Lecture 54 - Oscillator Design - IV

IIT Roorkee July 201
61 followers
Mar 4, 2024
Concepts Covered: Phase Noise, Roles of Linearity and Time Variance in Phase Noise, Reciprocal Mixing.
[Music]
[Music]
so welcome to RF trans design course we
are discussing about the phase no
so while discussing the phas noise we
have discussed that what is a unit that
is widely used in the uh in order to
find the phase noise that is a DBC per
Herz so if we want to understand the
phase Noise We again will come to the
one port view of the oscillator now in
oneport view what we already have
studied that there is a RLC resonance
Ator so this is RLC resonator and we
required
a uh active component which generate the
negative resistance so if we try to
consider the noisess magic box means
there is no noise noise generated and we
take a noise out which
is given by the if we try to draw
it this is a
inductor then there is a
capacitor
and
resistor you will have
a
noise which we have taken out noise
current I
n² by Delta F so this is noise current
dens
it and then we consider this as a
noisess
box
box now we will try to find out the uh
noise this is given by
lessons uh mean the scientist lessons
has derived this model of noise phasee
noise we we will get at a small
frequency of offset Delta
Omega
at
small frequency
offset Delta
Omega uh for LC
tank we can write a y
Omega + Delta Omega which is equal to G
if we consider the trans conductance C
plus
J Omega + Delta
Omega C + 1 by J Omega + Delta Omega
l so if if we try to multiply this on
this
side uh you you will get
equivalent Y which is
G
Plus
J
2 Omega Delta Omega IDE
by Omega + Delta Omega into
C so the why you can con convert into
the Jed so we'll see on other
slide so J which is equal to Omega
+ Delta
Omega = to 1
by we can write 1 by
g 1 by 1
+ J 2
Omega Delta
Omega by Omega + Delta
Omega into r
c so your
Q which is equal to Omega
RC you can
write J Omega + Delta Omega equal
to r 1 by G is = R
into 1 by 1 +
J Omega RC is given by
2
Q Delta Omega by Omega
+ Delta Omega
so if we consider the real part of
this is if we consider the
magnitude which is given by this will go
up
R Omega
KN Pi 2
Q Delta
Omega so this is a uh impedance that we
have find out that is R Omega by 2q
Delta Omega if if in this term Omega
plus Delta Omega Omega is higher so we
we can write a noise power spectral
density with Delta F equal
to
i²
by Delta f j squ so voltage can be given
by
this so this can be written as a 4 KT G
because we have considered the trans
conductance current into
r²
this r² Omega
Square by 2 Q Delta Omega
so this both G and R will be cancelled
out we can write a 4
KT
four me one of the R and G is cancel out
4 KT R Omega squ by 2
Q so this is also whole Square 2
Q Delta Omega
Square now what what is happening that
there is a possibility that your noise
phas uh what we have considered the
complete noise so noise can be divided
in terms of amplitude and pH so if half
of the portion of this noise is
contributed towards the phase then we
take the half of this and
that we we will get a phase component so
phase component component is given
by so your F component is equal to 2
KT 2 KT
R Omega KN
by 2 Q Delta Omega whole
Square So based on the lessons model L
Delta F Delta Omega
why we have considered Delta Omega
because here in the phase noise we are
interested for the Delta Omega offset as
I already discussed in the last class so
L Delta Omega equal
to we can write 10 log
of 2
KTR
Omega by 2 Q Delta Omega Square divide
by The Source noise which is given by V
s² and if we consider v² by R as a
signal
power which can be written as a 10 log
10 2
KT by P
signal Omega KN by 2 Q Delta Omega whole
Square so this is
how in this equation the only second
order has been considered so if you see
the
if you see this
model uh of phas noise then there is
a uh this two portion means this and
this has been captured so 2 KT by P
signal Omega by 2 Q Delta Omega squ so
this third component is coming due to
the flicker noise so this is happening
because of the thermal noise now due to
the flicker noise in the device this
third uh 1 by F because of 1 by F noise
this three times slope is coming and
that is uh uh that we need to add so in
order to do that uh here with this term
1 + Omega 2 Q Delta Omega one more term
can be added that is a 1 + Delta 1 by fq
by Delta
Omega so this is a uh third order
component and this is uh F which is a
empirical parameter which is uh added in
order to satisfy this otherwise this
green curve means the whatever we have
find out
this is representing this green curve so
in order to have this slope this
additional term has been added and this
empirical term F has been added so that
you we will get the uh phase noise model
which is accurate as uh we are getting
in uh measurement so this is how uh the
phase noise uh lessons model uh has been
developed and it is capturing all the
components of the phas
noise if we if you are interested more
uh of
understanding on the same phase noise
there are various phase noise models are
available in the literature you can go
through it the fundamental important
thing in the phase noise is it has a in
order to understand it is a the third uh
order slope second order slope and the
uh constant throughout the frequency as
a thermal noise so at Omega node
frequency up to the corner frequency of
the
moset or BJT any Active Components it is
going to have a third order slope after
that uh up to Omega by 2q it is a second
order and then it is a constant so this
is a model of phase noise and which you
can apply to uh any uh oscillator that
we have already
studied you can go through the
literature for more understanding on
phas nice so in the lessons model that
we have discussed uh there is a uh some
issues with the lessons model so we
already have discussed that how this
uh uh empirical term F came and that F
term is added uh due to the 1 by F noise
now this empirical term is uh not
defined in the lessons model and if uh
if if we try to improve the phase noise
uh the Improvement of phase noise is
depend upon the quality Factor Q so if
we increase the quality Factor Q we can
quality factor of the resonator we can
improve the phase noise however this if
we try to improve the Q uh with Active
Components and it might leads to
increase in the F which is the empirical
term defined so this uh there are many
places this model has a some invalid
assumption that is the F and that will
leads to designer to decide that what to
change and how to improve the phase
noise in the
oscillator so uh we try to uh see uh
another model of phase noise uh another
model how we can analyze the phase noise
compared to this lessons
model so in lons model there is a uh
linearity means we have consider linear
time invariant system but in reality uh
nonlinearity is a fundamental property
of real oscillator because without
nonlinearity we cannot have a
oscillation and it is also necessary for
uh amplitude limiting now what will
happen this results in a single
frequency sidal disturbance uh injected
into oscillator and it will rise of uh
rise to two equal amplitude side BS and
which are
symmetrical above the
carrier so linear time invariant system
cannot perform a frequency translation
and that's why what we need to consider
is a nonlinear system we already have
seen while discussing the RF mixer so uh
nonlinear mixing has uh uh often been
proposed to explain the phas noise so uh
the important Insight uh is that
disturbance or perturbation superimposed
on the main oscillator and they will
always be much smaller in magnitude than
the carrier here and we are when we are
doubling this disturbance it is doubling
the uh phase disturbance so uh it is
like uh when we are uh considering the
linear means in terms of the phase noise
so that phase disturbance is also
increasing so it is following the
superposition uh theorem where uh uh
addition means doubling the disturbance
is also doubling the phe
distribution uh linearity uh therefore
we can consider but we cannot consider
time invariant system so uh for uh the
model that we have discussed in the uh
lessons model that is considered the
linear time invariant system
but we should consider the linear time
variant model so we we'll discuss the
linear time variant
model so this is a example in which uh
there is a inductor and
capacitor which has infused the current
it now if we uh have uh this impulse
which is uh considered as a Delta
function T minus to and at what place we
are uh at add ing this impulse function
it will change uh the either amplitude
and phase accordingly so there are two
possibility that this impulse is
occurring at uh to where the maximum
amplitude is there and it will increase
the voltage amplitude by Delta V so
there is a change in
the uh voltage amplitude Delta V however
if the this uh impulse is uh happening
at the
uh zero Crossing and at the this uh tow
then what will happen it is contributing
to the change in the phase so what what
is uh Happening Here the amount of face
disturbance for a given injection pulse
depends on when the injection occurs so
here uh time invariance does uh fail to
hold so what we need to consider is a
linear time variant system so an
oscillator is a linear yet a Time
varying system it is a varying with the
uh
time now we will try to uh use the LTV
model means instead of LTI model to
define the phase
noise so uh in we have considered the
linear model so linearity correspond to
noise to phase conversion and as noise
is doubled phase is also doubled so here
LTI impulse
response is uh function of a two
argument uh that is observation time T
and at what time excitation is occurring
which is a to so we can write the
impulse response which is a uh HF T
comma to which is a function of T and to
and it is a variable uh which is a gamma
Omega T now what is this gamma this
gamma is nothing but impulsive sensitive
function this is very important to
understand for this uh LTV model now
what is this impulsive sensitive
function so if we consider this is a uh
unitless and and uh if we if we consider
the LC oscillator this impulse sensitive
function can be uh written as a it is a
periodic in nature it will change uh for
LC oscillator like sinusoidal and for
ring oscillator it has a uh impulse
Behavior like this now
here uh what is happening that this
function uh impulse sensitive function
has a maximum value if for LC oscillator
maximum value near the zero Crossing so
at zero Crossing it has a maximum value
you can see at this it has a maximum
value at uh uh Zer Crossing this it has
a maximum value and zero value at the
Maxima of oscillation so it is a zero
value at Maxima of
oscillation and uh this uh impulse
sensitive function is divided with the
um maximum charge displacement across
the capacitor and this is a unit step
function UT so this is how we can derive
the impulse response and it can be
written in this form it it is uh it is
divided by Q Max so that it can be a
unitless
function now what will uh what we'll try
to uh see that how this will be
converted in terms of the phase change
so when we try to find out the excess of
the phase due to arbitrary noise signal
through the superposition integral so
here the same integral uh whatever is
the this equation which is written here
this is a
uh this is
uh flow chart or block diagram this this
is a block diagram of the same equation
so it by Q Max is a input gamma Omega
not t which is a impulse sensitive
function is input which will generate
the function
Z which is this and uh when there is a
integration over the
time it will provide you the phase five
now this phase will be converted to the
voltage so you need to convert a phase
to voltage conversion with a uh phase
modulation and finally you will get the
output voltage VT so we can write uh 5 T
which is a output uh integration minus
infinity to Infinity H into I uh to D to
now uh we already know this function H5
T comma to we write 1 by Q Max maximum
charge minus infinity T it is a function
of gamma Omega T impulse sensitive
function into i d to so this is a f t
which is in relation with the impulse
sensitive function gamma Omega T now
this gamma Omega not t is a periodic in
nature
so because it is a periodic in nature we
can decompose into the foror series so
this gamma Omega T we can write in terms
of C by 2 and CN so this is a DC
quantity CN cos n Omega t plus Theta n
with this will change so it is a
multiple coefficient C 0 C1 CN and it is
a frequency component which is given by
cos Omega t + Theta so same uh same
block diagram can be re uh redrawn and
where uh each impulse function with the
uh each sinusoid uh can be added and
this is the final addition and we will
get the 5T and this five is converted to
the voltage so when we put this in this
equation we'll get a 5 T which is 1 by Q
Max C by 2 uh same thing uh
minus infinity to I to D to plus uh this
function now what is happening this five
uh we we need to convert into the
voltage so uh we have considered this
now we will uh apply the current
it a cus current it which is it equal to
IM cos M Omega plus Delta Omega T and
where Delta uh uh Delta Omega is much
much less than Omega not so 5T can be
written as IM CM sin Delta Omega T by 2
Q Max Delta Omega so now we came to the
function uh 5T which is a phase uh and
that phase we need to convert into the
voltage so the output voltage v o t
which is uh given as a COS Omega t t +
5T so this is a output voltage and where
5T is a given by this equation now
performing this phase to voltage
conversion and if we assume a small
amplitude disturbance uh we find that a
single tone injection results in a two
equal power side bends so we already
know that we have a two side bends one
is on this side one is on this side and
this is a center so this will have a uh
Delta Omega distance and we can write
the equal power side Bend and that is
also divided in uh if this is amplitude
that will be divided in a half half so
we can write a uh Power spectral density
for Delta Omega which is equal to 10 log
of Same imcm by because it is divid by 2
4 q Max Delta Omega whole
Square now this result
uh may be extended to General case of a
white noise source so here we can use
this uh White uh White Noise Source I
Square by Delta F and then instead of
considering the CM uh we consider the
all terms of CM m is equal to 0 to
infinity
cm² and this is a 4 q Max because now uh
this is also power quantity this is also
power quantity and this Square term is
coming here so 4 q Max s by Delta Omega
Square so this is a uh Delta Omega uh
whatever is the noise which is uh coming
back to the uh zero frequency now we
will try to analyze how uh this is
coming so what is happening this is a
thermal noise so this is a thermal noise
and and this Delta Omega is distributed
over Omega 2 Omega 3 omega and up to n
Omega KN and all these with uh C uh with
a component multiple this all multiple
component with this coefficient c not C1
C2 3 C2 or CN it is uh multiplied and it
is coming to the uh zero uh at the
frequency zero now what is happening
that this all and uh the 1X F noise that
is uh this is I in square by Delta F
noise and that 1X F noise is also uh
contributing at the zero
frequency now what is happening that uh
due to the phase
modulation whatever is coming at a zero
will be upconverted to the Omega not
frequency so because LC tank it will act
as a filter and it will tune only at one
frequency so all this uh noise component
and which is a down converted with the
different amplitude C 0 C1 C2 and
C3 it will
be uh up converted to the Omega not uh
by phas modulation so these components
of noise near integer multiples of the
carrier frequency all fold into Noise
near the carrier itself and it is clear
from figure that minimizing the various
coefficient CN so if we try to minimize
c0 C1 C2 C3 will minimize the phase
noise so this phase noise now what will
happen what happened in the lense model
we do not have any concrete proof and
concrete method to to reduce the phase
noise now because of this we got the uh
because of this impulse sensitive
function we come to know that we can
reduce the phase noise if we can reduce
this coefficient c0 to
CN and if you use a parel theorem this
uh is nothing but RMS value this uh all
summation can be derived in terms of RMS
value which is a 2 gamma r Ms Square uh
and that if we replace in that equation
the spectrum of 1x F Square region then
can be expressed by this equation I
Square by Delta F gamma RMS squ by 2 q s
into Delta Omega Square so this is the
same equation now this has been replaced
with the 2 gamma RMS square and that 2
two will be cancelled with the and we
get a two so now this model we help us
to understand there is no any empirical
term all are defined term and we can
have a uh we can handle the pH
noise now this uh up conversion of one
by up noise into uh uh closing phase
noise that I already told and Noise near
the carrier is particularly important in
communication system with very narrow
channel spacing and the allowable
Channel spacing uh
are frequently constrained by the
achievable phase noise however
unfortunately it is impossible to
predict a Clos in Phase noise correctly
with the LTI model and this problem now
is disappear if the new model is used
and specifically assume that the current
noise behave as a follow so 1X F region
we can write i² 1 by f is equal to i² 1
or Omega 1 by F by Delta Omega and if we
put this in that equation 1 by f is a 1
by F Corner frequency and
uh we obtain the following expression
for the noise in the 1 by F Cube region
so L Delta Omega we can
write equal to 10 log I n² by Delta f c
squ 8 Q Max Square Delta Omega square
and this is uh Omega 1 by a by Delta
Omega so this will be replaced with this
equation so the 1 by F Corner frequency
we can write by using this uh equation
that Delta Omega 1 by FB is equal to
Omega 1 by F into uh C squ so this is C
squ by 4 gamma RMS
squ so the this is uh we have considered
the same c not Square by gamma RMS
Square this when we write this Delta
Omega 1 by3 so this will have a uh this
term is in denominator for gamma and c
not square is nothing but the uh DC
quantity so that can be written as a
Omega 1 by F gamma DC by G gamma RMS
whole squ because c not Square we can
directly write this to gamma RMS
Square so that that can be written as a
gamma DC by gamma RMS whole square and
this will provide the insight into how
you can so what is the corner frequency
so if this is a page
noise
so that one by F Corner
frequency now this corner
frequency does not mean to be same as a
means Corner frequency after the uh
oscillator Corner frequency does not be
as same as a frequency of the uh circuit
or the device so the corner frequency of
the oscillator does not uh
same uh it it need not it need not to be
same as a uh 1X F noise Corner frequency
of the circuit and device so uh that can
be uh reduce means that we can improve
the 1 by F noise uh Corner frequency by
uh reducing the gamma DC uh component of
the impulse sensitive function uh and we
can improve the phase phase noise of the
oscillator so this model is this is a
linear time variant model and this model
help us to understand the phase noise
how we can improve the phase Noise by uh
V various method and in order to
understand the phase noise we should
have a understanding of this impulse
sensitive
function and uh which which can be uh
realized as a periodic uh in nature and
uh we can uh try to reduce the phase
Noise by uh having the by reducing the
DC term of the same so this is about the
phase noise we have discussed the two
model one is a linear time invariant
model and linear time variant model
now we'll discuss one of the uh
important effect of the phase noise that
is a reciprocal
mixing so what what is happening in
reciprocal mixing because uh up to now
what happened that while down conversion
we have considered that there is only
single tone which is appearing F not
this is a desired low frequency but due
to phase noise there are a uh the the
lello frequency looks like this and at
due to this noisy Alo phas noise due to
the phace noise there is other component
which is uh at this place at some Delta
Delta Omega distance now if there is
some unwanted signal or
interferer is also at the frequency just
of the I frequency desired I frequency
then what will happen that this loo and
this unwanted signal will down convert
to the base band because the ey
frequency is remain same and this uh is
called a reciprocal
mixing now this reciprocal mixing uh we
can uh in order to avoid we need to have
a
sufficient uh
Improvement in the phase noise me the
there should be a sufficient phase noise
so that some unwanted signal amplitude
mean will not uh down convert due to
this uh phas noise so that can be given
by so the maximum allowable phase noise
when we are designing the
oscillator uh the maximum allowable
phase
noise in order to achieve an adjacent
Channel rejection
for a uh or a selectivity of SB we can
write this is a equation so where L is a
phase noise equal to C uh dbm minus uh s
that is
a uh rejection uh and I is the
interference and then this is a
bandwidth this will be defined as a DBC
per Heartz now we'll take one example so
that we can
understand so in this example the GSM
cellular telephone as a standard
requires a minimum of 9 DB of rejection
of interfering signal level of minus 23
dbm at 23 mahz so what we want is a
rejection of 90p minimum 90b rejection
so this is nothing but s in that
equation and interfering signal level
has amplitude of minus 23 so this is
nothing but
I at 3 mahz from the carrier minus 33
dbm at 1.6 mahz so if you consider this
and then
uh there is a at 3 mahz it is a minus 23
so this 3
mahz as uh -
23
then uh 1.6 mahz has little bit low
amplitude that is minus uh 33 and- 43
dbm uh of
interference at6 mahz so for a carrier
level of- 99 dbm so this is a carrier
level which is - 99 dbm so what we need
to determine the required Alo local
oscillator phase noise at this carrier
frequency offsets and the channel
bandwidth is given as a 200 Koh because
bandwidth is also important here uh the
if a bandwidth is more then this uh
phase noise will
also be uh phase noise Improvement is
also required more so this
lfm uh is d DED so we can consider that
for 3 mahz offset it should have a minus
138 DBC per Herz and interfering signal
level isus 23 for 1.6 mahz offset it is
a interference signal is - 33
dbm and phase noise requirement is minus
128 so you can see that
as we are near to the carrier frequency
the phase noise requirement is little
bit relaxed and the interfering signal
level which is smaller compared to the
interference signal level which is at
Center frequency so uh if you have at 6
mega Hertz the interfering signal is
minus 43 dbm you need to have a at least
minus 1110 DBC per Hertz at the offset
frequency of 6 mahz so this is uh how we
can uh identify the phase noise in the
oscillator uh for uh reducing the
reciprocal
mixing we uh we have discussed about the
oscillator and Fise if you are
interested to uh have a more in detail
uh understanding you can uh refer this
reference and uh it will help you to
understand uh in more detail all this
material has been uh considered from
this references thank
[Music]
you
[Music]
a
