Lecture 59 - Basics of Phase Locked Loop - II

IIT Roorkee July 201
61 followers
Mar 4, 2024
Concepts Covered: Charge Pump, Integer - N Frequency Synthesizers, Fractional - N Frequency Synthesizers, Noise Shaping, Frequency Dividers, Dual-Mode Dividers.
[Music]
[Music]
uh welcome once again to Arab transer
design course we are discussing about
the basic of phase lock loop in last uh
class we have seen a type one PL and we
tried to derive the transfer function of
type 1 PLL now we are moving to the type
2
PL so we already have discussed that
when the drawback of type 1 PL is if
there is a change in the frequency then
it can not lock
the uh pace so if we want want to uh if
we have a frequency and phase both is a
different then we need a phase frequency
detector
so if we try to see that how what are
the condition for phase frequency
detector let's say there is a Omega a
frequency and Omega B frequency as shown
in the figure so this is a Omega a and
this is a Omega B
so if Omega a frequency is
high if Omega a is greater than Omega B
what will happen that you will have a QA
which is increased so if Omega
a is higher than Omega P your QA is
increased and the difference in the
phase will be observed by the QA and it
will start
increasing
now the moment uh QA is equal to QB so
there is a if Omega
a is equal to Omega
B then uh this
QA uh will have a constant change in the
pH because now there is a frequency
which is a constant there is a
possibility that the 5 a cannot be 5B
but that will uh same as a uh what we
have done in the type 1 P so either
QA or
QB will constantly give the output it is
depend upon how increase means how the
PA difference is now if there is a
change in the frequency let's see if
Omega
B is higher than Omega
a then QB will increase so uh in this
case in the figure which is shown there
is a two case only shown one is Omega a
which is greater than Omega B and Omega
a which is equal to Omega B so in the
both both the case your QB is a
constant there is a no change there is q
B is a constant but if there is a change
if instead of Omega a Omega B is greater
than Omega a q uh a will look like QB
and QB will look like q a and same will
be for the when Omega is equal to Omega
B so here a rising age on a yields a
rising age age on
QA if QA is low Rising age on B resets
QA
so uh if there is a rising age on B it
reset QA if QA is a high and this
circuit is a symmetric with respect to A
and B so uh either you give to a or b it
is a
symmetry so how we we can implement this
uh uh mechanism in the uh
circuit so if we try to see the straight
diagram the straight diagram look like
this when QA and QB is equal to Zer
there is a zero
State there is other two State either QA
is a high and QB is zero so there is a
one state which is a state zero where QA
is equal to 0 and QB is equal
to0 now as a means Omega a frequency is
increased then q a is equal to 1 QB is
equal to 0 it will REM remain high QA
will remain high up to when Omega a is
high when a is high the moment QB is
going to become High then uh it will try
to go to the state
zero if Omega B is high then QA is zero
and QB is equal to 1 and it will remain
constant it will start increasing QB if
if uh B is high Omega B is high and then
the moment Omega a is again going to
become High then it will come to the
state zero so what what it there is the
three logical states are necessary QA QB
is equal to 0 QA is 0 QB is 1 QA is 1 QB
is0 and to avoid a dependence of the
output upon the duty cycle of the input
the circuit should be realized by using
an H triggered sequential machine so if
we try to uh make a h triggered uh
sequential circuit the same PFD phase
frequency detector we can use a two
flipflop and end
gate so there is a uh this is a two flip
flop here there is a vdd is applied to
the uh D input and the clock clock is at
the a similarly uh there is a vdd and
clock so it will look same uh the D and
clock and the when the QA and QB both
are equal it will reset this pulse so
when QA and QB both are
zero it will reset this clock so q and
and QB are simultaneously high for
duration given by a total delay through
the end gate and reset the path so when
q and QB both are one it will reset the
path of the Flip Flop and the wids of
the narrow reset pulse on QA QB is equal
to the three gate delays so uh there is
a three gate delay one is uh this and
another is due to the uh two gates which
is there in the ledge one and ledge 2
and that total delay is depend upon this
three gate delay plus the delay of the
and
gate so this is how we can implement the
phase frequency
Detector by using a sequential circuit
so as a is increasing and uh what what
will happen that you will get a vdd
means uh D will be uh directly D will be
get at the output and as uh Omega B will
increasing you will get a QB which is
rised but the moment both are equal to 1
this will be reset so this is how this
uh P PFD will
work now once we have designed the PFD
what we need is a
uh we need to use uh same like uh type 1
PL the RC low pass filter and then you
need to uh amplify the difference
between these two and the Vu so using a
PFD in a pH loock Loop resolves the
issue of The Limited acquisition range
which was there in the type one and at
the beginning of the transition the PFD
act as a frequency detector
and then pushing V frequency towards the
input frequency and after that two are
sufficiently Clos BFD operates as a phe
detector so it will operate as a phe
detector so it will both it will do as a
frequency as well as a phase
detection so in order to do that uh
means we need in between a circuit which
can able to uh do this up and down
because there is a Q and QB so we need
to go for up in the frequency or down in
the frequency so we need a charge pump
so we need a circuit which is called a
charge pump which will increase the
voltage or decrease the voltage
depending upon the increasing of QA and
decreasing of the uh QA and QB so here
in charge pump uh this is uh
what will happen that when QA is high
then it will start charging this
capacitor and with the current and when
there is a uh QB which is high it will
start discharging so here switches S1 S2
are controlled by the up and down
respectively and the pulse width of
delta T on up turns S1 on for delta T
seconds allowing i1 current to charge C1
so your V out goes up by delta T into i1
by C1 because it is charging for only
Delta t period similarly a pulse on down
yields a drop in a V out if up and down
are asserted
simultaneously i1 simply flows through
the S1 and S2 to I2 creating a no change
in a output voltage if up and down are
asserted simultaneously so then then
your uh V out will remain a
constant so if you want to uh find out
the transfer function uh how how how do
we find a transfer function of this so
uh there is
a we can approximate the PFD or charge
pump Circuit of a uh figure which is
shown here as a current source and some
average value driving a
C1 uh then we calculate the average
value of a current source and the output
slope for an input input period of T in
so for input phase difference of Delta 5
radian what we get Delta 5 by 2 pi into
T in so seconds the uh T in seconds and
the average current is equal to IP P
Delta 5 by 2 pi and the average slope so
the average current is this if we try to
find out the average current and average
slope is IP Delta 5K by 2 pi d/
by C Delta Pi by 2 pi/ by
C1 so the heds transfer
function uh which we can write it down
which is IP by 2 pi
c1s into K VC s/ 1 + IP by 2 pi c1s into
kvco so this is a transfer function of
this charge pump PL such a loop ideally
forces the input phase error to zero
because uh there is a finite error and
it will lead to Unbound value of a v
control so what we need we first D the
transfer function of the PFD charge pump
and C1 Cascade and then we call a type
to PL because its open loop transfer
function contains a two poles at the
origin so now if you want to find the uh
transer function for
PFD uh then charge pump and C1 what what
we get there we can approximate this
waveform by a ramp so this is a uh how
the V control is increasing
and that can be considered as a ramp and
its slope is given by IP by 2 pi into CP
uh into Delta 5 so Delta V control
change in the is given by Delta 5 by 2
pi T in IP by C1 so if we write with
respect to time V control t equal to
Delta 5 by 2 pi IP by C1 into T into
UT uh because it is
uh its slope is also changing with the
time so T into UT so V control by Delta
5 S equal to IP by 2 pi into C1
s so if one of the integrator become a
lossy let's say if uh this uh capacitor
the system can uh be stabilize and this
can be accomplished by inserting a
resistor in series with a C1 so we need
to add a resistance R1 C1 that will
provide additional stability in the
charge pump PLL so this is also known as
a charge pump PLL because we are we
required a charge pump
circuit so now if we want to find out
the uh complete transfer function here
this is a v control where we have
approximated as uh linearly changing uh
with the time and above it there is a
reple which is uh so there is a v
control 1 and V control2 if we
decompose like a superposition theorem
and what we we'll get a v combination
total V control T is a combination of V
control 1 and V control 2 so where V
control 1 is Delta fin not by 2 pi IP by
C1 T UT plus Delta 5 by 2 pi I by C1
into R1 into UT so
this step of the height so this
initially what we have considered
because there is only capacitor now this
step of this height is depend upon this
value of
R1 so now the V control by Delta i s is
given by IP by 2 pi 1 by C1 s + R1 and
the transfer function HS is given by
this equation where we can we are
considering the R1 1 and C1 both so this
is a same uh similar transfer function
which we can uh find out the damping
factor and the if we compare with
the transfer function for a stability
that is s square + 2 Z Omega n s + Omega
n² so Z will be given by R1 by2 square
root of Ip into K vco and to C1 by 2 pi
so your Omega n natural frequency is
depend upon IP into k v by 2 pi into C1
so if we try to plot the open loop
transfer function uh the way we have
plot ploted the open loop transfer
function in the type 1 PLL so for type 2
PLL the uh there is a zero which is
added because R1 because of R1 and C1 so
if this Z is added uh without zero it
will look like this where it will be
constant uh around minus 180 but with
zero that will uh phase margin is uh
increasing and that will help to uh the
uh improve the stability in the circuit
so here we have tried to address both
the problem that we have seen in the
type 1 PL the that is the problem of the
phase and frequency and the problem of
the uh stability so that both problem
can be solved in the type 2
PL so there are uh more uh understanding
and details uh are required in order to
design the complete PLL so as far as uh
these course and uh this RF trans design
course is concerned the PLL is
uh we have tried to understand the PLS
just to have the basic idea if you are
interested for more in detail about to
understand about the PLL you can uh go
through the reference book which is
provided and that will help you to
understand and you can also uh refer the
lectures which are on the page lock loop
design so these are the two basic type
of PLL PLL 1 and pll2 once you have
understanding of the PLL design you can
also design a frequency synthesizer and
that frequency synthesizer will be
useful for generating a different
frequencies for a different bands that
we have studied GSM band and then CDMA
or uh different uh frequency band for
different application Wi-Fi bands so
that all can be designed uh mean
synthetizer can be designed for uh
different application and it will be
useful in the transceiver design so uh
uh we have discussed the PLL now uh
charge pump PLL in uh charge pump PLL uh
all the PLL that uh we have discussed
type one and type two so what is the
issue uh in reality that when we we want
to use this uh p in the uh actual RF
transceiver uh what we need is a our
frequency range so we need a frequency
synthesizer so why there is a need of
frequency synthetizer the first is the V
vco frequency range because uh we are
working at uh 900 MHz or 2.4 GHz or 5.6
GHz so different are a frequency and
that frequency are very high so the Vu
frequency is very high and it is in
gigahertz
range but in reality the Practical
reference that we use a crystal
oscillator they might be in kilohertz to
megaherz range so that is a limitation
of crystals which we are applying as a
reference and the uh vco frequency which
uh needs to be
tunable uh so that is the advantage that
vco frequency can be tuned uh with the
voltage at very high frequency but we
cannot tune the reference our reference
frequency is a fix so this reference
frequency uh cannot be tunable however
we output frequency can be tunable so
what is the solution the solution is why
can't we use the uh because the
reference frequency is small and the
output frequency which is high why can't
we divide the output with the number by
n so here we can divide F
out by
n now the as PFD uh we already have
discussed that phas frequency detect and
after that charge pump this PF required
the same frequency and only the phase
difference will be find out now in order
to uh do that we need this frequency
which is coming from this which is f out
by n so this is f out by n and this is f
reference frequency let's say F
reference frequency if we uh if this F
out by n and this F reference might not
be equal so we can put another divider
uh circuit here which is a divide by K
and we can divide F
reference by K so our F out by n should
be equal to
F by KY so this both the frequency
should remain same so what is the output
frequency
so output frequency we can write F out
which is equal to F
reference by K into n now this F
reference by K uh another uh important
thing in uh RF trans design is we are
going to use a different channel in the
single band uh for example in GSM
band uh has a 900 00
khz 900 mahz is a center
frequency and its Channel band
width is
around 200
khz so how many channels more than uh
whatever more than
125
more than
one uh 25
channels we can include so
here this uh Channel spacing between
these two is nothing but app R by K
because what we want to do we we want to
do the frequency control so when we do
the controlling uh from one frequency to
other frequency because uh we need to
select channel one channel two this
channel has a 200 khz
bandwidth and that channel uh bandwidth
is equal to F ref by K so that is a
channel frequency F Channel and our
output frequency is multiplication of n
into F channel so in that case if you uh
if your reference frequency is a small
then uh your n the uh n requirement of n
will be too high so that uh can be
decided based on the uh what is the uh
application or what is the standard that
you are
using now if we try to see the B plot of
the open loop gain GS as well as the P
Clos Loop transfer function HS so how
the GS can be so here GS can be given
by
kpfd that we have already defined
into uh K
vco
vco divide by
S
into
1 by
n so this is a uh GS transfer function
of
the uh this uh receiver and what what we
are adding is a 1 by n uh so this is a
GS and if we want to write
HS which is given by five
out
by 5
in equal
to
n 2 GS by 1 + GS so this is a open loop
transfer
function and this is a closed
loop transfer function now if we if we
see s uh there is
a one uh if type 1 PLL is there there is
only one pole if type 2 PLL is there uh
we we can add
FS which will add another pole so there
is a two pole and that because of that
two pole there is a slope which is a uh
40 DB per decade now another zero will
come into the picture in type two PLL at
Omega J and that will help us to improve
the stability that we already have
discussed and this uh once this zero
will come again it is a minus 20 DB per
decade and you you will get this is a
Omega U is a Unity gain
frequency now what is a uh catching
Point here is that our
n uh this if
GS
for uh frequency if GS is much much
higher than one for frequency Omega less
than Omega
U we can have a h j
Omega is equal to
n so the HJ is given by the n and it can
be written 20 log of N and this is a
transfer function of 1 by 1 + uh G and
uh which will uh which will be reflected
the G uh GS and H in 1 by 1 + G uh
transfer function so this is how the
transfer function of uh fractional and P
only the difference is here 1 by n and n
will be
added now what are the noise sources uh
in the synthetizer so if we want to see
if we want to see the noise sources in
the synthesizer if we try to draw
this what are the different noise
contribution that happens so uh
reference for low frequency reference
will dominant and it is a slope of
two and after that so
here uh
reference will dominant and then up to
the pl
bandwidth uh it will dominated by the
PLL again at very high frequency then
the PLL bandwidth the vco will dominate
the uh noise so this is a PLL noise
here and then after that if there is a
buffer then
buffer has a wider
bandwidth buffer will dominate so this
is how so here vco will dominant so this
is how the noise Source various noise
sources will be there in the
synthesizer now uh if we uh assume a
simple Loop filter uh we can give the
transfer function as this uh this is a
transfer function of uh uh fractional
and P and the two poles of the transfer
function we can write it this so for
phase noise uh orall phase noise there
are a different contributor uh if we
consider the loop filter vco and
reference and other things as ideal then
uh this is a curve of vco where uh at
low uh the phase noise is low at low
frequency it is
increasing with the frequency offset and
then again uh the phase noise will
decrease because of the low pass
behavior and as this is a reference
reference uh has a uh at frequency
offset reference will dominate The Phase
noise and then uh you can see this is a
reference which is dominating the phase
noise and after that what will happen
that uh vco will dominate so this is a
vco noise and this is a black curve is a
overall phase noise uh with the
frequency offset so this is about the uh
integer and uh fractional and synthesiz
integer and
synthetizer now uh integer and
synthetizer has some limitation so the
integer and frequency synthetizer has a
major drawback what is that major
drawback the reference frequency cannot
be higher than the channel spacing so
there is a channel spacing that we are
uh adding means by changing the uh
reference means if we want to change the
frequency uh it should that n should
change according to the channel
frequency now
that reference frequency uh cannot be
higher than the channel spacing so if uh
it it will limits our performance
because Channel spacing always will be
very small and that uh small Channel
spacing what will happen that the requir
uh we are dividing the output frequency
divide by n so you need a very high N
means n requirement is high because your
reference frequency is small and another
thing is when the reference frequency is
is less you required settling means it
required High settling time and the
capacitor and resistance value is
increasing so what will happen so this
intern sets upper bound on Loop
bandwidth and which cause sever
implication like size of the loop filter
capacitors and resistor uh synthetizer
sett settling time uh close in and far
out phase noise and ability to work with
any arbitrary Crystal oscillator because
you need a frequency which is equal to
the channel spacing so we cannot use any
Crystal oscillator and there are a
reference Spurs and amount of filtering
also is a uh impacted by a number of
n uh that is a how much division we are
doing so what is the solution if we can
do the
fractional uh division of the frequency
uh and that will help us to uh get rid
of the very uh get rid of the channel
spacing that minimum reference frequency
requirement and that can be done so
again the integer and frequency
synthetizer can be
redrawn uh here this uh n/ by n and n +
1 so how this will work for uh uh let's
say if there is a k cycle and for K
minus
1 uh there is a divide by n will happen
and after that uh for 1 time n + 1
division will happen so the overall uh
division will be in the fraction Alpha
so this is how we can make
the
this is how we can make the frequency
division so that will be uh around n +
alpha
n+ alpha and this Alpha can be given by
the value KY so we'll see how it works
so here uh uh as we have already seen
the main drw dra of simple approach
outline here in figure is a creation of
the fractional spare so how the
fractional SP will be generated because
this is a reference and this is a
divider uh this is a hypothetical
example dividing n for three cycle and n
+ 1 for four cycle so total K is equal
to
four so uh this is a reference this is a
div uh division this is U and D so what
will happen that uh here the up means uh
up and down so frequency is up and down
that we can find out so here the up is
happening when there is
a uh this difference of the reference
and division frequency and your control
voltage will start increasing so because
it it should start increasing the
frequency by uh voltage control
oscillator
but the as the frequency start
increasing what will happen after uh n
up to N Cycle this will happen and after
n + 1 means after up to three cycle it
will divide by n and after that so this
is 1 2 and three cycle it will divide by
n then it will divide by n by 2 so when
it is dividing and it will divide by n +
1 so when it is dividing by n + 1 there
is a uh one Spike will come and what
what happens that because of this
difference that again will start uh
increasing and what what it happens that
due to this uh it will generate a uh
Spurs uh spse signal that is called a
fractional Spurs so uh the PFD starts to
generate a down pulse to slow down the
vco which in turns causes the control
voltage to reduce on the fourth cycle
the vco is divid by n+1 resulting in the
divider output to be slower than the
reference and hence up signal is created
that rises the control voltage so what
will happen this will uh k k is equal to
four times this pattern repeats K is
equal to four cycle and it leading to a
V being modulated by a periodic signal
with a period of f by K or Alpha F which
creates a side Bend at Alpha F and it's
harmonic so in reality we want a signal
like this but there is a side band is
created and that will generate a
fractional Spurs like
this at uh same uh uh Alpha means for
period of alpha into
F reference so here uh how how can we
calculate this uh F out
so this division can be calculated
as so we are dividing by n for Kus 1 so
K
-1 into n + n + 1 / by K -1 +
1 so this will come as a n + 1 by
K and this first three cycle of divide
by n the divider frequency is given by F
n + 1 by K into F which is greater than
F so this is how the uh division will um
in happen in the fractional and
PL now in order to get rid of this
fractional Spurs what you need is a
uh because here we are making n * and n
+ 1 and if we can make the combination
of an n and n + one which will help us
to reduce this uh uh fractional Spurs
and one of the method that is widely
used is a noise shaping and this noise
shaping is uh one of the method by which
we can able to reduce this fractional
Spurs so one very common and elegant
technique to deal with the fractional
SPS is exploiting the noise shpping
concept which is widely used in a data
converter so the periodicity periodicity
of the control voltage could be broken
if the divider modulus uh is randomly
selected n and n+1 but still maintain
the desired average value so the
deterministics SPs appearing as a tones
will be converted to noise spreading
around the V signal with a total energy
remaining the same so F out which is
equal to F out n plus BT where BT is
given by Alpha plus QT and that F
division will be given by F out by n +
Alpha + QT which is equal to F out by n
+ Alpha and 1 + QT by n + Alpha so this
is a how we can make a uh find uh
fractional division by using this method
so this is what what is a fractional uh
and P now uh in fractional and
p uh you required different uh kind of
uh dividers and uh the that will divide
the fractional means that can be
consider that can be written as a MDD uh
du DMD
so so it is a dual modulus divide
and dual modulus divider can be made by
using a d flip
flop and combination of few
simple
GS so if we do this we can make a
various uh dual modular divider so uh
there is a uh this can also be extended
to the uh digital PLL so this is
how uh we will lead the um same PL
whatever we have designed uh whatever we
have discuss we can also make a digital
PLL so what is the limitation of analog
PLL so depending on the noise uh
requirement and the reference frequency
the loop filter size can be large and
the capacitor scales with the technology
and their size still does not reduce as
favorable as the digital Gates because
in digital we can use a digital Gates
and a large charge pump current is often
required to achieve a low inband phase
noise and power factor d uh phase
frequency detector uh and charge pump
linearity could be concerned for certain
application so uh there is an important
tradeoff between a noise contribution of
PL building blocks particularly vco and
Delta Sigma modulator so what we can do
is uh this block diagram will look same
only the difference is PFD will be
replaced with the TDC this is a time to
digital converter uh this is a phase
frequency detector will be converted to
time to digital converter it is it
required instead of RC it's required a
digital Loop
filter and then uh we
uh digital uh control oscillator can be
used and uh am AMD already it is a
digital circuit that is a uh modular
divider and then it will be provided to
the reference so this is how the uh
digital PLL can also be made by uh cons
by using the uh same analogy of a analog
PLL and the TDC is nothing but the uh
combination of ADC with the delay so uh
that will provide the time uh to digital
conversion so this is how the in all
these uh two three lectures we have
discussed about the basic PLL type 1 and
type 2 PLL we also discuss about the uh
frequency
synthesizer uh fractional and uh
synthesizer and the digital PL so if you
are interested for more in detail you
can go through the reference and uh
there is a separate lectures on the pl
are available you can go through the
same thank
[Music]
you
[Music]
