Bandwidth-Efficient Method for Adaptive Forward Error Correction on Wireless Local Area Network

1. Bandwidth-Efficient Method for Adaptive Forward Error Correction on Wireless Local Area Network • Co-Presenters: • David R. Pollard, Graduate Student, Eastern Illinois University • Dr. Ping Liu, Graduate Coordinator, Eastern Illinois University

2. Mobile Collaborative Computing Excellent Quality Transmissions

3. Audio Quality Decreased audio quality Transmission Channel Excellent audio quality

4. Bandwidth Must Be Optimized Absolute bandwidth = 12f-f = 11f f Effective bandwidth Data bits x 2 Redundancy assures audio qualilty 12f

5. Spread Spectrum Absolute bandwidth = 12f-f = 11f f Spread the Signal Over Wider Bandwidth Effective Bandwidth Avoid Noise Distortion 12f

6. Error Detection versus Error Correction • Retransmission of data packets: ARQ • Correction of data packets without retransmission: FEC

7. Block Code Retransmissions(ARQ Required for ‘detection’ codes) • Inadequate for wireless applications. Bit error rate (BER) high on wireless link. Propagation delay is very long. (Especially, satellites)

8. Bandwidth Must Be Optimized Absolute bandwidth = 12f-f = 11f f Effective bandwidth Data bits x 10 Redundancy consumes bandwidth 12f

9. Block Error Codeswhen used for ‘correction’ • (n-k) block code k = data bits n = codewords in bits vcodewords = f(vdatabits) where v is a vector

10. How much bandwidth is required for block code ‘correction’? • redundancy of the code = (n-k)/k • code rate = k/n • code rate of ½ = twice the bandwidth

11. How much bandwidth is required for block code ‘correction’? (Another example) • code rate of 2/5 = 2.5 times the bandwidth • if data rate = 1 Mbps then, output from the encoder must be 2.5 Mbps to keep up

12. Convolutional Codes for Correction • Generates redundant bits continuously. • Error checking and correcting continuously. • Better for collaborative computing and excellent audio quality.

13. What method provides the best correction and conserves bandwidth? • Dr. Philip McKinley, Pavilion Project, Michigan State University: • “The best bandwidth-efficient method for FEC on wireless LANs is one which is dynamically adaptive to channel loss behavior.”

14. Redundancy for Correction Determined by a Proactive Parameter: Alpha • For each group n: k(1+alpha) sent by proxy • When a receiver loses < alpha(k), recovers locally. • When a receiver loses > alpha(k), NAK sent to proxy for k(1+alpha) retransmission.

15. Packets Required by a Specific Receiver Receiver L(1+alpha) N A K Dispatcher N A K Processor

16. Alpha Increase άincis based on observation of NAK behavior for each group. άinc = = M * L/k L = requested parity packets M = small integer

17. Alpha Decrease • άdec prevents high ά • In the absence of NAKs, ά is reduced until one or more NAKs are received. • άinc resumes

18. Objective for ά For optimal audio signals and most efficient use of bandwidth in collaborative sessions, keep k(1+alpha) slightly higher than needed.

19. Conclusion • Unlimited wants for mobile collaborative computing. • Limited bandwidth for excellent quality audio transmissions.

20. Conclusion (continued) • Error correction using ARQ insufficient for wireless communications. • Static forward error correction requires high usage of bandwidth.

21. Best Method: Spread Spectrum and Adaptive Forward Error Correction Alpha Increase Absolute bandwidth = 12f-f = 11f Alpha Decrease f Avoid Alpha Increase Noise Distortion Alpha Decrease Optimize 12f Bandwidth Usage