Sei sulla pagina 1di 77

Simplified MIMO receiver architectures

Rohit Budhiraja

MIMO Wireless Communications (EE677)

March 8, 2017

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 1


Matched filter1
Our system model
y = Hx + n

1 Reference for todays class: Tse and Vishwanath Sec 8.3 onwards
MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 2
Matched filter1
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1

1 Reference for todays class: Tse and Vishwanath Sec 8.3 onwards
MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 2
Matched filter1
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi +n
i =1,i 6=k
| {z }
inter-stream interference for kth stream

1 Reference for todays class: Tse and Vishwanath Sec 8.3 onwards
MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 2
Matched filter1
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi +n
i =1,i 6=k
| {z }
inter-stream interference for kth stream

Matched filter maximizes the kth stream SNR by ignoring inter-stream interference

1 Reference for todays class: Tse and Vishwanath Sec 8.3 onwards
MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 2
Matched filter1
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi +n
i =1,i 6=k
| {z }
inter-stream interference for kth stream

Matched filter maximizes the kth stream SNR by ignoring inter-stream interference
Matched filter wk for kth stream hk

1 Reference for todays class: Tse and Vishwanath Sec 8.3 onwards
MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 2
Matched filter1
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi +n
i =1,i 6=k
| {z }
inter-stream interference for kth stream

Matched filter maximizes the kth stream SNR by ignoring inter-stream interference
Matched filter wk for kth stream hk
X
nt
hHk y = ||hk ||2 xk + hHk hi xi + hHk n
i =1,i 6=k

1 Reference for todays class: Tse and Vishwanath Sec 8.3 onwards
MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 2
Matched filter1
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi +n
i =1,i 6=k
| {z }
inter-stream interference for kth stream

Matched filter maximizes the kth stream SNR by ignoring inter-stream interference
Matched filter wk for kth stream hk
X
nt
hHk y = ||hk ||2 xk + hHk hi xi + hHk n
i =1,i 6=k

Performs well when noise dominates the inter-stream interference

1 Reference for todays class: Tse and Vishwanath Sec 8.3 onwards
MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 2
Matched filter1
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi +n
i =1,i 6=k
| {z }
inter-stream interference for kth stream

Matched filter maximizes the kth stream SNR by ignoring inter-stream interference
Matched filter wk for kth stream hk
X
nt
hHk y = ||hk ||2 xk + hHk hi xi + hHk n
i =1,i 6=k

Performs well when noise dominates the inter-stream interference


Perfoms poorly at regime when interference dominates the noise

1 Reference for todays class: Tse and Vishwanath Sec 8.3 onwards
MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 2
Matched filter1
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi +n
i =1,i 6=k
| {z }
inter-stream interference for kth stream

Matched filter maximizes the kth stream SNR by ignoring inter-stream interference
Matched filter wk for kth stream hk
X
nt
hHk y = ||hk ||2 xk + hHk hi xi + hHk n
i =1,i 6=k

Performs well when noise dominates the inter-stream interference


Perfoms poorly at regime when interference dominates the noise
Generalized matched filer expression is HH
1 Reference for todays class: Tse and Vishwanath Sec 8.3 onwards
MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 2
Zero-forcing/decorrelator (1)
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi +n
i =1,i 6=k
| {z }
inter-stream interference for kth stream

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 3


Zero-forcing/decorrelator (1)
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi +n
i =1,i 6=k
| {z }
inter-stream interference for kth stream

Decorrelator cancels the the inter-stream interference by ignoring the noise

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 3


Zero-forcing/decorrelator (1)
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi +n
i =1,i 6=k
| {z }
inter-stream interference for kth stream

Decorrelator cancels the the inter-stream interference by ignoring the noise


Decorrelator wk for kth stream

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 3


Zero-forcing/decorrelator (1)
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi +n
i =1,i 6=k
| {z }
inter-stream interference for kth stream

Decorrelator cancels the the inter-stream interference by ignoring the noise


Decorrelator wk for kth stream
Should be orthogonal to the vectors h1 , , hk1 , hk+1 , hnt

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 3


Zero-forcing/decorrelator (1)
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi +n
i =1,i 6=k
| {z }
inter-stream interference for kth stream

Decorrelator cancels the the inter-stream interference by ignoring the noise


Decorrelator wk for kth stream
Should be orthogonal to the vectors h1 , , hk1 , hk+1 , hnt
nt
X
yk = wkH y = wkH hk xk + wkH hi xi +wkH n
i =1,i 6=k
| {z }
=0

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 3


Zero-forcing/decorrelator (1)
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi +n
i =1,i 6=k
| {z }
inter-stream interference for kth stream

Decorrelator cancels the the inter-stream interference by ignoring the noise


Decorrelator wk for kth stream
Should be orthogonal to the vectors h1 , , hk1 , hk+1 , hnt
nt
X
yk = wkH y = wkH hk xk + wkH hi xi +wkH n
i =1,i 6=k
| {z }
=0

= wkH hk xk + wkH n

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 3


Zero-forcing/decorrelator (2)
Signal power of kth stream is

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 4


Zero-forcing/decorrelator (2)
Signal power of kth stream is
E [wkH hk xk (wkH hk xk )H ] = wkH hk E [xk xkH ]hH w = Pk |wkH hk |2

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 4


Zero-forcing/decorrelator (2)
Signal power of kth stream is
E [wkH hk xk (wkH hk xk )H ] = wkH hk E [xk xkH ]hH w = Pk |wkH hk |2

Noise power of kth stream is

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 4


Zero-forcing/decorrelator (2)
Signal power of kth stream is
E [wkH hk xk (wkH hk xk )H ] = wkH hk E [xk xkH ]hH w = Pk |wkH hk |2

Noise power of kth stream is


H H H H H H 2
E [wk n(wk n) ] = wk E [nn ]wk = N0 wk wk = N0 ||wk ||

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 4


Zero-forcing/decorrelator (2)
Signal power of kth stream is
E [wkH hk xk (wkH hk xk )H ] = wkH hk E [xk xkH ]hH w = Pk |wkH hk |2

Noise power of kth stream is


H H H H H H 2
E [wk n(wk n) ] = wk E [nn ]wk = N0 wk wk = N0 ||wk ||

SNR of kth stream is


Pk |wkH hk |2
N0 ||wk ||2

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 4


Zero-forcing/decorrelator (2)
Signal power of kth stream is
E [wkH hk xk (wkH hk xk )H ] = wkH hk E [xk xkH ]hH w = Pk |wkH hk |2

Noise power of kth stream is


H H H H H H 2
E [wk n(wk n) ] = wk E [nn ]wk = N0 wk wk = N0 ||wk ||

SNR of kth stream is


Pk |wkH hk |2
N0 ||wk ||2

Applying Cauchy-Schwartz inequality on the numerator of SNR

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 4


Zero-forcing/decorrelator (2)
Signal power of kth stream is
E [wkH hk xk (wkH hk xk )H ] = wkH hk E [xk xkH ]hH w = Pk |wkH hk |2

Noise power of kth stream is


H H H H H H 2
E [wk n(wk n) ] = wk E [nn ]wk = N0 wk wk = N0 ||wk ||

SNR of kth stream is


Pk |wkH hk |2
N0 ||wk ||2

Applying Cauchy-Schwartz inequality on the numerator of SNR


Pk |wkH hk |2 P||wk ||2 ||hk ||2 Pk ||hk ||2
2
2
=
N0 ||wk || N0 ||wk || N0

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 4


Zero-forcing/decorrelator (2)
Signal power of kth stream is
E [wkH hk xk (wkH hk xk )H ] = wkH hk E [xk xkH ]hH w = Pk |wkH hk |2

Noise power of kth stream is


H H H H H H 2
E [wk n(wk n) ] = wk E [nn ]wk = N0 wk wk = N0 ||wk ||

SNR of kth stream is


Pk |wkH hk |2
N0 ||wk ||2

Applying Cauchy-Schwartz inequality on the numerator of SNR


Pk |wkH hk |2 P||wk ||2 ||hk ||2 Pk ||hk ||2
2
2
=
N0 ||wk || N0 ||wk || N0

We know that inequality is met with equality when wk = chk and c can be set as 1

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 4


Zero-forcing/decorrelator (2)
Signal power of kth stream is
E [wkH hk xk (wkH hk xk )H ] = wkH hk E [xk xkH ]hH w = Pk |wkH hk |2

Noise power of kth stream is


H H H H H H 2
E [wk n(wk n) ] = wk E [nn ]wk = N0 wk wk = N0 ||wk ||

SNR of kth stream is


Pk |wkH hk |2
N0 ||wk ||2

Applying Cauchy-Schwartz inequality on the numerator of SNR


Pk |wkH hk |2 P||wk ||2 ||hk ||2 Pk ||hk ||2
2
2
=
N0 ||wk || N0 ||wk || N0

We know that inequality is met with equality when wk = chk and c can be set as 1
We cannot choose wk = hk as it should be designed to be orthogonal to vectors
h1 , , hk1 , hk+1 , hnt
MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 4
Zero-forcing/decorrelator (2)
Signal power of kth stream is
E [wkH hk xk (wkH hk xk )H ] = wkH hk E [xk xkH ]hH w = Pk |wkH hk |2

Noise power of kth stream is


H H H H H H 2
E [wk n(wk n) ] = wk E [nn ]wk = N0 wk wk = N0 ||wk ||

SNR of kth stream is


Pk |wkH hk |2
N0 ||wk ||2

Applying Cauchy-Schwartz inequality on the numerator of SNR


Pk |wkH hk |2 P||wk ||2 ||hk ||2 Pk ||hk ||2
2
2
=
N0 ||wk || N0 ||wk || N0

We know that inequality is met with equality when wk = chk and c can be set as 1
We cannot choose wk = hk as it should be designed to be orthogonal to vectors
h1 , , hk1 , hk+1 , hnt except when h1 to hnt are orthogonal
MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 4
Zero-forcing/decorrelator (3)

Maximum rate of kth stream is

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 5


Zero-forcing/decorrelator (3)

Maximum rate of kth stream is


!
Pk |wkH h|2
Ck = log 1+
N0 ||wk ||2

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 5


Zero-forcing/decorrelator (3)

Maximum rate of kth stream is


! !
Pk |wkH h|2 Pk ||hk ||2
Ck = log 1+ < log 1+
N0 ||wk ||2 N0

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 5


Zero-forcing/decorrelator (3)

Maximum rate of kth stream is


! !
Pk |wkH h|2 Pk ||hk ||2
Ck = log 1+ < log 1+ = CSIMO
N0 ||wk ||2 N0

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 5


Zero-forcing/decorrelator (3)

Maximum rate of kth stream is


! !
Pk |wkH h|2 Pk ||hk ||2
Ck = log 1+ < log 1+ = CSIMO
N0 ||wk ||2 N0

Pnt
Sum rate is k=1 Ck

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 5


Zero-forcing/decorrelator (3)

Maximum rate of kth stream is


! !
Pk |wkH h|2 Pk ||hk ||2
Ck = log 1+ < log 1+ = CSIMO
N0 ||wk ||2 N0

Pnt
Sum rate is k=1 Ck
Decorrelator wk should be orthogonal to vectors h1 , , hk1 , hk+1 , hnt

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 5


Zero-forcing/decorrelator (3)

Maximum rate of kth stream is


! !
Pk |wkH h|2 Pk ||hk ||2
Ck = log 1+ < log 1+ = CSIMO
N0 ||wk ||2 N0

Pnt
Sum rate is k=1 Ck
Decorrelator wk should be orthogonal to vectors h1 , , hk1 , hk+1 , hnt

Decorrelator wk should be in the left nullspace of matrix k = [h1 , , hk1 , hk+1 , hn ]


H t

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 5


Zero-forcing/decorrelator (3)

Maximum rate of kth stream is


! !
Pk |wkH h|2 Pk ||hk ||2
Ck = log 1+ < log 1+ = CSIMO
N0 ||wk ||2 N0

Pnt
Sum rate is k=1 Ck
Decorrelator wk should be orthogonal to vectors h1 , , hk1 , hk+1 , hnt

Decorrelator wk should be in the left nullspace of matrix k = [h1 , , hk1 , hk+1 , hn ]


H t
Decorrelator wk can therefore be designed by finding the left nullspace of H k

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 5


Zero-forcing/decorrelator (3)

Maximum rate of kth stream is


! !
Pk |wkH h|2 Pk ||hk ||2
Ck = log 1+ < log 1+ = CSIMO
N0 ||wk ||2 N0

Pnt
Sum rate is k=1 Ck
Decorrelator wk should be orthogonal to vectors h1 , , hk1 , hk+1 , hnt

Decorrelator wk should be in the left nullspace of matrix k = [h1 , , hk1 , hk+1 , hn ]


H t
Decorrelator wk can therefore be designed by finding the left nullspace of H k
k is nr (nt 1) = nr nt + 1
Dimension of left nullspace of H

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 5


Zero-forcing/decorrelator (3)

Maximum rate of kth stream is


! !
Pk |wkH h|2 Pk ||hk ||2
Ck = log 1+ < log 1+ = CSIMO
N0 ||wk ||2 N0

Pnt
Sum rate is k=1 Ck
Decorrelator wk should be orthogonal to vectors h1 , , hk1 , hk+1 , hnt

Decorrelator wk should be in the left nullspace of matrix k = [h1 , , hk1 , hk+1 , hn ]


H t
Decorrelator wk can therefore be designed by finding the left nullspace of H k
k is nr (nt 1) = nr nt + 1
Dimension of left nullspace of H
To design wk , dimension of left nullspace of Hk > 0

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 5


Zero-forcing/decorrelator (3)

Maximum rate of kth stream is


! !
Pk |wkH h|2 Pk ||hk ||2
Ck = log 1+ < log 1+ = CSIMO
N0 ||wk ||2 N0

Pnt
Sum rate is k=1 Ck
Decorrelator wk should be orthogonal to vectors h1 , , hk1 , hk+1 , hnt

Decorrelator wk should be in the left nullspace of matrix k = [h1 , , hk1 , hk+1 , hn ]


H t
Decorrelator wk can therefore be designed by finding the left nullspace of H k
k is nr (nt 1) = nr nt + 1
Dimension of left nullspace of H
To design wk , dimension of left nullspace of Hk > 0

nr nt + 1 > 0

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 5


Zero-forcing/decorrelator (3)

Maximum rate of kth stream is


! !
Pk |wkH h|2 Pk ||hk ||2
Ck = log 1+ < log 1+ = CSIMO
N0 ||wk ||2 N0

Pnt
Sum rate is k=1 Ck
Decorrelator wk should be orthogonal to vectors h1 , , hk1 , hk+1 , hnt

Decorrelator wk should be in the left nullspace of matrix k = [h1 , , hk1 , hk+1 , hn ]


H t
Decorrelator wk can therefore be designed by finding the left nullspace of H k
k is nr (nt 1) = nr nt + 1
Dimension of left nullspace of H
To design wk , dimension of left nullspace of Hk > 0

nr nt + 1 > 0
nr > nt 1

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 5


Zero-forcing/decorrelator (3)

Maximum rate of kth stream is


! !
Pk |wkH h|2 Pk ||hk ||2
Ck = log 1+ < log 1+ = CSIMO
N0 ||wk ||2 N0

Pnt
Sum rate is k=1 Ck
Decorrelator wk should be orthogonal to vectors h1 , , hk1 , hk+1 , hnt

Decorrelator wk should be in the left nullspace of matrix k = [h1 , , hk1 , hk+1 , hn ]


H t
Decorrelator wk can therefore be designed by finding the left nullspace of H k
k is nr (nt 1) = nr nt + 1
Dimension of left nullspace of H
To design wk , dimension of left nullspace of Hk > 0

nr nt + 1 > 0
nr > nt 1
nr nt

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 5


Zero-forcing/decorrelator design

Figure: Bank of decorrelator, each estimating parallel data streams

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 6


Zero-forcing/decorrelator design

Figure: Bank of decorrelator, each estimating parallel data streams

General expression for decorrelator is (HH H)1 HH

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 6


Zero-forcing/decorrelator design

Figure: Bank of decorrelator, each estimating parallel data streams

General expression for decorrelator is (HH H)1 HH


Recall our system model is y = Hx + n. Decorrelator is the left inverse of H.

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 6


Zero-forcing/decorrelator design

Figure: Bank of decorrelator, each estimating parallel data streams

General expression for decorrelator is (HH H)1 HH


Recall our system model is y = Hx + n. Decorrelator is the left inverse of H.
Left inverse exists for a tall matrix with nr nt
MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 6
Zero-forcing/decorrelator performance

Figure: Achievable rate performance of decorrelator for nt = nr = 8 iid Rayleigh fading channel

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 7


Zero-forcing/decorrelator with successive interference cancellation

Figure: Decorrelator-SIC

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 8


Zero-forcing/decorrelator SIC performance

Figure: Ratio of rate to capacity for nt = nr = 8

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 9


Matched filter and decorrelator performance comparison

Figure: Ratio of rate to capacity for nt = nr = 8

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 10


Simplified receiver architectures MMSE (1)
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 11


Simplified receiver architectures MMSE (1)
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi + n
i =1,i 6=k
| {z }
zk

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 11


Simplified receiver architectures MMSE (1)
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi + n
i =1,i 6=k
| {z }
zk

= hk xk + zk

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 11


Simplified receiver architectures MMSE (1)
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi + n
i =1,i 6=k
| {z }
zk

= hk xk + zk

Here zk is colored Gaussian noise with covariance matrix Kzk = E[zk zH


k ]

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 11


Simplified receiver architectures MMSE (1)
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi + n
i =1,i 6=k
| {z }
zk

= hk xk + zk

Here zk is colored Gaussian noise with covariance matrix Kzk = E[zk zHk ]
H
nt nt
X X
Kzk = E[zk zH
k ] = E hi xi + n hi xi + n
i =1,i 6=k i =1,i 6=k

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 11


Simplified receiver architectures MMSE (1)
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi + n
i =1,i 6=k
| {z }
zk

= hk xk + zk

Here zk is colored Gaussian noise with covariance matrix Kzk = E[zk zHk ]
H
nt nt
X X
Kzk = E[zk zH
k ] = E hi xi + n hi xi + n
i =1,i 6=k i =1,i 6=k

nt nt
X X
= E hi xi + n xiH hH
i +n
H

i =1,i 6=k i =1,i 6=k

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 11


Simplified receiver architectures MMSE (1)
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi + n
i =1,i 6=k
| {z }
zk

= hk xk + zk

Here zk is colored Gaussian noise with covariance matrix Kzk = E[zk zHk ]
H
nt nt
X X
Kzk = E[zk zH
k ] = E hi xi + n hi xi + n
i =1,i 6=k i =1,i 6=k

nt nt
X X
= E hi xi + n xiH hH
i +n
H

i =1,i 6=k i =1,i 6=k



nt h i
X
= E hi xi xiH hH
i
+ E nnH
i =1,i 6=k
MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 11
Simplified receiver architectures MMSE (1)
Our system model
y = Hx + n
nt
X
= hi xi + n
i =1
nt
X
= hk xk + hi xi + n
i =1,i 6=k
| {z }
zk

= hk xk + zk

Here zk is colored Gaussian noise with covariance matrix Kzk = E[zk zHk ]
H
nt nt
X X
Kzk = E[zk zH
k ] = E hi xi + n hi xi + n
i =1,i 6=k i =1,i 6=k

nt nt
X X
= E hi xi + n xiH hH
i +n
H

i =1,i 6=k i =1,i 6=k



nt h i nt
X X
= E hi xi xiH hH
i
+ E nnH = Pi hi hH
i + N0 Inr
i =1,i 6=k i =1,i 6=k
MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 11
Simplified receiver architectures MMSE(2)
Recall any covariance matrix is a positive definite matrix

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 12


Simplified receiver architectures MMSE(2)
Recall any covariance matrix is a positive definite matrix
1 1
Kzk is a positive definite matrix (with SVD Kzk = UUH ) and Kz2k = U 2 UH

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 12


Simplified receiver architectures MMSE(2)
Recall any covariance matrix is a positive definite matrix
1 1
Kzk is a positive definite matrix (with SVD Kzk = UUH ) and Kz2k = U 2 UH
Our system model

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 12


Simplified receiver architectures MMSE(2)
Recall any covariance matrix is a positive definite matrix
1 1
Kzk is a positive definite matrix (with SVD Kzk = UUH ) and Kz2k = U 2 UH
Our system model
y = hk xk + zk

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 12


Simplified receiver architectures MMSE(2)
Recall any covariance matrix is a positive definite matrix
1 1
Kzk is a positive definite matrix (with SVD Kzk = UUH ) and Kz2k = U 2 UH
Our system model
y = hk xk + zk

We want to whiten the noise so that we can user matched filter to maximize SINR of xi

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 12


Simplified receiver architectures MMSE(2)
Recall any covariance matrix is a positive definite matrix
1 1
Kzk is a positive definite matrix (with SVD Kzk = UUH ) and Kz2k = U 2 UH
Our system model
y = hk xk + zk

We want to whiten the noise so that we can user matched filter to maximize SINR of xi
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 12


Simplified receiver architectures MMSE(2)
Recall any covariance matrix is a positive definite matrix
1 1
Kzk is a positive definite matrix (with SVD Kzk = UUH ) and Kz2k = U 2 UH
Our system model
y = hk xk + zk

We want to whiten the noise so that we can user matched filter to maximize SINR of xi
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk


y = k xk + zk
h

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 12


Simplified receiver architectures MMSE(2)
Recall any covariance matrix is a positive definite matrix
1 1
Kzk is a positive definite matrix (with SVD Kzk = UUH ) and Kz2k = U 2 UH
Our system model
y = hk xk + zk

We want to whiten the noise so that we can user matched filter to maximize SINR of xi
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk


y = k xk + zk
h

Covariance matrix of zk
H
E[zk
zk ]

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 12


Simplified receiver architectures MMSE(2)
Recall any covariance matrix is a positive definite matrix
1 1
Kzk is a positive definite matrix (with SVD Kzk = UUH ) and Kz2k = U 2 UH
Our system model
y = hk xk + zk

We want to whiten the noise so that we can user matched filter to maximize SINR of xi
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk


y = k xk + zk
h

Covariance matrix of zk
 
H 1 H
1
zk ] = E Kzk 2 zk zk Kzk 2
E[zk

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 12


Simplified receiver architectures MMSE(2)
Recall any covariance matrix is a positive definite matrix
1 1
Kzk is a positive definite matrix (with SVD Kzk = UUH ) and Kz2k = U 2 UH
Our system model
y = hk xk + zk

We want to whiten the noise so that we can user matched filter to maximize SINR of xi
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk


y = k xk + zk
h

Covariance matrix of zk
 
1 1 1
h i 1
H H H
zk ] = E Kzk 2 zk zk Kzk 2 = Kzk 2 E zk zk Kzk 2
E[zk

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 12


Simplified receiver architectures MMSE(2)
Recall any covariance matrix is a positive definite matrix
1 1
Kzk is a positive definite matrix (with SVD Kzk = UUH ) and Kz2k = U 2 UH
Our system model
y = hk xk + zk

We want to whiten the noise so that we can user matched filter to maximize SINR of xi
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk


y = k xk + zk
h

Covariance matrix of zk
 
1 1 1
h i 1 1 1
H H H
zk ] = E Kzk 2 zk zk Kzk 2 = Kzk 2 E zk zk Kzk 2 = Kzk 2 Kzk Kzk 2
E[zk

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 12


Simplified receiver architectures MMSE(2)
Recall any covariance matrix is a positive definite matrix
1 1
Kzk is a positive definite matrix (with SVD Kzk = UUH ) and Kz2k = U 2 UH
Our system model
y = hk xk + zk

We want to whiten the noise so that we can user matched filter to maximize SINR of xi
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk


y = k xk + zk
h

Covariance matrix of zk
 
1 1 1
h i 1 1 1
H H H
zk ] = E Kzk 2 zk zk Kzk 2 = Kzk 2 E zk zk Kzk 2 = Kzk 2 Kzk Kzk 2 = Inr
E[zk

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 12


Simplified receiver architectures MMSE(3)
Resultant system model is
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 13


Simplified receiver architectures MMSE(3)
Resultant system model is
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk


y = k xk + zk
h

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 13


Simplified receiver architectures MMSE(3)
Resultant system model is
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk


y = k xk + zk
h
H H h
H zk
h ky = h k k xk + hk matched filtering

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 13


Simplified receiver architectures MMSE(3)
Resultant system model is
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk


y = k xk + zk
h
H H h
H zk
h ky = h k k xk + hk matched filtering
1 1 1
hH 2 hH 2 k xk + hH K 2z
k Kzk y = k Kzk h k zk k

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 13


Simplified receiver architectures MMSE(3)
Resultant system model is
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk


y = k xk + zk
h
H H h
H zk
h ky = h k k xk + hk matched filtering
1 1 1
hH 2 hH 2 k xk + hH K 2z
k Kzk y = k Kzk h k zk k

1 1 1 1 1 1
hH
k Kzk
2 Kzk 2 y = hH 2 2 H 2 2
k Kzk Kzk hk xk + hk Kzk Kzk hzk

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 13


Simplified receiver architectures MMSE(3)
Resultant system model is
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk


y = k xk + zk
h
H H h
H zk
h ky = h k k xk + hk matched filtering
1 1 1
hH 2 hH 2 k xk + hH K 2z
k Kzk y = k Kzk h k zk k

1 1 1 1 1 1
hH
k Kzk
2 Kzk 2 y = hH 2 2 H 2 2
k Kzk Kzk hk xk + hk Kzk Kzk hzk
1 1 H 1
hH
k Kz y = hH
k Kz hk xk + hk Kz zk
k k k

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 13


Simplified receiver architectures MMSE(3)
Resultant system model is
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk


y = k xk + zk
h
H H h
H zk
h ky = h k k xk + hk matched filtering
1 1 1
hH 2 hH 2 k xk + hH K 2z
k Kzk y = k Kzk h k zk k

1 1 1 1 1 1
hH
k Kzk
2 Kzk 2 y = hH 2 2 H 2 2
k Kzk Kzk hk xk + hk Kzk Kzk hzk
1 1 H 1
hH
k Kz y = hH
k Kz hk xk + hk Kz zk
k k k

vkH y = vkH hk xk + vkH zk

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 13


Simplified receiver architectures MMSE(3)
Resultant system model is
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk


y = k xk + zk
h
H H h
H zk
h ky = h k k xk + hk matched filtering
1 1 1
hH 2 hH 2 k xk + hH K 2z
k Kzk y = k Kzk h k zk k

1 1 1 1 1 1
hH
k Kzk
2 Kzk 2 y = hH 2 2 H 2 2
k Kzk Kzk hk xk + hk Kzk Kzk hzk
1 1 H 1
hH
k Kz y = hH
k Kz hk xk + hk Kz zk
k k k

vkH y = vkH hk xk + vkH zk


1
where vk = Kz hk
k

Weight vector vk is called MMSE combiner

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 13


Simplified receiver architectures MMSE(3)
Resultant system model is
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk


y = k xk + zk
h
H H h
H zk
h ky = h k k xk + hk matched filtering
1 1 1
hH 2 hH 2 k xk + hH K 2z
k Kzk y = k Kzk h k zk k

1 1 1 1 1 1
hH
k Kzk
2 Kzk 2 y = hH 2 2 H 2 2
k Kzk Kzk hk xk + hk Kzk Kzk hzk
1 1 H 1
hH
k Kz y = hH
k Kz hk xk + hk Kz zk
k k k

vkH y = vkH hk xk + vkH zk


1
where vk = Kz hk
k

Weight vector vk is called MMSE combiner

vk = K1
z hk = k

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 13


Simplified receiver architectures MMSE(3)
Resultant system model is
1 1 1
Kzk 2 y = Kzk 2 hk xk + Kzk 2 zk
| {z } | {z } | {z }

y
hk
zk


y = k xk + zk
h
H H h
H zk
h ky = h k k xk + hk matched filtering
1 1 1
hH 2 hH 2 k xk + hH K 2z
k Kzk y = k Kzk h k zk k

1 1 1 1 1 1
hH
k Kzk
2 Kzk 2 y = hH 2 2 H 2 2
k Kzk Kzk hk xk + hk Kzk Kzk hzk
1 1 H 1
hH
k Kz y = hH
k Kz hk xk + hk Kz zk
k k k

vkH y = vkH hk xk + vkH zk


1
where vk = Kz hk
k

Weight vector vk is called MMSE combiner


1
nt
X
vk = K1
zk hk = Pi hi hH
i + N0 Inr hk
i =1,i 6=k

MIMO Wireless Communications (Rohit Budhiraja, IITK) Channel capacity 13

Potrebbero piacerti anche