DAAD Workshop on Embedded System Design Skopje, October 2009 for Maximum Reliability and Performance Zoran Stamenković IHP, Frankfurt (Oder) Germany
Problem Definition MIMO techniques in wireless networks Enable an increased capacity Enable a large number of supported users Require more complex architectures compared to SISO Increase the power consumption, system size and costs Signal processing is performed in the digital baseband Transceiver needs as many parallel operating paths as antennas are used for communication
MIMAX Objectives Development of an innovative low power and ultra compact MIMO transceiver which is reconfigurable with respect to the application demands and channel properties
I MIMAX Approach Shifting the spatial signal processing to the analogue RF domain reduces the number of receiving and transmitting paths to a single one Hardware overhead consists only of additional antennas and a modified RF front-end compared to a SISO system I D A Vector modulator PA Band-pass LNA + VM Low-pass A D EPP Base band low-pass sin LO Vector modulator PA Band-pass LNA + VM LNA + VM sin LO cos Base band EPP Q D A low-pass cos Vector modulator Vector modulator PA PA Band-pass Band-pass LNA + VM Low-pass A D Q EPP RF control unit SPI + control signals SPI + control signals RF control unit EPP
MIMAX Features Inside IEEE802.11a Baseband
MIMAX Baseband Processor MAC User 1 w1 CQI PER User 2 w2 PER CQI... IHP BASEBAND PROCESSOR EPP port UC BASEBAND MIMAX MODULES MIMAX FRAME I MIMAX FRAME II Baseband Processor Freq. Offset Estimation?f IFFT/FFT MIMO Channel estimation LTS SF SISO Channel Estimation (LS) Data Decoding User k wk PER CQI BIDIRECTIONAL REGISTERS w Weight Correction QT QR w ~ G MIMAX RF Weights MIMAX CONTROL EPP port LTS: Long training symbols SF: Signal Field (MIMAX frame II) CQI: Channel Quality Indicator PER: Packet Error Rate RF Control Unit
MIMAX Baseband Board
MIMO Channel Estimation One OFDM training symbol sent n T x n R times with different beamforming weights H h = w H w k = 1, K, N ; k R k T c
Optimal Beamforming Weights Starting point: Algorithm: Estimated n T x n R SISO channels at each of N c subcarriers Calculate the dominant eigenvectors (power method) to find the MaxSNR beamforming weights (suboptimal) LMS algorithm to find the optimal MinMSE weights Finally, the frequency offset is corrected and the optimal weights are sent to MAC processor RF Control Unit compensates RF impairments
Beamforming Weights Calculation
Beamforming Block Step A: Create 52 column vectors x k where the i-th element of x k is the sample of the k-th subcarrier for the i-th equivalent SISO channel The size of x k is n T x n R (16 if 4 antenna MIMAX terminals are used)
MIMAX Performance Data rate of 12 Mbps Data rate of 54 Mbps
MAC Processor Architecture EJTAG Interface GPP (ASIC) MIPS 4KEp Processor Core HW Accelerator (FPGA) Transmission FCS generation Antenna weights CardBus Interface CardBus Interface Cache Memory Controller Asynchronous Memory and IO Bus Channel state Backoff procedure NAV control Reception FCS check Address filter ACK generation MIPP Interface to PHY External RAM External Flash Others 16 Timers 64 bit system time
MAC-PHY Interface
Slave Side of MIPP
MAC Layer Testing RAM/ Flash Host Interface MAC Processor Baseband Processor PHY link emulator optionally replaces PHY layer and radio interface with digital link (FPGA) PHY Link Emulator Host Interface MAC Processor Baseband Processor RAM/ Flash WLAN Modem 1 AD/DA Converters AD/DA Converters WLAN Modem 2 Analog Front End Radio Air Interface Analog Front End
MIMAX MAC Board
Conclusion and Outlook MIMAX is an RF-MIMO WLAN IEEE802.11a compliant Modified analogue RF front-end Great reduction of digital hardware Modified digital baseband processor New baseband board Minor modifications of MAC processor Modified MAC board Simulation results show a significant optimization gain
Links for More Information http://www.ict-mimax.eu http://www.ihp-ffo.de/~stamenko