Orthogonal Basis in Wireless Data: Lessons Unlearned

1. Lessons Unlearned in Wireless Data Rajiv Laroia Qualcomm Flarion Technologies

2. Lessons Unlearned • All orthogonal bases are equivalent • CDM, TDM and OFDM • Cellular channel model is y=hx+n • OFDM is a physical layer technology • TDM is optimal for downlink data • Reuse 1 is the most efficient for data © 2004 Qualcomm Flarion Technologies

3. OFDM Modulation Data bits f=N/T f=2/T f=1/T T Cyclic prefix OFDM symbol © 2004 Qualcomm Flarion Technologies

4. Tone Orthogonality © 2004 Qualcomm Flarion Technologies

5. Orthogonality • Aren’t all orthogonal basis equivalent? • What about Eigenbasis? Sinusoids are Eigenfunctions of all linear time invariant systems. • Sinusoidal orthogonality is preserved under multipath delay spread. • Other basis, e.g., Walsh functions, are not. • Sinusoids are nature’s ‘chosen’ functions • Many advantages above physical layer © 2004 Qualcomm Flarion Technologies

6. Tones 1/T Time OFDM Physical Layer Design • High-speed downlink and uplink based on OFDM • no in-cell interference • no equalization for multipath delay-spread Resource Orthogonality (>35 dB) © 2004 Qualcomm Flarion Technologies

7. Lessons Unlearned - Channel Model 80 dB 0 dB SNR = 13 dB SNR = 0 dB Large dynamic range! © 2004 Qualcomm Flarion Technologies

8. Channel Model Fading (multipath) plus noise is the traditional wireless model • Good enough for point-to-point links • Not good enough in multi-user mobile environment WHY NOT? © 2004 Qualcomm Flarion Technologies

9. Channel Model • Channel (h) uncertainty introduces additional noise • The power of this noise is proportional to signal power. Hence called ‘Self Noise’ • Noise power N=NT+ αP Self noise is a fundamental property of mobile wireless systems © 2004 Qualcomm Flarion Technologies

10. Channel Estimation F In a mobile environment, channel knowledge is intrinsically imperfect because there is only a finite energy available to estimate it. T © 2004 Qualcomm Flarion Technologies

11. Channel Model • Still fading channel - Gaussian noise N=NT+ αP • No difference for point-to-point. • No difference once power is set. • No difference to receiver. • Big difference for multi-user power allocation. • Big difference when self noise is not cross-user: increases dynamic range. © 2004 Qualcomm Flarion Technologies

12. Multi User Power Allocation • Transmit to two users A & B simultaneously (at different powers) xA+xB • Receiver for user A: • CDMA (Walsh basis) N=NT+ α(PA+PB) • Self noise is fixed if total transmit power is fixed • OFDM (Eigenbasis) N=NT+ αPA • Self noise depends on user signal power © 2004 Qualcomm Flarion Technologies

13. SNR and Self noise Without signal-dependent noise SNR With signal-dependent noise Transmit power © 2004 Qualcomm Flarion Technologies © 2004 Qualcomm Flarion Technologies 13

14. Channel Estimation • Average channel requires 2 parameters; • pilot snr • null-pilot snr F Pilots Null pilots T © 2004 Qualcomm Flarion Technologies

15. Self Noise Implications for OFDM • Large dynamic range of multiuser power allocation • Better snr – higher capacity • Many more • Superposition coding © 2004 Qualcomm Flarion Technologies

16. Superposition Coding R2 Timesharing C2 Superposition C2 C1 R1 R2 Timesharing Superposition C2 C2 C1 R1 © 2004 Qualcomm Flarion Technologies

17. Classical Superposition Coding • Regular information for stronger receiver is superposed on protected information Regular info Protected info © 2004 Qualcomm Flarion Technologies

18. Receiver Algorithm • Joint decoder is too complex • Successive decoding involves cancellation of protected signal © 2004 Qualcomm Flarion Technologies

19. Impact of Imperfect Cancellation • Cancellation is often imperfect, e.g., due to imperfect channel estimation • Residual self-noise affects all degrees of freedom © 2004 Qualcomm Flarion Technologies

20. Superposition Coding Traditional superposition by cancellation (subtraction) is vulnerable to channel estimate errors. © 2004 Qualcomm Flarion Technologies

21. Superposition Coding Traditional superposition by cancellation (subtraction) is vulnerable to channel estimate errors. © 2004 Qualcomm Flarion Technologies

22. Lessons Unlearned QPSK is the right constellation for relatively low rate wireless communication. QPSK Constellation © 2004 Qualcomm Flarion Technologies

23. What is optimal ? © 2004 Qualcomm Flarion Technologies

24. What is practical ? • Capacity calculations support the idea. © 2004 Qualcomm Flarion Technologies

25. Better than QPSK? 5 Point Constellation © 2004 Qualcomm Flarion Technologies

26. Practical version for OFDM … … QPSK is 2 bits per symbol. One out of 4 symbols (2bits) is QPSK (2 bits) = 1 bit per symbol. © 2004 Qualcomm Flarion Technologies

27. Practical version for OFDM Conditional distribution of position and phase. Performs as well as QPSK/LDPC for low (1/4) rate codes. © 2004 Qualcomm Flarion Technologies

28. Practical version for OFDM So What ? Conditional distribution of position and phase. Performs as well as QPSK/LDPC for low (1/6) rate codes. © 2004 Qualcomm Flarion Technologies

29. Zero symbol has no self noise! • No cancellation of protected code • Full superposition gain available for users with very different snrs © 2004 Qualcomm Flarion Technologies

30. Lessons Unlearned OFDM is a physical layer technology What are some other advantages of OFDM? • Granularity of resource allocation • Better MAC layer, QOS • Better link layer, low delay • Flash signals for cell identification © 2004 Qualcomm Flarion Technologies

31. Flash Signaling • High power concentrated on one or more tones for a short time. • Capacity achieving for fading channels at very low data rate, or very wide band. • Achieves minimal Eb/No requirement. © 2004 Qualcomm Flarion Technologies

32. Beacon Tone • Beacon is a special downlink symbol in which power of a single tone (beacon tone) is significantly (e.g., 26 dB) higher than average per-tone power • Beacon is so strong that it could never be mistaken to be anything produced by Gaussian noise process • Beacon tone occurs once every ~100,000 symbols • Negligible overhead and interference impact © 2004 Qualcomm Flarion Technologies

33. Beacon Tone • Beacon can be easily detected prior to timing or frequency synchronization or channel estimation • Exploit unique property of sinusoid tones (impossible for Walsh codes) • Almost no additional computational complexity (no chip-level search required) © 2004 Qualcomm Flarion Technologies

34. Use of Beacon Tone • Information conveyed in beacon tone includes • Carrier location • Cell/sector ID • Symbol level timing • Some uses of Beacons • Detect a candidate base station long before pilots are visible • Estimate path loss from cell • Make hand-off decisions © 2004 Qualcomm Flarion Technologies

35. Beacon Interference • Beacons provide impulsive noise • Decode signal using saturation or reversal metrics in decoder • Automatic cancellation (erasure) • Protection against impulse noise • Little impact on Gaussian noise performance Saturation metric -1 1 Reversal metric -1 1 Decoder metrics © 2004 Qualcomm Flarion Technologies

36. Conclusions • The World welcomes technological improvement. • If you join a wireless start-up you have a good chance of getting rich. Many interesting things unlearned © 2004 Qualcomm Flarion Technologies