Optimal Sequence Allocation and Multi-rate CDMA Systems Krishna Kiran Mukkavilli, Sridhar Rajagopal, Tarik Muharemovic, Vikram Kanodia
Motivation 4 3rd Generation Comm. Systems –Multimedia(Data, Voice, Video) –Multiple Rate Comm. 4 Multi-Rate Detection 4 Users entering/leaving the system 4 Optimal Sequence Allocation to achieve Capacity.
Outline 4 Conventional CDMA multiuser system 4 Discussion of multirate systems –Methods of multirate CDMA access –Performance of multiuser detectors 4 Interference avoidance 4 Application to variable number of users
Multi-rate CDMA systems 4 Multi code access (MC) –Give more Codes 4 Variable spreading length (VSL) –Change Spreading Length 4 Variable chip rate(VCR) –Change Chip Frequency
Multi code (MC) 4 Higher Rate users assigned more codes 4 Data transmitted in parallel 4 “Virtual User” Concept Same Spreading for all users.
Multi Code User Rate R User Rate 2R Code 1 Code 2 Code 3 T
Variable spreading length(VSL) 4 Higher Rate Users allocated smaller spreading lengths 4 For detection, rate of slowest user is considered. 4 More bits of higher rate users detected per bit of lower rate users 4 For detection, put 0’s
Variable Spreading Length User Rate R User Rate 2R 2T T
Variable Chip Rate(VCR) 4 User allocated different chip rates 4 Larger Bandwidth required 4 Requires more RF hardware –Oscillators 4 Not practical for implementation
Variable Chip Rate User Rate R User Rate 2R T 2T
Implementation Aspects 4 VSL and VCR have a sparse correlation matrix 4 VCR requires larger bandwidth 4 MC requires more codes 4 VSL proposed for next generation systems
Multiuser Detectors 4 Maximum likelihood detector (MLD) 4 Conventional single user detector (SUD) 4 MMSE detector 4 Decorrelating detector
Simulations 4 Four users –2 users at rate R –2 users at rate 2R 4 Random Codes 4 Spread length 32 for low rate user bits 4 Channels –AWGN –Fading - Jakes Model
Investigate... 4 Performance of multiuser detectors 4 Near far problem in detectors 4 Performance of high rate and low rate users in MC and VSL systems –All users with equal power –Users with unequal power
SNR BER BER comparison for different detectors in multi code system MLD MMSE Decorrelator Single user detector
BER comparison for detectors with unequal powers SNR BER MLD Equal Power MLD Unequal Power SUD Equal Power SUD Unequal Power
SNR BER Comparison of Different Rate Users in MC and VSL High rate MC High rate VSL Low rate MC Low rate VSL
VSL System 4 Virtual user from high rate user –lower spreading length –lower interference (other virtual users are orthogonal) 4 High rate user –interference from same number of virtual users with lower spread length
Variable Spreading Length User Rate R User Rate 2R 2T T
SNR BER Near Far effect for Different Rate Users in MC and VSL Low rate MC Low rate VSL High rate MC High rate VSL
Results 4 Multi Code –High rate and low rate users have same performance (both BER and NFR) 4 VSL –Low rate users have bad BER and NFR –High rate users’ performance is similar to multicode access system.
Interference Avoidance in Wireless Multiuser Systems 4 Interference Avoidance send where there is less noise 4 Fixed modulation - traditional approach –TDMA –FDMA –CDMA –CWMA 4 Future wireless systems - dynamically adapt to the changing interference pattern
Preliminaries for Multiuser Systems 4 System model: Capacity region: R1R1 R2R2 X1X1 XMXM N Y (class notes pg. 5-12)
Iteration Number Total Square Correlation Total Square Correlation vs Iteration Number Total Square Correlation Optimum Lower Bound
Preliminaries for Multiuser Systems 4 Sum Capacity: –W : channel bandwidth – P i : power of i-th user – N 0 : noise power spectrum (class notes pg. 5-12)
Multiuser Spread Spectrum Systems 4 System model Y sMsM X1X1 XMXM N s1s1
Multiuser Spread Spectrum Systems 4 Sum Capacity: –Optimum sequences maximize Sum Capacity 4 Total Square Correlation (TSC): – –Max. Sum Capacity Min. TSC
Eigen-Algorithm 4 Iterative reduction of TSC –User k updates his spreading sequence –Rayleigh quotient –Choose s k to be eigenvector with smallest eigenvalue
SNR BER Performance comparison of optimal codes with random codes Random Code Allocation Optimal Code Allocation
SNR BER BER Performance with an Incoming User Random Code to new user Iteration for new user only Optimal Code Allocation
Conclusions 4 Significant improvement in performance with optimal codes 4 Iterative algorithm compatible with user dynamics 4 Good sub-optimal schemes for user addition 4 Can be combined with the multi- rate schemes