Fundamentals of MIMO Wireless Communication

Prof. Suvra Sekhar Das

Department of Electronics and Communication Engineering
Indian Institute of Technology, Kharagpur

Lecture -12
Small Scale Propagation Received Signal Correlation (Contd.)

Welcome to the lecture on Fundamentals of MIMO Wireless Communications. We are

currently discussing the signal correlation properties and we have gone half way through the
derivation of signal correlation. So, we start from the point where we left in the last class.

(Refer Slide Time: 00:40)

So, we had this particular expression, and we started off with the expression here received

band pass signal correlation as .

(Refer Slide Time: 00:47)

We expanded each of this expressions and it all multiplications use the trigonometry
identities following which we collected like terms and finally, we have the expression where

we have along with the certain coefficient and along with a whole
set of coefficients. So, we have gathered the terms with t together and we have also gathered
the terms without the t. So, we have these two sets of four functions with coefficients of
which two are with t and two are without t. So, at this point we will make a very important
assumption or and this is a very valid assumption details of which we may look into later
lectures, but as of now we assume that this particular signal r(t) is ​Wide Sense Stationary​.
(Refer Slide Time: 01:57)

We will assume a wide sense stationarity for this band pass signal and that is quite valid and
it is proposed by a particular model known as wide sense stationary uncorrelated scattering in
a famous paper by Bellow in the 60’s. So, we will simply use that result over here and see

what we get if you assume wide sense stationarity, what we are defining over here is .

So, going by that, this - if it is a wide sense stationary process then it will be
independent of t and will be dependent only on Δt . So, if we use this particular set then what
we are left with is these functions which are functions of t should have the coefficient which
is 0. So, if we take that what we get is from this particular one we would have

because if this has to be 0 then this must be equal to

this.

Secondly, what we have is the other function of time is the . So, if it is wide
sense stationary this should have 0 coefficient and going by that we are going to have

So, if we now apply these solutions into this expression what will be left with is this first
equation there is a first expression; that means, this term has gone to 0 because this is wide
sense stationary. Similarly this term has also gone to 0 and top of which we have results that

. So, with this two being equal we can use this in the
first equation; that means, we are going to use this particular expression in substitute over
here.

So, we are going to get because this is equal to this and this two will cancel
out with this half. So, we will be left with . So, that is one of the
terms we are going to left with. Now this term is also going to 0 which gives these is equal to
minus of this, if we take that and replace it over here then we are going to get hQ hI it equal

to minus of this term. So, we will be getting 2 times minus of this term, minus. So, again the
2 and this half cancels out and remains .

So, now we have . Now we

can write and similarly . So, this

So, looking at this expression or if you would derive the expression, you would also get it as

and then we would write

where . So, we have these three important expressions which describe the signal
correlation.

So, through the last expression what we see is that the signal correlation in the band pass of
the pass band can be related to that of the signal in the baseband equivalent through the
carrier frequency, and the real part of it just like we have earlier written that

. So, this expression is similar and what we can see is that again the
translating from passband to baseband and baseband to passband is identical and hence it is
related through the f c and the corresponding functions of baseband or passband. So, we can
study the baseband equivalent or the complex envelope and we could still characterize the
channel.

So, as of now what we have over here is the passband correlation function is related to the I
component of the base band correlation function with cos and the cross components. So, this
autocorrelation component and this is a cross correlation component along with the

and corresponding .
So, with this we would now proceed to study this and because they
characterize because only after understanding, only after expanding these
components we will be able to completely characterize , because characterize
in terms of . So, from this point we proceed with our description of . So, in
order to study which is autocorrelation of the complex envelope what the baseband
equivalent would first begin with the autocorrelation of the I components.

(Refer Slide Time: 10:57)

So, what we have as . At this point we would iterate the assumptions

which we have made earlier and they were is uniformly distributed; that means,

why this was so because had a term amongst others which is

and we said that since f c is very very high when it is close to GHz or few hundreds of MHz

small change in would result in big changes in . So, it is a reasonable assumption to

make to be randomly distributed we can make the assumption is uniformly distributed in
the range .

We would also make the assumption at and are independent for n =/ m . Now, this is
again part of assumption that we made wide sense stationarity because there is this full model
is wide sense stationary uncorrelated scattering which leads to basically that the doppler shift
in the different direction or the nth paths are independent of each other. We will try and see
them later. As if now we will use these assumptions which are common and which are
fundamental in helping us derive these expressions.

So, we try to calculate this expression

and again we will follow the same techniques; that means, we will expand it.

So, we have these two terms, we have broken them into two terms for the ease of our steps
that we are supposed to take.

And if you look at this particular first term, if you look at the first term we have made some

assumptions that and are independent for n =/ m . So, we have created that set of

summation where n =/ m and if they are independent we have an expectation of cos

times expectation of this and if is uniformly distributed between [0, 2π] we can say that
this whole term expectation of this term goes to 0 and that term goes to 0. So, finally

, and will be left with only the second term. So, if

we proceed with the second term what we are going to have is again this

. So, this could be again broken up into two parts; that means,

and one which is the difference.

(Refer Slide Time: 15:56)

So, when we have that will be getting

. So, will be left with this term and if we look at . So, that is what will be
left in this particular expression.

And again as we bring the expectation operator inside this term

. So, what will be left with is

as . So,
. Now we are going to make some assumptions again or if we look into the details, so that we

are able to derive the expression correctly One thing we would remember is that .
So, if we have the expectation operator on this, so what we would be getting is

or in other words .

So, we can say this over here. Now on this expectation operator, at this expectation is to be
taken up across θn and to do that we basically need p(θn ) . So, p(θn ) is basically probability
of ray the nth path arriving at a particular θ and we make again this assumption that means,
what we are talking about is there is a transmitter there is a receiver. So, rays coming from
different directions is the first path, the second path, the third path. So, the first path is the
probability that the first path comes to a particular direction.

Now we have been saying the probability of this ray coming or the nth path is is the
distribution and that will be the same for all the paths. We do not know whether the first path
comes from this direction or that direction or the second path comes from this particular

direction. So, we would say that for all paths and that would
help us in evaluating this particular expression. So, when we take a look at this first part what

would we get is .

So, if we now look at this, this expression is independent of small n this expression is
independent of small n and this gets n times added. So, we would have n multiplied and this
would result in

.
And we have seen earlier that we have got . So, we have evaluated this which is

equal to this one we only have to remind to calculate and it is kind of from this

derivations you can almost guess what would be the ; so,

(Refer Slide Time: 21:18)

This is a kind of clear because the coefficient that you have with this is .
So, I mean that is kind of an indication, but if you derive through the steps you have going to
arrive at this.

So, with this we have been able to define . Now since is defined so we
can then define . So, what remains for us to do is to evaluate this expressions for E θ .
So, basically this expectation if we can work out this expectation over theta then we have

solved the complete expression of . So, if we go ahead and again our comes toward

. So, this is basically the 2D isotropic scattering model that is what we

are referring to continuously and since this was first proposed by Clarke’s we would usually
call this Clarke’s model.

This is how well known that means the probability of rays coming from all directions is that
the probability of rays coming from different directions is uniform now this is very, very
important. So, we just need to remember this is the Clarke’s model of 2D isotropic scattering

is a very, very famous model and if we follow this and we try to evaluate what we
get is

Now, we have the term that this has to be calculated. So, to calculate this we need to have
two things - one is p(θ) and also we need to identify what is the gain of the antenna as a
function of θ . Because sometimes there are directional antennas and sometimes Omni
directional antennas again in the 2D isotropic scattering model we will assume that the gain

of the antenna with respect to direction ; that means, it is the same gain in all
directions. So, this actually gives us a general method of calculating the correlation function.

So, you will be using and will be using in our case. So, we could get
(Refer Slide Time: 26:37)

Where J 0 is the 0th order Bessel function of the first kind.

So, what we have derived to this point is the autocorrelation or the component of
delta t which has come out to be Bessel function for the first kind of 0th order and the
expression would look like, the figure would look like the one we have here.

(Refer Slide Time: 28:05)

As you can see clearly this is the x axis this is the x axis which is indicating in this particular
figure this figure is taken from Stuber. So, this notation for our example will be having f m Δt
on the x axis. So, that is basically normalized because what we have on the x axis is

. So, if you look at this expression and we take f m Δt then we could choose
different values over here and we could plot it and they all map to one figure. So, this is x

axis is having f m Δt . This is an autocorrelation function which is normalized by . So, that

the max value is at 1, we could clearly see that what we have in the x axis is just J 0 function.
This curve is very very important because at a certain point you can clearly see that the
correlation which is 0 and from this we will be defining what is known as the coherence time
of the signal that is the time for which signal is correlated to itself.

We stop our discussion at this point and we will continue the next lecture and describe what
is the coherence time and also we look at the correlation of hQ and hI components. So, that

we are able to define the autocorrelation or de-correlation of the received signal which would
help us in understanding or characterizing the channel partially for small scale propagation
models.

Thank you.