SIGNAL PROCESSING FOR COMMUNICATIONS

Transcription

1 Introduction ME SIGNAL PROCESSING FOR COMMUNICATIONS Alle-Jan van der Veen and Geert Leus Delft University of Technology Dept. EEMCS Delft, The Netherlands 1

2 Topics Multiple-antenna processing Radio astronomy Space-time coding for wireless communications Localization based on multiple antenna measurements Time-varying channels Making current broadband systems applicable in mobile environments (DVB, WiFi,...) Underwater communications Ultra wideband communications 2

3 Ultra Wideband Communications time [ns] 3

4 What is UWB? Definition Bandwidth > 20% of the central frequency carrier-less transmission Bandwidth > 500 MHz power spectral density GSM GPS WLAN UWB noise floor frequency 4

5 How does an UWB signal look like? Traditional narrowband signal (e.g., FM modulation) UWB signal (e.g., time-domain pulses with pulse position modulation) 5

6 Why is UWB interesting? (1) Truly high data rates (> 100 Mbps) Comparison of Short Range Wireless Spatial Capacities Spatial Capacity (kbps/m2) UltraWideBand 0.2 mw, 10 m 1000 kbps/m2 IEEE a 200 mw, 50 m 55 kbps/m2 Bluetooth 1 mw, 10 m 30 kbps/m2 IEEE b 50 mw, 100 m 1 kbps/m2 (Source: D. Leeper, Scientific American, May 2002) 6

7 Why is UWB interesting? (2) Overlays existing spectral use; can be globally used; free Devices can be smaller/cheaper (single-chip radio) Narrowband: mixer filter sampler digitizer digital processing LPF A/D DSP fc local oscillator UWB: gating digitizer A/D digital processing DSP timing generator offset adjust 7

8 Why is UWB interesting? (3) Fading resistant, even with only one antenna Narrowband: UWB: 8

9 Why is UWB interesting? (4) High-accuracy positioning possible (cm) T delay τ W r(τ) correlator 0 τ = T 9

10 Research challenges Antennas: good UWB antennas are bulky! Circuits: making well defined narrow pulses, timing of the pulses Signal processing: synchronization, dealing with multipath and multi-user interference 10

11 UWB signal modulation Using time-domain pulses with pulse-position modulation time [ns] Instead of modulating a carrier, the data is transmitted using a sequence of very narrow time-domain pulses. The time-offset of each pulse is controlled by the data and a user-dependent pseudorandom code (like CDMA). Pulse shape and code are designed to yield a smooth spectrum without sharp peaks. 11

12 UWB versus CDMA UWB CDMA symbol: T c T c symbol: T c s k (t) = 0= 1= s k (t) =(±1 ± j)p(t) code: code: c = c = chip symbol period chip symbol period convolution (spreading): convolution: y(t) = y(t) = 12

13 UWB detection scheme Detector for a single user and a channel without delay spread CDMA correlate (MF) correlate with code detect symbol T c T s s k h h -h h UWB time gating correlate (MF) detect symbol T s s k T s code 13

14 UWB detection scheme multipath Detector for multipath channel corr. receivers combiner detect symbol T s h 1 s k h 2,τ τ T s delay estimate code gain estimates h 1,h 2 In the presence of multipath, a bank of correlators with slightly different time offsets allows to combine the energy (RAKE) Optimal combining needs channel estimates (delays and gains) Very short delays give pulse distortion: needs different solution 14

15 UWB detection scheme multi-user Multi-user detector? user 1 code 1 T s s 1 user 2 code 2 code 1 multi-user interference With multiple users and short codes (high rates), the output of the correlator contains multi-user interference (MUI) Obtaining synchronization is harder In CDMA, MUI is cancelled using a multi-user detector; a similar solution may apply here (need to know all codes and channels) 15

16 Signal processing research issues Main signal processing issues are caused by the extremly short pulse duration Direct digital processing of the ns pulses is not practical Research questions Modulation and codes to facilitate detection, synchronization. Problem: don t want to sample at 10 GSps, 16 bits and process this digitally don t want synchronization at picosecond level Channel estimation algorithms, multipath diversity combining Problem: don t want to acquire a rich multipath channel at full resolution (sub-ns) Multi-user separation algorithms Problem: initial acquisition 16

17 Transmit Reference UWB The best receiver is a matched filter, but matching has to be done in analog domain using an unknown pulse (distorted by the channel) A transmit-reference UWB scheme was introduced by Hoctor-Tomlinson Two pulses transmitted: one used as reference, second carries the information Both pulses undergo the same multipath channel Matched filter through correlating first pulse with the second ( dirty template ) No channel estimation needed Easier synchronization and detection 17

18 Transmit Reference UWB τ h(t) τ 2 reference t + W T s detect symbol τ 1 t s k t + W T s τ 2 t Still problems if high data rates or user concentrations are needed (pulses overlap) 18

19 Applications Applications should target the specific advantages of UWB: global use; licence-free high data rates or small/low cost accurate location estimation better penetration through obstacles Currently, likely applications are short-range high data rate ad hoc network ( follow-up of bluetooth) RFID: tagging objects (replacing bar-codes), asset tracking through-wall radar imaging, medical monitoring 19

20 Applications First chipsets/demonstrators/products are available TimeDomain/RadarVision: through-wall motion detection TimeDomain/PulsON (demonstrator): 10 Mbps over 10 m, 150 kbps over 100 m Ubisense and Multispectral: tracking/tracing, 30 cm accuracy over 50 m 20

21 Conclusions Impulse UWB is an interesting new paradigm for signal processing Receivers based on sampling at Nyquist are not practical Transmit-reference systems are practical Additional signal processing required to improve TR-UWB Industry (MBOA) focuses on multi-band solution, which is OFDM on steroids Freescale (Motorola): video streaming, USB 500 Mbps over 2 m, 100 Mbps over 10 m 21