- Home
- Technology
*Multiuser detection new*

prev

next

out of 31

View

81Download

0

Embed Size (px)

A project reported submitted in partial fulfillment of the requirements of the course EEL6503: Spread Spectrum and CDMA, Fall 2001

Submitted by Arun Avudainaygam

Project Website: http://arun-10.tripod.com/mud/mud.html

LINEAR AND ADAPTIVE

LINEAR MULTIUSER DETECTION

IN CDMA SYSTEMS

SECTION 0

Introduction

Multiuser detection is a technology that spawned in the early 80s. It has now developed into an important, full-fledged field in multi-access communications. Multiuser Detection (MUD) is the intelligent estimation/demodulation of transmitted bits in the presence of Multiple Access Interference (MAI). MAI occurs in multi-access communication systems (CDMA/ TDMA/ FDMA) where simultaneously occurring digital streams of information interfere with each other. Conventional detectors based on the matched filter just treat the MAI as additive white gaussian noise (AWGN). However, unlike AWGN, MAI has a nice correlative structure that is quantified by the cross-correlation matrix of the signature sequences. Hence, detectors that take into account this correlation would perform better than the conventional matched filter-bank. MUD is basically the design of signal processing algorithms that run in the black box shown in figure 0.1. These algorithms take into account the correlative structure of the MAI.

Figure 0.1 A typical multiuser detector

0.1 Overview of the project This project investigates a couple of different approaches to linear multiuser detection in CDMA systems. Linear MUDs are detectors that operate linearly on the received signal statistic i.e., they perform only linear transformations on the received statistic. This report assumes that the reader has a basic knowledge of probability theory and random processes and is familiar with the fundamental concepts of spread spectrum and CDMA systems.

0.2 Organization of the report This report has been typeset in the 2-column landscape format to help preserve continuity in the mathematical development in some sections. The reader need not keep flipping pages to keep up with the large number of equations that are involved in certain sections. The report is orgainzed as follows: The next sub-section introduces the system model that will be used

throught this project. Section I presents the conventional method of demodulating

mutually interfering signals: the matched filter (which treats the MAI as AWGN).

Section II studies the decorrelating detector which takes the matched filter one step further by taking into account the correlative structure of the MAI.

Section III presents the minimum mean square error (MMSE) linear detector which is a compromise between the matched filter approach and the decorrelating detector. Two different adaptive implementations (least mean square or LMS and blind adaptation) are implemented and the convergence properties are studied.

0.3 System model The K-user discrete time basic synchronous CDMA model has been used throughout the development of this project. The case of antipodally modulated user information (BPSK modulation) spread using BPSK spreading is considered. To make the project realizable in the time allocated, a very small spreading sequence of length 31 was used. A preferred pair [1,2] of m-sequences generated by the primitive polynomials and were used for all the 2-user scenarios. For the number of users greater than 2, gold codes generated by the 2 m-sequences

described above were used. See Appendix A for a list of codes in this project. Unless otherwise mentioned, in all the simulations perfect power control is assumed (i.e, the received amplitudes of all the users are assumed to be the same. The signal at the receiver is given by

=

+=K

1kkkk n(t)(t)sbAy(t) (0.1)

, where sk is the signature waveform of the kth user (sk is normalized to

have unit energy i.e., < sk ,sk > =1). For BPSK spreading with an m-sequence of length 31, the signature waveform is defined as

=

=

30

0kcTkk )kT-(tpa(t)s , T=bit period, Tc =chip interval (0.2)

where ak represents the normalized spreading sequence. bk is the input bit of the kth user, bk {-1,1}. Ak is the received amplitude of the kth user. n(t) is additive white gaussian noise with PSD No .

Since synchronous CDMA is considered, it is assumed that the receiver has some means of achieving perfect chip synchronization. The cross-correlation of the signature sequences are defined as

=

==

N

1kjijiij (k)(k)ssss (0.3)

where N is the length of the signature sequence (31 in our case). The cross-correlation matrix is then defined as R={ij}

i.e.,

=

KKKK

K

K

21

212221

11211

R .(0.4)

R is a symmetric, non-negative definite, toeplitz matrix.

SECTION I

TheMatched-Filter

Bank

This section introduces and analyses the matched filter bank detector which was the conventional and most simplest way of demodulating CDMA signals (or any other set of mutually interfering digital streams). The matched filter also forms the front-end in most MUDs and hence understanding the operation is crucial in appreciating the evolution of MUD technology. 2.0 Receiver Operation In conventional single-user digital communication systems, the matched filter is used to generate sufficient statistics for signal detection. In the case of a multi-user system, the detector consists of a bank of matched filters (each matched to the signature waveforms of different users in the case of CDMA). This is shown in figure 1.1. This type of detector is referred to as the conventional detector in MUD literature. It is worth mentioning that we need exact knowledge of the users signature sequences and the signal timing in order to implement this detector.

The decision statistic a the output of the Kth matched filter is given by

(t)dty(t)sy0

kk =T

...(2.1)

where y(t) and sk(t) is given by (0.1) and (0.2). Expanding (2.1),

(t)dtsn(t)(t)sbAy0

k

K

1jjjjk

+==

T

.(2.2)

Using (0.3)

=

+=K

1jkjkjjk nbAy ..(2.3)

where (t)dtn(t)sn0

kk =T

.(2.4)

Since 11=1, (2.3) simplifies to

=

++=K

kj1j

kjkjjkkk nbAbAy ..(2.5)

The 2nd term in (2.5) is the MAI. The matched filter treats the MAI just as white noise. The noise variance at the output of the matched filter is given by

Similarly, the noise covariance can be shown to be

ijoji N)nE(n = ..(2.7) Hence the noise covariance matrix can be defined as { } Rnn oijijoT NN]E[ == ..(2.8) where R is given by (0.4) and n=[n1,n2,,nk]T. Stacking up (2.5) for all the users we get

Figure 1.2 A matched filter bank

(2.6)

+

=

K

2

1

K

2

1

2

1

21

212221

11211

K

2

1

n

nn

b

bb

A00

0A000A

y

yy

KKKKK

K

K

.......(2.9)

In matrix notation we have, y=RAb+n(2.10)

Figure 2.2 shows the error rate performance of the bank of matched filters. The simulation scenario is explained in section 0.3. It is observed that as the MAI increases (the number of users increases) the performance becomes poor. This is because the detector ignores the cross-talk between users (the MAI) as white noise. Good MUDs, as described in the next few sections, take into the account the correlative property of the cross-talk.

2.1 Limitations of the conventional detector Although {y1,y2,,yk} are sufficient statistics for detecting {b1,b2,,bk}, yk is not a sufficient statistic for detecting bk. The conventional detector makes the mistake of making this assumption (yk is a sufficient statistic for detecting bk) by ignoring the MAI as background noise. This is one reason for the poor performance of the matched filter bank when the number of users is large. Another serious limitation of the conventional detector is that it is seriously affected by the near-far problem. This causes a significant degradation in the system performance even when the number of users is very small. This fact will now be illustrated with an example. This example is a condensed version of a scenario described in [3]. Adapting (2.5) to the 2 user scenario we get,

1122j2111 nbAbAy ++= ..(2.11) It is now obvious that the bit error probability for user 1 is

+

+=

o

1221

o

1221s N

AAQ

NAA

Q21P

1(2.12)

The probability of bit error is then readily upper bounded as

o

1221s N

AAQ

21P

1.(2.13)

The fact that Q is a monotonically decreasing function was used to get the upper bound. If the interferer is not dominant i.e., A212A1, the bound becomes greater than half. Consider the case when there is no noise in the system (i.e., No=0) and the interferer

is dominant, then (2,13) gives 21P

1s= . This is because the polarity of the

matched filter outputs is now governed by b2 rather than b1. Hence we see that in the absence of noise, though highly hypothetical, the matched filter receiver reduces to flipping a coin and deciding the output bits. This is an undesirable feature of the conventional detector (may perform better in the presence of noise than in the absence of noise).

Figure 1.2 B.E.R performance of the matched filter bank detector (perfect power control)

2.2 The conventional detector as a