1. Tripping the light fantastic: Optical Wireless Communications Sarah Kate Wilson Santa Clara University

2. Outline Some history What’s different now? Challenges with Optical Wireless How does it work? Research challenges Summary

3. Indoor lighting and communication

4. History In the beginning, there was light

5. Optical communications TV remotes 1979, Gfeller and Bapst, Wireless diffuse indoor communications Joe Kahn Mohsen Kavrehad And many others

6. Late 80’s early 90’s Wireless optical research Mostly infrared Why didn’t we do it then?

7. 20 years ago – Requirements and Background Wireless LAN was new Data needs were meager Mosaic and Netscape 1994 Data needs opened up only after Internet Browsers were developed

8. What we know now about Wireless LANs Then and Now Then: Had to justify the need for wireless LANs One technology at a time Limited data rates Voice was the key technology; not data Now Wireless LAN is everywhere Cellular providers using Wireless LAN to supplement service Text, data, videos etc.

9. Technology 20 years ago CDMA and TDMA were the technologies of choice Orthogonal Frequency Division Multiplexing (OFDM) was just being investigated for Wireless LAN Choices for Wireless LAN were radio frequency or infrared LEDs existed, but…

10. Outline Some history What’s different now? Challenges with Optical Wireless How does it work? Research challenges Summary

11. What’s different now? The need for speed and data

12. What’s different now technically? LEDs brighter, faster, better New types of modulation 5G mm wave push Heterogeneity is key More avenues for data

13. 5G mm wave Data needs are massive Spectrum is limited Push to use millimeter wave lengths Lots of spectrum Short reach Will require heterogeneity Current methods and spectrum working with new methods and spectrum

14. Indoor lighting can now use LEDs: No additional transmission energy needed

15. Outline Some history What’s different now? Challenges with Optical Wireless How does it work? Research challenges Summary

16. Optical Wireless data Visible Light versus Infrared Visible lightInfrared Can use light fixturesAdditional transmission energy Downlink only?Uplink and downlink Light levels safeNeed to restrict maximum level

17. How does optical stack up with RF? Parameter5GHz60 GHzOptical 10 meter free-space loss44 dB66 dB86 dB Additional loss due to oxygen05-12 dB0 Range (10 mW)~35 meters~5.5 meters<5 meters Delay spread250 ns (in a big arena) < 20 ns How fast can the signal change? (at 3 m/sec trot) 50 Hz600 Hza few Hz

18. Visible Light Communications Does not replace radio frequency communications Complementary An additional way to satisfy the need for more data

19. Visible Light Communications Intensity Modulation (IM) not coherent Direct Detection (DD) Bandwidth limited by the Response of the LEDs and detectors The arrangement of the source and user Line-of-sight versus non-line of sight

20. Directed transmission – works well with line of sight but fails if the transmitter and receiver are not aligned Diffuse transmission – received electrical power falls off as the fourth power of distance (because of intensity modulation) Dispersion due to reflected signals

21. Indoor lighting and communication

22. Concept Figure from the Boston University Smart Lighting Group

23. Some scenarios

24. Some people who have contributed to Optical Wireless Including but not limited to: Joseph Kahn Dominic O’Brien Harald Haas Steve Hranilovic TDC Little

25. How did I get interested? 2006 Globecom; Jean Armstrong OFDM Physical Layer Fatigue Opens up a whole new way of thinking

26. My interest in OFDM: Started in 1993 Channel Estimation Wireless LAN Synchronization Scheduling

27. Outline Some history What’s different now? Challenges with Optical Wireless How does it work? Research challenges Summary

28. What makes it fun? Modulation is totally different Real and positive signal Handover issues Interference issues

29. Visible Light Communications Short range Intensity-modulated (IM) Direct Detection –Baseband modulation –Rate of change NOT a function of carrier frequency –Channel and signal is real and positive

30. Optical Wireless Single-carrier Pulse amplitude modulation: no Light, Light, brighter, brightest – 2 bits OFDM solutions: Must modify to fit the real, positive channel

31. Single-carrier frequency Power

32. OFDM frequency Power

33. Single-Carrier

34. Multicarrier: conceptually a series of chords

35. OFDM Uses an FFT to modulate the tones Cyclic prefix –Preserves orthogonality –Prevents interblock interference Power assignment, bit-loading

36. OFDM and optical wireless Must be real: Hermitian symmetric Must be positive DC –biasing Reserve carriers to mitigate negative parts and bias Asymmetrically-Clipped Optical (ACO)- OFDM Consider bandwidth versus power

37. ACO-OFDM Asymmetrically-Clipped Optical (ACO) Wireless OFDM Armstrong and Lowery, “Power Efficient Optical OFDM,” Electronics Letters, 2006. Only modulate on the odd tones; then clip. More power efficient than DC biasing! BUT half the bandwidth of DC biasing.

38. ACO-OFDM 32 tones

39. ACO-OFDM 16 QAM,128 tones

40. ACO-OFDM

41. Why does ACO-OFDM work? Or just read the paper by Wilson and Armstrong 2009 Odd tones implies odd symmetry –If x_n >0 then mod(n+N/2) points away we have -x_n Complex exponential of 2(k+1)(n+N/2) is equal to the negative of the complex exponential of 2(k+1)n Adding a negative to a negative makes a positive! the DFT of the clipped signal is equal to the DFT of its complement.

42. Rate vs. Power Want fast rate Want low power Traditionally increasing power by 6 dB will add 2 bits per dimension E.g. 2 bits – QPSK, need an additional 7 dB for 4 bits – 16 QAM Increasing bandwidth by 2 doubles the rate

43. Comparing Single Carrier and ACO-OFDM : Joint work with D.F. Barros and J.M. Kahn – 1 bit per symbol

44. Why do people like OFDM? Flexibility No equalization Simpler filters at the receiver

45. Challenges with OFDM Peak-to-average power ratio LEDs Detectors Dimming Driving the LEDs Different levels changing quickly Drive all LEDs together? Treat each LED like a bit and turn on/off individually

46. ACO-OFDM 16 QAM,128 tones

47. LED configurations

48. Change intensity via individual LEDs

49. Things to consider Physical layer modulation Efficiency; Peak-to-average power Intensity modulation restrictions Bandwidth Heterogeneity Multiple light sources Handover Scheduling

50. Some recent work Joint work with Lam, Elgala and Little of Boston University Find a way to use the “un-used” carriers in ACO-OFDM Spectral and Energy Efficient (SEE) OFDM

51. SEE OFDM Given N subcarriers ACO modulates N/4 subcarriers SEE OFDM then adds additional components 1 st : Modulate N/8 of the unused subcarriers 2 nd : Then add N/16 of the rest of the unused subcarriers Etc Use successive decoding at the receiver

52. SEE OFDM

53. Scheduling for OFDM Multiple users in a cell – all want access Scheduling for RF OFDMA systems –Multi-user diversity –Fairness But real and positive channel changes how we approach it

54. Single-carrier Scheduling

55. Scheduling with RF-OFDM one time instant… Frequency User 1 User 2 User 3

56. Scheduling in RF OFDM versus optical OFDM RF Users send their best clusters The base-station scheduler assigns clusters of spectrum based on reported SNRs Very little chance of collision Optical Optical wireless channel is low-pass All users will pick their lowest frequency Collisions!!

57. What to do? Pick clusters semi-randomly –The user has a certain desired rate –Find all clusters that can deliver that rate –Send best two cluster indices that are ‘good enough’

58. Scheduling for Optical-OFDM In RF scheduling –Cluster subcarriers –Send back index of cluster with best SNR and/or most need In Optical-OFDM –The best cluster is most likely the lowest frequency cluster –All users will be trying to send the same cluster

59. Single-carrier Scheduling

60. Scheduling with RF-OFDM one time instant… Frequency User 1 User 2 User 3

61. What to do? Pick clusters semi-randomly –The user has a certain desired rate –Find all clusters that can deliver that rate –Send best two cluster indices that are ‘good enough’

62. Full-buffer Throughput

63. Conclusions Visible Light Communication is an exciting new field Opportunities in both the physical and upper layers of communications Spectrum is limited; demand is growing Future unknown applications will demand more innovation Talk to people at workshops/conferences Can lead to exciting research

64. Peak-to-Average Power ratio reduction Peak-to-Average Power Ratio is a problem in OFDM Good Solution: Use Single-Carrier Frequency Domain Multiplexing with ACO- OFDM (Joint work with Acolatse and Bar- Ness)

65. Peak-to-Average Power ratio reduction High Peaks can play havoc with power- limited LEDs NameMethodNumber of subcarrierr s Number of repetitio ns ACO- SCFDE Modulate odd of half subcarriers N/41 RCO- OFDM Modulate half subcarriers N/22 DQO- OFDM Modulate All subcarriers N4

66. SCFDE and ACO-OFDM SCFDE combined with ACO-OFDM ckck FFT N/4 Map to odd Subcarriers Hermitian symmetry IFFT Clip Negative Values

67. .1 W Power

68. 1 W Power

69. Conclusions Higher frequency for the last few meters Higher rates; less interference Optical -- Real channel; Low Doppler Difference in modulation and channel leads to neat problems

70. References P. Smulders, “Exploiting the 60 GHz Band for Local Wireless Multimedia Access: Prospects and Future Directions”, IEEE Comm. Magazine, January 2002. R. Daniels and R. Heath,” 50 GHz Wireless Communications: Emerging Requirements and Design Recommendations,” IEEE Vehicular Technology Magazine, September 2007 V. Chandrasekhar, J. Andrews, A. Gatherer, “Femtocell Networks: A Survey,” IEEE Communications Magazine, September 2008. C. Anderson and T. Rappaport, “In-Building Wideband Partition Loss Measurements at 2.5 and 60 GHz,” IEEE Transactions on Wireless Communications, May 2004. J. Armstrong and A. Lowery, “Power efficient optical OFDM”, Electronics Letters, March 2006. J. Kahn, W. Krause and J. Carruthers, “Experimental Characterization of Non-Directed Indoor Infrared Channels”

71. ACO-OFDM

72. Simulations Channels with random number of taps, random tap positions Compare equal constellation to low-frequency bit-loading over 5000 different channels for each SNR Equal number of bits/word on each channel

73. Scheduling with RF OFDM More than one value to choose from

74. ACO-OFDM Scheduling

75. 4-PAM vs. ACO-OFDM 2 bit/symbol, Equal or Unequal Symbol Rates, With CSI ACO-OFDM at 2R s best on all channels. ACO-OFDM at R s worst on all channels.

76. 4-PAM vs. Three OFDM Techniques 2 bit/symbol, Equal or Unequal Symbol Rates, With CSI ACO-OFDM (or PAM-DMT) at 2R s best on all channels. ACO-OFDM (or PAM-DMT) and DC-OFDM at R s worse than OOK on all channels.

77. Optical wireless scenarios ( courtesy J. Kahn )

78. System Design Parameters Values of B e are in MHz.

79. OFDM without Channel State Information

80. OOK vs. ACO-OFDM 1 bit/symbol, Without Channel State Information OOK outperforms all OFDM techniques at moderate-to-low outage probabilities. For OFDM, ceiling bounce loading is best at low outage probabilities.