MULTI-LEVEL MODULATION IN THE INDOORS LEAKY FEEDER. ENVIRONMENT J. M. Torrance, T. Keller and L. Hanzo Dept. of Electr. and Comp. Sc., Univ. of Southampton, SO17 1B4, UK. ‘Tel: +44 17 03 59 31 25, Fax: +44 17 03 59.30 45 Email: jmt94rGecs.soton.ac.uk, tk@ecs.soton.ac.uk and lh@ecs.soton. sGecs.soton.ac.uk itp: /rice ABSTRACT signal i transmitted from a series of slots acting as an- snd superposition of the paths results in fad of up to 25 dB below the mean signal level. A model {a presented for the propagation path botween the leaky feeder Bate Station (BS) antonna and a ingle omni= irectlonal recelver antenna, which was verified using ‘measurements. ‘The Bit Error Rate (BER) performance fof such a tystem is Investigated for 1, 2 and 4 bits per symbol, coherent and non-coherent modulation schemes ‘These are compared with the performance of @ more con- ‘ventional indoor BS antenna 1. INTRODUCTION ‘This paper presents a model simulating the propagation path between a leaky feeder transmitter and a singe om ‘irectional receiver antenna in an indoors environment. The ‘model is verfed by comparing simulated and measured im- pulee responses, The model is then used to simulate the up And down links for one user in a DECT (Digital Buropean, Cordless Telephone) system. This is compared with the performance of a single omni-directional to another single ‘omni-directional antenna, in the same enviroument. The BER performance of 1, 2 and 4 bits per symbol coherent ‘and noncahereat svadulation schemes are then presented for ‘oth antenna arrangements 2. LEAKY FEEDER 24. Propagation modal Leaky feeders area type of distributed antennas. They were originally conceived to provide subterranean radio propagar tion in train tunmels and coal mines. "They have receatly been considered 4s an alternative to conventional anton ‘ae for indoors micro-ells [1 They are constrarted From coaxial cable where the ovter shield is perforated with Series of oles, The coaxial eable is typically about ane hun ‘dred metres long and can be threaded through a building ‘offering radio illumination in away that would require many individual omni-directional antennae. Fabrication costs are relatively cheap and deployed ineligntly, leaky feeders a ‘reduce network inferstructure cost. ‘The propagation model presented considers the path between a leaky feeder and a (0-7803-3157-5196 $5.00 © 1996 IEEE c.uk single omni-directional antenna without taking account of the detailed geometry of the building upon the radiation from each alot of the leaky feeder, The model is restric ted to two dimensions, only the relative signal strength is considered and reciprocity is assumed. ‘The leaky feeder * tused was constructed of a 0.5" diameter copper foil outer fand a foam polyethylene dielectric inner. It had trans versal slots preudo-randomly distributed with miniamam and maximum spacing of 0.01 m and 0.3 m, respectively ‘The feeders linear axial attenuation x was 0.184B/m0 at a frequency of 2 GHz, the relative phase velocity 3 was (088. ‘The model uses a ray-tracing technique in conjune- tion with Devasirvathamn's [2] madel for indoors path loss. ‘Bach transverse leaky feeder sot is approximated by the radiation profile ofa dipole antenna. The leaky feeder was ‘modelled as a group of radiating dipoles lying on the sais of a evo dimensional plane, with the input at and N slots at coordinates (D(n},0) for n= 1,2, 0, N. Let P= Ae’, an idealised impulse of amplitude A and phase , propagate along the leaky feeder from x = 0 at a ve. locity of e- 8 ms~, where cis the speed of light in free space. Te impulse will take ¢(n) = D(n)/(Be) seconds to travel to alot n, The attenuated and phase-rotated impulse at this point is described in (1), where is the free space vwaralengeh: Stn) = 10" Actto+ 222RL2)) mn ‘The propagation distance from the n* slot to a receiver at (Re Ry) is given by (2) for n= 1,2)... Ve = Dia)? +B ® ‘A free space propagation velocity is assumed once the im ‘Pulse has left the cable and hence the impulse takes a(n) Rin) fe seconds to travel from the n™ slot to (Ray By) By treating each slot ac an independent dipole, t can be assumed that the radiation pattern of each soe relates the trauamitted power In a ven direction, to the cosine ‘of the angle, 0, between the direction of propagation and perpendicalar to the eable. Tecan be seen that the received limpulse at the antenna is given by (8), where M(R(n)) is, the amplitude path les from slot n to the receiver antenna Rin) “i anthors would He to expres their gratitude to Profesor ‘Stele for intating the projet, to Mr David Stewart fand Terry Mltcel fr tht lp ding it execution abd to Mr ‘Davies a RES Hanover for the supply ofthe Leaky Feeder cable. 1554at position (Re, Ry) Qn) = S{n) -co0(6)- A(Rn)) (8) Devasirvatham's model is used in the following form: M(Q)= fF 0] i: @ Using a similar technique to [3] for calculating the sum of aa ensemble of paths and Vetting represent time, can be ‘cen that the impulse response igen by no-{ A series of impulse response measurements were made from ‘aleaky feeder to an omni-diretional antenna, using a Swept ‘Time Delay Cross Correlator (STDCC). Impulse responses calculated using (5) are of a much higher resolution than the measured results and must be post processed in order ‘to enable a reasonable comparison, Since the average auto- correlation W'(n) of the STDCC pseudo-noise sequence is not an ideal impulse-like autocorrelation function, it follows from [4] that the STDCC's estimated output signal is given by: Qa) t= tn) +400) 0 otherwise © ds w= D> A) -We=#) "The average auto-correlation is used because in practice the STDCC receiver wll not be perfetiy aligned with the ‘transmitter and this wil resule im the receiver's sampling points not always falling on the auto-correlation peaks. The STDCC receiver partly avoids this problem by taking the average of every two measurements, ‘The post processing ‘st fae this into account. From {6] we ean ave at (7), ‘which is an expression forthe auto-correlation of the PN quence ¥(t): CC) ve m= (m+1)k if [kl <1 otherwise, @ a hence we have vq -M@evee o Finally, asthe STDCC only tas restion of 20n, ‘he simlated imple response s averaged nto 20m bin ich allow the goertion of comparable amber of sample Ponts fer both the smslated and teased resus Figure 1 shows a measured impulse response and sima- Inted impulse response using the fee space path lo Devasivatham's model after postprocesing. The Figure portrays a comparison at (yyy) eal to 162) m. The Fee nace model predit an pele reopone wth higher ‘stay epred han that measured, wile Devaar’ ‘ode! flows the sheared erat more cow.” Further omparions beewees the mearaed and modeled senaroe ‘towed tha when the receiver was very cose (es tha 1 rete) tothe cable the modes were less acura. 1585, Figure 1: Measured and simulated impulse responses between a leaky feeder positioned along the y = 0 and a Single omni-directional antenna positioned at (15,3) Figure 2: ‘Three dimensional plot ofthe received signal level, difference between an omni-directional antenna positioned at (50,0) and a 100 m leaky feeder positioned along the =O axe 2.2, Pathlose In this Setion we contrasted the patloss ofthe leaky feeder with that ofa conventional omidirectionl antenna by plot- tng the difference between their received signal levels as & function ofthe receiver's positon, while using an identical transmitted power. The three-dimensional plot of Figure 2 shows that the leaky feeder has a higher average received ‘signal level than a single omni-directional antenna and it can offer up to 50 dB reduction in propagation lose com- pared with the single omnidirectional antenna. By con trast, the single omni-directional antenna never offers more than’ 10 dB benefit, which is limited to the narrow valley im the centre of the Figure. Figure 3 shows more clearly hat the leaky feeder gives an improved SNK for over 73% of the floor space compared with a single omni-directional antenna,Figure 8: Contour plot of the recived signal love difference Detween an omni-directional antenna positioned st (50,0) and a 100 m leaky feeder positioned along the y ‘The area between the O dB contour-lines represents the re- ion where the omai-directional antenna’s path lssis lower - Delay Spread ‘The root mean squared (RMS) delay spread, A, is defined [ete Dy? mat] T= Moa where D isthe mean delay. The mean delay, Dy and the RMS delay spread, 4, can bo calculated for the simulated ‘impulse response given in (5). Figures 4 and S show these valle for “10 m'< 2 < 110 mand Om - Qn). (19) ‘This addition was performed for each transmitted or re- ceived symbol 2.5. Fading Characterisation, ‘The fading profile between the leaky feeder and the single antenna can be sean in Figure 6. Each marker inthis FigureFigure 6: Mean path los before and after power normalisa- tion of 15 dBm" caleulated in blocks of 100 ms for the up and down slots ofthe frst user in a DECT system along the y= 10 axis with a leaky feeder along the y = 0 axis, mobile velocity of 0.5 ms“! represents the average power of the symbols that constitute 10 up link and 10 down link slots. The 15 dBm~" linear attenuation ofthe leaky feeder is explicit in the Figure, but in order to consider probability density fanction (PDF) the fading, this factor is removed from the fading by normal- ising, The effect of normalisation is also seen in Figure 6, “Having removed the 15 dBm™~ linear attenuation factor from the channel, a histogram of the fade depths can be gen~ exated. In preference to calculating a single histogram for the whole channel, 1 m (or 2s) sogments were considered and plotted in Figure 7. From this itis clear that the PDF of fade depths is similar over the distance range of Om <2 < 100 m. In tho multipath environment of the ran- domly distributed slots there is no. dominant line-of-sight (LOS) path, hence simple logic would dictate that the fad- ing is Rayleigh. However, comparison of the experimental PDFs with a Rayleigh distribution reveals that the fades are less deep, since the fade depth is reduced by the mildely dispersive channel 8, MODEM PERFORMANCE, Im our further investigations the BER. performance of co- herently and noncoherently detected 1, 2 and 4 bits per symbol modems was studied in the indoors envionment of the leaky feeder, assuming a DECT-like TDD system sig- naling at 1152 kBaud irrespective of the number of mod- ‘lation levels, Since the symbol interval length is nearly 1s, the 0.03, dispersion seen in Figure 5 allows us to consider the channel to be essentially non-dispersive and Gerive BER results using the leaky feeder fading envelope of Figure 6, These will be contrasted with the BERS of the conventional omnidirectional antennae. For noncoher- tent modulation there were 421 symbols per frame, the fist symbol being the reference for demodulation ofthe fst data ‘Sibel, while for coherent modulation there ware 422 sy bbols per frames with a pilot eymbol both at the start and end of the data block, The pilot symbols were used to es timate and hence compensate for channel fading. Perfect 1587 Figure 7: Segmented PDF of fading profile taken every Im, along the y = 10 axis with a leaky feedor along tho y = ‘axis, mobile velocity of 0.5 ma~*, considering the up and down slots of the first user in a DECT system. ll measure- ‘ments were made after normalising for the linear attenuation factor clock and carrer recovery were assumed ‘The simulated BER experiments were conducted forthe leaky feeder channel as desribed in Section 24. The fad ing PDF is similar along the longth of the leaky foder, as shown in Figure 7. Removing the 15 4B bias fom the sims lated leaky chanel before performing the BER experiments, permits the generation of SNR versus BER. curves that are Independent of the position ofthe omni-directional receiver enna relative to Gu leaky feeder. In reference 7] wo have Shown that fr reasonable velocities the performance of in fly interpolated pilot symbol asisted modems (PSAM) wl not deteriorate significantly with respect to more com> Plex interpolators. ‘Therefore, given that the chanel can be assumed to lekave in a narrowbaad fashion, the BER performance of the various modulation schemes is only de- Pendent upon the mean SNR. ‘The BER performance of the modulation schemes was vated fr Both the leaky feeder channel model and for ‘ conventional omni-directional scenario using the best-case Gaussian chaasel and the worst-case Rayleigh channel. ‘The Rayleigh channel had a normalised Dopplr frequency qc ‘alent to that of the leaky feeder channel. ‘The BER per- Tormance over the leaky feeder channel is expected to be between the upper and lower bound of the couveational omni-directional antenna Figure § shows the performance ofall the modulation schemes through the leaky feeder chansel. The usual ac- Tonys are used, namely Binary Phase Shift Keying (BPSK) 4nd its diferent version DBPSK, Quaternary Phase Shift ‘Keying (QPSK) and DQPSK as wel as Quadrature Amp- litude Modulation (7] with eoherent maximum minimura distance square (SQ) constellation andthe non-coherent star (ST) constallaion. As expected, the coherent schemes per- fon better than the equivalent diferental, noncoberent ‘ehoman, where th diffrence is greater with more moda lation levels. Figuros 9 and 10 show the performace in Rayleigh and Ganssian channels forthe corresponding omni- ectional schemes. As expected, the leaky feeder BER.Figure 8: BER for a mobile travelling at 0.5ms~* along the 1y = 10 axis with a leaky feeder along the y =O axis. Results are the average BERS for up and down-link slots of the fist user in a DECT system. Differential modulation deploys a singe plot atthe beginning of a DECT frame and coherent modulation deploys PSAM and a pilot at the beginning and end of the frame curves are between that ofa Gaussian channel and a Rayleigh channel 16QAM performs between 7 and 15 dB better over the conventional Gaussian scenario than in the leaky feeder chan nel. Tn order to compensate for the performance degrada- tion induced by the inherent multipath characteristies ofthe leaky feeder, the SNR benefit of the leaky feeder must ex- ‘eed these values in stationary situations. From Figure 3 it can be observed that the leaky feeder does offer more ‘than 15 dB improved SNR over most of the illuminated floor space. In the more typical case of uon-LOS Rayleigh ‘channels the benefit of the leaky feeder will be even sore Drevalant, A range of further intresting conclusions can >be gleamed from the detailed examination of Figures $10, 4. CONCLUSIONS ‘A simulation model of the leaky feeder channel bas been presented and verified by impulse response measurements. ‘Tho BER performance of the channel between a mobile receiver and the leaky feeder has be considered for viti= ‘ous modulation schemes. ‘The leaky feeder typically offers higher average SNR than a single omnidirectional anteuns, ‘which is associated with inherent multi-path propagation, 5. REFERENCES [1] Kevin J. Byo, “Loaky-feedare for cordless communica: ton in the office,” in TEBE Burocon, pp. 387-390, IEEE, 1038. D. M. Devasirvatham, “Multi-frequency propagation ‘measurements and models in a large metropolitan com- ‘mercial building for personal communications,” in IEEE ‘PIMC, pp. 98-103, IEEE, 1901. K.W.Cheung, RD. Murch and CC. Ling, “Phe fad- ing, characteristics of distributed antennas for indoor Q a 1588 Figure 10; BER in Rayleigh Channel with normalised Dop- pier frequency equivalent to a mobile velocity of 0.5ms in a DECT-Iike system ‘wireless systems,” in Fourth IEBE International Con- ference on Universal Personal Communications Record, bp. 377-381, TEEB, 1995, J.D. Balcomb, H. B. Demuth and E. P. Gyftoponlos, “SCross-correlation method for measuring the impulse re- sponse of reactor systems” Nuclear Srience and Engin- ering, vol 1, pp. 159-166, 1961. 3.D. Parsons, D. A. Demery, A. M. Turkmani, “Sound ing techniques for wide-band mobile radio channels: A. review? IBE Proceedings-1, vl. 188, no. 5, pp. 437-440, 1991 W.T. Webb, L. Hanzo: Modern quadrature amplitude ‘modalation:' Principles and applications for fixed and. wireless channels, IEEE Press-Pentech Press, 10904, ISBN 0-7278.1701-6, p 557 J. M, Torrance and L. Hanzo, “Comparative study of pi- lot symbol assisted modem schemes," in IEE Conference ton Radio Receivers and Associated Systems, pp. 36-41, IEE, 1995. a fo] @