The Effect of Multipath Propagation on the Performance of DPIM on Diffuse Optical Wireless Communications

Transcription

1 The Effect of Multipath Propagation on the Performance of DPIM on Diffuse Optical Wireless Communications 1 Z. Ghassemlooy a, A. R. Hayes a and N. L. Seed b a-optical Communications Research Group, School of Engineering, Sheffield Hallam University, Pond St., Sheffield, S1 1WB. U.K. b- Electronic Systems Research Group, Dept. of Electronic and Electrical Engineering, University of Sheffield, Mappin Street, Sheffield, S1 3JD. U.K. z.f.ghassemlooy@shu.ac.uk Web site:

2 What is Optical Wireless Communication? Optical carrier, rather than the radio/microwave (305 THz). Outdoor point-to-point links up to 5km, Indoor systems a few meters. First suggested for indoor communications in Wireless networks offer increased mobility and flexibility. Provides higher bandwidths which is difficult to obtain in the already congested radio frequency spectrum. Have much more stringent power requirement due to eye safety.

3 3 OWC - Advantages Abundance of unregulated bandwidth 00 THz in the nm range No multipath fading intensity modulation and direct detection High security Higher capacity per unit volume (bps/m 3 ), due to neighbouring cells sharing the same frequency Cost effective at rates near 100 Mbps Small cell size At nm and 1550 nm absorption effects are minimal.

4 4 OWC- Disadvantages Multipath dispersion this is the main bandwidth limitation up to a few tens of MHz. Limited range: ambient noise is the dominant noise Difficult to operate outdoor High power requirement SNR can vary significantly with the distance and ambient noise Costly

5 Optical Wireless Vs. Radio Wireless 5 Property Radio Optical Implication for Optical Bandwidth regulated? Yes No. Approval not required. World-wide compatibility Passes through Yes No. Inherently secure walls?. Carrier reuse Multipath fading Yes No Simple link design Multipath dispersion Dominant noise Yes Yes Problematic at high data rates Other users Background light Short range

6 Classification of OW Links 6 Directed Hybrid Non-directed TX TX TX Line-of-sight RX RX RX Non-line-of-sight (Diffuse) TX RX TX RX TX RX

7 Current Standards and Bodies 7 1- Infrared Data Association (IrDA) 1993, by HP, IBM, and Sharp (now 130 members) 9.6 kbps 4 Mbps Line of sight links: Coverage area 1- meters - IEEE standard for wireless LANs - The IR physical layer specifies: nm range,. bit rates of 1 or Mbps using diffuse propagation

8 Safety Classifications for a Point Source Emitter Total power in a 5cm Lens (mw) class 3B class 3A class class 1.5mW 0.5mW class 3B class 3A class 1 45mW 8.8mW class 3B class 3A class 1 50mW 10mW class 3B class 3A class visible 880 indoor infra-red Wavelength (nm) Source:BT

9 Optical Power Spectra of Ambient Light Sources IR The average power received from ambient light sources is much larger than the desired signal (Typically 30dB if no optical filtering is used.) 9

10 10 Modulation Schemes Selection Criteria The average optical power emitted: limited by eye safety regulations, and therefore, modulation techniques are compared in terms of the average received optical power required to achieve a desired BER at a given data rate. This favours modulation schemes with a high peak to average power ratio. The bandwidth efficiency: typical infrared receivers use large area photodetectors (eg. 1 cm ) which limits the receiver bandwidth. Resistance to multipath dispersion effects Amenability to multiple access the last two are best achieved by subcarrier multiplexing, but at the cost of increased power.

11 Modulation Tree 11 Pulse Modulation Analogue Digital Pulse Time Pulse Shape Pulse Time PSM Isochronous Anisochronous PAM Isochronous Anisochronous PWM PPM PIM PIWM PFM SWFM DPPM MPPM DPWM PCM DPIM DPIWM difppm DH-PIM RZ RB AMI Manchester NRZ NRZ(L) NRZ(I) Miller code

12 Digital Pulse Time Modulation Schemes 1 Frame 1 Frame Frame 3 OOK PPM T b DPIM DH-PIM Time T s

13 Digital PIM 13 Source Data OOK NRZ 4-PPM 4-DPIM Symbols No gs 1gs gs

14 Code Properties 14 PPM slot rate: R s, PPM = log L L R b DPIM slot rate: R s, DPIM = L log avg L R b where Lavg = ( L + 1) ( ) L + 3 ( L + 5) ngs 1gs gs If a guard band is used, in order to accommodate the additional slot(s) in each symbol, the slot rate must be increased in order to maintain the same average bit rate. In the absence of multipath propagation, the presence of a guard band gives a slight improvement in power efficiency due to the reduced average duty cycle.

15 Average Data Rate and Bandwidth Requirement 15 Data Rate Bandwidth Requirement OOK-NRZ R b R b L R log L L-PPM R b b L-DPIM A (Average) L-DPIM M (Maximum) R b (average) ( + ) L 3 log L ( L + 1) ( L + 1) Rb ( + ) L L 3 log R b R b

16 Normalised Average Data Rate and Bandwidth Requirement Normalised average data rate DPIM M DPIM M PPM DPIM A PPM Normalised bandwidth requirement No. of bits per symbol, M Normalised average data rate Normalised bandwidth requirement

17 Probability of Symbol Error for L-DPIM 17 Assuming the pulse indicating the start of the symbol is correctly detected, the pulse indicating the start of the following symbol must be detected correctly and all preceding slots must be empty. If the maximum symbol length is reached and the pulse indicating the start of the following symbol has not been detected, a one is assigned at random to one of the L slots. If more than one pulse is received over the maximum symbol length, this is interpreted as being more than one symbol. 1 L 1 n 1 L 1 PSE DPIM = 1 ( 1 P01 )( 1 P10 ) + P01( 1 P10 ) L n= 0 L ( L + 3) P R T P 01 = P 10 = 1 Erfc 4 ave qi b s

18 Diffuse System Multipath propagation 18 Non-LOS TX RX LOS TX RX

19 Delay Spread & Normalised Delay Spread 19 Power per unit time, h RMS delay spread Impulse Response Total delay spread Excess delay time, t DS( RMS) ( t µ ) h ( t) The total delay spread is the time interval during which reflections with significant energy arrive. The rms delay spread is the standard deviation of the delay of reflections, weighted proportional to the energy in the reflected waves. Normalised delay spread is a dimensionless parameter defined as the RMS delay spread T DS(RMS) divided by the bit duration T b. T = u = h ( t) th h dt ( t) dt dt ( t) dt 1 where mean delay µ is given by:

20 Ceiling Bounce Model 0 The multipath channel is modelled using the ceiling bounce model developed by Carruthers and Kahn. The channel impulse response is given by: h( t, a) = D = 6a 6 ( t + a) a u( t) where u(t) is the unit step function and a is related to the RMS delay spread, D, by: As we are interested in the effects of multipath propagation only, the optical path loss, G o, is set to unity, i.e. G o = h( t, a) dt 0 = 1

21 Block Diagram of the Unequalised DPIM(NGB) System 1 n-slot DPIM sequence DPIM encoder Transmitter filter p(t) L avg P avg Multipath channel h(t) R n(t) Unit energy filter r(t) (matched to p(t)) Ts sample Optimum threshold detector Compute Ps in (n-1)th slot Ts = Tb log L L avg L avg = ( L +1)

22 Bit Error Rate (BER) and Packet Rrror Rate (PER) In PPM, an error is confined to the symbol in which the error occurs and therefore, it is possible to convert from the probability of symbol error to bit error rate. In DPIM, errors are not confined to one symbol, therefore the packet error rate will be used as a measure of performance signal wrong-slot error erasure error false-alarm error * * *

23 Impulse Response 3 The discrete-time equivalent impulse response of the cascaded system, which is given by: c k = P( t) h( t) r( t) r= kt s Unless the channel is non dispersive, c k contains a zero tap, a single precursor tap and possibly multiple postcursor taps. On a non dispersive channel, the optimum sampling point occurs at the end of each bit period. On dispersive channels, the optimum sampling point changes as the severity of ISI changes. In order to isolate the power penalty due to ISI, two assumptions are made: Firstly perfect timing recovery optimal decision threshold.

24 4 Impulse Response No ISI ISI Output from c k No ISI Any pulse transmitted before bit position 1 and in position 6 onwards has no effect on bit position 4. Case Only bits in positions 1 to 5 result in ISI which affects bit position 4. Therefore it is only necessary to consider every combination of bits 1 to 5 in order to calculate the average BER for bit position 4. Case Case 3 Case 4 Penultimate bit bi

25 Slot Error Rate 5 For any m-slot DPIM(NGB) sequence, denoted as b i, the input to the threshold detector for the (m-1) th slot period, in the absence of noise, is given as: y i = L RP b avg avg i c k k = m Bit error for (m -1)th slot of sequence bi, Average probability of slot error ε i Q = Q P slot = p i ε all i y i α N 0 α y i N 0 Pi = Probability of occurrence of a given m-slot sequence i if if b b i i = 1 = 0

26 6 Segment of m slots: {s i,1, s i+,...,s i+m } Start Yes Next slot = 1? No No Leading slot = 0? Yes Count No. of 0's Count No. of 0's No Acceptable? Acceptable? No Yes Calculate prob of occurrence Yes Calculate prob of occurrence '1' detected at start of full DPIM symbol At least 1 more slots? No Yes Next slot = 0? Set prob = 1 & multiply No Yes Calculate no. of 0's Calculate prob of occurrence & multiply All done Yes Acceptable? No Yes '1' detected? No Calculate prob of occurrence & multiply Invalid segment prob = 0.

27 Packet Error Rate (PER) 7 Determine the length of sequence required, n. (N o. of slots spanned by h(t)) Generate every possible valid sequence of length n slots, and the probability of occurrence for each sequence. Calculate the probability of slot error for the (n-1) th slot in each sequence, using the optimum sampling point and optimum threshold level. Calculate the overall probability of slot error by multiplying the probability of slot error for each sequence by the probability of occurrence for that sequence and sum together. Compute PER and adjust P avg until 10-6 target is reached. Repeat for various values of nds, order L and number of guard slots. ( ) LavgD log L 1 P PER 1 slot

28 PIM (NGB) normalised optical power requirement Vs. normalised delay spread L = 4 L = 8 L = 16 L =

29 PPM System Block Diagram 9 Input bits PPM encoder Transmitter filter p(t) LP avg Multipath channel h(t) R n(t) Unit energy filter r(t) (matched to p(t)) sample sample Optimum threshold detector Choose largest Output slots Output symbols

30 PPM normalised optical power requirement Vs. normalised delay spread for different values of L employing threshold detector 30 L= 4 L= 8 L= 16 L= 3

31 DPIM optical power penalty Vs. normalised RMS delay spread L = 4 L = 8 L = 16 L =

32 DPIM(NGB) normalised optical power requirement versus normalised delay spread optimum α1 1 α L = 4 L = 3 Threshold

33 Adding a Guard Band to Reduce the Effects of Multipath Propagation 33 4-DPIM Symbols ngs 1gs gs One or more guard slots may be added to each symbol immediately following the pulse. The receiver would then ignore these slots (i.e assigns 0 ), hence it would not be possible to falsely detect a pulse in these slot(s), provided that the pulse initiating the symbol is correctly detected. The price paid for this is an increased bandwidth requirement or a reduction in the average data rate.

34 DPIM with Guard Slots 34 1 Guard slot: L = ( L + 3) avg Guard slot: L = ( L + 5) avg On its own, a reduction in slot duration would result in increased optical power requirements, since the ISI would affect a greater number of slots. Therefore, in order for the guard band to achieve a net reduction in optical power requirement, the reduction in power due to the presence of the guard band must outweigh the increase in power due to the reduced slot duration

35 DPIM -GS - Detection scenarios 35 Since the probability of slot error for the m th slot in a sequence is affected by the detection of the (m-1) th slot, there are 4 possible detection scenarios: If b =1 and is correctly detected, bm m 1 must be a zero and is automatically bˆm assigned a zero. Therefore, there is no chance of an error in slot m. if b =1 m 1 but is falsely detected as a zero, bm must be a zero but bˆm is not automatically assigned a zero. Therefore, an error will occur if bˆ m =1. if b = 0 m 1 and is correctly detected, bm could be either a one or zero, and the (m-1) th slot has no effect on the probability of error in slot m. if bm 1 = 0 but is falsely detected as a one, m is automatically assigned a zero, but an error will occur if. b m =1 bˆ

36 DPIM normalised optical power requirement Vs. normalised delay spread with and without guard bands for L= 4, 8, 16, and NGS 1GS GS NGS 1GS GS NGS 1GS GS 6 4 NGS 1GS GS

37 Eye Diagrams for 4-DPIM (1gs) & 3-DPIM (1gs) for Channels with NDS of 0.01 and x NDS = 0.01 NDS = 0.1 x 10-3 Each eye diagram is based on a sequence length of 7 slots and an average optical transmit power of 1W. L = 4 L = x x

38 Packet Error Rate Vs. Average Received Irradiance 38 Packet length = 104 bits, Data rate = 1 Mbps, Background power = - 10 dbm/cm 1.00E E-01 OOK-NRZ PER 1.00E E-03 8-DPIM A 8-PPM 16-DPIM A 1.00E PPM 1.00E Average received irradiance (dbm/cm )

39 Average Optical Power Requirement Vs. Bandwidth Average optical power requirement normalised to that required by OOK-NRZ to achieve a PER of 10-6 for packet lengths of 1 kbyte, on a channel limited by AWGN. Bandwidth requirement normalised to OOK-NRZ. Normalised Average Optical Power Requirement (db) NRZ RZ (0.5) RZ (0.33) Normalised Bandwidth Requirement OOK DPIM with no guard slot DPIM with 1 guard slot PPM with threshold detector PPM with MAP detector RZ (0.5)

40 40 Conclusions New modulation techniques has been proposed with built in frame synchronisation, higher transmission capacity and reduced bandwidth requirement compared to OOK and PPM PIM schemes are more resistant to the effects of multipath propagation compared with the same order PPM due to its longer slot duration For low values of normalised delay spread, PPM has lower power requirement than DPIM (NGB). However, DPIM offers lower ISI power penalty, and consequently as delay spread increases, it offers a lower power requirement. The addition of a guard band further improves the performance of DPIM in the presence of ISI. 3-DPIM offers lower ISI power penalty than 16-PPM, and it requires lower bandwidth.