Mustafa K. Gurcan - Imperial College London
Irina Ma - Imperial College London
Hamdi Joudeh - Imperial College London
Qinxin Liu - Imperial College London
At Imperial College London, we conduct research and educational activities in wireless communications including cellular transceiver optimisation; the application of multiple input, multiple output (MIMO) systems; and resource allocation for 3G, 4G and next-generation networks. The focus of our work is on methods and technologies for next-generation MIMO systems, which trade-off diversity gain to improve spatial multiplexing. This provides a measure of the data throughput which can be transmitted by the MIMO system. The theoretical capacity for the high-speed downlink packet access (HSDPA) and long-term evolution (LTE) systems for 3G and 4G define the upper bounds of data throughput.
Algorithmic simulations and theoretical calculations are an effective part of mobile radio system design. However, the transition to real signals and physical hardware is a crucial part of system verification. To truly test the performance of our spatial multiplexing techniques, we built a rapid prototyping system to capture throughput performance with real signals.
Our system implementation was based our system on NI PXI Express RF and FPGA modular instruments in the NI LabVIEW software environment. Using NI tools, we combined real-time functionality with flexibility and a shortened development time.
Rapid Prototyping for Spatial Multiplexing Gain Measurements
To evaluate the performance of methods and technologies under realistic transmission channel conditions, we configured a rapid prototyping and test-bedding system to operate with an FPGA-based real-time channel emulator. The channel emulator mimics the operations of wireless transceivers under real-world operating conditions by applying multipath reflections, attenuation and interference to the radio transmission signals at the receiver.
The hardware of the rapid prototyping platform included an NI PXIe-7966R FlexRIO FPGA module with an NI 5791 RF module installed on an NI PXIe 1062Q chassis. We divided the test bed and rapid prototyping system into two parts:
- The host processing unit was on a PXI controller and ran on a Windows OS. It handled high-complexity floating-point operations and baseband operations such as signal generation, symbol mapping/demapping and modulation/demodulation for different wireless standards. We developed the source code and graphical user interface in LabVIEW with help from the NI Modulation Toolkit and the LabVIEW Digital Filter Design Toolkit.
- The channel processing unit operated in real-time FPGA channel emulator mode built using the LabVIEW FPGA Module. The channel processing unit handled the computationally intensive signal processing calculations such as digital upconversion and digital downconversion. The channel processing unit also handled multipath fading channel emulations specified by the standardisation organisations.
The traditional approach for integrating numerous hardware targets required expertise in several programming languages and tools. Uniquely, LabVIEW allowed us to create both Windows-based and FPGA applications in a single, unified development environment. We created and modified FPGA applications without requiring expertise in highly complex hardware description languages. The seamless integration between NI hardware and LabVIEW was comprehensible and easy to use.
The rapid prototyping system is currently being used to test the spatial multiplexing performance of HSDPA and LTE MIMO downlink transmissions for single-user applications. We reused a 2x2 MIMO spatial multiplexing HSDPA system with a total of 16 channelization signature sequences with a processing gain of 16 across two spatial streams for multiplexing gain evaluations. Each signature sequence carrying up to 6 bits per symbol and a total of 240 by 10³ symbols per second was transmitted, delivering a maximum total rate of 46 Mbit/s. The required transmission bandwidth was 5 MHz. A linear Minimum Mean Square Error (MMSE) equaliser, both with and without interference cancellation receivers, produced the signal-to-noise ratio versus the throughput curves for the HSDPA system.
We also evaluated the throughput performance of a 2x2 LTE MIMO system with 25 resource blocks and 300 frequency bins. Each frequency bin carried up to 6 bits per symbol across a 0.5 ms duration time slot in which each time slot contained 6 symbols transmitted in sequence. This gave the maximum throughput for the LTE system of approximately 43 Mbit/s. The prototype system used a 512-point inverse Fast Fourier Transform (iFFT) to generate the transmitted symbol with the sampling rate of 7.68 MS/s. The required transmission bandwidth was 5 MHz with a 0.25 MHz guard band on each side. The throughput curves were produced for the well-known singular value decomposition method for different signal-to-noise ratios.
Flexible, Efficient Design
Our research is important to identify methods suitable to meet growing data rate expectations of mobile users by using available bandwidth more efficiently. The prototyping system based on PXI Express not only offers the power required for our current research but, because it is software defined, it is flexibility enough that is can evolve to meet our future research requirements.
Mustafa K. Gurcan
Imperial College London
Imperial College London, Exhibition Road, South Kensington
London SW7 2BT
Tel: 02075 946264