A FLEXIBLE MIMO TESTBED WITH REMOTE ACCESS Christian Mehlf¨ uhrer, Stefan Geirhofer, Sebastian Caban, and Markus Rupp Institute of Communications and RF Engineering Vienna University of Technology Gusshausstr. 25/389, 1040 Wien, Austria Email: {chmehl, sgeirhof, scaban, mrupp}@nt.tuwien.ac.at Web: http://www.nt.tuwien.ac.at/rapid prototyping ABSTRACT While wireless systems with multiple transmit and receive antennas have been thoroughly investigated over the past ten years, experimental work to corroborate theoretical results is lagging behind. A major reason for this situation is given by the fact that strong theoretical researchers often have no expertise in implementation issues. Also, building an experi- mental system is very time-consuming and still it is typically focussed on one transmission mode only. By the time the experimental setup is ready, hundreds of new ideas already exist that cannot be handled by the just finished setup. We thus propose a flexible MIMO testbed that allows to connect directly onto MATLAB code and thus shortens experimental times considerably. A few examples for UMTS HSDPA and for Space-Time Codes strengthen our approach. 1. INTRODUCTION With the theoretical understanding of the nature of multi- ple antenna systems in a scattering environment by Foschini, Gans [1] and Telatar [2], an enormous potential for high spectral efficiency was found. However, now, more than 10 years later, still Multiple-Input, Multiple-Output (MIMO) products are missing. It is thus not simply adding a few more transmitters and receivers in order to take advantage of this potential. Many real-time experiments including physical, wireless channels are needed to understand the true advan- tages. Since our knowledge of wireless channels in partic- ular for MIMO transmissions is rather poor, simulations in- cluding mathematical channel models cannot give us final answers. Simulation methods based on MATLAB and C- programming often simplify the transmission scenarios and the true problems become hidden. For example, the im- pact of nonlinear power amplifiers which is very typical for WCDMA as well as OFDM with their very high Crest factors is typically neglected in simulations. Also quantization ef- fects of the AD and DA converters are not included. Simula- tions are typically based on floating-point (double) precision while products have to work on low cost 16bit fixed-point processors or on dedicated HW with even lower precision. It is thus of utmost importance to come up with real- time experiments including physical wireless channels and equipment much like the one used later in products in or- der to identify potential show-stoppers as early as possible and de-risk emerging new technologies. In this paper we This work has been funded by the Christian Doppler Laboratory for Design Methodology of Signal Processing Algorithms. present such a flexible testbed suitable for real-time experi- ments. In Section 2 we give a short overview of the hardware involved. Section 3 then wraps up the design methodology utilizing a MATLAB interface to the hardware that allows for quickly adapting MATLAB simulation code unto the RF front end. And finally Section 4 presents a few measurement re- sults obtained by this testbed, concerning the UMTS HSDPA mode and Space-Time Codes for MISO transmissions. 2. MIMO PLATFORM The main idea of our MIMO testbed (Fig. 1) is the so-called MATLAB I NTERFACE, a powerful, easy to use, and still flexi- ble interface to the MIMO testbed hardware. Using this inter- face, neither hardware programming skills nor a lot of time is needed to transmit and receive complex baseband data sam- ples directly out of MATLAB via a radio frequency (RF) air interface from any PC in the local area network (LAN). This allows for a great variety of real-time experiments with min- imum effort. • Utilizing MATLAB and optional DSP+FPGA boards, the testbed user creates complex digital baseband data sam- ples (2×14bit) for up to four transmit antennas on his own PC. These data samples and a set of transmit options (e.g. the desired IF frequency) are then transferred to the transmit PC via the MATLAB I NTERFACE. Note that the data transfer via the MATLAB I NTERFACE does not require real-time ability since large amounts of transmit data can be stored on the internal harddisk of the trans- mit PC. The stored data samples can be transmitted in real-time at a certain instant. • In the transmit PC, the four streams of digital baseband data samples are buffered, interpolated, digitally upcon- verted to the desired intermediate frequency (IF), and at last converted to the analog domain by four 14bit 200MHz DACs. • IF filtering, analog upconversion to the 2.45GHz ISM band, and power amplification are carried out in the next step. • The channel may either be – a physical cannel operated by eight λ / 4 -monopole an- tennas at 2.45 GHz – or channel emulators (Spirent TAS4500 FLEX), al- lowing for repeatable experiments with a wide range of definable channel parameters. • Low noise amplification, RF preselection, analog RF to IF downconversion, and IF filtering are carried out in the next step. Copyright 2005 EURASIP. Accepted for publication in the proceedings EUSIPCO 2005, Antalya, 2005