Phase-Locked Loop: Budapest University of Technology and Economics Laboratory 2 - Experiment 9

Budapest University of Technology and Economics
Laboratory 2 - Experiment 9
PHASE-LOCKED LOOP
Basic loop conguration Operation principle of phase-locked loop Linear operation of the PLL
Loop equations and nonlinear baseband model Linear baseband model Transfer functions PLL with active loop lter (Most commonly used PLL conguration) Stability considerations
An example for PLL application: Coherent FM demodulator

Dept. of Measurement and Information Systems G eza KOLUMBAN 1
References for phase-locked loop: [1] G. Kolumb an, Phase-Locked Loops Article in The Encyclopedia of RF and Microwave Engineering, K. Chang, (Ed.), vol. 4, pp. 37353767, Wiley, New York, 2005. [2] G. Kolumb an, Phase-Locked Loops Article in The Encyclopedia of Electrical and Electronics Engineering, J. G. Webster, Ed., vol. 16, pp. 158188, Wiley, New York, 1999. Phase-locked loop is one of the most commonly used circuit in both telecommunication and measurement engineering. Depending on the operation principle of loop components we distinguish Analog Digital Hybrid phase-locked loops. Only the analog phase-locked loop (APLL) is discussed in this course. For the sake of simplicity, we will call this circuit PLL
ANALOG PHASE-LOCKED LOOP Circuit conguration: Phase detector (PD) is an analog multiplier All loop components are analog circuits Mathematical model: Operation of analog phase-locked loop is modeled by an ordinary dierential equation Conditions: For the sake of simplicity, only the noise-free case is studied here We assume that the only source of nonlinearity is the phase detector, the other loop components are assumed to be linear
Dept. of Measurement and Information Systems G eza KOLUMBAN
BASIC LOOP CONFIGURATION PLL block diagram showing inputs and outputs for various applications

A few important applications: Demodulation of FM and PM signals FM modulator Carrier recovery
The PLL is a nonlinear feedback system that tracks the phase of input signal The basic PLL conguration contains a Phase detector (PD) Time-invariant linear loop lter and Voltage-controlled oscillator (VCO); the oscillator to be synchronized
QUALITATIVE CHARACTERIZATION OF LOOP COMPONENTS Phase detector (PD): Analog multiplier PD produces an error signal that is proportional to the phase error, i.e., to the dierence between the phases of input and output signals of the phase-locked loop Loop lter: Low-pass lter It is characterized by its transfer function F (s) Low-pass lter suppresses the noise and unwanted PD outputs. It determines the dynamics of phase-locked loop Voltage-controlled oscillator (VCO): VCO generates a sinusoidal signal The instantaneous VCO frequency is controlled by its input voltage
OPERATION PRINCIPLE OF PHASE-LOCKED LOOP Part I Basic loop conguration PLL block diagram

Voltages appearing in the loop are also shown

Phase detector (PD) compares the phase of the input signal s(t, ) against the and produces an error signal vd(t) phase of the VCO output r (t, ) This error signal is then ltered, in order to remove noise and other unwanted components of the input spectrum The sum of lter output vf (t) and an additive external control voltage ve(t) controls the instantaneous VCO frequency
OPERATION PRINCIPLE OF PHASE-LOCKED LOOP Part II Basic loop conguration PLL block diagram

Voltages appearing in the loop are also shown

A nonzero output voltage must be provided by the PD, in order to tune the VCO frequency to the input one if the input frequency diers from the VCO center frequency Consequently, the PLL tracks the phase of input signal with some phase error. However, this phase error can be kept very small in a well-designed PLL
IMPORTANT PLL CHARACTERISTICS Part I Acquisition and Tracking In every application, the PLL tracks the phase of the input signal. However, before a PLL can track, it must rst reach the phase-locked condition In general, the VCO center frequency 0 diers from the frequency i of the input signal Therefore, rst the VCO frequency has to be tuned to the input frequency by the loop. This process is called frequency pull-in Then the VCO phase has to be adjusted according to the input phase. This process is known as phase lock-in Both the frequency pull-in and phase lock-in processes are parts of acquisition which is a highly nonlinear process and is very hard to analyze After acquisition the PLL achieves the phase-locked condition, where the PLL tracks the input phase. Under this phase-locked condition, the VCO frequency is equal to the input frequency
Pull-in Range P = |i 0| is the maximum initial frequency dierence between the input and VCO center frequencies both in positive and negative directions, for which the PLL eventually achieves the phase-locked condition. The pull-in range is related to the dynamics of the PLL Lock-in Range L = |i 0| is the frequency range over which the PLL achieves the phase-locked condition without cycle slips, i.e., < e(t) < during the entire lock-in process Hold-in Range Suppose the phase-locked condition has been achieved in the PLL. Now vary the input frequency i slowly and the VCO frequency will follow it. The hold-in range H = |i 0| is determined by the lower and upper values of i, for which the phase-locked condition is lost. The hold-in range represents the maximum static tracking range and is determined by the saturation characteristics of the nonlinear loop elements of the PLL
LOOP EQUATIONS AND NONLINEAR BASEBAND MODEL PLL block diagram

F (s) denotes the transfer function of the loop lter In order to write the dierential equations in compact form, the operation of dierentiation d/dt in the time domain will be denoted by the multiplication of the Heaviside operator p Note, the Heaviside operator is valid in the time domain, while s denotes the complex frequency . If the transfer function F (s) of a linear network is given in the complex frequency domain s then the transfer function in operator form may be expressed as F (p) = F (s)|s=p
DEVELOPMENT OF LOOP EQUATIONS In the equations to be developed, the time variable t is suppressed for conciseness where it does not cause misunderstanding Input signal Let the phase (t) of input signal s(t, ) be expressed with respect to the VCO center frequency 0 as = 0t + i Then the input signals becomes s(t, ) = 2A sin = 2A sin(0t + i)
where A(t) describes the amplitude modulation of input signal and i(t) is the input phase modulation, i.e., the PM of the input signal. Note that i(t) also incorporates the input frequency error i = i 0
Output signal of voltage controlled oscillator (VCO) VCO output is the output signal of phase-locked loop. Since the VCO phase t) tracks the phase (t) of input signal we call it loop estimate of (t). It is ( expressed with respect to the VCO center frequency as = 0t + o Then the VCO output is obtained as = r (t, ) = 2Vo cos(0t + o) 2Vo cos
In the above equations, o(t) and Vo denote the phase and rms amplitude of VCO output, respectively
12
Transfer function of voltage controlled oscillator (VCO) The frequency of a voltage controlled oscillator is determined by the VCO control voltage vc(t). The instantaneous VCO frequency referenced to 0 varies linearly with the control voltage vc(t) d d do 0 = [0t + o(t)] 0 = Kv vc dt dt dt where Kv is the VCO gain in rad . Note if vc (t) = 0 then the VCO frequency is Vs equal to the center frequency 0
13
Transfer function of phase detector (PD) Block diagram of a phase detector

Note a PD consists of An analog multiplier

A low-pass lter
The analog multiplier in the PD multiplies the input signal = 2Vo cos(0t + o) s(t, ) = 2A sin(0t + i) and VCO output r (t, ) and produces both the dierence- and sum-frequency terms. The low-pass lter eliminates the sum-frequency component. The PD output is obtained as } = AVo sin(i o) = AVo sin e = KdA sin e vd = F ILT {s(t, )r (t, ) where the phase error is dened by and Kd = Vo, a dimensionless quantity, is the gain of PD
e(t) = i(t) o(t)
Properties of phase detector vd = KdA sin e Phase detector is a nonlinear device Its output depends on the dierence of input and VCO phases Its output also depends on A(t), i.e., on the AM of input signal Loop lter and adder The VCO control voltage vc(t) is the sum of the loop lter output vf (t) and external control voltage ve(t) vc(t) = vf (t) + ve(t) = F (p)vd(t) + ve(t) where F (p) = F (s)|s=p and p =
d dt
is the Heaviside operator

15
Equations we obtained up to this point

do = Kv vc = po = Kv vc dt vc = F (p)vd + ve vd = KdA sin e e = i o
LOOP EQUATIONS
o = = Kv Kv F (p) Kv Kv KdF (p) Kv vc = vd + ve = A sin e + ve p p p p p Kv KF (p) A sin e + ve p p
where K = KdKv denes the loop gain in rad/Vs

e = i o = i Kv KF (p) A sin e ve p p
16
Loop equations e = i o KF (p) Kv Kv o = A sin e + ve = [F (p)KdA sin e + ve] p p p
NONLINEAR BASEBAND MODEL

Recall:
vd vf vc
= = =
KdA sin e F (p)vd vf + ve
17
Properties of nonlinear baseband model
Real input and output signals:
2A sin(0t + i) s(t, ) = r (t, ) = 2Vo cos(0t + o)
Note: Baseband model contains only low-pass signals because the carrier has been removed Input and output signals of baseband model are the input i and output o phase modulations Real input and output signals do not appear in the baseband model they have to be calculated from i and o Since the VCO can generate only angle modulated signals, only angle modulated signals can be produced by the PLL Because of the nonlinear PD characteristic, this model is nonlinear, consequently, its analysis must be performed in time domain. Transfer function concept may not be used
LINEAR OPERATION OF PLL The linear operation of PLL assumes that Phase-locked condition has been achieved and is maintained Phase error remains in the neighborhood of its quiescent value, i.e. we may linearize the PLL using the small-signal approximation Mathematical background of linearization: Taylor series representation Steps of linearization 1. Determination of the quiescent point 2. Approximation of nonlinear characteristic by its tangent (Linear term in the Taylor series approximation)
19
Determination of quiescent point If a PLL operates in steady-state and all its input signals are constant then the PLL is operating in the quiescent point Let the PLL loop equation rearranged as
e = i o = i KF (p) Kv A sin e ve p p pe = pi KF (p)A sin e Kv ve
Under steady-state conditions, all signals are constant, but a constant input frequency error may be present
e (t) ve (t) i (t) = = = ss ve0 (i 0)t + i0 = i t + i0
Since the Heaviside operator means derivation d/dt in the time domain we get
0 = i KF (0)A sin ss Kv ve0
0 = i KF (0)A sin ss Kv ve0
From which the quiescent point ss of PLL is obtained as

ss = sin1 i Kv ve0 KF (0)A
where F (0) is the dc gain of loop lter Note: To get the quiescent point, a nonlinear dc analysis had to be performed To get the best system performance, the quiescent value of phase error has to be set to zero, i.e.,
ss = 0
It can be achieved if the dc gain of loop lter goes innite F (0) . This conditions may be satised by the most commonly used active loop lter. In the remaining part of discussion we assume that an active loop lter is used
Mathematical background of linearization: Taylor series approximation

1 dg ( ) y = g ( ) = g (ss ) + 1! d 1 dg n ( ) ( ss) + + n! d n ss
ss
( ss)n + . . .
Only the linear term is considered in the small-signal model

y = y g (ss ) = 1 dg ( ) 1! d
ss
( ss) =
dg ( ) d
ss
where
dg ( ) d
point ss, y and are called perturbations If ss = 0 and g (0) = 0 then the variables and y , and their perturbations and y , respectively, become identical. Consequently, we obtain
y= dg ( ) d
ss
ss
is the tangent of the nonlinear function f ( ) at the quiescent
22
Linearization of nonlinear baseband model (Determination of the small-signal model) The only nonlinear loop component is the phase detector
vd = KdA sin e
Since ss = 0 and vd(0) = 0, vd = vd and e = e. If during the operation the phase error always remains in the neighborhood of ss then we may linearize the phase detector
vd = KdA sin e KdAe
Substituting sin e e in the nonlinear loop equation, the linear loop equations are obtained as
e = i o o = KF (p) Kv Kv Ae + ve = [F (p)KdAe + ve ] p p p
23
Linearized loop equations

e = i o KF (p) Kv Kv o = Ae + ve = [F (p)KdAe + ve ] p p p
LINEAR BASEBAND MODEL
Recall:
vd vf vc
= = =
KdAe F (p)vd vf + ve
Based on the linear baseband model, the transfer functions may be developed To show explicitly the dependence of PLL parameters on the amplitude of input signal, A is not lumped with Kd
TRANSFER FUNCTIONS A linear (and only a linear) system may be characterized by its transfer functions. The transfer function expresses the output signal of the linear system as a function of an input signal Transfer function gives the response of a linear system to an arbitrary input in closed form A linear system may have many inputs and outputs, transfer functions may be developed between each pair of output and input The transfer functions may be expressed starting from Loop equations
Linear baseband model applying the rules of block diagram algebra
25
An example: Express o(s) as a function of i(s) in the complex frequency domain s Step 1: Linearized PLL loop equations in the time domain
o = Kv Kv KF (p) Ae + ve = [F (p)KdAe + ve ] p p p e = i o
Step 2: Transformation of the signals and system into the complex frequency domain s by means of Laplace transform
p=s i (s) = L {i (t)} ... Vc (s) = L {vc (t)}
Recall: Fourier transform can be determined from the Laplace transform by substituting s = j 2f
Step 3: Development of the transfer function H (s)

KF (s) Kv o(s) = Ae(s) + Ve (s) where Ve (s) = 0 s s e(s) = i(s) o(s) o(s) = 1+ AKF (s) [i(s) o(s)] s
AKF (s) AKF (s) o (s) = i(s) s s AKF (s) i (s) = H (s)i(s) s + AKF (s)
o (s) =
where H (s) is the closed-loop transfer function

PLL transfer functions

e(s) o(s)
= =
[1 H (s)] i(s)
Kv Ve (s) s Kv Ve (s) s
H (s)i (s) + [1 H (s)]
Note: Only two transfer functions Error function [1 H (s)] Closed-loop transfer function H (s)
are required to characterize completely the PLL
28
Closed-loop transfer function (Low-pass characteristic) AKF (s) H (s ) = s + AKF (s) Error function (High-pass characteristic) s 1 H (s ) = s + AKF (s) Parameters of closed-loop transfer and error functions are determined by Loop gain K = KdKv Transfer function of loop lter F (s) and, unfortunately, by Amplitude (and if there is any, the AM) A(t) of input signal
In the majority of applications, this dependence on A(t) is not allowed. Solution: An AGC circuit preceding the PLL is used to x the amplitude of input signal
PLL IMPLEMENTED WITH ACTIVE LOOP FILTER Circuit diagram of active loop lter
If an ideal operational amplier (op amp) is used then the transfer function of loop lter is
F (s) = 1 + s2 , s1
where
1 = R1 C and 2 = (R1 + R2 )C
Due to the innite dc gain of ideal op amp, F (0) and, consequently, the steady-state phase error ss is equal to zero
Closed-loop transfer function (PLL implemented with active loop lter)

2 2ns + n H (s ) = 2 2 s + 2ns + n
where the natural frequency n of the loop is dened by n = AK 1
and the damping factor of the PLL is dened by = 2n 2
PLL implemented with an active loop lter is a second-order, type-two feedback system. Unfortunately, both the natural frequency n and damping factor depend on A(t) which may be an AM or the eect of a time-varying channel
Frequency response of the PLL implemented with ideal loop lter
o(f ) = H (s)|s=j 2f i(f ) = H (f )i(f ) Transfer response has a low-pass characteristic to the input PM Parameter is the damping factor (0.3 2)

5 (a) 0
(d) (c)
10
(b)
15
1 0 1
10
10

10
Recall: The real input and output signals measured in a built PLL may be calculated from i and o s(t, ) = 2A sin(0t + i) r (t, ) = 2Vo cos(0t + o)
Error function (PLL implemented with active loop lter) s2 1 H (s ) = 2 2 s + 2ns + n where the natural frequency n of the loop is dened by
n = AK 1
and the damping factor of the PLL is dened by

2n = 2
Note again, both the natural frequency n and damping factor depend on A(t). This dependence may be prevented by an AGC circuit preceding the PD. The duty of AGC is to remove A(t) caused by either AM or introduced by the time-varying channel
Error response of the PLL implemented with ideal loop lter

(a)
e(f ) = 1 H (s)|s=j 2f i(f )

(b) (c)
= [1 H (f )]i(f ) Error response has a high-pass characteristic to the input PM Parameter is the damping factor (0.3 2)
10 (d)
20
30
10
10

10
Recall: i(t) and e(t) cannot be measured in a built PLL. The signals that may be measured in a built PLL are calculated from i and e s(t, ) = 2A sin(0t + i) vd(t) = KdA sin e
STABILITY CONSIDERATIONS PLL baseband model
where the error signal is

e(s) = i(s) o(s)
Conclusion: PLL is a negative feedback system which may become unstable Mathematical background of stability analysis Transient response is determined by the characteristic equation A system is stable, if it does not generate an output without an input signal. Transient response is determined by the characteristic equation. A system is stable if all roots of characteristic equation have a negative real value The characteristic equation is equal to the denominator of closed-loop transfer function H (s)
STABILITY CONDITION The characteristic equation is equal to the denominator of closed-loop transfer function H (s) Consequently, a linear system is asymptotically stable if all poles of its transfer function, that is, the roots of the denominator of closed-loop transfer function H (s) are in the left side of the s-plane A necessary and sucient condition for the stability of a linear feedback system is that all the poles of the closed-loop transfer function lie in the left half s-plane
36
Stability of PLL implemented with active loop-lter Closed-loop transfer function

2 2ns + n H (s ) = 2 2 s + 2ns + n
Characteristic equation (denominator of closed-loop transfer function)

2 s2 + 2ns + n =0
The two poles of PLL implemented with an active loop lter always lie in the left half-plane, consequently, this circuit is unconditionally stable Even if the amplitude A(t) of input signal varies and changes the closed-loop parameters n and , the PLL implemented with an active loop lter remains always stable
An example for PLL application: COHERENT FM DEMODULATOR FM waveform:

s(t) = Ac sin 2fc t + 2kf
t 0
m( )d
Input of the FM demodulator

s(t, ) = 2A sin(0 t + i)
To simplify the problem, assume 0 = 2fc Input FM

t
i (t) = 2kf
0
m( )d
Output of the FM demodulator

vc (t)
38
Basic equations in the complex frequency domain:

o (s) = H (s)i (s)
t
i(t) = 2kf
0
m( )d o(s) =
1 i (s) = 2kf M (s) s
Kv Vc (s) s
Development of FM demodulator output in the complex frequency domain:

Vc (s) = s s s 1 2kf o (s) = H (s)i (s) = H (s)2kf M (s) = H (s)M (s) Kv Kv Kv s Kv
If the maximum modulation frequency is much less than the PLL natural frequency and
n max{fm } << fn = 2 = H (s) 1 2kf Vc (s) M (s) Kv
39
FM demodulator output in the complex frequency domain:

Vc (s) = 2kf M (s) Kv
FM demodulator output in the time domain:

vc (t) = 2kf m(t) Kv
Coherent FM and PM demodulation by an analog phase-locked loop

40

Phase-Locked Loop: Budapest University of Technology and Economics Laboratory 2 - Experiment 9

Uploaded by

Copyright:

Available Formats

Phase-Locked Loop: Budapest University of Technology and Economics Laboratory 2 - Experiment 9

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Phase-Locked Loop: Budapest University of Technology and Economics Laboratory 2 - Experiment 9

Uploaded by

Copyright:

Available Formats

Budapest University of Technology and Economics

An example for PLL application: Coherent FM demodulator

Budapest University of Technology and Economics

Budapest University of Technology and Economics

Dept. of Measurement and Information Systems G eza KOLUMBAN

Budapest University of Technology and Economics

A few important applications: Demodulation of FM and PM signals FM modulator Carrier recovery

Budapest University of Technology and Economics

Budapest University of Technology and Economics

Voltages appearing in the loop are also shown

Budapest University of Technology and Economics

Voltages appearing in the loop are also shown

Budapest University of Technology and Economics

Budapest University of Technology and Economics

Budapest University of Technology and Economics

LOOP EQUATIONS AND NONLINEAR BASEBAND MODEL PLL block diagram

Budapest University of Technology and Economics

Budapest University of Technology and Economics

Dept. of Measurement and Information Systems G eza KOLUMBAN

Budapest University of Technology and Economics

Dept. of Measurement and Information Systems G eza KOLUMBAN

Budapest University of Technology and Economics

Transfer function of phase detector (PD) Block diagram of a phase detector

e(t) = i(t) o(t)

Budapest University of Technology and Economics

is the Heaviside operator

Dept. of Measurement and Information Systems G eza KOLUMBAN

Budapest University of Technology and Economics

Equations we obtained up to this point

do = Kv vc = po = Kv vc dt vc = F (p)vd + ve vd = KdA sin e e = i o

where K = KdKv denes the loop gain in rad/Vs

Dept. of Measurement and Information Systems G eza KOLUMBAN

Budapest University of Technology and Economics

Loop equations e = i o KF (p) Kv Kv o = A sin e + ve = [F (p)KdA sin e + ve] p p p

NONLINEAR BASEBAND MODEL

KdA sin e F (p)vd vf + ve

Dept. of Measurement and Information Systems G eza KOLUMBAN

Budapest University of Technology and Economics

Properties of nonlinear baseband model

Real input and output signals:

2A sin(0t + i) s(t, ) = r (t, ) = 2Vo cos(0t + o)

Budapest University of Technology and Economics

Dept. of Measurement and Information Systems G eza KOLUMBAN

Budapest University of Technology and Economics

Budapest University of Technology and Economics

0 = i KF (0)A sin ss Kv ve0

From which the quiescent point ss of PLL is obtained as

Budapest University of Technology and Economics

Mathematical background of linearization: Taylor series approximation

Only the linear term is considered in the small-signal model

is the tangent of the nonlinear function f ( ) at the quiescent

Dept. of Measurement and Information Systems G eza KOLUMBAN

Budapest University of Technology and Economics

Dept. of Measurement and Information Systems G eza KOLUMBAN

Budapest University of Technology and Economics

Linearized loop equations

LINEAR BASEBAND MODEL

Budapest University of Technology and Economics