An Adaptive PatternedSubcarrier Allocation Algorithm for LowComplexity Multiuser OFDM System

1
An Adaptive Patterned-Subcarrier Allocation
Algorithm for Low-Complexity Multiuser OFDM System

• Youngok Kim, Haewoon Nam, Sanghyun Chi,
• and Baxter F. Womack
• Dept. of Electrical and Computer Engineering
• The University of Texas at Austin
• VTC 2005
• Dallas, TX
• Sep. 25-28, 2005

2
Introduction

• Multiuser OFDM
• ? multiple access technique sharing the
  subcarriers with multiple
• users
• Fixed v.s. Adaptive subcarrier allocation
  algorithm
• Different users experience mutually independent
  fading
• Given equal power per subcarrier, the channel
  capacity is proportional to the channel gain
• Given channel state information (CSI), capacity
  is enhanced by an adaptive SA algorithm
• FFT is employed for subcarrier identification

3
System Model

• N subcarriers, M users, and K(N/M) subcarriers
  per a user

Subcarrier Identification and Selection (SIS)
Block diagram of conventional multiuse OFDM
4
System Model

• Frequency selective fading channel
• Transmitter knows channel state information
  (CSI)
• Receiver knows subcarrier allocation (SA)
• information
• Adaptive modulation is considered for capacity
• enhancement
• N is power of two

5
Problem Statement

• Each user uses only K subcarriers out of N
  subcarriers
• Full N-FFT is used for the subcarrier
  identification
• For each user, the conventional SIS process is
  inefficient
• ? Unnecessary Computational Cost!!
• Focus on SA for high system capacity and SIS
  process for
• low computational complexity
• ? Find an efficient SIS process that is
  computationally
• efficient while high system
  capacity is preserved

6
Proposed System

• Given CSI, capacity is enhanced by an adaptive
  SA
• algorithm with patterns
• The SIS process is performed via a Partial FFT
• ? Selected-Subcarrier FFT Method (SSFFT)
• The proposed SSFFT replaces the N-FFT and the
  selection
• processes of the conventional system
• The SSFFT consists of smaller FFTs and linear
  transition
• matrixes

(Not a Full FFT)
7
Proposed System
Base station transmitter
User 1 data
Subcarrier 1
Subcarrier Allocation And Encoding
IFFT
Add cyclicprefix
User 2 data
Subcarrier 2


User M data
Subcarrier N
User 1 CSI feedback
User 2 CSI feedback

User m CSI feedback
Subcarrier 1
Remove cyclicprefix
Decoder
SSFFT
Subcarrier 2
Dr or Dr
K-FFT
Dc or Dc
User m data

Subcarrier K
User m receiver
Proposed system with the SSFFT scheme
8
Adaptive Patterned SA

• Given CSI, subcarriers with high channel gains
  are
• assigned to each user
• The ratio M ( N/K) is expressed as M pq
• where p and q are prime 2 factors
• The assigned subcarriers (patterns) for each
  user
• ? are equally q spaced when modulo-process is
  applied
• ? do not include Kq spaced subcarriers
• By changing q value, find the maximized system
  capacity
• ? q1 or M (case 1) and q2 to M/2 (case 2)

9
SSFFT Method
N x N DFT matrix
q spaced subcarriers
K x N DFT matrix
10
SSFFT Method
K x N DFT matrix
Extract every p-th column
K x K DFT matrix
SK,N consists of M SK,K
Note
11
SSFFT Method
K x K DFT matrix K,K
Linear Transition
K-FFT matrix
or
12
SSFFT Method
or
where
13
SSFFT Method
Example N16, M4, q1, Im0,1,2,3 ? K4, p4
14
Computation Complexity
Exponentially increase
15
Simulation Results

• The Comparison of the overall system capacity
• Fixed v.s. Adaptive w/o patterns v.s. Proposed
• Simulation Environment
• Equal amount of power on each subcarrier
• Equal number of subcarriers on each user
• The capacity of i-th subchannel of user m is
  defined as
• where B is a bandwidth, hm,i is the channel
  gain
• M 8 , N 1024, and B 10 MHz

16
Simulation Results - continued

• Frequency selective Rayleigh fading channel with
  an exponential power delay profile

17
Simulation Results - continued

• Performance comparison
• The proposed schemes outperform the fixed
  allocation scheme at same SNR
• closes to that of an adaptive scheme w/o
  patterns

18
Conclusion

• With patterned SA and SSFFT method,

• High system capacity is achieved !!

• Low computational complexity is achieved !!

19

Thank you !!
20
K-FFT Matrix
21
Modulo-index Set
Example N16, M4, q1, Im0,1,2,3 ? K4, p4
Kq 4 Thus, the same SSFFT process can be
allied If the index set is same after modulo
process Ex) Im0,1,2,3, 0,1,2,7, 0,1,6,3,
4,1,2,3, . . .