2.8)

Chapter 3

Successive Approximation Register

ADC
This chapter presents a literature review of SAR ADC. First, successive
approximation algorithm is explained and different architectures of SAR ADC
are investigated. In the following, the operations of the sub-modules of the
SAR ADC are described.

3.1 Successive Approximation Algorithm

Successive Approximation Register ADC is a proper choice for low power
applications. SAR ADC employs a successive approximation algorithm to
convert analog input to a digital code successively. In other words, one bit is
determined in each clock cycle using binary search algorithm.
In order to investigate the operation of the SAR ADC, consider a 4 bit ADC.
As shown in Figure 3.1, in the first clock cycle DAC voltage is set to half of
by setting the code to 1000, then the input voltage is compared to
and based on the comparison result, MSB is defined. If

the MSB

will not be changed and will remain at one, otherwise the MSB is reset to
zero. So here MSB (D3) remains at one. In the next clock cycle the DAC
input is set to 1100 and again
since

is compared to

. D2 retains its value

. For the next bit the DAC input is set to 1110. Based on

comparison, D1 is reset to zero since

and finally for defining LSB,

the DAC input is set to 1101. D0 is one because

. Therefore, the

analog input is converted to the digital code 1101 in four clock cycles.

Figure 3.1. 4-bit SAR ADC operation

3.2 Two Different Architectures of SAR ADC

3.2.1 Separate DAC and Sample and Hold Circuit
Figure 3.2 illustrates the block diagram of a SAR ADC including a sample
and hold circuit which is a common architecture for SAR ADC. In this
configuration, the comparator offset voltage can be modeled as a voltage
source in series combination with the output of the sample-and-hold circuit
which implies the addition of the offset to the analog input. As a result, an
offset appears in the overall characteristic. Therefore, the linearity of the A/D
converter is not affected by the comparator offset voltage [6]. The drawback
of this approach is that it consumes high power due to separate S/H circuit.

10

Figure 3.2. SAR ADC architecture

3.2.2 Charge Redistribution Architecture

This architecture encompasses a capacitive DAC which also operates as a
sample and hold. The block diagram of the charge redistribution SAR ADC is
illustrated in Figure 3.3. The D/A converter usually contains a binaryweighted capacitor array. In each conversion, first the analog input is sampled
and stored in the capacitor array and then the output of the DAC is compared
to VCM for N clock cycle. The output of the DAC successively follows the
VCM voltage at the comparators input and reaches VCM at the end of each
conversion.

Figure 3.3. SAR ADC with capacitive DAC

11

The main advantage of this configuration is its low power consumption due to
inherent sample-and-hold operation inside the capacitive DAC.

3.3 Sub-blocks of SAR ADC

3.3.1 Sample and Hold
In general, Sample and hold circuit (SHC) contains a switch and a capacitor.
In the tracking mode, when the sampling signal is high and the switch is
connected, it tracks the analog input signal. Then, it holds the value when the
sampling signal turns to low in the hold mode. In this case, sample and hold
provides a constant voltage at the input of the ADC during conversion.
Regardless of the type of S/H (inherent or separate S/H), sampling operation
has a great impact on the dynamic performance of the ADC such as SNDR.

3.3.2 Successive Approximation Register

Successive Approximation Register (SAR) control logic determines each bit
successively. The SA register contains N bit for an N-bit ADC. There are 3
possibilities for each bit, it can be set to 1, reset to 0 or keeps its value.
In the first step, MSB is set to 1 and other bits are reset to 0, the digital
word is converted to the analog value through DAC. The analog signal at the
output of the DAC is inserted to the input of the comparator and is compared
to the sampled input. Based on the comparator result, the SAR controller
defines the MSB value. If the input is higher than the output of the DAC, the
MSB remains at 1, otherwise it is reset to 0. The rest of bits are determined
in the same manner. In the last cycle, the converted digital word is stored.
Therefore, an N-bit SAR ADC takes N+1 clock cycles to perform a
conversion. Different architectures of SAR logics are implemented and
compared in Chapter4.

3.3.3 Comparator
Comparator is the only analog block of a SAR ADC and performs the actual
conversion. It compares the analog sampled input to the analog output of the
DAC and generates digital output of 0 or 1 which will be used in the SAR
logic. Accuracy and speed of the comparator are two important factors. The
comparator need to resolve voltages with small differences. The offset voltage
12

of the comparator employed in SAR ADC is translated to the transfer

characteristic of the ADC thus will not affect the linearity of ADC. The
comparators are discussed in detail in Chapter5 and implementation of
different comparators and comparison between them are presented in
Chapter6.

3.3.4 Digital to Analog Converter

The digital to analog converter (DAC) converts the digital word at the output
of the SAR logic to an analog value. Then in the comparator, this value is
compared to the input signal.
In the capacitive DAC with inherent Sample and Hold, the sampling operation
is performed by DAC and is called charge redistribution DAC. Nowadays,
charge redistribution DACs are commonly used. They consume less power
and induce less mismatch errors compared to the resistive based DAC. Charge
redistribution DAC has fast conversion time. Moreover, they are fabricated
easily [6], [7]. In the following three different architectures of capacitive DAC
are presented.

Binary-Weighted Capacitor Array

An N-bit binary-weighted capacitor array is shown in Figure 3.5. It consists of
scaled binary capacitors, i.e., 2N-1C, 2N-2C 2C, C, C. The last capacitor is a
dummy that has equal value as the LSB capacitor. Thus, the total value of the
capacitors is 2NC.

Figure 3.5. Binary-weighted capacitor array DAC [8]

13

First, in the reset phase all the bottom plates are grounded. During the
redistribution mode in which the actual conversion is performed, based on the
provided digital code, the switches are connected to either Vref or ground.
The occupied area and power consumption of the BWC is increased with the
increase of the resolution [8].

Two-Stage Weighted Capacitor Array

Two-stage capacitor array was first proposed in [9] to mitigate the large
capacitance size in BWC array. Figure 3.6 illustrates a TWC array. In this
approach the BWC array is divided into two smaller BWC and a coupling
capacitor is added between two parts. The value of the coupling capacitor is
determined in Equation 3.1 [8].
(3.1)
This architecture reduces occupied area and power consumption.

Figure 3.6. Two-stage capacitor array DAC [8]

C-2C Capacitor Array

C-2C ladder is an extension of TWC array. Figure 3.7 shows a C-2C ladder
example. In this configuration, the values of the capacitors are drastically
reduced. As a result, this type of DAC can achieve higher speed while
consuming less power. The power consumption increases linearly, in contrast
to BWC in which the power rises exponentially. The main drawback of this
14

configuration is the degradation of the linearity due to parasitic capacitances

[10].

Figure3.7. C-2C ladder [8]

15

Chapter 4

Successive Approximation Register

(SAR) Control Logic
In this chapter, SAR control logic operation is described and then two
conventional structures of SAR are reviewed and implemented in 65nm CMOS
technology. In the following, delay line based controller is investigated.
Finally, a comparison between them in terms of power, speed, and voltage
scalability is presented.

4.1 SAR Logic Operation

Successive approximation register ADC implements the binary search
algorithm using SAR control logic. In general, there are mainly two
fundamentally different approaches to designing the SAR logic. The first one
which is proposed by Anderson consists of a ring counter and a shift register.
At least 2N flip flops are employed in this kind of SAR [14]. The other, which
is proposed by Rossi, contains N flip flops and some combinational logic [15].
SAR control logic determines the value of bits sequentially based on the result
of the comparator. SAR operation can be represented using a sequential Finite
State Machine which is illustrated in Table 4.1. Each conversion takes 12
clock cycles. In the first clock cycle, SAR is in the reset mode and all the
outputs are zero. In the next ten clock cycles, data is converted and each bit is
determined sequentially. The last cycle is for storing the results of the
complete conversion.

16

Table 4.1: Finite State Machine algorithm

Cycle

Sample

D9

D8

D7

D6

D5

D4

D3

D2

D1

D0

comp

0
1
2
3
4
5
6
7
8
9
10
11

1
0
0
0
0
0
0
0
0
0
0
0

0
1
a9
a9
a9
a9
a9
a9
a9
a9
a9
a9

0
0
1
a8
a8
a8
a8
a8
a8
a8
a8
a8

0
0
0
1
a7
a7
a7
a7
a7
a7
a7
a7

0
0
0
0
1
a6
a6
a6
a6
a6
a6
a6

0
0
0
0
0
1
a5
a5
a5
a5
a5
a5

0
0
0
0
0
0
1
a4
a4
a4
a4
a4

0
0
0
0
0
0
0
1
a3
a3
a3
a3

0
0
0
0
0
0
0
0
1
a2
a2
a2

0
0
0
0
0
0
0
0
0
1
a1
a1

0
0
0
0
0
0
0
0
0
0
0
a0

a9
a8
a7
a6
a5
a4
a3
a2
a1
a0
-

4.2 SAR Logic Type1

SAR architecture shown in Figure 4.1 is presented in [14] and is commonly
used in SAR ADCs due to its straightforward design technique. This control
logic encompasses a ring counter and a code register. The ring counter is in
fact a shift register.
For each conversion, in clock cycle 0, the EOC signal is high and all Flip
Flops outputs are reset to zero, and for the rest of cycles EOC is low. In the
next clock cycle, the most significant Flip Flop is set to one which
corresponds to MSB of the digital word to the DAC. Then the counter shifts
1 through the Flip Flops from MSB to LSB.

17

Reset
CLK
D

SET

CLR

SET

CLR

SET

CLR

SET

CLR

SET

CLR

COMP
D

SET

CLR

SET

CLR

D9

SET

CLR

D8

SET

CLR

D7

SET

CLR

D0

Figure 4.1. SAR block diagram

In each clock cycle, one of the outputs in the ring counter sets a Flip Flop in
the code register. The output of this Flip Flop which is set by the ring counter
is used as the clock signal for the previous Flip Flop. At rising edge of the
clock, this Flip Flop loads the result from the comparator. Figure 4.2 shows
the transient response. At the end of each conversion, EOC signal turns to
high. This type of SAR logic, converts each sample in 12 clock cycles.

18

Figure 4.2. Transient response of the SAR logic type 1

The Flip Flops which are employed in this structure are set-reset D-FFs. For
low power purpose, transmission gate based Flip Flops are used [17].
Minimum size transistors with double length are chosen for improving the
power performance. The schematic of the DFF is illustrated in Figure 4.3. In
order to decrease the leakage power even more while simultaneously
maintaining the speed, high threshold voltage transistors are used in the non-

19

critical paths and low-VT transistors in the critical path. Thus, this dual
threshold approach provides high performance Flip-Flops [18].
clkb

clk
set

reset

clkb

clk

clk

clkb

reset
set

Figure 4.3. Transmission gate based Flip-Flop

Figure 4.4 illustrates total power consumption of the SAR logic versus
frequency.

Figure 4.4. Average power vs. frequency

20

The measured leakage power is 707pW which is %58 of the total power at
sampling frequency of 1kS/s.
The architecture shown in Figure 4.1 has some advantages. First, it is low
power due to low signal transition. Furthermore, since this logic is iterative, it
can be extended to higher resolutions by just extending the shift registers.
In order to decrease the power even more, supply voltage can be reduced.
Figures 4.5 and 4.6 illustrate the power consumption and propagation delay
under different supply voltages, respectively.

Figure 4.5. Power consumption vs. scaled VDD

21

Figure 4.6 Propagation delay vs. scaled VDD

4.3 SAR Logic Type2

For an N bit SAR ADC, digital control logic needs at least N flip-flops to
perform conversion. Therefore, a non-redundant SAR employs minimum
number of flip-flops. These FFs both guess and store the converted result.
For its operation, at the initialization step (step 0), SAR assigns MSB to 1 and
the other bits to 0. This word is equivalent to

after being applied to D/A

converter. At step1, the SAR control logic makes its decision based on the
output of the comparator. If it is high, MSB remains 1 otherwise SAR changes
MSB to 0. Thus the value of MSB is defined now. Simultaneously SAR sets
next MSB to 1. Applying this word to DAC again, input voltage is compared
to the DAC output and SAR defines the value of this next MSB based on the
result of the comparator. Therefore, successive approximation logic
determines all the bits sequentially. Figure 4.7 shows 10-bit SAR logic. The
design is based on the logic proposed by Rossi [15].

22

Figure 4.7. Successive approximation logic [15]

As it can be seen from the Table 4.1, there are three possibilities for each bit.
1. Shifting right
2. Taking the comparator results
3. Memorization mode
As illustrated in Figure 4.7, non-redundant successive approximation control
logic contains an N-bit shift register and OR chain. Each Flip Flop can take
the result of the comparator, the output of the previous Flip Flop, or the value
of the OR gate in each step.
For its operation, at initialization state the first Flip Flop on the left is set to 1
and the rest of Flip Flops are reset to 0. This condition is provided by an
external control signal called start signal. In the next steps, one of three
possible inputs is taken for each Flip Flop. Since there are three possible
inputs, a multiplexer is required. Inside of the Flip Flop is shown in Figure 4.6
(b). Although a three input 3:1 mux is required, a 4:1 mux is employed for
implementation of the SAR for simplicity. A and B are the control signals of
the multiplexer. The appropriate input is selected according to the table 4.2.

23

Table 4.2

A B
1 - Memorization
0 1
Data load
0 0
Shift right
The conversion is terminated by applying high voltage to the OR chain.
Consequently, SAR enters the storing mode. In general, at the end of each
conversion, converted result is stored in the SAR. This is performed by
connecting the output of the last Flip Flop to the OR chain. Therefore, end of
conversion is defined by the least significant Flip Flop on the right.

Figure 4.8. Transient response of SAR logic type 2

24

Figure 4.8 shows the transient response of the SAR control logic. Each
conversion takes 11 clock cycles in this type of SAR controller.
Figure 4.9 presents the total power consumption of SAR versus frequency. In
order to consider clock power in the total power, some drivers are used for
clock distribution.

Figure 4.9. Average power vs. frequency

The leakage power for this type of SAR is 824pW which is %55 of the total
power at 1kS/s.
This architecture is iterative; therefore by extending the sequence of FFs
higher resolutions can be achieved. Start signal should be adjusted manually
for different clock frequencies. However, this problem is mitigated by adding
a counter to the structure. The behavior of the circuit by applying different
supply voltages at the sampling frequency of 1kS/s is illustrated in Figures
4.10 and 4.11.

25

Figure 4.10. Power consumption vs. scaled VDD

Figure 4.11. Propagation delay vs. scaled VDD

4.3.1 Non-redundant SAR with counter

A counter is added to the non-redundant SAR in order to generate the start
signal automatically. Non-redundant SAR architecture contains a sequence of
three-input Flip-Flops, a 4 bit counter and a decoder. The schematic view is
shown in Figure 4.12.

26

comp

d1 clk
d2
o
d3
rst
s1 s0

d1 clk
d2
o
d3
rst
s1 s0

D9

d1 clk
d2
o
d3
rst
s1 s0

d1 clk
d2
o
d3
rst
s1 s0

d1 clk
d2
o
d3
rst
s1 s0

D8

Decoder

D1

D0

clk
Counter
rst

Figure 4.12. Schematic view of the control logic

There is a 4:1 mux inside of each FF block which is shown in Figure 4.7(b).
This digital logic is similar to non-redundant SAR in [15]. The differences are
the 4-bit counter and a decoder which are added to the schematic. There are
some minor changes in the finite state machine which is implementing the
SAR operation.
Each bit has 4 possibilities in the new structure:
1.
2.
3.
4.

Storing the previous state (memorization)

Value of the previous Flip Flop (right shift)
Result of the comparator
Assigning 0 or 1

For a complete conversion 12 clock cycles are required. At initialization state

all the outputs are reset to zero. The start of the conversion is identified by the
end of conversion (EOC) signal. As mentioned above, SAR has a 10-bit shift
register. Each D-FF has four external inputs and hence a 4:1 mux. The
multiplexer has two control bits and functions according to the truth table
below.

27

Table 4.3

A
0
0
1
1

B
0
Memorization
1 Assigning 1 or 0
0 Comparator value
1
Shift right

The 4-bit counter generates control bits A and B, then the decoder defines the
counter value. D0 to D9 are the digital output data of the A/D converter. The
transient response of the SAR logic is illustrated in Figure 4.13. As it can be
seen from the figure this types of SAR, converts a sampled data in 12 clock
cycles.

28

Figure 4.13. Transient response of non-redundant SAR with counter

29

Figure 4.14 represents the total power consumption of the SAR versus
frequency. The drawback of this configuration is that the logic is not iterative,
thus by changing the resolution the entire decoder needs to be changed.

Figure 4.14. Average power vs. frequency

The leakage power of this structure is 1.4nW which is %48 of the total power
at 1kS/s.
The circuit was simulated under different supply voltages and the results of
the power consumption and propagation delay are shown in Figures 4.15 and
4.16, respectively.

30

Figure 4.15. Power consumption vs. scaled VDD

Figure 4.16. Propagation delay vs. scaled VDD

31

4.4 Delay Line Based SAR Control Logic

Delay line circuits are used for clock generation purposes. Typically, a wide
range of adjustment is desired to achieve different clock frequencies from
high to low. An adjustable delay line encompasses a string of variable delay
elements. Variable delay components are voltage or current controlled
elements. In fact, RC time constant of each delay element determines its
delay. Hence, varying the total resistance or capacitance leads to different
delays [19].
Delay elements are vulnerable to the noise of power supplies which causes
jitter in the generated clock. Therefore, designing delay elements with wide
adjusting range and low jitter which are also equipped with power supply
noise rejection circuits is desired. There are several kinds of delay elements.
In this thesis work some of them which consume less power are explored.

4.4.1 Inverter Based Delay Line

Inverter delay line is constructed from a chain of inverters in a row. Figure
4.17 (a) shows an inverter delay element. The required control signals for
SAR algorithm are generated by employing some gates and alternately
delaying rising and falling times. This is accomplished by increasing the
channel length of NMOS or PMOS transistors in turns. These sequential
pulses control some parts of SAR [16].
Inverter delay line has some difficulties in its implementation. For example,
large delays cannot be realized by inverter delay line since many stages are
required, which is not feasible and also increases jitter at the output
proportional to the square root of the number of stages.

32

C1
a)
A
B

b)

C1

Figure 4.17. Delay lined based SAR controller a) inverter delay element b)
waveforms at A, B, and C1 nodes

After sampling phase, ADC starts to convert the sampled input data. Sample
signal is an external signal which is inverted, see voltage A, as shown in
Figure 4.17 (b). When A is low, B is high and then voltage A with a certain
delay turns to 1. Therefore, a pulse C1 at the output of the NAND gate is
generated which is active low and is a control signal for SAR algorithm. For
increasing the width of these control pulses, the rising and falling times should
be increased alternately. In this configuration when A becomes high, the
NMOS transistor in the inverter is on and PMOS transistor is not conducting.
The large length of NMOS causes the capacitance at the output of the inverter
discharge slowly and B turns to low voltage slowly and C1 becomes high
again. In order to delay the rising edge, a PMOS transistor with large channel
length should be used and for delaying a falling edge an NMOS transistor
with large channel length can be employed. One of the drawbacks of this
configuration is that the delay value of each delay element is not precise
because of the process and temperature variation as well as power supply
noise. However, the width of these control signals should be large enough for
DAC and comparator to settle. Therefore, this inaccuracy of delay does not
affect the ADC performance.
The generated control signals are shown in Figure 4.18.

33

Figure 4.18. Produced control signal from delay line

Since the PMOS and NMOS transistors in a simple inverter will not conduct
simultaneously, there is no static power consumption, which is the main
advantage of the inverter based delay line. However, there are some periods of
time in which both transistors are in the saturation region and conduct,
causing short circuit current flows between power rails. Short circuit may also
occur when one of the transistors is in the saturation region while the other is
in the triode region. This delay line can provide an adjustable delay by using a
multiplexer. As a result, it is a digitally controlled system which is not power
efficient and precise compared to the analog control systems. Thus analog
control delay elements are desirable [20].

4.4.2 Current-Starved Inverter Delay Line

For having a voltage controlled delay element, current controlled transistors
can be added in series to the inverter. This structure is called current-starved
inverter which is shown in Figure 4.19.
Delay can be adjusted by changing the current through the inverter. For
instance, larger delay is achieved as a result of decreasing the current.
Changing the control gate voltage in fact varies the resistance of the transistor.
Therefore, Current Starved Inverter is a resistive controlled delay element.
34

VDD
VBP
M9

M13

M17

M8

M12

M16
Vout

Vin

M7

M11

M15

VBN
M6

M10

M14

Figure 4.19. Current-starved delay element

4.4.3 Capacitive-Controlled Delay Element

An example of a capacitive-controlled delay element is shown in Figure 4.20.
M1 is in the triode region and acts as an adjustable resistance. This transistor
decouples an explicit output capacitance [19].
Vin

Vout

VC

Figure 4.20. Capacitive controlled delay element

The maximum delay that can be obtained with capacitive-controlled delay

elements is limited to
. Thus, achieving a large delay
mandates a large number of delay elements.

35

4.5 Comparison
Widths of transistors are minimum size since the operating frequency is quite
low for biomedical applications and lowering power consumption is the
primary target. However, for decreasing the leakage, length is chosen to be 1.5
times the minimum length. It should be mentioned that in all the schematics
presented in this work, transistors are high VT which leads to lower leakage
power. The logics are compared more or less in the same situation.
All the outputs are loaded with NMOS transistors as switches in the capacitor
array DAC. The rising time of the clock is 1ns. However, it can be increased
up to 2ns to achieve lower dynamic power with preserving the performance.
External asynchronous signals to the logic, including reset and start signals,
are synchronized through two D-FFs. All clock signals are buffered.
According to the simulation for a fan-out of 12 Flip Flops, a simple two
inverter in a row can be used as a driver with proper aspect ratio. Figure 4.21
illustrates the power consumption of the three successive approximation
logics versus frequency. From the simulation results, SAR logic with a
sequencer and a code register is more power efficient.

Figure 4.21. Average power vs. frequency

36

Figure 4.22 shows a comparison between the power consumption of the three
logics versus scaled supply voltage at 1KS/s.

Figure 4.22. Average power vs. scaled supply voltage

Delay line can be used to control the SAR. In general, delay lines consume
more power compared to the conventional SAR control logics. Therefore,
conventional SAR logic type1 is employed in the designed ADC.

37

Chapter 5

Comparator Design
A brief description of a comparator is described and then the performance
metrics of a comparator are discussed. Finally, basic comparator types are
presented and a comparator suitable for low power applications is chosen.

5.1

Comparator

Comparator is one of the building blocks of A/D converters. In some sorts of

ADCs such as flash ADC, the comparator has a great impact on the
performance of the whole ADC. However, in some cases the error produced
by the comparator can be compensated in the followed digital processing
block. Figure 5.1 illustrates the comparator and the voltage transfer function
of an ideal as well as a real comparator [6]. A comparator generates a logic
output high or low based on the comparison of the analog input with a
reference voltage. In an ideal comparator, with infinite gain, for input voltages
higher than the reference voltage, the comparator outputs logical one and for
the input voltages lower than the reference voltage it produces zero at the
output.
Vin+

Vin-

Vout

Comp

Vout

Vout

VOH

VOH
VIL
Vin+ - Vin-

Vin+ - VinVOL

VIH
VOL

a) ideal comparator
b) real comparator
Figure 5.1. Input-output characteristic of an ideal and real comparator

However, in reality due to finite gain, comparator results in one when Vin >
Vref + VIH , and is set to zero when Vin < Vref + VIL .

38

A typical comparator architecture containing a pre-amplifier and a

regenerative latch is illustrated in Figure 5.2 [6].

Clock

Vin+

Vout+

Pre-Amp

Regenerative Latch

Vin-

Vout-

Figure 5.2. Typical block diagram of a comparator

5.2

Performance Metrics of Comparators

Speed, accuracy, low power consumption and wide input common mode
range are some design considerations for comparators, which define
performance metrics of a comparator.

5.2.1 Resolution
The minimum input voltage difference which is detectable by a comparator is
called resolution. Noise and input referred offset are considered as limiting
factors of the resolution.
In an A/D converter the minimum required resolution is denoted as V LSB. For
instance, in an N-bit ADC, the comparator should be able to detect one LSB
which is VLSB =

5.2.2 Propagation Delay

Propagation delay defines speed of the comparator, i.e. how fast the
comparator decides. It also affects the ADC speed. As shown in Figure 5.3,
rising propagation delay is measured between the transition points of the input
and output signal when they reach 50% of the signal levels. According to the
definition, propagation delay time is the average time of the rising and falling
propagation delay times.
(5.1)
The propagation delay is shown in Figure 5.3.

39

Figure 5.3. Rising and falling propagation delays [28]

Propagation delay of a clocked comparator is measured between the 50%

transition points of the clock signal and the differential output signal (
).

5.2.3 Comparison Rate

Comparison rate is the highest frequency in which a comparator results in a
correct value and is defined by the overdrive recovery test. Assume the input
difference of the comparator toggles from a large value to a small one, if the
comparator responds correctly to this large voltage variation at its input, the
comparator has recovered from an overdrive at this frequency. Comparison
rate depends on the speed or basically on the recovery time of the regenerative
latch and the preamplifier. [6]

5.2.4 Input-Referred Offset

The static offset at the input of the comparator originates from mismatches in
identical devices such as input pair transistors. Therefore, the behavior of the
comparator might arbitrarily change. In other words, the comparator may
result in high voltage at its output when the input voltage is less than the
reference voltage and also it may result in low voltage when the input voltage
is higher than the reference. In the presence of offset voltage, the inputreferred offset can be modeled as is shown in Figure 5.4. According to this
model the comparator decides based on the following equations:

40

Vin

+
Comp

Vref

Vout

Voff

Figure 5.4. Comparator with offset modeling

If Vin > Vref +Voff then Vout = VH

If Vin < Vref + Voff then Vout = VL

This non-ideality of the comparator can adversely affect the accuracy of the
comparator, and consequently decrease the resolution of the ADC.
Input-referred offset can be reduced by the aid of some techniques and/or by
using a pre-amplifier in the comparator architecture at the expense of
complexity and more power consumption.

5.2.5 Kickback Noise

Kickback noise is in fact the voltage disturbance at the input nodes of the
comparator due to large variation of the voltage at internal nodes. These
switching effect of the latch are coupled to the input nodes via parasitic
capacitances and vary the input voltage. This disturbance is known as
kickback noise.
There are some ways to reduce the impact of the kickback noise. The effect
can be alleviated by employing a pre-amplifier before latch and also isolating
the input nodes by using cascode transistors as switches.

5.2.6 Metastability
Metastability is an error which occurs in a comparator containing a latch.
Considering a comparator with two operation phases, reset or pre-amp phase
and regeneration phase, the latch is assigned a certain time referred to as
regeneration time which is half of the clock period. This means that a latch
41

generates a logic level in this time. When the input voltage is close to the
reference voltage, the latch takes more time to produce a logical level which
results in a metastable state. In other words, a metastable situation occurs
when the latch is not able to switch to a valid logical level, zero or one, in the
regeneration time constant and reaches an intermediate value. Therefore, an
important issue in latch design is to compute the probability of this
occurrence. Metastability causes errors in comparators and can adversely
affect the accuracy of the comparator and the ADC.

5.3

Comparator Architectures

There are various comparator types. In this thesis three basic types of
comparators are described. The comparator topology is chosen based on the
specific application. In the following, appropriate comparator architecture for
low power application is selected.

5.3.1 Open-Loop Comparator

Open-loop comparator is in fact a high gain amplifier with differential input
and single ended output with large swing. A two stage op-amp without
compensation is an excellent realization of open-loop comparator which is
illustrated in Figure 5.5. Since precise gain in not required in comparator, no
compensation technique is needed.
VDD
M3

M4
M6

ViN

M1

M2

Vout

ViP

CL
VB1

M5

M7

Figure 5.5. Open-loop comparator

42

This comparator has two poles which compared to the one-pole

implementation of the open loop comparator achieves higher speed. In one
stage amplifier which has one pole, increasing the gain and simultaneously
maintaining the large bandwidth, fu, poses limitation on the gain bandwidth,
fc, which results in speed reduction of the comparator. In order to mitigate this
problem two stage comparator has been proposed. Cascading two stages with
lower gain in each stage leads to higher speed however it introduces more
input-referred offset voltage. Furthermore, this configuration consumes more
power; therefore this type of comparator is not suited for high speed A/D
converters as well as low power ADCs.

5.3.2 Latched Comparator Following Pre-amplifier

In this configuration the pre-amplifier attenuates the latch offset by its gain.
(5.2)
This configuration of comparator benefits from low input-referred offset and
thus is capable of detecting small voltage differences at its inputs.
According to the Equation 5.2 for input referred offset, increasing preamplifier gain results in much lower offset voltage at the input. However,
increasing the gain reduces the speed of the comparator. Thus, this comparator
exhibits a speed-gain tradeoff. In this architecture, pre-amplifier is a single
pole, one stage amplifier.
The relation between unity gain frequency, gain, and 3-dB cutoff frequency is
defined as below:
(5.3)
Frequency response for a single pole amplifier is shown in Figure 5.6 [28].

43

Figure 5.6. Frequency response of a single pole amplifier

If maintaining fu is desired when the gain is increased, fc or the pole value will
decrease and consequently reduce the comparator speed.
An example of a comparator including latch and preamplifier is illustrated in
Figure 5.7.
VDD
M5

M3

M4

M6
Vo
+
CLK

M9
M1

M2

M7
VBias

M8

Mtail

Figure 5.7. Comparator schematic including latch and pre-amplifier

Static latch consumes static power which is not attractive for low power
applications. Comparator delay is demonstrated in Equation 5.4. It is in fact
the delay time of a latched comparator, which is described in the next section.
The only difference is that the pre-amplifier gain added to the equation.
44

(5.4)

5.3.3 Dynamic Latched Comparator

Latch only comparators are clocked comparators. They operate based on
amplification and positive feedback. An example of a latched only comparator
is shown in Figure 5.8.

Figure 5.8. Dynamic Latched Comparator circuit [26]

Figure 5.9 shows the operation of the comparator. There are two operation
phases, reset phase and regeneration or evaluation phase. In the reset phase,
the output nodes are charged to supply voltage or discharged to the ground
depending on the architecture of comparator. During the reset phase, the
comparator tracks the input, and then in the regeneration phase the positive
feedback produces a digital value at the comparator output. One of the
advantages of dynamic latched comparators is their power efficiency since
they only consume power in regeneration phase and there is no static power
consumption in the reset phase.

45

Figure 5.9. Latched Comparator Operation [28]

Dynamic latched architecture is the most power efficient comparator,

however, it introduces large input referred offset which makes it unappealing
for high resolution ADCs. This effect can be reduced by increasing the width
of input transistors in differential pair. Employing offset cancelation
techniques in the comparator implementation is also an effective approach to
mitigate this problem. Furthermore, the offset voltage can be reduced by using
a pre-amplifier which is previously described as a latched comparator with
preamplifier. However, in all the mentioned methods, offset reduction is
achieved at the cost of more power consumption.
Latched comparators are fast and are suited to be used in high speed ADCs. In
order to derive the delay equation, the latched comparator can be modeled as a
single pole comparator with positive feedback. The delay time of this
comparator is calculated as below [28]:
(

(5.5)

46

Chapter 6

Implementation of Dynamic Latched

Comparators
In this chapter, different architectures of dynamic latched comparators, aimed
for low power application are designed, and then their performances are
measured.

6.1 Simple Architecture of Dynamic Latched

Comparator
Figure 6.1 illustrates an example of dynamic latch comparator presented in
[18]. When clock signal is low, in reset phase, the reset transistors are
switched on and charge the output nodes as well as drain nodes of the input
transistors to supply voltage. Current source transistor (Mtail) is off, thus,no
current flows through the circuit during the reset phase. When clock signal
turns to high, reset transistors are disabled and the tail transistor is turned on.
The inverters of the cross coupled circuitry inside the latch receive different
amount of current depending on the input voltage and start to regenerate the
comparators output. In other words, the drain voltage of each input transistors
(M1, M2) start to discharge from VDD to ground with different rates depending
on their input voltage. Once the drain voltage of either of the transistors drops
below VDD-Vth, NMOS transistor of the inverter is turned on and output node
starts to discharge and positive feedback is activated. Once the output node
reaches VDD- |Vth| the PMOS transistor of the other inverter is turned on.
Consequently, the output voltage is regenerated and after regeneration phase
one of the outputs is 1 while the other is 0.
In this configuration, the PMOS switch transistors (M5, M7, M8, and M10)
which charge drain of input transistors to VDD during the reset phase, increase
the time period that input transistors are in saturation region during
regeneration phase. As a result, higher gain is attained.
The drain nodes of input transistors have large voltage variations causing a
large kickback noise. Furthermore, the variation of the operation region of
input transistors causes voltage variation at their gates producing kickback
noise. In the reset mode, differential input pair transistors are off. When the
comparator enters the evaluation phase, VDS is high and transistors are in
47

saturation region, then these nodes are discharged to zero in the regeneration
phase causing transistors to enter the triode region [18].

VDD

M5

Clock

M7

M5

M6

Clock

M10

Vo-

Vo+

M3

Vin+

M8

M4

M1

M2

Clock

Vin-

Mtail

Figure6.1. N-type Latched Comparator Circuit

This implementation is not appropriate for low voltage applications since it is

a stack of four transistors demanding excessive headroom voltage which is not
feasible in low voltage technologies.
Moreover, for enhancing the comparator speed the current through the latch
must be increased via increasing the tail current. Thus, the size of tail
transistor should be increased which in turn, increases the current of
differential input pair transistors. Consequently, the time period in which the
input transistors are in saturation region decreases leading to a lower
comparator gain. As a result, the effect of input-referred offset voltage is more
significant. In other words, higher gain and thus lower offset voltage can be
attained at the cost of speed reduction. Therefore, there is a tradeoff between
speed, offset voltage and power consumption [23].
48

6.1.1 Simulation Results of N-type Latched Comparator

To verify the functionality and evaluate the performance, the comparator was
simulated in Cadence and the results are shown in the following figures. Two
inverters are inserted at the outputs of the comparator to recover the voltage
level at the outputs of the comparator. In order to hold the result of the
comparator during the reset phase, a NOR type Set-Reset Latch is used in this
test bench. Figure 6.2 shows the SR latch.

in1

out

in2
Figure6.2. NOR type SR latch

In 65nm technology the minimum width and length of a CMOS transistor are
135nm and 60 nm, respectively. Sizing of the input transistors has a great
impact on offset voltage. As shown in (6.1), increasing the size of the input
transistors results in reduction of the input referred offset voltage.

(6.1)

In this design, the input transistors are three times larger than the minimum
width in order to decrease the input-referred offset voltage.
As it is mentioned above, the tail width is set to the minimum width in order
to minimize the power consumption and also achieve higher gain which
results in lower offset voltage. In order to decrease the leakage power, the
length of transistors is chosen two or three times larger than the minimum
length.
By employing both high VT and low VT transistors, lower leakage power is
achieved without any speed reduction. Table 6.1 shows the aspect ratio of the
transistors of the comparator.

49

Table6.1. Optimized Size of Transistors

Component Width(m) Length(m) Transistor type

Mtail

0.135

0.12

High VT

M1,2

0.405

0.18

Low VT

M3,4

0.270

0.12

Low VT

M5,6

0.540

0.12

High VT

M7,8,9,10

0.135

0.12

High VT

The comparator was simulated with VDD= 1V, clock frequency= 100 kHz,
input frequency = 5 kHz with full swing, and temperature = 27C.
The comparator operation is shown in Figure 6.3.

Figure6.3 Transient response of the comparator

In order to measure the input referred offset, the device mismatch is

considered and a method which relies on Monte Carlo analysis is used. For
this purpose a simple test bench is created. The input voltage is a ramp which
is sampled with a sample and hold. The obtained staircase signal is applied to
the input of the comparator and a DC threshold voltage is applied to the other
input of the comparator [24]. The test bench circuit is shown in Figure6.4.

50

S&H
+

Data
Load

VCM=VDD/2

Figure6.4. Test bench circuit for offset simulation

First, Monte-Carlo simulation is run for 500 times and the output results are
stored in a file. Then in MATLAB, by performing some post-processing on
the stored results, the input referred offset is evaluated. Figure 6.5 shows the
probability of the occurrence of 1 at the output versus the input voltage. The
statistical properties of the comparator such as threshold voltage and offset
can be extracted from this curve. Offset is defined as the deviation from the
threshold voltage.

Figure6.5. Monte-Carlo simulation Results

The simulation results are presented in table 6.2.

51

Table6.2. Simulation results

Performance Metric
Value Unit
Power Consumption
276
pW
Input-Referred Offset voltage 8.1
mV
Propagation Delay
2.1
ns
Since this comparator aimed for low power and low speed application, delay
is in the last priority. The results show that the power consumption is quite
low. The offset voltage is higher than one LSB which is reasonable for
dynamic latch comparator. Further reduction of offset voltage is achievable by
using offset cancelation techniques at expense of higher power consumption
and more complexity.
In order to evaluate the scalability of the comparator with supply voltage, the
comparator is simulated with different supply voltages. The power
consumption and propagation delay of the comparator versus different supply
voltages are given in table6.3.

Table6.3. Simulation results for scaled VDD

VDD (V) Power Consumption (pW) Propagation Delay (ns)

1
0.8
0.6
0.4

276
216
159
107

2.1
5.3
31.5
1400

According to the results, decreasing VDD reduces the power consumption and
increase the delay of the comparator. For VDD = 0.4 V the propagation delay is
1.4 s which is quite large so this architecture does not perfectly work under
0.4V.
This comparator shows a relatively constant offset voltage over different
supply voltages.

6.2 Energy Efficient Dynamic Two-Stage Latched

Comparator
To mitigate the problems of the previous structure, dynamic two-stage latched
comparator has been proposed. This comparator is energy efficient and fast
and has lower input referred offset voltage. An example of two-stage
comparator is illustrated in Figure 6.6 [16].

52

VDD

clk
FN

FP

ViP

ViN

clk
VDD

VoN

VoP

clk

clk

Figure6.6. comparator schematic

The first stage is a voltage amplifier and the second stage is a latch. During
the reset phase when clock signal is low, PMOS transistors in the first stage
charge F nodes to VDD and turn off the latch stage. In this phase, the output
nodes are reset to zero through NMOS switches in the latch stage. This
architecture is power efficient and fast due to low capacitance at F nodes
which are mainly drain diffusion capacitances of NMOS and PMOS
transistors connected to these nodes.
When clock turns to high in regeneration phase, the tail transistor is turned on
and amplification in the first stage is initiated. F nodes start to discharge
through differential input pair transistors in the first stage. These nodes are
discharged with different rates proportional to the input voltage. Once either
of the output nodes of the first stage (F nodes) drops to an amount around Vth
of the input transistors of the latch stage, these transistors are switched on and
amplification starts in the second stage. Gradually, the output voltage
increases and positive feedback system is activated and generates output level
of high and low voltage in the regeneration phase.
53

The first stage consumes power only when the parasitic capacitances at F
nodes are discharged. The Second stage dissipates power until the rail-to-rail
output voltage is generated.

6.2.1 Simulation Results of Two-Stage Latched Comparator

The comparator was simulated in Cadence to examine the functionality and
performance. The results of the simulation are presented in the following. In
this comparator, two inverters are added to recover the voltage level at the
output of the comparator. A NAND type SR latch is also used to hold the
comparator result during the reset phase. NAND type SR latch is shown in
Figure 6.7.

in1

out

in2
Figure6.7. NAND type SR latch

The transistors are sized to meet the requirements targeting minimum power
consumption. First, all the transistors are minimum size with double length to
mitigate the leakage problem. Then for improving the performance in terms of
offset voltage and delay, input transistors are chosen to be three times larger
than the minimum size. The optimized transistor sizes are presented in table
6.4.
In order to further decrease the leakage power while maintaining the speed of
the comparator, High VT transistors are employed occasionally.

54

Table6.4. Aspect ratio of the transistors

Component

Width(m) Length(m) Transistor type

Mtail

0.135

0.12

High VT

M1,2

0.405

0.18

Low VT

M3,4

0.135

0.12

Low VT

M5,6

0.135

0.12

High VT

M7,8(pmoslatch)

0.540

0.12

High VT

M9,10(nmoslatch)

0.270

0.12

Low VT

M11,12

0.135

0.12

Low VT

The comparator was simulated under VDD= 1V, Clock frequency= 100 kHz,
input frequency = 5 kHz with full swing, and temperature = 27C.

For measuring input referred offset voltage, Monte-Carlo simulation is run for
500 times and the same test bench as the previous case is used. Figure 6.8
shows the results of the Monte-Carlo simulation after performing postprocessing in MATLAB. Offset can be extracted from this curve.

Figure 6.8. Monte-Carlo Simulation Results

The simulation results are presented in table6.5.

55

Table6.5. Simulation results

Performance Metric
Value Unit
Power Consumption
257 pW
Input-Referred Offset voltage 7.2
mV
Propagation Delay
1.27
ns

Compared to the latch-only comparator, two-stage latch comparator exhibits

lower offset and higher speed while consuming almost the same power. As
mentioned before, further reduction of offset voltage is achievable by using
offset cancelation techniques at expense of higher power consumption and
more complexity.
Scalability with different supply voltages is verified by simulating the circuit
with different VDD. Power consumption and delay of the comparator versus
different supply voltages are given in table6.6.

Table6.6. Simulation results over the scaled supply voltage

VDD (V) Power Consumption (pW) Propagation Delay (ns)

1
0.8
0.6
0.4

257
199
144
94

1.27
2.7
19
657

As VDD decreases, power consumption decreases while propagation delay

increases. Compared to the latch-only comparator, this structure is functional
at VDD = 0.4 V and the propagation delay is 657 ns which is smaller than the
delay of the previous comparator.
This comparator shows a relatively constant offset voltage for different supply
voltages.

6.3

Modified Two-Stage Latched Comparator

In this architecture, two inverters are inserted between the two stages of the
conventional two-stage dynamic latched comparator in order to strengthen the
voltage signal at Di nodes providing higher regeneration speed. In addition to
the advantage of the conventional two-stage dynamic latched comparator such
as power efficiency, high speed, and low kickback noise, this architecture
provides lower input-referred offset. The comparator schematic is shown in
Figure 6.9 [23].

56

VDD

VDD

M18

M17

VDD

M11

M13

M12

M9

M14

M15

M16
M10

Vout+

VoutSW-

SW+
M7

M8

Di-

Di+
M6

M5

VDD

Di-

Di+

M3

Vin+

M4

M2

M1

Vin-

Clock
MTail

Figure 6.9. Transistor-level schematic of comparator2

In the reset phase when clock signal is low, PMOS transistors in the gain stage
are on (M3, M4) and charge the capacitances of Di nodes to VDD and
subsequently the Di nodes are discharged to ground, thus, there is no static
path and no static power dissipation during reset phase. The Di nodes are
discharged to ground and the PMOS transistors of the regeneration stage turn
on and charge the output nodes as well as regenerative nodes i.e. drain of the
NMOS transistors (M5 and M6) to VDD.
During evaluation phase when clock signal turns to high, the Di nodes
discharge through input and tail transistors to the ground with different rate,
depending on the input voltage. While Di nodes are discharged, Di nodes
start to charge from 0 to VDD with different rate. Once either of Di nodes
reaches Vth, the NMOS transistor (M5 or M6) in the second stage is switched
on, then the other transistor is also turned on.
As a result, latch is activated and regenerates the digital voltage at the output
from the small VDi.

57

6.3.1 Simulation Results of Modified Two-Stage Latched

Comparator
This comparator was simulated in Cadence. In this structure for recovering the
output voltage, two inverters are added at the output and like previous
comparator a NAND type SR latch is used to keep the comparator result
during reset phase.
Sizes of the transistors are optimized to meet the required specification. The
priority is the power consumption. The component sizes are presented in table
6.7.
In order to further decrease the leakage power while maintaining the speed of
the comparator, in some places High VT transistors are employed. Increasing
the sizes of the inverters between the stages, results in lower propagation
delay.
Table6.7. Aspect ratio of the transistors

Component Width(m) Length(m) Transistor type

Mtail

0.135

0.12

High VT

M1,2

0.405

0.18

Low VT

M3,4

0.135

0.12

Low VT

M5,6

0.135

0.12

High VT

M7,8

0.27

0.12

High VT

M9,10

0.135

0.12

Low VT

M11,12,15,16

0.135

0.12

Low VT

M13,14

0.54

0.12

High VT

M17,18

0.27

0.12

Low VT

The comparator was simulated under VDD= 1V, Clock frequency= 100 kHz,
input frequency = 5 kHz with full swing, and temperature = 27C.
For measuring input referred offset voltage, Monte-Carlo simulation is run for
500 times and same test bench is used as for the previous one. Figure 6.10
shows the results of the Monte-Carlo simulation after performing postprocessing in MATLAB. Offset can be extracted from this curve.

58

Figure6.10. Monte-Carlo Simulation Results

The simulation results are presented in table below:

Table6.8. Simulation results

Performance Metric
Value Unit
Power Consumption
1.2
nW
Input-Referred Offset voltage 5.2
mV
Propagation Delay
0.92
ns
In compare to the two-stage latch comparator, this architecture shows lower
offset and higher speed while consuming more power.
Scalability with different supply voltages is verified by simulating the circuit
with different VDD. Power consumption and delay of the comparator versus
different supply voltages are given in table 6.6.

Table6.6. Simulation results over the scaled supply voltage

VDD (V) Power Consumption (nW) Propagation Delay (ns)

1
0.8
0.6
0.4

1.2
0.529
0.274
0.130

0.92
2.1
15
382

59

This comparator also shows a relatively constant offset voltage for different
supply voltages. The propagation delay and power consumption of the
comparators are illustrated in the figures below:

Figure6.11. Propagation delay comparison

Figure6.12. Power consumption comparison

60

Since the priority is the power consumption, the last comparator consumes
more power while the first two comparator designs are ultra-low power, but
the two-stage comparator has lower offset, which makes it a good candidate
for the ADC in pacemaker.

61

Chapter 7

Implementation of SAR ADC and

Performance Evaluation
In this chapter, first the implementation of 10-bit SAR ADC in 65nm CMOS
process technology at 1kS/s and supply voltage of 1V is investigated followed
by the design of 10-bit charge redistribution DAC, then test bench and
simulation setup for verification of the performance is discussed and finally,
simulation results for low supply voltage as well as high temperature are
presented to evaluate their influence on power consumption.

7.1 Implementation of SAR ADC

First, the SAR A/D converter was simulated in high level model. Then, each
block was replaced with its transistor level design. Based on the results
obtained in the previous chapters the suitable architectures for the blocks were
chosen. In this work, a dynamic two-stage comparator (see Chapter 6 for more
details), a SAR controller containing a sequencer and a shift register, SAR
logic type 1(Chapter 4), and charge redistribution DAC were selected in order
to meet the low power requirements. In the following, the design of D/A
converter is provided.

7.1.1 Design of 10-bit D/A Converter

In this thesis a 10-bit charge-redistribution DAC with BWC array was
implemented in 65nm CMOS process. Figure 7.1 shows the block diagram of
the 10-bit DAC.
There are three phases of operation to perform a conversion:
First, in the sampling phase switch S1 connects the top plats of all capacitors
to the VCM and the bottom plates are connected to VIN. Thus, the input
voltage is sampled on the capacitor array.
During the hold phase, S1 is opened and the rest of switches connect the
bottom plates to ground therefore a charge of VIN + VCM is stored in the
capacitor array.

62

VCM
Top Plate Switch

VCM

DACOUT

512Cu

D9

256Cu

Cu

Dummy

D8

D0

Sample

Bottom Plate
Switch

VREF
VIN
GND

Figure7.1. DAC schematic view

In redistribution phase, the digital code determines the status of the switches
and the actual conversion is performed in this phase. In the beginning of the
conversion, D9 is high so the MSB capacitor is connected to VREF. At this step
the output voltage of the DAC is equal to VIN + VCM + 0.5VREF and is
compared to VCM. Based on the comparator result, D9 remains connected to
VREF if the comparator output is one, or change the connection to ground
when the result of the comparator is zero. Thus, the MSB is defined. Next, D8
is connected to VREF. Depending on the value of D9, Vout-DAC is VIN + VCM +
0.5D9.VREF + 0.25VREF and is compared to VCM. All the bits are generated
successively and Vout-DAC in the last step is defined as below:

(7.1)

Binary-Weighted Capacitor Array

The linearity of ADC is restricted by the linearity of the DAC which is caused
by the capacitor mismatch. Therefore, choosing an appropriate value for the
unit capacitance is vital. Reducing the unit capacitance value improves the
linearity but deteriorates the noise performance at the same time due to
thermal noise. The minimum value of the unit capacitor is limited by several
factors including

thermal noise, capacitor matching and the value of the

parasitic capacitances [11].

63

A unit capacitance of 20fF is chosen in this design. The values of the other
capacitors in the capacitor array are defined based on the unit capacitance.

Switches Implementation
The operation of the bottom plate switches are presented in table 7.1. The
block diagram of the switch can be obtained from the table.
Table7.1. Bottom switch operation

Sample Digital data VBottom_plate

1
x
VIN
0
1
VREF
0
0
GND
NMOS switches can properly pass a zero while PMOS switches can pass a
strong one. On the other hand, CMOS transmission gate combined both
features of NMOS and PMOS switch and is capable of properly passing both
zero and one. It also benefits from low on-resistance [25].
In this design, VREF is set to VDD; therefore PMOS switch is used for VREF.
NMOS switch is employed to ground the bottom plate. Since Vin varies from
0 to 1, CMOS TG is chosen for Vin. Figure 7.2 (a) shows the block diagram of
bottom plate switches [12].

64

M3

M4
Sample

Data

VREF

Sample

Samplebar

VIN

Data
Samplebar

a)
Samplebar

VCM
M1

M2

b)

Figure7.2. Schematic views of a) bottom plate switch and b) top plate switch

The leakage current contribution of the top plate switch is significant in this
low speed design since most of the time this switch is off and it only turns on
during the sampling phase. The leakage current of the top plate switch
adversely affect the linearity of the DAC and consequently the linearity of
ADC [13]. In order to alleviate this problem, a stack of two PMOS switches
are used in series; this switch is depicted in Figure 7.2 (b).The sizes of the
switches are presented in table 7.2.
Table 7.2. Sizing of the switches

Transistor

W(m)

L(m)

M3
M4
TG_NMOS
TG_PMOS
Inv_NMOS
Inv_PMOS
NOR_NMOS
NOR_PMOS
NAND_NMOS
NAND_PMOS
M1,2

0.135
0.54
0.135
0.54
0.135
0.45
0.135
0.81
0.27
0.45
0.135

0.06
0.12
0.06
0.06
0.12
0.12
0.12
0.12
0.12
0.12
0.06

65

7.2 Performance Evaluation

Figure 7.3 represents the test bench for ADC performance measurement. As
indicated in the figure, the 10-bit output data is stored in a Verilog-A file.
Then, the stored data is read in MATLAB and post-processing operations are
performed to measure the dynamic performance of ADC including SFDR,
SINAD and ENOB.

VDD

VCM
FN

FP

INP

INN
CL
K

VDD

SN

CLK

SP

CLK

Top Plate Switch

VCM
512Cu 256Cu

2Cu

Cu

Cu

Reset
CLK

Latch

VREF
VIN

Result

GND

Data
Load
D9

D8 D7 D6 D5 D4 D3 D2 D1 D0

Figure7.3. SAR ADC test bench for measuring dynamic performance

7.2.1 Power Consumption measurement

The ADC is simulated with VDD= VREF = 2.VCM = 1V, and sampling frequency
of 1kS/s. The input signal is a full swing sinusoidal with fin= 450Hz. The
simulation results under 27C and 80C temperatures are presented in table
7.3.

66

Table7.3. Power consumption of different blocks of ADC

DAC

Power Consumption at
27C (nW)
10.4

Power Consumption at
80C (nW)
11.1

SAR
Comparator
Clock Power
Total

1.54
0.26
0.208
12.408

2.028
0.620
0.224
13.97

Block

The total power consumption of the implemented SAR ADC including the
clock power is almost 12.4nW. The distribution of power consumption
between different blocks of ADC is shown in figure 7.4.

Figure7.4. Power distribution

As shown in the Figure 7.4, DAC consumes the largest amount of power
among other blocks which is %84. As discussed above, the unit capacitance in
the DAC is chosen to be 20 fF. After DAC, SAR control logic with %12,
clock power and comparator both with %2 consumes the largest amount of
power respectively.
In order to testify the voltage scalability of the ADC, it is simulated with
scaled supply voltage. Table 7.4 presents the power consumption of ADC
blocks. The ADC can properly operate with scaled supply voltage down to
0.5V and consumes a minimum power of 6.28nW.

67

Table 7.4. Power consumption (nW)

Block

VDD=1V VDD=0.8V VDD=0.6V VDD=0.5V

DAC

10.4

7.6

6.12

5.44

SAR
Comparator
Clock Power
Total

1.54
0.26
0.208
12.408

1.1
0.2
0.157
9.057

0.79
0.15
0.109
7.17

0.64
0.11
0.085
6.275

7.2.2 Dynamic Performance Evaluation

As described in Chapter 2 the dynamic performance is evaluated by
calculating the SFDR, SINAD, and ENOB of ADC. For this purpose, a full
swing sinusoidal wave with fin = 30.273438 Hz which is based on coherent
sampling is applied to the input of ADC. The simulation is performed to
achieve 1024 samples with sampling frequency of 1kS/s. Then, Fast Fourier
Transform of the stored output data is executed in MATLAB and by
performing some post processing the SFDR, SINAD, and ENOB are
measured.

68

Figure7.5. FFT of the ADC output for Fin=30.273438Hz

The FFT of 10-bit ADC output is shown in Figure7.5. The simulation results
predict that the ADC have SINAD=60.29, SFDR=79.89, and achieves 9.72 of
ENOB which are reasonable for schematic level simulation.
Energy per conversion-step can be calculated using FOM definition of ADC
which is given by Equation 7.2 [16].
(fJ/conversion-step)

(7.2)

Table 7.5 summarize the performance parameters of the designed SAR ADC.

69

Table 7.5. Performance parameters of ADC

Performances

Result

Unit

Process Technology
Supply Voltage
Resolution
Sampling Frequency
Power Consumption
SINAD
SFDR
ENOB
FOM

65nm
1
10
1
12.4
60.29
79.89
9.72
14.7

V
bits
kS/s
nW
dB
dB
bits
fJ/conversion-step

Finally, a result comparison with other works is presented in table 7.6.

Table 7.6. Comparison with other published works

Performances
Technology
Supply Voltage (V)
Resolution
Sampling
Frequency(S/s)
ENOB
Power
Consumption(W)
FOM (fJ/conversionstep)

This
work
65nm
1
10

[30]

[26]

[31]

[16]

0.18m
1
12

90nm
1
9

0.18m
1
8

65nm
1
10

1k

100k

20M

400k

1M

9.72

10.55

7.8

7.31

8.75

12.4n

25

290

6.15

1.9

14.7

167

65

97

4.4

70

Chapter 8

Summary
This thesis presents implementation of a 10-bit SAR ADC operating at 1kS/s
and supply voltage of 1 V in 65nm CMOS technology. The power
consumption of 12.4nW is achieved. The ADC employs a chargeredistribution DAC, a dynamic two-stage comparator, and a SAR control logic
containing a sequencer and a ring counter. The ADC exhibits good
performance and achieves an FOM of 14.7fJ/conversion-step with ENOB of
9.72 bit.
In this work, after a deep study on different possible structures of SAR logic,
they are implemented and compared in terms of power consumption and
speed. Comparison results obtained in Chapter 4 indicate that the designed
conventional SAR logic with a sequencer and a ring counter, consumes the
lowest power of 1.2nW at 1kS/s. Thus the power consumption of the SAR
control logic is significantly reduced and consumes only %12 of the total
power.
Designing the comparator is a crucial part of ADC design. In this work,
comparator performance metrics as well as several types of comparators are
studied, such as open loop comparator, pre-amplifier preceding a latch
comparator, and dynamic comparator. Based on these studies, dynamic
comparators consume lower power compared to the other approaches.
Therefore, diverse architectures of dynamic comparators are implemented and
compared regarding power consumption, speed, and accuracy. Consequently,
the dynamic two-stage comparator is selected to be used in the designed ADC.

71

References
[1] J. Werner, M. Meine, K. Hoeland, M. Hexamer, and A. Kloppe,Sensor
and control technology for cardiac pacing, Transactions of the Institute
of Measurement and Control, 2000.
[2]

L. S. Y. Wong, S. Hossain, A. Ta, J. Edvinsson, D. H. Rivas, and H.

Ns. A Very Low-Power CMOS Mixed-Signal IC for Implantable
Pacemaker Application, IEEE Journal of Solid-State Circuits, vol. 39,
no. 12, Dec 2004.

[3] P. Lowenborg, Mixed-Signal Processing Systems, Linkoping University,

2006.
[4] Kent. H. Lundsberg, Analog-to-Digital Converters Testing, 2002.
[5] H.Khurramabadi ADC Converters (Lecture 13). UC Berkeley Course,
Analog-Digital Interfaces in VLSI Technology EE247. 2006
[6] B. Razavi, Principles of Data Conversion System Design, WileyInterscience,
IEEE Press, 1995.
[7] J. L. McCreary and P. R. Gray, All-MOS Charge Redistribution Analogto-Digital Conversion Techniques-Part I, IEEE Journal of Solid-State
Circuits, vol. SC-10, no. 6, December 1975.
[8]

A. Rodriguez-Perez, M. Delgado-Restituto, and F. Medeiro, Power

efficient ADCs for Biomedical Signal Acquisition, IMSE-CNM and
University of Seville.

[9] Y. S. Yee, L. M. Terman, and L. G. Heller, A Two-Stage Weighted

Capacitor Network for D/AA/D Conversion, IEEE Journal of SolidState Circuits, vol. SC-14, pp. 778781, August 1979.
[10]

L. Cong, Pseudo C-2C Ladder-Based Data Converter Technique,

IEEE Transactions on Circuits and Systems, vol. 48, no. 10, October
2001.

[11] M. Saberi, R. Lotfi, K. Mafinezhad, W. A. Serdijn, Analysis of Power

Consumption and Lineary in Capacitive Digital-to-Analog Converters
Used in Successive Approximation ADCs, IEEE Transactions on
Circuits and Systems, vol. 58, no. 8, August 2011.
72

[12] D. Zhang, Design and Evaluation of an Ultra-Low Power Successive

Approximation ADC, Master thesis, Linkoping University, 2009.
[13] D. Zhang, A. K. Bhide, and A. Alvandpour, Design of CMOS Sampling
Switch for Ultra-Low Power ADCs in Biomedical Application, in
proceedings of the 28th Norchip conference IEEE,Nov.2010.
[14] T.O. Anderson, Optimum Control Logic for Successive Approximation
Analog-to-Digital Converters, Computer Design, vol. 11, no. 7, pp. 8186, 1972.
[15]

A. Rossi and G. Fucili, Nonredundant successive approximation

register for A/D converter, Electronics Letters, vol.32, no.12, pp.10551057,1996.

[16] M. Van Elzakker, E. van Tuijl, P. Geraedts, D. Schinkel, E. Klumperink

and B. Nauta, A 1.9W 4.4fJ/Conversion-step 10b 1MS/s ChargeRedistribution ADC, ISSCC Dig. Tech. Papers, Feb. 2008.
[17] S. T. Oskuii, Comparative Study on Low-Power High-Performance
Flip-Flops, Master thesis, Linkoping University, 2003.
[18] Pedro M. Figueiredo and Joao C. Vital, Kickback Noise Reduction
Techniques for CMOS Latched Comparators, IEEE Transactions on
Circuits and Systems, vol.53, no.7, July 2006.
[19] C. K. Yang, Delay-Locked Loops An Overview, Phase-Locking in
High-Performance Systems, IEEE Press, 2003.
[20] W. J. Dally and J. W. Poulton, Digital Systems Engineering, PhaseLocking in High-Performance Systems, Cambridge University Press,
1998.
[23] H. Jeon, Low-Power high-speed low-offset fully dynamic CMOS
latched comparator, Northeastern University, 2010.
[24] A. Graupner, A Methodology for Offset Simulation of Comparators,
The Designer Guide Community, Oct. 2006.
[25] J. M Rabaey, A. Chandrakasan , and B. Nikolic, Digital Integrated
Circuits A Desigsn Perspective, Prentice Hall, 2nd edition, 2003.

73

[26]

J. Cranninckx and G. Van der Plas, A 65fJ/conversion-step 0-to50MS/s 0-to-0.7mW 9b Charge-Sharing SAR ADC in 90nm Digital
CMOS, ISSCC Dig.Tech.Papers, Feb. 2007.

[27] L. Li and J. Hu, A Transmission Gate Flip-Flop Based on DualThreshold CMOS Techniques, Circuits and systems, IEEE Press, Aug.
2009.
[28] H.Khurramabadi ADC Converters (Lecture 21). UC Berkeley Course,
Analog-Digital Interfaces in VLSI Technology EE247. 2006
[29] S.H. Yang, K.S. Lee, S. Kim, Y.M. Lee, Charge-redistribution DAC
with double bit processing in single capacitor, Electronic Letters, vol.
45, no. 5, March 2011.
[30] N. Verma and A. P. Chandrakasan, A 25uW 100kS/s 12b ADC for
Wireless Micro-Sensor Applications, IEEE ISSCC, pp. 222-223,
February 2006.
[31] H-C. Hong, and G-M. Lee, A 65-fJ/Conversion-Step 0.9-V 200-kS/s
Rail-to-Rail 8-bit Successive Approximation ADC, IEEE Journal of
Solid-State Circuits, vol. 42, no. 10, October 2007.

74