Presentation is loading. Please wait.

Presentation is loading. Please wait.

By Nikita Goel Prerna Mayor Sonal Ambwani

Similar presentations


Presentation on theme: "By Nikita Goel Prerna Mayor Sonal Ambwani"— Presentation transcript:

1 By Nikita Goel Prerna Mayor Sonal Ambwani
STUDY OF DS-CDMA SYSTEM AND IMPLEMENTATION OF ADAPTIVE FILTERING ALGORITHMS By Nikita Goel Prerna Mayor Sonal Ambwani

2 OBJECTIVES Extensive Analysis of the Adaptive Algorithms in MATLAB and LabVIEW and comparison of the Algorithms on various points such as convergence, BER performance. The basic signal model chosen is that of a multi-user DS-CDMA system. Implementation of the Algorithms in C language. Design of a suitable GUI for the system. Interfacing the TI-DSP kit with the computer using the C codes.

3 Understanding the Signal Model
We are dealing with a DS-CDMA system with multi-user communication. (In a CDMA system, all the users transmit in the frequency spectrum simultaneously and are coded using spread-spectrum techniques.) However, we are interested in the communication of a single user of interest. The other users become ‘Interferers’. Key idea of the project: To develop online, adaptive algorithms which are recursive in nature to process real time data and work towards the minimization of the MEAN-SQUARE ERROR (MMSE) between the received signal and the desired response.

4 BASIC BLOCK DIAGRAM ANTENNA ARRAY Where, the received signal rk (t) = xk (t) + ik (t) + nk (t) for k=1,2,…M. M= Number of antenna array elements in the receiver. Wk *= Adaptive tap weights ( called adaptive because the real-time received data rk(t) is unknown or random in practical cases and a stochastic approach is required to estimate it ) Where, the received vector rk (t) = xk (t) + ik (t) + nk (t) for i=1,2,…M. M= Number of antenna array elements in the receiver. Wk *= Adaptive tap weights ( called adaptive because the real-time received data rk(t) is unknown or random in practical cases and a stochastic approach is required to estimate it )

5 Received Vector R(t)=X(t) + I(t) + N(t) Weighted signal y(t) = WH R(t)
After arranging the M received signals rk(t) and the M tap weights Wk * in the form of vectors R(t) and WH respectively, the following mathematics is performed: Received Vector R(t)=X(t) + I(t) + N(t) Weighted signal y(t) = WH R(t) Error e(t) = y(t) – d(t) , where d(t) is the desired response or the pilot signal which directly correlates with the user of interest. Mean Square Error= |e(t)|2 The weight vector W is derived such that it minimizes the Mean-Square Error, |e(t)|2

6 Why Adaptive Algorithms?
Fast and considerably reduce system overheads as data can be processed online. Process real-time and random data (Online). Tend to the WMMSE in the mean-square sense with probability 1. Adapt easily to the communication system data. Follow a general recursive pattern.

7 LMS ( Least Mean Squares)
Wn+1 = Wn –μ rn (rn H Wn – dn*) Wlms Wmmse (in the mean square sense) Advantages: simplicity in implementation stable and robust performance against different signal conditions Disadvantage: Relatively slow Convergence ( but that can be overcome by using normalised LMS)

8 CMA ( Constant Modulus Algorithm)
Wn+1 = Wn –μ( rn rn H Wn( | Wn H rn|2 – A2 )) p(t) Advantage: Blind, Online scheme ( no pilot signal) Tb Disadvantages: Needs a constant modulus signal of interest Algorithm will not work for power-controlled CDMA (wireless) system

9 RLS ( Recursive Least Squares)
Key Idea : βn-k |e(k)|2 is to be minimized. Wn+1 = Wn –( R-1rn ( rn H Wn – dn* )) 0<β<1 (usually close to 1). Advantages: Very Fast Disadvantage: Increase in computational complexity ( something we realized while writing the C programs and the LabVIEW codes)

10 EXPERIMENTAL DATA Number of users (K) = 5 Number of Array Elements (M) = 12 Direction of Arrival for the user of interest (the LOOK angle)=60o Direction of Arrival of the Interferers = [-80o,-15o, 0o, 40o] Number of iterations or data points considered (N)=1000

11

12

13 EXPERIMENTAL RESULTS Observe the peak at 60o ( the LOOK angle) and nulls at [-80o,-15o, 0o,40o] the angles of arrivals of the interferers. The RLS curve converges fastest and the best to the MMSE curve.

14 deciphers which bit was transmitted by the user-of-interest.
The above diagram depicts the adaptive algorithm at the receiver which essentially deciphers which bit was transmitted by the user-of-interest. The received vector r is passed through a linear filter characterized according to the adaptive filtering technology used (LMS, RLS, CMA etc) . The bit transmitted is decided by performing a hypothesis testing on Sign(W T *r), i.e if Sign(W T *r), >0 then a +1 was transmitted, and if it is <0 then -1 was transmitted.

15 The figure depicts the BER versus the user-of-interest SNR
Convergence: RLS converges best and fastest to the MMSE

16 GUI The GUI has been implemented in
LabVIEW, because of the user-friendly nature of the interface. We have also coded the algorithms in LabVIEW, because of the novelty of the idea. contd..

17 About LabVIEW It is an out and out graphical programming tool with an excellent and user-friendly interface. Terminology: A program in LabVIEW is called a VI (Virtual Instrument). The graphical programming is done on the Block-Diagram and the user interface is called the Front Panel.

18 A Look at the GUI …where the user can select the signals’ angles of arrivals and the operating SNR (in dB)

19 …where the Online Adaptive Algorithm can be chosen
…where the Online Adaptive Algorithm can be chosen. On clicking on one of the control buttons, thepower beam-pattern and the BER curves can be obtained.

20 BER plots from LabVIEW

21

22

23 A look at the TMS-320-C6211 DSP Board

24 The DSP Board Interfaced with the Code Composer Studio that executes the
C codes.

25 Results as seen after executing the LMS C code on the TI DSP
Weight Vector Wlms

26 Weight Vector Wlms

27 OBJECTIVES ACHIEVED The above described algorithms have been implemented in MATLAB. The codes have also been implemented in LabVIEW, and the GUI has been developed in LabVIEW. Stand-alone codes have been written in C. Comparative analysis has been carried out by varying the number of iterations N, changing the direction of arrivals of the user of interest and the interferers. Bit-Error-Rate performance for the three algorithms, and the convergence issues have been compared. Successfully Interfaced the TMS-320-C6211 DSP kit with the C codes.

28 Additional Work Done MATLAB analysis of some variants of LMS like, sign-LMS, Constraint LMS, etc has been done. Analysis of a Space Division Multiple Access (SDMA) system in MATLAB, LabVIEW and C. A simple Joint Space-Time Multiple Access system has been considered.

29 A preview of the Joint Space -Time System
It is SDMA combined with the DS-CDMA system, i.e. there is an Antenna Array at the receiver which exploits the spatial characteristics of the user of interest and the interferers. We have not considered Multipath fading and Rayleigh fading. We have an AWGN channel, and AWGN channel fading is taken into consideration.

30 Joint ST : The 3-D space and time plots
The DS-CDMA signals have a 12 bit signature sequence generated by a PN generator. T=transmitted bit period for the user-of-interest and the interferers Ts =bit period of each signature bit (chip period) Thus, T=12* Ts There are 12 antenna array elements at the receiver. We have plotted the power beam pattern with respect to each signature bit, thus obtaining a 3-D plot.

31 3-D Power Beam Patterns for Joint ST systems
LMS

32 CMA

33 RLS


Download ppt "By Nikita Goel Prerna Mayor Sonal Ambwani"

Similar presentations


Ads by Google