Therefore, for any user J , there should be only one link established to any of the potential base stdons, namely, N
CXZJ= 1 i=l
drywall partitions
concrete services core
Conclusions: An alternative wireless system planning strategy that uses linear programming has been introduced. Although the use of linear expressions to define objective and constraint functions is a prerequisite to this strategy, it has been shown that it is possible to use performance measures such as SIR in optimisation by making a few reasonable assumptions. The results presented have shown that the proposed strategy is capable of delivering useful solutions for a base station placement problem. Although only a simple problem is considered as an illustration of the proposed strategy, more complex problems can be treated, e.g. allowing mobiles to connect to more than one base station simultaneously. The real benefit of the proposed strategy over nonlinear heuristic approaches (such as GSA) is more obvious for large problems. Heuristic techniques cannot be guaranteed to find the optimal solution. In contrast, the proposed strategy solves the problem deterministically and the optimal solution can always be found.
0 IEE 2001
1 June 2001
Electronics Letters Online No: 20010749 DOI: IO. 1049/el:2OOIO749
J.K.L. Wong, M.J. Neve and K.W. Sowerby (Department of Electrical and Electronic Engineering, School of Engineering, The Ukiversity of Auckland, Private Bag 92019, Auckland, New Zealand) Fig. 1 Floor plan of building used in example problem (dimension 18.5 x 18.5 m)
E-mail:
[email protected]
0 users f potential base station sites
References
Example: For the indoor environment in Fig. I , we aim to select a number of base station sites from a group of 12 potential sites and determine how 15 users should be allocated to these base stations.
In this example, a CDMA system with a processing gain of 128 is assumed (G, = 128). With an operating frequency of 1.8GHz, = lpw) the receiver sensitivity is assumed to be -90dBm and a protection ratio of approximately 9dB (Q = 8) is used. All antennas are assumed to be omni-directional and all the base stations transmit at = 1 mW. Each base station is allowed to serve a maximum of five users (Blm= 5) and an arbitrarily high base station cost is assumed ( W = 50). For this frequency, a database of path loss data has been obtained from measurements. The problem has been solved using a branch-and-bound method incorporating the simplex algorithm [7]. For comparison, a heuristic algorithm, guided simulated annealing (GSA) has also been used to solve the problem [I]. Its objective function maximises the SIR of each link as defined in eqn. 1 and the number of base stations to be considered is fixed at four to allow direct comparison with the results generated by the proposed strategy.
Table 1: Base station (BS) sites and user allocations for proposed strategy
I I
BS site number
I
1 6-
1 I
User allocations (SIR level in dB) l(l5.1), 2(13.7), 13(13.8), 14(14.6), 15(15.0) 9fl6.1). lO(16.3). ll(16.3). 12(16.2) .,. . .. ,
I
l?
Table 2 Base station (BS) sites and user allocations for GSA algorithm
I BS site number I 1
4 5
9
User allocations (SIR level in dB) 1(16.3), 2(16.3), 3(16.0), 15(16.3) 1l(13.3). I2(16. I)., 13(15.9). , , _ , 14(13.6) _ . , IO(4 1.8) 4(14.0), 5(13.9), 6(14.0), 7(14.l), 8(13.8),9(11.4)
I
Results: Tables 1 and 2 summarise the base station sites to be used and the allocation of mobile users for the proposed strategy and the GSA algorithm, respectively. These results give a minimum SIR for the proposed strategy of 13.7dB compared to a corresponding value of 11.4dB for the GSA. Although both cases satisfy the threshold constraint, the solution obtained using the proposed strategy offers an improvement over GSA and is guaranteed to be optimal given the initial constraints (unlike GSA).
ELECTRONICS LETTERS
16th August 2001
WONG, J.K.L.,NEVE, M.J., and SOWERBY, K.w.: ‘Wireless personal communications system planning using combinatorial optimisation’. Proc. 10th Virginia TecWMPRG Symp. Wireless Personal Communications, June 2000, pp. 179-190 NEVE, M.J., and SOWERBY, K.w.: ‘Optimising the performance of indoor wireless communications systems’. Proc. IEEE Vehicular Technology Conf., 1999, Vol. 2, pp. 968-972 CHEUNG,K.w.,and MURCH, R.D.: ‘Optimising indoor base-station locations in coverage- and interference-limited indoor environments’, IEE Proc. Commun., 1998, 145, (6), pp. 4 4 4 5 0 MOLINA, A., ATHANASIADOU, G.E., and NIX, A.R.: ‘The automatic location of base-stations for optimised cellular coverage: a new combinatorial approach’. Proc. IEEE Vehicular Technology Conf., 1999, Vol. 1, pp. 606-610 WRIGHT. M.H.: ‘Optimization methods for base station placement in wireless applications’. Proc. IEEE Vehicular Technology Conf., 1998, Vol. 1, pp. 387-391 LEE, C.Y., and KANG, H.G.:‘Cell planning with capacity expansion in mobile communications: atabu search approach’,IEEE Trans. Veh. Technol., 2000, 49, (9,pp. 1678-1691 HILLJER, F.S., and LIEBERMAN, G.J.: ‘Introduction to Operations research’ (McGraw-Hill, New York, 1990) RAPPAPORT. T.s.: ‘Wireless communications - principles and practice’ (Prentice-Hall Inc, 1996) KIM, K.: ‘Handbook of CDMA system design, engineering, and optimisation’ (Prentice-Hall Inc, 2000)
Vol. 37
Radiofrequencytransmission of 32-QAM signals over multimode fibre for distributed antenna system applications D. Wake, S.Dupont, C.Lethien, J-P.Vilcot and D. Decoster High frequency, narrowband radio transmission over multimode fibre is shown to be possible with performance comparable to singlemode fibre using low-cost lasers and photodiodes. This allows the use of existing fibre cable infrastructures within buildings for the realisation of cost effective distributed antenna systems.
Zntroduction: Distributed antenna systems (DAS) are being used increasingly in public and private buildings such as shopping centres and corporate office’blocks to provide enhanced capacity and coverage of cellular radio services. DAS may be implemented using either coaxial cable or optical fibre, the latter having the advantages of small size, low weight and virtually length inde-
No. 17
1087
pendent loss. It would be advantageous to make use of the preinstalled fibre infrastructure that most large buildings possess for local area network (LAN) applications in order to reduce the system cost for fibre-based installations. Most fibre-based DAS products use singlemode fibre (SMF) to avoid modal dispersion, which degrades the high frequency performance of multimode fibre (MMF). In most cases, however, the SMF cables must be installed separately; a recent survey has shown that SMF represents less than I% of fibre cables within buildings [I]. One fibre-based DAS product has been designed for MMF transmission by using downconversion of the radiofrequency carrier signal to a lower (IF) frequency which falls within the MMF bandwidth [2]. This system also transmits a global reference tone for stabilising and locking the local oscillators which are required to upconvert the signal at the remote antenna units. This adds considerable complexity and cost to the installation. MMF with a 6 2 . 5 core ~ diameter has a minimum specified bandwidth-length product of 5 0 0 M H z . h at a wavelength of 1300nm as recommended by ISODEC 11802. It is clear that only very short fibre spans will have a bandwidth able to support even current second generation (2G) cellular radio systems operating at 900 and 1800MHz. Next generation systems such as third generation (3G) cellular and high performance radio LANs (HIPERLAN) operate in even higher frequency bands; 2 GHz for 3G and 5 GHz for HIPERLAN. However, Raddatz et al. have shown recently that reliable passband transmission is possible over MMF beyond the 3 dB bandwidth limit [3]. This is because the impulse response of MMF contains a series of delta functions with different arrival times, which gives rise to high frequency components with considerable amplitude in the frequency domain. They have demonstrated successful data transmission for high-speed LAN applications at frequencies much higher than the bandwidth limit. In this Letter we aim to show that successful radio frequency transmission using complex digital modulation schemes can be supported over MMF using, low-cost lasers and photodiodes for DAS applications.
HIPERLAN) is less than 1dB. Fig. 2 shows the impulse response of the lkm MMF reel. The solid line shows the response using standard FCiPC connectors and adaptor between the SMF laser pigtail and the MMF reel. It is clear that most of the optical power is contained within one group of modes, which is expected to be low-order from earlier work on SMF-MMF coupling [4]. The dotted line shows the response obtained when the SMF to MMF launch between laser and fibre reel is offset by around 10pm using butt-coupled cleaved fibres and micropositioners. In this case a high-order mode group is also excited, which confirms the low-order dominance of the FCRC coupling arrangement. This low-order dominance was maintained with a variety of different F C P C adaptors, which demonstrates the repeatability of the launch conditions for this type of physical contact connector.
time, ns
1783121
Fig. 2 Impulse response of 1 km MMF using DFB laser source
using FCIPC connectors and adaptor _ _ _ _ using 10 pn offset launch ~
.... ...... ......
....... .....
2
....
0
.... ......
...... ...... .... m..r..
m -2 U 0-
i4 2-6
-a -10
0
2
4
frequency, GHz
6
178311
a a
b
1783/31
Fig. 1 Frequency response of M M F fibre reels using DFB laser sourre Fig. 3 Constellation and eye diagramsfor 32-QAM
(i) lOOm (ii) 500m (iii) 1 km
Coaxial cable (reference) b 1 km MMF transmission U
Frequency und impulse response meumrenzents: Two types of laser were used for the optical links, both operating at a wavelength of
13OOnm and both with SMF pigtails. The first was a high performance distributed feedback (DFB) device with integrated isolator and the second was a low-cost Fabry-Perot (FP) device, designed as a product for DAS applications. The photodiode used was p i w e d with MMF. Three different fibre reels of 50 pm core diameter MMF were used, with lengths of 100m, 500 m and 1 km. A recent survey [l] showed that more than 80% of in-building fibre spans are less than 300m, so the fibre spans used in our experiments are much longer than those typically found. Network analysers were used to measure frequency and impulse responses of these fibres and a vector signal analyser was used to perform the digital modulation experiments described in the next Section. Fig. 1 shows the frequency response of the fibre reels up to a frequency of 8 GHz using the DFB laser as the optical source. The worst-case response occurs for the 1km MMF reel and is around -7dB (electrical) over this frequency range. Variation of response within each of the frequency bands of interest (for 2G, 3G and
1088
32-QA AM transmission experiments: To demonstrate that we can obtain good quality radiofrequency transmission over MMF we used a vector signal analyser (VSA) set for 32-QAM modulation and demodulation. This complex digital modulation scheme requires a signal-to-noiseratio of more than 25 dB and is therefore a good test of the fibre performance. In our experiments we chose carrier frequencies of 1.5GHz (which gives the worst-case response for the 1 km reel of MMF) and 2GHz (3G band). The symbol rate was set to 2 MsymboUs, which corresponds to a transmission bit rate of 10Mbitis. Fig. 3 shows the constellation and eye-diagrams for 32-QAM transmission; Fig. 3a shows results for a short coaxial cable link as reference and Fig. 3b shows results for l k m MMF using the DFB optical source. It is clear that transmission over the fibre reel is not significantly degraded compared to the short coaxial cable. Error vector magnitude (EVM) is the most widely used metric for assessing digital transmission quality since it is often more convenient to measure compared with bit error ratio. In Table I the
ELECTRONICS LETTERS
16th August2001
Vol. 37
No. 17
EVM measurement results are given as a function of link type and carrier frequency. The results show that the optical links that use the DFB laser and SMF do not degrade the EVM compared to the coaxial cable reference. When the link uses 1km MMF, the degradation is extremely low at 2 GHz and only slightly higher at 1.5GHz (where the fibre response is lowest). The results are worse when the F P laser is used, which may be due to the absence of an isolator for this device. The worst-case transmission occurs for 1 km M M F at 1.5GHz using the low-cost F P laser, where the EVM is 3.3%).This value, however, is still very low compared to what is required for successful 32-QAM transmission. The minimum SNR required for 32-QAM is 25dB, which corresponds to an EVM of more than 17%. Furthermore, the maximum EVM allowed in the UMTS specifications (one of the 3G standards) is also over 17% [5] for QPSK modulation. It is clear that even for worst-case transmission, the use of 1km M M F and low-cost lasers do not degrade EVM significantly.
Table 1: Error vector magnitude as a function of link type and carrier frequency
Fibre: SMF patchcord Fibre: 1.1km rcel SMF Fibre: 1 km reel MMF Coaxial cable reference
Error vector magnitude, % 1.5GHz 2.0GHz FP DFB FP DFB 1.2 0.7 0.8 0.6 1.2 0.7 2.1 0.6 3.3 1.2 1.8 0.7 0.7 0.5
Conclusions: We have shown that fibre of the type that is preinstalled in most large buildings can be used for radiofrequency transmission with complex digital modulation schemes above the 3 dB bandwidth using low-cost lasers and photodiodes. The penalty compared to using specially-installed fibre is negligible for 32QAM systems. Fibre-based DAS can therefore utilise existing infrastructure without the added complexity of frequency translation and global reference transmission, which means that low-cost installation will be possible.
0.3 dB minimum noise figure at 2.5 GHz of 0.13 pm Si/Si0.58Ge0.42 mMODFETs M. Enciso, F. Aniel, P. Crozat, R. Adde, M. Zeuner, A. F o x and T. H a c k b a r t h RF and microwave noise perfonnances of strained Si/Sio.58Gq,42 n-MODFETs are presented for the frst time. The 0 . 1 3 gate ~ devices have de-embedded fT = 49 GHz, Anox= 70 GHz and a record intrinsic gt. = 700 mSimm. A de-embedded minimum noise figure NF,,,,, = 0.3dB with a 41 Q noise resistance R, and a 19dB associated gain G, are obtained at 2.5GHz, while NF,,,, = 2.0dB with G,, = l0dB at 18GHz. The noise parameters measured up to I8GHz and from 10 to 180mAimm with high
gain and low power dissipation show the potential of SiGe MODFETs for mobile communications. Introduction: Very good microwave performances have been
reported for n-channel SiiSiGe-based MODFETs at 300K and low temperature [l, 21; however, only noise predictions by simulation are available for them [2, 31. We present the first experimental R F and microwave noise investigation at 300 K of n-type strained SiiSiGe MODFETs. The four noise parameters of 0.13 pm mushroom gate MODFETs were studied over a frequency range from 1 to 18 GHz and over a current range from 10 to 18OmAimm. Experimental: The epilayer structure SiiSi, 58Geo42 ,is grown by
solid source molecular beam epitaxy (MBE) on a high resistivity lOOOC2cm p-type Si substrate, starting with a relaxed SiGe buKer layer, the Ge content of whch is graded from 0 to 42% The core of the active structure is a 9 nm-thick biaxially strained Si chaniiel embedded between two undoped 3 mn Si, s s G42~ spacers and two Sb-doped SiGe carrier supply layers. The doping level ofthe upper supply layer is 2 x 101ycn-3,while it is 2 x 1018cm-3for the lower supply layer. The mushroom Schottky gate with 0.13 pn footprint, n: topology and 10Opn width is pdttemed by e-beam lithography and is located symmetrically within a 2 p n source-drain space. The noise parameter determination exploits the noise figure variation against frequency [4] for only two impedance sources, a 50Q and an offset short line presented at the input transistor.
Acknowledgments: The authors would like to thank Professor
Seeds of University College London. for the loan of some of the optical devices, and C. Segunot and N. Rolland of IEMN, for the loan of some of the measurement equipment. D. Wake would also like to thank the Centre National de la Recherche Scientifique for funding his stay at IEMN. 0
IEE 2001
22 May 2001
Electronics Letters Online No: 20010748 DOI: IO.1049/el:20010748
D. Wake, S. Dupont, C. Lethien, J-P. Vilcot and D. Decoster (Institut d’Electronique et de Microdiectronique du Nord, U M R CNRS 8520, UniversitP des Sciences et Technolofiies de Lille. Avenue Poinuare, 59652 Villeneuve d X s c q Cedex, France)
- 20
E-mail:
[email protected] -I5
m U
.-i
References
m
-10
pm .-0
1 FLATMAN. A., and MUSK, R.: ‘In-premises optical fibre deployment’. IEEE 802.3 High Speed Study Group Plenary Week Meeting,
5:
a
Montreal, 1999
-5
GEORGES, J.B., CUTRER, D.M., and LAU, K.Y.: ‘Distribution O f radiofrequency signals through low bandwidth infrastructures’. US Patent 5,765,099, 1998 3 RADDATZ, L., HARDACRE, D., and WANG, s.-Y.:‘High bandwidth data transmission in multimode fibre links using subcarrier multiplexing with VCSELs’, Electron. Lett., 1998. 34, pp. 686-688 4 HAAS, z., and SANTORO. M.A.: ‘A mode-filtering scheme for improvement of the bandwidth-distance product in multimode fiber systems’, J. Lightwave Technol., 1993, 11, pp. 1125-1130 5 3GPP TS 25.106 version 4.0.0: ‘UTRA repeater; radio transmission and reception’. 2001
2
ELECTRONICS LETTERS
16th August 2001
Vol. 37
0
5
I
I
10
15
frequency, GHz b
10 20 -~
1045111
Fig. 1 Equivalent noise resistance R,,. minimum noise figure NE,,, and available gain G , against Jrequency j b r 0 . 1 3 p gutr irngth n-type Si/ SiGe MODFET at V D , ~= 1.25 V and Ins = 70.7mA/mm a R, against frequency b NF,,,, and G,, against frequency
No. 17
1089
