Adaptive Multicarrier CDMA Space-Time Receivers - Adaptive Signal Processing in Wireless Communications - page 290

and amplitudes, received power, CFO). The fourth module is a null-constraint generator
common to all in-cell users. It gathers the data decisions and channel parameter esti-
mates from the second and third modules dedicated to each in-cell user-carrier pair in
order to reconstruct the total in-cell signal vector I n . Then to each combiner, say of the
desired user-carrier pair as illustrated in Figure 10.1b , it passes on the associated null
constraint (i.e., I k,n = I n - s k,n U n,k,n ) calculated with the least computations by simple sub-
traction from I n of the desired signal contribution from the corresponding user-carrier
pair.
The implementation of an adaptive closed-loop CFOR* jointly with multicarrier and
multiuser detection (here by MC-ISR) requires careful attention regarding the order in
which these two tasks should be processed. Indeed, conventional operation of CFOR at
an early processing stage† prior to interference suppression would require (on the uplink
only) as many independently CFO-compensated observations and interference null con-
straints as received in-cell users, thereby resulting in a tremendous complexity increase.
Here we develop an efficient post-interference-suppression CFOR scheme by splitting
the MC-ISR combining operation of equation (10.29) into two steps, an observation-
cleaning projection and an MRC combining, and by inserting CFO compensation in
between as follows:

d

d

Y = Π Y , (10.47)

Π ,, kn kn , n

d d d

 

∆∆ f = f

−1 δ ,

+ f (10.4 8)

n n n

d





d d

− 2π j ∆ f nT

n

Y = Y e , (10.49)

Π ,, kn Π ,, kn

d d

ˆ ˆ

d

V = Π V , (10. 50)

Π ,, kn kn , Π ,, kn

H

d

ˆ 

d

VY

= Π

,, kn Π ,, kn

ˆ

d

s . (10. 51)

kn ,

2

d

ˆ

V

Π ,, kn

The cleaning projection of equation (10.47) results in an almost interference-free
observation Y d ∏ ,k,n and allows for CFO estimation and compensation in equations (10.48)
and (10.49), respectively, using the CFOR module of the single-user MC-STAR (refer
to [ 32 ] and [ 48 ] for details on how to estimate the CFO adjustment term in equation
(10.49)), and for MRC combining in equation (10.51) using the projected estimate of the

* In contrast to open-loop structures, closed-loop CFOR reduces the channel time variations and
greatly improves their tracking [32].
† Usually CFOR is embedded in the RF chain or plugged to the preprocessor output.