PHASE LOCKED LOOP

(Design and Implementation)

A Project Report

submitted by

SNEHIL VERMA (14700)

in partial fulfilment of the requirements

for the award of the degree of

BACHELOR OF TECHNOLOGY

DEPARTMENT OF ELECTRICAL ENGINEERING

INDIAN INSTITUTE OF TECHNOLOGY KANPUR.
APRIL 2017
 ACKNOWLEDGEMENTS

My sincere thanks to Prof. S. Qureshi for mentoring me and making sure that I achieved
practical results. He helped me learn focusing on the core ideas, and completely under-
standing the results. I would also like to thank Prof. S.S.K. Iyer and Prof. B. Mazhari,
who came to my talk, asked crucial questions, and gave advice on further improvements
of the design.

i
 ABSTRACT

KEYWORDS: Phase Locked Loop (PLL) ; Phase Frequency Detector (PFD) ;

Charge Pump (CP) ; Low Pass Filter (LPF) ; Current Starved Volt-
age Controlled Oscillator (CSVCO) ; Frequency Divider (FD)

This thesis presents a design for clock generating circuitry using PLL techniques. A
simple design of CPPLL is followed by design of linear CSVCO. Feedback is provided
through a divide-by-2 frequency divider. The reference signal is 4 MHz square wave
from a crystal oscillator and the technology used is 180 nm (SCL PDK). The design is
optimised for low area and low power consumption.

ii
 CONTENTS

ACKNOWLEDGEMENTS i

ABSTRACT ii

1 INTRODUCTION 1

2 Phase Frequency Detector & Charge Pump 2

2.1 PFD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Charge Pump . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3 Current Starved Voltage Controlled Oscillator & Frequency Divider 7

3.1 CSVCO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.2 Divide-by-2 Frequency Divider . . . . . . . . . . . . . . . . . . . . 11

4 Stability Analysis (LPF) 12

4.1 Transfer Function of VCO . . . . . . . . . . . . . . . . . . . . . . 12
4.2 Transfer Function of PFD/CP/LPF combination . . . . . . . . . . . 12
4.3 Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

5 RESULTS 15

6 CONCLUSION 17
 CHAPTER 1

INTRODUCTION

Concept of PLL was introduced in 1930s and since then the scope of PLL application
has attracted many designers. With the advancement of the technology new designs,
new problems and non idealities are emerging and hence the motivation to work in this
field. PLL has vast application in the area of electronics and communication. PLL
can be used for clock generation for a microprocessor, as a frequency synthesizer in a
mobile, etc. In this thesis we would focus on the clock generation aspect of the PLL.

Figure 1.1: Block Diagram

The basic structure of the PLL can be understood from the block diagram above.
The following chapters will explain each of these components, but the main focus will
be on their design aspects.
 CHAPTER 2

Phase Frequency Detector & Charge Pump

2.1 PFD

PLL is a feedback system that keeps the input signal aligned, w.r.t phase, to the refer-
ence signal. There are various implementations of PLL. Here we choose to implement
Charge-Pump PLL as type I PLLs have tight trade-offs (between ωLP F and ζ, the damp-
ing ratio) and have limited acquisition range. In order to remedy the acquisition prob-
lem, frequency comparison is also done in addition to the phase detection, as shown
in Fig 2.1. For the periodic signals it is possible to merge both, frequency and phase,
feedback loops and this role is played by Phase Frequency Detector.

Figure 2.1: Aided Acquisition using Frequency Detection

The operation of PFD can be easily understood from Fig 2.2. Lets assume Qa and
Qb are zero initially. Thus when input A rises (before input B), this results in Qa being
high but Qb still low. The circuit will remain in this state until B rises, at which point
Qa will return to zero and Qb will still be zero. Similar goes for Qb and Qa when input
B rises before input A. It can be easily observed that in Fig 2.2 (a) Qa represents the
phase difference (φA - φB ) between the inputs whereas in Fig 2.2 (b) Qa represents the
frequency variation (ωA - ωB ) between the inputs.
 Figure 2.2: PFD Operation

PFD can be implemented as shown in Fig 2.3 i.e. two edge triggered resettable D
flip flops with their D inputs tied to logic one.

Figure 2.3: PFD Implementation Figure 2.4: PFD at Gate Level

Figure 2.5: PFD Realisation

3
2.2 Charge Pump

Since the difference between the average values of Qa and Qb is of interest to us, there-
fore the outputs of Qa and Qb can be directly passed through a LPF and sensed differen-
tially, as shown in Fig 2.6. However, a more conventional way is to introduce a Charge
Pump between a PFD and LPF.

Figure 2.6: PFD followed by LPF

A charge pump consists of two switched current sources that may pump charge in or
out of the loop filter according to the logic states of Qa and Qb . I1 and I2 are generally
equal and referred as Up and Down currents respectively.

Figure 2.7: PFD along with Charge Pump

Charge pump can be easily implemented using MOSFETs as shown on the Fig 2.8
(a). Here M 1 and M 2 operates as current sources, and M 3 and M 4 operate as switches.
An immediate drawback of this implementation can be seen as the inputs to the charge
pump are reaching CP at different times, due to the delay caused by the inverter. This
error can suppressed by inserting a transmission gate between Qb and the gate of M 3,
equalizing the delays [Fig 2.8 (c)].

4
Figure 2.8: (a) Implementation of Charge Pump (b) Effect of skew between Qa
and Qb (c) Suppression of skew by transmission gate

5
 The design presented above can be further improved to tackle the problem of Boot-
strapping. This problem originates due to finite capacitance seen at the drains of the
current sources. Suppose, as illustrated in Fig 2.9, S1 and S2 are off. This allows CX
to discharge to ground and CY to charge to VDD . In the next instant, when both S1 and
S2 are on, VX ≈ VY ≈ Vcont (assuming the voltage drop across S1 and S2 to be zero).
If the phase error is zero and ID1 = ID2 , even if CX = CY , change in VX and VY will
not be same and would result in sudden jump of Vcont , which should remain constant
ideally. Therefore this creates a problem. The design solution of this problem will not
be discussed here.

Figure 2.9: Bootstrapping

6
 CHAPTER 3

Current Starved Voltage Controlled Oscillator &

Frequency Divider

3.1 CSVCO

Introduction

An oscillator produces periodic output, generally in the form of voltage. Most ap-
plications require to control the output frequency of the oscillator. Here is where VCO
comes into picture. VCO provides us the capability to control the frequency of the
oscillator by tuning its input voltage i.e VCO is a voltage to frequency converter.

Design and Operation

VCO is designed according to the design parameters mentioned in Table 3.1.

Table 3.1: VCO Design Specifications

Power Supply (VDD ) 1.8 V

Centre Frequency 8 MHz
Inverter Delay 2.18ns
Technology 180nm (SCL PDK)

There are several architectures for designing VCO. Here we choose to design and
analyze Current Starved VCO. Current Starved architecture is preferred as it is less
sensitive to the voltage variation in VDD . The basic structure of VCO is designed using
Ring Oscillator and further to implement the current starved architecture current mirrors
are used.
Figure 3.1: Building block of Ring Oscillator (with Current Mirrors)

Figure 3.2: Schematic of CSVCO

8
 The operation of CSVCO is similar to that of a ring oscillator. MOSFETs M2 and
M3 operate as inverter while M1 and M4 operate as current sources, as illustrated in
Fig 3.2. The current through M5 and M6 is mirrored into each inverter stage and is
controlled by VinV CO . Hence the inverter stages are starved for current.

Frequency of Oscillation

Total capacitance at the drains of M2 and M3 is given by

Ctot = Cout + Cin (3.1)

′ 3 ′
Ctot = Cox (Wp Lp + Wn Ln ) + Cox (Wp Lp + Wn Ln ) (3.2)
2
5 ′
Ctot = Cox (Wp Lp + Wn Ln ) (3.3)
2
where Cout and Cin simply represent the input and output capacitances of the inverter.
The time taken to charge Ctot for zero to VSP (switching point of the inverter) with
constant current ID4 is given by
VSP
t1 = Ctot (3.4)
ID4

While the time taken to discharge Ctot from VDD to VSP is given by
VDD − VSP
t2 = Ctot (3.5)
ID1

Since ID1 = ID4 = ID , therefore

Ctot .VDD
t1 + t2 = (3.6)
ID

Hence the oscillation frequency of a CSVCO with N (odd >3) number of inverter stages
is
1 ID
fosc = = (3.7)
N (t1 + t2 ) N.Ctot .VDD

Linearization

Many applications require VCO to be linear i.e. output frequency is linearly related
to the input voltage. This can implemented in the following manner:

From Eq. (3.7) it can be observed that fosc linearly varies w.r.t ID . Therefore it is
required that the current flowing through the MOSFETs varies linearly w.r.t the input
voltage of VCO. This can be achieved by designing as shown in the Fig 3.3. The width
of M5R is made wide so that its VGS is always approximately VT N . This ensures that

9
drain current is linearly proportional to VinV CO .

Figure 3.3: Linearizing VCO gain

Simulation and Results

Full schematic of the Current Starved VCO is shown in Fig 3.4. Fig 3.5 represents
the linearity of the CSVCO.

Figure 3.4: Current Starved VCO

10
 Figure 3.5: VinV CO vs Output frequency

The CSVCO is designed and simulated using SCL PDK 180nm CMOS Technology.
The voltage supply (VDD ) is taken to be 1.8V. The output frequency ranges from 0 MHz
to 16.172 MHz, centred around 8 MHz, linearly w.r.t the input of VCO (VinV CO ). The
maximum power consumption by CSVCO is 181.161µW.

3.2 Divide-by-2 Frequency Divider

Divide-by-2 frequency divider is used to provide feedback form CSVCO to PFD. It is

simply implemented using D flip flops, as shown in the Fig 3.6. This structure reduces
the time jitter of the VCO.

Figure 3.6: Divide-by-2 Frequency Divider

11
 CHAPTER 4

Stability Analysis (LPF)

4.1 Transfer Function of VCO

Ideally VCO should be capable of transferring input voltage to output frequency lin-
early.
ωout = ω0 + KV CO Vcont (4.1)

Here, ω0 represents the intercept corresponding to Vcont = 0 and KV CO denotes the

gain of the VCO (rad/s/V). The achievable range ω2 − ω1 , is called tuning range of the
VCO.

Figure 4.1: Ideal VCO transfer function

Due to the non ideal behaviour, the transfer function of VCO is modelled as

φout KV CO
= (4.2)
Vcont s

4.2 Transfer Function of PFD/CP/LPF combination

The combination of PFD, CP and LPF (as shown in Fig 2.7) doesn’t represent a linear
system however we model their behaviour linearly assuming the ramp approximation,
as shown in Fig 4.2.

Figure 4.2: Step response of PFD/CP/LPF combination (ramp approximation)

Let us assume the period of the input is Tin and the charge pump provides the current
of ±IP . As shown in the Fig 4.2, initial phase difference is zero and at t = 0 step phase
of φ0 is applied to input B i.e. ∆φ = φ0 u(t). As a result QA continues to produce
pulses of width φ0 Tin /2π. This resulted in raising of output voltage by (IP /CP ) times
the QA pulse width. Therefore

IP
Vout (t) = t.φ0 u(t) (4.3)
2πCP
Impulse response is therefore given by its derivative

IP
h(t) = u(t) (4.4)
2πCP

Hence the transfer function of the combination is given by

Vout IP 1
(s) = (4.5)
∆φ0 2πCP s

4.3 Stability Analysis

Now let us construct the linear model of the charge pump PLL. This would yield the
open loop transfer function as

φout IP KV CO
(s) = (4.6)
φin 2πCP s2

13
 Figure 4.3: Linear model of CPPLL

It can be easily observed that the the closed loop response will have two imaginary
poles which would result in instability of the system. Hence to stabilize the system it
is suggested to add a resistor RP in series with the capacitor CP in the low pass filter.
But since the charge pump drives the series combination of RP and CP , each time a
current is injected in the loop filter, the control voltage experiences a large jump. Hence
to minimize the ripple effect, a small capacitor C2 (∼ 0.2 - 0.1 times CP ) is added in
parallel to their series combination.

Hence the final LPF is modified as shown in Fig 4.4 in order to attain and enhance
stability.

Figure 4.4: Improved LPF to attain stability and reduce ripple in VinV CO

14
 CHAPTER 5

RESULTS

The full schematic of the PLL is shown in Fig 5.1.

Figure 5.1: Complete PLL schematic

The PLL is working as expected and producing an output frequency of 8 MHz.

Settling time of the PLL (±5%) is around 33µs and Lock time is around 125µs which
is practical.

Figure 5.2: Waveforms observed from the PLL

Figure 5.3: Output Frequency

16
 CHAPTER 6

CONCLUSION

The PLL circuit is designed and simulated using SCL PDK 180nm CMOS Technology.
The voltage supply (VDD ) is taken to be 1.8V. The output frequency of the CSVCO
ranges from 0 MHz to 16.172 MHz, centred around 8 MHz, linearly w.r.t the input
of VCO (VinV CO ). Maximum power consumption by CSVCO is 181.161µW. After
performing stability analysis, the values of RP , CP and C2 , used in the LPF, are taken
to be 1kΩ, 40pF and 4pF respectively. The PLL is stable and is producing an output
frequency of 8 MHz. Settling time of the PLL (±5%) is around 33µs and Lock time is
around 125µs.
 REFERENCES

1. Mr. Madhusudan Kulkarni and Mr. Kalmeshwar N Hosur, “Design of a Linear

and Wide Range Current Starved Voltage Controlled Oscillator for PLL”, Inter-
national Journal on Cybernetics & Informatics ( IJCI), Vol.2, No.1, (2013).

2. Floyd M Gardner, “Charge-Pump Phase-Lock Loops”, IEEE Transactions on

Communications, Vol. COM-28, No.11, (1980).

3. Deog-kyoon Jeong, Gaetano Borriello, David A. Hodges And Randy H. Katz,

“Design of PLL-Based Clock Generation Circuits”, IEEE Journal of Solid-State
Circuits, Vol. SC-22, No.2, (1987).

4. Behzad Razavi, “Design of Analog CMOS Integrated Circuits”, McGraw Hill

International Edition, (2001).

18