Long-Range Channel Prediction for Adaptive OFDM Systems

1
Long-Range Channel Prediction for Adaptive OFDM
Systems

• I. C. Wong, A. Forenza,
• R. W. Heath and B. L. Evans

2
Adaptive OFDM

• Adapt modulation, coding, or power in each
  subcarrier at the Transmitter (Tx) in order to
  maximize throughput
• Adaptation based on current channel state
  information (CSI) being fed back to the Tx
• Problem Outdated CSI Souryal Pickholtz, 2001
• Effect very relevant in mobile situations
• How do I minimize the impact of this delay?

3
Wireless Channel Prediction

• Long-range prediction (LRP) Duel-Hallen, et.
  al. 2000
• Used an FIR Weiner prediction filter
• Designed for flat-fading channels
• Key Idea Downsampling the observed channel
  coefficients

4
Application of LRP to OFDM

• Briefly investigated in Forenza Heath, 2002
• Directly predict channel for each of the N
  subcarriers
• Valid since each subcarrier is a flat-fading
  narrowband subchannel
• Storage needed for pN previous channel
  coefficients ck and pN prediction coefficients
  dk
• Used Burgs algorithm to compute predictor
  coefficients

5
Low-Complexity LRP for OFDM

• Pilot-tone Prediction
• Perform LRP on the Npilot pilot tones only
• Since Npilot lt N, less computation and storage
  needed (e.g. Npilot 8 N 256 for 802.16e )
• Use the same Wiener predictor for the subcarriers
  nearest to the pilot carrier

Pilot


Data Carriers
6
Low-Complexity LRP for OFDM

• Time Domain channel tap Prediction
• Perform LRP on the L Npilot time domain channel
  taps, and thus further reduce complexity
• It can be shown that MMSE predictor for the time
  domain taps also minimize MSE for frequency
  domain

tn
t1
t0
7
Simulation Parameters (IEEE 802.16e)
Parameter Value Parameter Value
N 256 BW 5 MHz
Guard Carriers (7) 0-27 201256 Fcarrier 2600 MHz
Channel Model ETSI Vehicular A Mobile Velocity 75 kph
Prediction Order 75 Downsampling rate 25 (4fd)
8
Channel Prediction Example
9
Performance comparisons
10
Conclusion

• LRP for OFDM systems can be accomplished by
• Prediction on all the tones
• Prediction on pilot tones
• Prediction on the time domain channel taps
• Time-domain prediction gives better MSE
  performance, specially in the presence of channel
  estimation error
• Future work Adaptive prediction with Weiner
  smoothing