Ultra-Wideband (UWB 2): Physical Layer Options and Receiver Structures Outline of Course Fundamentals 1. Fundamentals of short/medium range wireless communication 1 – – – 2. Fundamentals of short/medium range wireless communication 2 – – – 3. digital transmission systems equivalent baseband model digital modulation and ML-detection fading channels diversity MIMO wireless Fundamentals of short/medium range wireless communication 3 – Multicarrier modulation and OFDM Systems I: OFDM based broadband access 4. WLAN 1: IEEE 802.11g, a 5. WLAN 2: IEEE 802.11n 6. Vehicular Networks Systems II: Wireless short range access technolgies and systems 7. UWB 1: Promises and challenges of Ultra Wideband Systems 8. UWB 2: Physical Layer options 9. Wireless Body Area Network case study: UWB based human motion tracking 10. The IEEE 802.15x family of Wireless Personal Area Networks (WPAN): • • • Bluetooth, ZigBee, UWB Systems III: RF identification (RFID) and sensor networks 12. RFID 1 13. RFID 2 14. RFID 3 15. Summary and Conclusions Communication Technology Laboratory Wireless Communication Group 2 Outline • Physical Layer Options – UWB Impulse Radio – Direct Sequence UWB – UWB Multiband • Receiver Structures – RAKE Receiver – Transmitted Reference Receiver – Energy Detector • Appendix – UWB Multiband – IEEE 802.15.3a Multipath Model 3 Ultra-Wideband Impulse Radio (UWB-IR) 4 UWB-IR: Modulation and MA Options - Modulation schemes: PPM, BPSK (BPAM), PAM, OOK, … - MA schemes: TH-MA, DS-MA, … 5 Peer-to-Peer Scenario: • In the following, we discuss UWB-IR modulation schemes in peer-to-peer communication: One transmitter One receiver • Only one transmitter and one receiver • No interferer No need for a MA scheme. Picture from [Weisenhorn, IZS, 2004] 6 Similarities Among UWB-IR Systems: • Application of very short duration pulses with 2ns Tp 0.13ns , occupying a very large bandwidth of 500MHz Bp 7.5GHz. • In contrast to UWB-MB, the whole band is used in one block. • Each symbol consists of Tp pulses g (t ) Repetition coding g (t ) Time • One pulse per frame ( Tf ) • Very low duty cycle 7 Most Popular UWB-IR Modulations: Binary Pulse Position Modulation (BPPM) Symbol ‘0’ Symbol ‘1’ Binary Pulse Amplitude Modulation (BPAM) (Binary Phase Shift Keying (BPSK)) Symbol ‘-1’ Time Time Symbol ‘1’ s t N f 1 g t kT jT k j 0 s f bk , Ts N f Tf , Tf 2 and bk 0,1. • Modulation of pulse position • Extension to any M-ary PPM possible with: Tf M s t N f 1 b g t kT k j 0 Ts N f Tf and k s jT f bk 1 • Modulation of pulse polarity Pictures from [Giannakis, CEWIT, 2003] 8 Other Types of PAM: Pulse Amplitude Modulation (PAM) On-Off Keying (OOK) Symbol ‘1’ Symbol ‘0’ Time Symbol ‘2’ s t Symbol ‘1’ N f 1 b g t kT jT k j 0 k Ts N f Tf and s Time f s t N f 1 b g t kT k j 0 bk 1,2 k s jT f Ts N f Tf and bk 0,1 • Modulation of amplitude • Extension to any M-ary PAM possible Pictures from [Giannakis, CEWIT, 2003] 9 Example of BPPM: s t N f 1 g t kT jT s k j 0 s(t) t Tf bk 0 s(t) bk 1 bk g(t) • transmitting • transmitting f Ts t Tf Ts Ts N f T f i.e. Ts 4 T f T f M i.e. T f 2 10 Example of BPAM: s t N f 1 b g t kT k j 0 s(t) • transmitting jT f g(t) t Ts s(t) bk 1 s Tf bk 1 • transmitting k t Tf Ts Ts N f Tf i.e. Ts 4 Tf 11 Uncoordinated Multiple Access Scenario: Multiple access (MA) scheme required to reduce interference! [Weisenhorn, IZS, 2004] 12 Direct Sequence Spread Spectrum (DSSS): Conventional Principle DSSS signal Data signal „Chip“ sequence Time domain Frequency domain Data signal convolution * Pseudo-Random sequence Spread data signal 13 Direct Sequence in UWB-IR (1): DS data signal „Chip“ sequence Data signal Spectral Lines due to Rep. Coding 1 1 1 1 Time domain Frequency domain Data signal * Pseudo-Random sequence Randomized data signal 14 Direct Sequence in UWB-IR (2): s ( A) t b g t kT jT c k s ( B) N f 1 (A) k f ( A) j Note: can also be combined with PPM N f 1 t bk g t kTs jTf c(jB) B k c c j 0 s j 0 ( A) j User A specific binary pseudo-random sequence (PN) of length Nf (B) j User B specific binary pseudo-random sequence (PN) of length Nf User A User B Tf ... ... Time Pictures from [Giannakis, CEWIT, 2003] 15 DS-UWB Compared to DSSS: • DS in UWB-IR is very similar to DSSS in conventional systems: – Data bit is spread over multiple consecutive pulses. – Pseudo-random code is used to separate users (MA). – Spectrum is smoothed very efficiently. but: – In UWB-IR-DS the code rate equals the pulse rate. Spectrum is not significantly spread by the DS. 16 Time-Hopping Multiple Access: s ( A) t s( B) t ( A) ( A) b g t kT jT c k s f j Tc k j 0 N f 1 b g t kT jT k c c N f 1 ( B) k j 0 s f Note: can also be combined with PPM c(j B )Tc ( A) j User A specific Tf / Tc -ary pseudo-random sequence (PN) of length N f (B) j User B specific Tf / Tc -ary pseudo-random sequence (PN) of length User A Nf User B Tf 5 Tc ... ... Time Pictures from [Giannakis, CEWIT, 2003] 17 Time-Hopping Properties: • Data bit is spread over multiple consecutive pulses. • Pseudo-random code is used to separate users (MA). • User separation also possible in non-coherent receivers such as the energy detector. • Spectral smoothening not as effective as with DS. 18 UWB Receivers Outline Matched filter Receiver structures Rake Transmitted reference Energy detector 20 Communication Technology Laboratory – Wireless Communications Group 20 Introduction Pulse based UWB Transmitter and receiver for UWB are said to be very simple due to no need of Mixers, RF Oscillators and PLLs For transmitters this assumption holds probably But receivers are probably more complex as often assumed since energy has to be captured from all multipaths 21 Communication Technology Laboratory – Wireless Communications Group 21 Matched Filter Transmission of a single pulse s(t) with duration T n(t) is a white Gaussian noise process of zero mean and power spectral density N0/2 Receiver consists of a linear time-invariant filter g(t) and a sampler n(t ) t= T single pulse s (t ) g (t ) y (t ) y (T ) receiver The matched filter g(t) = s(-t+T) is a time reversed and delayed version of the input signal s(t). It maximizes the SNR at the sampling instant T. 22 Communication Technology Laboratory – Wireless Communications Group 22 UWB System with Multipath Channel TX RX s (t ) 1 t TP r (t ) = h(t )* s (t )+ n(t ) h(t ) 1 - 1 - 1 (noiseless) 23 Communication Technology Laboratory – Wireless Communications Group 23 Matched Filter for Multipath Channel I Optimum receiver: correlator or matched filter g (t )= s (t )* h(t ) n (t ) s (t ) h (t ) r (t ) N g (t )= s (t )* h (t )= s (t )* å hi ×d(t - t i )= i= 1 r t y N å hi ×s (t - t i ) i= 1 y1 t s t 1 h1 + y yN t s t N hN 24 Communication Technology Laboratory – Wireless Communications Group 24 Matched Filter for Multipath Channel II r t y1 t s t 1 h1 + y yN t s t N r t hN y1 s t 1 h1 s t N + y yN correlator in each branch „RAKE“ hN 25 Communication Technology Laboratory – Wireless Communications Group 25 ARAKE (All RAKE) Optimum receiver with unlimited resources Combines all N resolved multipath components Number of resolvable components N increases with bandwidth => large number of RAKE fingers 0.6 N yA hk r t s t k dt 0.4 k 1 t 0.2 0 -0.2 -0.4 -0.6 -0.8 0 500 1000 1500 2000 Communication Technology Laboratory – Wireless Communications Group 26 26 SRAKE (selective RAKE) Also referred as selection combining (SC) Only subset of resolved multipath components is processed Selects the L strongest paths Better performance than a single path receiver Requires the knowledge of the instantaneous values of all multipath components 0.6 0.4 yS hk r t s t k dt 0.2 kL 0 t L indices of L strongest paths -0.2 -0.4 -0.6 L=6 -0.8 0 500 1000 1500 27 2000 Communication Technology Laboratory – Wireless Communications Group 27 Impact on the design of WBAN´s Number of RAKE fingers using a SRAKE Antennas placed on the front side of the body in 15cm steps Collecting 75% of the whole energy (front side measurements) 2 fingers @ 15cm 20 fingers @ 90cm 1 0.8 P/Ptotal Short distance multihop increases the energy that can be captured with a simple RAKE 0.6 15cm 30cm 45cm 60cm 90cm 0.4 0.2 0 5 10 15 Number of strongest paths 20 28 Communication Technology Laboratory – Wireless Communications Group 28 PRAKE (Partial RAKE) Sometimes also referred as nonselective combining (NSC) Collects the energy from the M first multipath components These multipath components must not be the best, e.g. in NLOS environment Compared to SRAKE no selection mechanism is required Needs only to find the first M multipath components => complexity reduction 0.6 0.4 0.2 M yP hk r t s t k dt 0 k 1 t -0.2 -0.4 -0.6 -0.8 0 M=6 29 500 1000 1500 2000 Communication Technology Laboratory – Wireless Communications Group 29 Selective nonselective Combining (SC-NSC) Only the strongest path is tracked The K-1 paths following the strongest path are chosen for the remaining path delays SC-NSC is better suited for NLOS channels (where the direct path with the shortest delay, i.e. the first path, is attenuated) than PRAKE/NSC since the strongest path can be tracked 0.6 0.4 0.2 yP 0 k0 K 1 k k0 hk r t s t k dt t k0 arg max hk -0.2 k -0.4 -0.6 K=6 -0.8 0 500 1000 1500 2000 Communication Technology Laboratory – Wireless Communications Group 30 30 Conclusions on Rake Receivers ARAKE is an optimum receiver Realization of a matched filter High complexity Complexity reduction by using only a fraction of all paths Performance degradation Channel estimation necessary Amplitudes and delays have to be known Simpler receiver structures without channel estimation would be desirable 31 Communication Technology Laboratory – Wireless Communications Group 31 Transmitted Reference Receiver 32 Communication Technology Laboratory – Wireless Communications Group 32 Principles of Transmitted Reference Systems 2 pulses (=1 doublet) are transmitted for one symbol 1st pulse is the reference pulse, which is used as template 2nd pulse is the data pulse Implicit channel estimation since both pulses pass the same channel Channel has to be invariant over 1 doublet only BPF required for noise reduction Noisy template for correlation Information rate usually drops by 50 % since half of the pulses are used as reference 33 Communication Technology Laboratory – Wireless Communications Group 33 TR PAM I Information in the amplitude of the data pulse Reference pulses a 1 TP a0 t T TP T Data pulses 34 Communication Technology Laboratory – Wireless Communications Group 34 TR PAM II TX energy higher since two pulses are needed for 1 bit Correlation of reference and data pulse r (t ) BPF T Tp T T Tp d r t r t T dt T T Performance depends on the time of integration Tp Performance degradation if inter-pulse interference exists 35 Communication Technology Laboratory – Wireless Communications Group 35 Integration Duration I BER performance depends on the integration duration If integration duration is too short, not enough energy can be captured If integration duration is too long, the CIR is decayed so much that the noise term gets dominant Channel models from IEEE 802.15.3a LOS (Line of Sight) NLOS (Non-Line of Sight) LOS NLOS 1.2 0.2 0.15 0.9 0.1 0.6 0.05 0.3 0 -0.05 0 -0.1 -0.3 -0.6 0 -0.15 20 40 60 [ns] 80 100 -0.2 0 20 40 Communication Technology Laboratory – Wireless Communications Group 60 [ns] 80 100 36 36 Integration Duration II Body area network measurements around the torso In general, shorter integration duration for LOS links than for NLOS 37 Communication Technology Laboratory – Wireless Communications Group 37 Energy Detector 38 Communication Technology Laboratory – Wireless Communications Group 38 Energy Detector I Energy detector (ED) collects energy from multipaths Integrates the energy of the receive signal Non-coherent receiver structure No antipodal signaling possible, e.g. BPSK Usually used with pulse position modulation (PPM) No explicit channel estimation is necessary Begin and end of the integration interval has to be known 39 Communication Technology Laboratory – Wireless Communications Group 39 Energy Detector II Pulse Position Modulation (2 PPM) a0 kT a 1 k 1T TI TI TP r (t ) BPF t 2 TP kT T p kT kT TI Tp kT TI kT TI Tp dk r t dt 2 kT TI kT T p r 2 t dt kT 1, if d k 0 aˆk 0, otherwise Integration of noise on the position where no pulse is located 40 Communication Technology Laboratory – Wireless Communications Group 40 Performance Comparison Real measured channels around the human body (15cm distance quasi LOS) 0 10 TR MF ED -1 BER 10 -2 10 -3 10 -4 10 0 2 4 8 6 E /N b 10 12 14 0 Performance of TR and ED similar but about 6 dB worse than MF 41 Communication Technology Laboratory – Wireless Communications Group 41 Conclusions TR and ED are much simpler than an ARAKE No explicit channel estimation necessary Position of the receive signal in time domain has to be known accurately Performance is worse than ARAKE Performance strongly depends on integration duration Integration duration is for LOS channels usually shorter than for NLOS channels 42 Communication Technology Laboratory – Wireless Communications Group 42 Appendices UWB Multiband (Certified Wireless USB) IEEE 802.15.3a Multipath Model 43 Ultra-Wideband Multi-Band (UWB-MB) 44 UWB-Multiband OFDM • • • • Spectrum is divided into sub-bands Serial transmission over the sub-bands Application of TF codes for piconet separation Strongly promoted by industry (Wireless USB, WiMedia, MBOA) 45 ECMA-368 (MB-OFDM Standard): Basic Idea • Split overall spectrum into 14 bands of 524MHz bandwidth • Serial transmission of OFDM symbols over the bands • OFDM symbol: – 128 point FFT/IFFT independent of data rate – Modulation: QPSK or DCM • Information is coded across several bands (TF codes) to achieve frequency diversity and piconet separation. • Zero-padded suffix: – Robustness against multi-path – Time to switch band 46 ECMA-368: Bandplan • Overall band of 7.5GHz is split into14 bands: – Bandwidth: 524MHz – Separation: 524MHz • Bands are grouped into 5 band groups • Several TF codes for each band group several piconets • Band groups are managed by FDMA: – Better SOP performance [ECMA-386, 2006] 47 Code Map of Band Group 1: Fixed Frequency Interleaved Channels (FFI) Time Frequency Interleaved Channels (TFI) [ECMA-386, 2006] 48 EMCA-368: Rate Independent Parameters • 8 OFDM tones are set to zero. [ECMA-386, 2006] 49 ECMA-368: Rate Dependent Paramters [ECMA-386, 2006] 50 QPSK versus DCM: • Quadrature Phase Shift Keying (QPSK) • Dual Carrier Modulation (DCM): – Two different 16-QAM mappings – Two different carriers – Frequency diversity Subcarrier 1 4 bits Subcarrier 50 [ECMA-386, 2006] 51 Should One Go Multi-Band? • • Pros – Flexible band selection • • – – – – Easy to fit to spectral masks NBI mitigation Power efficient Implementation by COTS Suitable for IC integration Suited and strongly promoted for HDR systems (e.g. Wireless USB) Cons – – – – Not low complexity Not low power High rate sampling Small advantage over other systems, e.g. 802.11n [Giannakis, CEWIT, 2003] 52 IEEE 802.15.3a Multipath Model 53 IEEE 802.15.3a Multipath Model (1) The proposed model uses the following definitions: Tl = the arrival time of the first path of the l-th cluster k,l = the delay of the k-the path within the l-th cluster relative to the first path arrival time Tl = cluster arrival rate = ray arrival rate, i.e., the arrival rate of path within each cluster . t 54 IEEE 802.15.3a Multipath Model (2) • • • • Log-normal (rather than Rayleigh) distribution for the multipath gain magnitude Independent fading assumed for each cluster as well as each ray within the cluster Real valued passband model Target channels: CM 1 (LOS 0-4m), CM 2 (NLOS 0-4m), CM 3 NLOS 4-10m, CM 4 (Extreme NLOS) Discrete time impulse responses: L K hi (t ) X i ki ,l (t Tl i ki ,l ) l 0 k 0 Xi ki ,l Shadowing coefficient Tl i Delay of the lth cluster ki ,l Delay of the kth multipath component relative to the lth cluster arrival time Multipath gain coefficients Tl i 55 IEEE 802.15.3a Multipath Model (3) • Shadowing coefficient • Multipath gain coefficients – Multipath amplitude sign – Ray power Exponential decay 20log10( X i ) Normal(0, x2 ) k ,l pk ,l l k ,l ( k , l n1 n2 ) / 20 l k ,l 10 n1 Normal(0, 12 ) Cluster fading n2 Normal(0, 22 )Ray fading pk ,l 1 (equiprobable) 20log10(l k ,l ) Normal( k ,l , 12 22 ) k ,l 10ln( 0 ) 10Tl / 10 k ,l / ln(10) ( 12 22 ) ln(10) 20 2 / E l k ,l 0 e Tl / e k ,l • Poisson cluster and ray arrival p Tl Tl 1 exp Tl Tl 1 , 0 mean energy of the first path of the first cluster l 0 p k ,l ( k 1),l exp k ,l ( k 1),l , k 0 56 IEEE 802.15.3a Multipath Model (4) Target Channel Characteristics Mean excess delay (nsec) ( m ) CM 1 CM 2 CM 3 5.05 10.38 14.18 RMS delay (nsec) ( rms ) 5.28 8.03 14.28 CM 4 25 NP10dB NP (85%) 24 36.1 35 61.54 Model Parameters (1/nsec) (1/nsec) 1 (dB) 0.0233 2.5 7.1 4.3 3.3941 0.4 0.5 5.5 6.7 3.3941 0.0667 2.1 14.00 7.9 3.3941 0.0667 2.1 24.00 12 3.3941 3.3941 3.3941 3.3941 3.3941 3 3 3 3 Model Characteristics Mean excess delay (nsec) ( m ) 5.0 9.9 15.9 30.1 RMS delay (nsec) ( rms ) 5 8 15 25 NP10dB NP (85%) Channel energy mean (dB) Channel energy std (dB) 12.5 20.8 -0.4 2.9 15.3 33.9 -0.5 3.1 24.9 64.7 0.0 3.1 41.2 123.3 0.3 2.7 2 (dB) x (dB) 57