1. Sujan Rajbhandari LCS2006 1 Convolutional Coded DPIM for Indoor Optical Wireless Links S. Rajbhandari, N. M. Aldibbiat and Z. Ghassemlooy Optical Communications Research Group, School of Computing, Engineering and Information Sciences, The University of Northumbria, Newcastle upon Tyne, U.K. Web site: http://soe.unn.ac.uk/ocr

2. Sujan Rajbhandari LCS2006 2 Optical Wireless Communication  Definition : a telecommunication technology that uses light propagating in free space to transmit data between two points. [ http://en.wikipedia.org/wiki/Free_Space_Optics.]  Also popularly known as free space optics (FSO) or Free Space Photonics (FSP) or open-air photonics.

3. Sujan Rajbhandari LCS2006 3 Optical Wireless – Advantages  Unregulated bandwidth, free for commercial and personal use.  200 THz bandwidth in the 700-1500 nm range.  No multipath fading.  Availability of low cost optical transmitter and receiver.  Small cell size.  Can not penetrate through wall- same frequency can be utilized in adjacent rooms.

4. Sujan Rajbhandari LCS2006 4 Practical Implementations - Issues  Intense ambient noise.  Average transmitted power is limited due to eye safety.  Do not penetrate wall, thus a need for infrared access point.  Large area photo-detectors – limiting the bandwidth.

5. Sujan Rajbhandari LCS2006 5 Digital Modulation Techniques for OWC  Modulation scheme adopted should have one or two of the following characteristics:  power efficient – Since the maximum power that can be transmitted is limited because of eye safety.  bandwidth efficient – particularly in non-line of sight configurations  Types  On-Off Keying (OOK), Pulse Position Modulation (PPM), Digital Pulse Interval Modulation (DPIM), Dual Header Pulse Position Modulation (DH-PIM), Differential Amplitude Pulse-Position Modulation (DAPPM)

6. Sujan Rajbhandari LCS2006 6 Digital Modulation Techniques for OWC

7. Sujan Rajbhandari LCS2006 7 DPIM  DPIM is an anisochronous pulse time modulation technique.  A symbols starts with a pulse followed by k empty slots. 1≤ k≤ L and L = 2 M.  Guard slots can be added to provide resistance to ISI arising from multipath propagation.

8. Sujan Rajbhandari LCS2006 8 DPIM – contd.  For DPIM with a guard band of g guard slots DPIM(gGS) the minimum and maximum symbol durations are gT s and (L+g)T s, respectively, where T s is the slot duration where T b is the bit duration and L avg is the mean symbol length (no. of slots).

9. Sujan Rajbhandari LCS2006 9 Error Performance of DPIM The slot error rate for DPIM with no guard slot, P se(0GS) The slot error rate with 1 guard slot, P se(1GS)

10. Sujan Rajbhandari LCS2006 10 DPIM- Comparison with other modulation schemes  Bandwidth efficient compared to PPM.  Built-in slot and symbols synchronisation.  Higher transmission capacity compared to PPM.  Resistance to effect of multipath propagation compared to PPM

11. Sujan Rajbhandari LCS2006 11 Why use Error Control Coding ?  Improves the reliability of system.  Improves the Signal to Noise ratio (SNR) required to achieve the same error probability.  Efficient utilization of available bandwidth and power.

12. Sujan Rajbhandari LCS2006 12 Convolutional Coded DPIM  Linear block codes like Hamming code, Turbo code and Trellis coding are difficult (if not impossible ) to apply in PIM because of variable symbol length.  So either convolutional code or modification of convolutional codes are only alternatives because convolutional encode act on serial input data rather than block.

13. Sujan Rajbhandari LCS2006 13 The convolutional Coding State diagram (3,1,2) convolutional encoder. ½ code rate and constraint length = 3 Generator function g1 = [111] and g2 = [101]

14. Sujan Rajbhandari LCS2006 14 Error performance  Viterbi ‘Hard ‘ decision Decoding  The Chernoff upper bond on the error probability is: where P se is the slot error probability of uncoded DPIM.

15. Sujan Rajbhandari LCS2006 15 CC-DPIM(2GS) Speciality  2 empty slots at in all the symbols so that memory is cleared after each symbol.  Trellis path is limited to 2.  No need to use Viterbi algorithm instead we can use simple look-up table.

16. Sujan Rajbhandari LCS2006 16 Look-up Table  Consider received sequence to be {00 00 10 11 00}  The closest match to the sequence in the look-up table is {00 11 10 11 00} i.e. correct decision!

17. Sujan Rajbhandari LCS2006 17 System Block Diagram PIM Demodulator PIM Modulator Convolutional Encoder Optical Transmitter Matched Filter Threshold Detector + Shot Noise n(t) Output Bits Input Bits Viterbi Decoder Sampler Optical Receiver h(t)

18. Sujan Rajbhandari LCS2006 18 CC-DPIM : Upper Error bound Difficult to ascertain exact Hamming distance of an convolutional encoder. Union bound is utilised to evaluate the performance. The simulation result is expected to be less than but close match to the error bound.

19. Sujan Rajbhandari LCS2006 19 Performance comparison of CC-DPIM with different guard slots DPIM(2GS) offers an improvement of 0.5 dB and 1dB in SNR compared to DPIM(1GS) and DPIM(0GS). A code gain of 4.8 dB achieved at slot error rate of 10 -4.

20. Sujan Rajbhandari LCS2006 20 Performance of DPIM for different bit resolution A code gain of ~4.9 dB, 4.8 dB and 4.5 dB for M= 5, 4 and 3, respectively at P se of 10 -4. Code gain increases as P se decreases.

21. Sujan Rajbhandari LCS2006 21 Comparisons with other modulations The performance of CC-DPIM(2GS) close to CC-DH-PIM 1 with formal requiring 1 dB more SNR.. CC-DPIM performances better than uncoded PPM

22. Sujan Rajbhandari LCS2006 22 Conclusions  Convolutional coded DPIM offered an improvement of 4.5dB compared to uncoded DPIM.  CC-DPIM(2GS) performed better than CC-PIM(1GS) and DPIM(0GS).  Performance of CC-DPIM is very close to performance of CC-DH-PIM 1  Simple implementation when using 2 Guard slots instead of 1 or no guard slot in DPIM, since no need for Viterbi decoding algorithm

23. Sujan Rajbhandari LCS2006 23 Thank you!