Digital Signal Processing Reference
In-Depth Information
1
α
Σ
−
1
/
2
Λ
−
1
θ
=
(using Eq. (17.18))
h
Substituting from Eq. (17.10) we therefore have
1
α
Λ
−
1
/
2
Λ
−
1
/
2
θ
Λ
g
=
.
(17
.
20)
c
Now, the diagonal matrix
Λ
−
1
/
θ
is unitary because the diagonal elements
have the form
e
−jθ
k
/
2
.
Thus, if we replace
Σ
f
and
Λ
g
in Fig. 17.3(a) as
follows:
Σ
f
→
Σ
f
×
Λ
−
1
/
2
θ
Λ
g
→
Λ
g
×
Λ
1
/
2
θ
and
,
the system performance is unaffected, that is, neither the transmitted power
nor the MSE is changed. From Eq. (17.20) we have
1
α
Λ
−
1
/
2
Λ
g
×
Λ
1
/
2
θ
=
,
c
whereas from Eqs. (17.17) and (17.10) we have
Σ
f
×
Λ
−
1
/
2
θ
=
α
Λ
−
1
/
2
c
.
This shows that the optimal system of Fig. 17.3(a) can be replaced with the
system shown in Fig. 17.4. Equation (17.19) is obtained by replacing
σ
h,k
with
|
C
[
k
]
|
in Eq. (12.66).
17.3.2 Pure-MMSE transceiver
For the MMSE transceiver without zero-forcing constraint the diagonal matrices
Σ
f
and
Σ
g
were derived in Sec. 13.6. These are reproduced below with the
substitution
P
=
J
=
M.
First, the diagonal matrix
Σ
f
at the transmitter is
⎡
⎣
⎤
⎦
√
q
00
0
...
0
√
q
11
0
...
0
Σ
f
=
σ
q
σ
s
.
(17
.
21)
.
.
.
.
.
.
√
q
M−
1
,M−
1
0
0
...
Here
⎧
⎨
D
1
σ
h,k
σ
h,k
−
0
≤
k
≤
K
−
1
q
kk
=
(17
.
22)
⎩
0
otherwise,
where
K−
1
p
σ
q
1
σ
h,k
+
k
=0
D
=
(17
.
23)
K−
1
1
σ
h,k
k
=0
Search WWH ::
Custom Search