MIMO Channel Measurements for an Indoor Office Environment

Transcription

1 MIM Channel Measurements for an Indoor ffice Environment Paul Goud Jr. 1, Christian Schlegel 1, Robert Hang 1, Witold A. Krzymien 1,, Zachary Bagley 3,4, Shayne Messerly 3, Paul Watkins 3, Viswanathan Rajamani 3 University of Alberta 1, TRLabs, University of Utah 3, L3 Communications Inc. 4 pgoud@ee.ualberta.ca Abstract - A mobile 4x4 multiple input multiple output (MIM) system has been constructed and used to perform wireless measurement campaigns on non-line of sight (NLS) channels in an indoor office setting. The MIM system uses orthogonal 5 kchips/s Walsh coded signals that are filtered and simultaneously transmitted at 916 MHz. Two identical campaigns were conducted: one with antenna spacing of λ/4 at both the transmitter and receiver stations and another with antenna spacing of λ/. The complex MIM channel matrices obtained in the campaigns were used to calculate the theoretical capacity for a 4x4 system. ur measurements show a slightly higher average channel capacity for the λ/ case as compared to the λ/4 case (. bits/use vs bits/use).the 9 th percentile channel capacity is 18. bits/use for the λ/ spacing and 16.5 bits/use for the λ/4 spacing. I. INTRDUCTIN MIM (multiple input multiple outpu wireless systems hold the promise of providing much higher spectral efficiencies than conventional systems [1]. By using techniques such as space-time coding and space-time layering, a rich scattering environment can provide multiple coexistent radio channels. To better understand the spectral efficiencies that can be achieved, a detailed study of the MIM channel for typical environments is necessary. A flexible and mobile MIM testbed containing four transmit antennas and four receive antennas has been developed in the icre HCDC (High Capacity Digital Communications) laboratory at the University of Alberta, Canada. In this paper, we present the results of our recent MIM channel measurement campaigns for non-line of sight channels in an indoor office environment. For our measurement campaigns, we radiated orthogonal Walsh coded signals that had a chip rate of 5 kchips/s and were filtered by a square-root raised-cosine filter that had a passband cutoff of 5 khz. The four signals were transmitted simultaneously on the same frequency band. This simultaneous transmission method is in contrast to the spaced sinusoid [] and virtual antenna array [3] techniques employed by other research teams. The MIM channel complex gain matrices are used to calculate their theoretical capacity and decoupling properties. This paper is structured in the following manner: Section II gives an overview of the system hardware and the implemented algorithms. Section III discusses the validation test and the measurement accuracy of our system. Details of the measurement campaign location and procedure are presented in Section IV. The calculated capacity results are given in Section V. Finally, Section VI contains some interpretations of the results and conclusions. II. SYSTEM VERVIEW A. System Hardware The transmitter and receiver stations each consists of an FPGA (field programmable gate array) development board for baseband processing, a custom RF (radio frequency) module (for upconversion or downconversion) and a custom antenna array structure. A PC is used to process captured data from the receiver FPGA board and calculate the theoretical channel capacity. A block diagram of the hardware components is shown in Figure 1. Figure 1 - Block diagram of the system hardware All the FPGA development boards are the GVA-9 model, manufactured by GV & Associates Inc. Each GVA-9 board contains four Analog Devices AD976 DACs, four Analog Devices AD943 ADCs, two ilinx Virtex-E and a ilinx Spartan-II FPGAs. Four-channel RF modules were custom made for this project by SignalCraft Technologies Inc. Each upconversion module receives the spread signals at an IF of 1.5 MHz and further upconverts them to the ISM Supported in parts by the Alberta Informatics Circle of Research Excellence (icre), the Natural Sciences and Engineering Research Council (NSERC) of Canada, the Canadian Foundation for Innovation (CFI) and the National Sciences Foundation (NSF) under grant ECS

2 Normalized Correlation band (9-98 MHz). The signals are also amplified to a maximum output power of dbm per channel. The down-conversion module is the mirror image of the up-conversion one: the four signals received from the antennas are amplified and downconverted from the ISM band to 1.5 MHz. n both sets of RF modules, only one local oscillator is used at each frequency stage to generate the signals that mix with the four input signals. This ensures that no phase shifts are introduced between the four channels by the modules. For our measurement campaigns, broadside antenna arrays are used at both the transmitter and receiver stations. Each array consists of four monopole antennas that are mounted on conductive sheets. Such a design makes them effectively behave as vertically polarized centre fed dipoles. Two sets of arrays have been constructed: one with antenna spacing of λ/4 and another with antenna spacing of λ/. with the same characteristics as the modulation pulse shaping filter. An involved parallel synchronization algorithm [4] is used to synchronize the reference signals in the receiver with the received signals. Each received signal is correlated with all four coded signals in order to create complex 4x4 correlation matrices for each sample instant. The individual matrix elements are squared and subsequently all 16 values are added together. A plot of the summed squared correlation values will be periodic (the period of the spread signals is 64 µs) and will contain one prominent peak every period (Figure 4). The sampling instant, which corresponds to correct synchronization in every period, is determined by selecting the peak squared correlation value. The correlation matrix at the selected synchronization instant is equal to the MIM channel gain matrix G( and is used to calculate the theoretical channel capacity. B. System Algorithms An FPGA image has been developed for our transmitter station which generates and simultaneously transmits four filtered 5 kchips/s spread spectrum signals (see Figure ). The spreading sequences used are orthogonal Walsh codes of length 3 multiplied by a pseudorandom m-sequence to provide good spectral and correlation properties. The chip pulse has a square-root raised-cosine shape with a cutoff frequency of 5 khz and a roll-off factor of.31. The four filtered signals are upconverted digitally to an IF of 1.5 MHz before being sent to the DACs. Figure 3 Block diagram of receiver signal processing chain Figure Block diagram of transmitter signal processing chain (single channel) At the receiver station (see Figure 3), the four parallel downconverted signals are received at an IF of 1.5 MHz from the RF downconverter module. The signals are sampled simultaneously by the ADCs on the FPGA board at 5 Msps. The four signals are digitally downconverted to baseband by the FPGA. Blocks of complex data samples are captured and sent to the PC for further processing in Matlab. The Matlab program decimates the signals to.5 Msamples/s (5 samples per chip) and passes them through a square-root raised-cosine matched filter Code Phase (chips) Figure 4 Summed squared correlation response for one code sequence period

3 C. MIM Channel Capacity Calculation A MIM system with n T transmitters and n R receivers can be depicted as shown in Figure 5. The signal at the output of the receive antenna at time t is given by the equation: ρ y ( = H ( ( Z( n n T + (1) Where ρ is the average received SNR, H is the normalized channel gain matrix (the elements of the matrix are H ij (), ( is the transmitted space-time signal and Z( is the sampled noise at the receiver. For the system described above, the instantaneous capacity of the MIM channel is calculated as [1]: ρ * C( = log det( I nr + H ( H ( ) () n where I nr is the identity matrix and H( * represents the conjugate transpose of H(. T where G( is the Frobenius norm of G(. For the situation where n T = n R, the maximum MIM channel capacity will occur when H(n) is a diagonal matrix with equal values (i.e. uncoupled equal power paths are present from the transmitter to the receiver). For a 4x4 MIM system operating on a channel with ρ = db, the maximum capacity can be calculated to be 6.6 bits/use after the channel gain matrix has been normalized to create H(. The maximum capacity for a SIS channel can be calculated using Shannon s Theorem (3) which yields 6.7 bits/channel use. III. SYSTEM VALIDATIN TEST AND RESULTS In order to assess the error in the channel gains obtained with our system, measurements were made for an uncoupled line-of-sight (LS) MIM channel [6]. This LS channel was created by removing the antenna arrays and connecting the four RF outputs from the transmitter to the four RF inputs of the receiver with cables. Attenuators were also inserted in each connection to prevent damage to the receivers. For this uncoupled LS channel, the diagonal elements (h 11, h, h 33, h 44 ) of the channel gain matrix correspond to the connected paths and should have the same magnitude. All the off-diagonal values (h 1, h 13, h 14, h 1, h 3, h 4, h 31, h 3, h 34, h 41, h 4, h 43 ) represent the non-existent cross paths and should be zero. The deviation in the measured values from this expected result represents the system error. Figure 5 MIM system model For a single input single output (SIS) communication system (n T = 1 and n R = 1), capacity equation () becomes the Shannon s Theorem [5]: The normalized results for 8 measurements are plotted in Figure 6 and demonstrate that the error introduced by the system is low. Measured gains of all connected paths are within 1 db of each other. The power measured in each of the non-existent cross channels is at least db below the power measured for connected paths. C( = log (1 + ρ H ( ) ) (3) t where H( is the normalized channel power transfer characteristic [1]. The channel gain matrices G( obtained from our measurement system were normalized in order to create the H( matrices. This normalization was done using the following equation: G( nrnt H ( = (4) G Figure 6 Squared channel gain values for uncoupled LS test

4 Probability Capacity > Abscissa IV. MEASUREMENT CAMPAIGN DETAILS The fifth floor of the ECERF (Electrical & Computer Engineering Research Facility) building on the University of Alberta campus was selected as the location for the channel measurement campaigns. The ECERF building is constructed of steel, concrete and drywall which is typical of many modern office buildings. Eight non-line-of-sight (NLS) points on the floor were selected as measurement locations for the mobile receiver station. These locations were marked so that precisely the same ones were used for both campaigns. The MIM transmitter was positioned on a bench in the HCDC lab. A map of the floor showing the location of the transmitter and receiver positions is shown in Figure 1. At each measurement point, a set of nine measurements was taken (see Figure 7). The measurements formed a grid pattern with a spacing of λ/4. The transmitter and receiver gain settings were constant for an entire set of measurements. λ/4 λ/4 in Figure 9. The mean and 9 th percentile MIM channel capacities are shown in Table I. TABLE I Measured Channel Capacities: ρ = db Campaign Measured: λ/4 Spacing Measured: λ/ Spacing Bin Probability Simulated: Gaussian Channel λ/4 Spaced Array λ/ Spaced Array Simulated Gaussian Mean Capacity (bits/use) 9 th Percentile Capacity λ/4.4. λ/ Channel Capacity (bits/use) Figure 7 Receiver measurement locations Figure 8 Histograms of the measured channel capacities: ρ = db V. MEASURED CAPACITY RESULTS nce we had obtained channel gain data in our measurement campaigns, we normalized the matrices according to (4) and calculated the MIM channel capacities with (). In order to create an interesting comparison, a set of complex 4x4 MIM channel matrices was generated on a computer. Each element of these matrices was generated using an independent Gaussian random number generator with zero mean and a variance of.5 per dimension. The capacities of the simulated channels were calculated using () after being normalized. Channel gain matrices generated in this fashion have been used in several MIM capacity studies [1,7]. Histograms of the measured capacities for our 4x4 MIM system operating on a channel with ρ = db are shown in Figure 8. The graphs correspond to the two different antenna separation cases and the simulated channel case. The bin size for the histogram is.5 bits/use. The corresponding complementary cumulative distribution functions (CCDFs) are shown Channel Capacity (bits/use) λ/4 Spaced Array λ/ Spaced Array Simulated Gaussian Figure 9 CCDFs of measured channel capacities: ρ = db

5 VI. CNCLUSINS ne interesting observation results from comparing the calculated capacities for the λ/4 and λ/ cases. The capacity for the λ/ case is about 1.5 bits/ channel use higher. Since no other differences were present between the two measurement campaigns, it can be concluded that wider antenna separation results in a higher channel capacity. A second conclusion can be drawn by comparing the λ/ measured results in Table I to the maximum 4x4 MIM and SIS channel capacities. It can be seen that the average capacity for our indoor office setting (. bits/use) is equal to 3% of the SIS capacity (6.7 bits/use) and is about 4% lower than the maximum MIM capacity (6.6 bits/use). A third conclusion can be obtained by comparing the capacities of the simulated MIM channels and the experimentally measured channels. The channel capacities for the Gaussian channels are about bits/use higher. The reason for this difference may be the presence of a small specular component in the measured channels. Acknowledgement The authors gratefully acknowledge the support provided by the Natural Sciences and Engineering Research Council (NSERC) of Canada, the Alberta Informatics Circle of Research Excellence (icre), the Canadian Foundation for Innovation (CFI) and the National Science Foundation under grant References [1] G.J. Foschini and M. J. Gans, n limits of wireless communications in a fading environment when using multiple antennas, Wireless Personal Communications, vol. 6, no. 3, pp , March [] J. Ling, D. Chizhik, P. Wolniansky, R. A. Valenzuela, N. Costa and K. Huber, MIM measurements in Manhattan, in the Proc. IEEE Vehicular Technology Conference (VTC -Fall), Vancouver, Canada, Sept.. [3] A. F. Molisch, M. Steinbauer, M. Toeltsch, E. Bonek and R.S. Thoma, Capacity of MIM systems based on measured wireless channels, IEEE Journal on Selected Areas in Communications, vol., pp , April. MIM channel capacity measurements using the VT-STAR architecture, in the Proc. IEEE Vehicular Technology Conference (VTC -Fall), Vancouver, Canada, Sept.. [7] D. Chizhik, F. Rashid-Farrokhi, J. Ling, A. Lozano, Effect of antenna separation on the capacity of BLAST in correlated channels, IEEE Communications Letters, vol. 4, no. 11, pp , November MIM Transmitter Station - MIM Receiver Measurement Locations Upper Number - Measured Capacity for λ/ Array Lower Number - Measured Capacity for λ/4 Array Figure 1 Floor map for measurement campaigns [4] Patent pending. [5] R.E. Ziemer and W.H. Tranter, Principles of Communications, Houghton Mifflin Company, Boston, pp. 5 53, [6] R. Gozali, R. Mostafa, R.C. Palat, P.M. Robert, W.G. Newhall, B.D. Woerner and J.H. Reed,