Receivers, Antennas, and Signals Professor David H. Staelin 6.661 Fall 2001 Slide 1 A1 Subject Content A Human Processor Radio Optical, Infrared Acoustic, other C Human Transducer Electromagnetic Environment B Processor Transducer Communications: A → C (radio, optical) Active Sensing: A → C (radar, lidar, sonar) Passive Sensing: B → C (systems and devices: environmental, medical, industrial, consumer, and radio astronomy) 6.661 Fall 2001 Slide 2 A2 Subject Offers • Physical concepts • Mathematical methods, system analysis and design • Applications examples • Motivation and integration of prior learning 6.661 Fall 2001 Slide 3 A3 Subject Outline • Review of signals and probability • Noise in detectors and systems; physics of detectors • Receivers and spectrometers; radio, optical, infrared • Radiation, propagation, and antennas • Signal modulation, coding, processing and detection • Communications, radar, radio astronomy, and remote sensing • Parameter estimation 6.661 Fall 2001 Slide 4 A4 Review of Signals Signal Types to be Reviewed: • Pulses (finite energy) • Periodic signals (finite energy per period) • Random signals (finite power, infinite energy) 6.661 Fall 2001 Slide 5 A5 Pulses v(t) ∞ −∞ Have Finite Energy : ∫ 2 v( t ) dt < ∞ Define Fourier Transform: ∆ V( f ) = ∆ v( t ) = ∞ ∫ v( t )e − j2 πft N ω dt [volts/Hz = volt sec ] −∞ ∞ + j2πft V ( f ) e df [volts ] ∫ −∞ Define notation " ↔" for Fourier Transform : e.g. v(t) ↔ V( f ) Dimensions must only be self consistent; e.g. v(t) can be dimensionl ess, volts, meters, newtons, etc. 6.661 Fall 2001 Slide 6 B1 Energy Spectral Density S(f) ∆ V( f ) = S( f ) ∆ V(f) 2 ∞ ∫ v( t )e − j2πft ω −∞ S(f) can have dimensions of : sec2 if t is time and v is dimensionless m2 if t is distance and v is dimensionless (volts/Hz)2 if t is time and v(t) is volts ( ) where (v/Hz)2 = v 2 /sec (Hz ) = (v / sec )2 Joules/Hz if S(f) is dissipated in a 1-ohm resistor 2 by v(t) volts, where Joules = volts • sec/ohm Et cetera 6.661 Fall 2001 Slide 7 B2 Autocorrelation Function ∞ v( t )v ∗ ( t − τ)dt −∞ R( τ) ∆ ∫ [v 2 sec ]or [J], etc. Claim: R( τ) ↔ S( f ) ∞ ∞ − j2πfτ S( f ) ? ∫ R( τ)e dτ = ∫ −∞ −∞ = {∫ } • {∫ ∞ v(t)e− j2πft dt −∞ { } ∞ ∗ − j2f τ v ( t ) v ( t ) dt e dτ − τ ∫−∞ ∆ t′ } t − t′− dt′ ∞ v ∗ (t′)e+ j2πft′ dt′ −∞ Reverses sign of - dt′ for t = constant 2 S( f ) = V( f ) • V ∗ ( f ) = V( f ) Q.E.D. Therefore: ∞ R( τ) = S( f )e + j2πfτ df −∞ ∞ 2 ∞ R(0) = v ( t ) dt = S( f ) df Parseval' s Theorem −∞ −∞ ∫ ∫ 6.661 Fall 2001 Slide 8 ∫ B3 Compact Transform Notation [v ] v( t ) ↔ V ( f ) ↓ ↓ ↓ 2 R( τ) ↔ V( f ) ∆ S( f ) ↔ [v / Hz] ↓ [v 2 sec ] ↔ [v / Hz]2 [Joules] ↔ [J / Hz] 6.661 Fall 2001 Slide 9 If power is dissipated in a 1-ohm resistor B4 Define “Unit Impulse” ε uo ( t ) ∆ δ( t ) where lim uo ( t ) dt = 1, uo ( t ) = 0 for t > 0 − ε ε →0 ∫ t un −1( t ) ∆ u ( t ) dt −∞ n ∫ Ramp Step uo ( t ) Impulse t 0 6.661 Fall 2001 Slide 10 u−1( t ) 0 t u− 2 ( t ) 0 Slope = +1 t B5 Define “Convolution” ∞ a( t ) b(t − τ) dt −∞ a( t ) ∗ b( t ) ∆ ∫ a(t) = c( τ) a( t ) ∗ b( t ) = c( τ) b(t) * t 0 6.661 Fall 2001 Slide 11 τ t 0 0 B6 Useful Transformation Pairs for Pulses ∞ a( t ) e − j2πft dt −∞ a( t ) ↔ A (f ) A( f )∆ ∫ uo ( t ) = δ( t ) ↔ 1 1 ↔ uo (f) Have ∞ energy (treated as special pulses) a′( t ) ↔ jωA (f ) un ( t ) ↔ ( jω)n a( t ) e jωo t ↔ A (f − fo ) a( t − t o ) ↔ A ( f ) e − j ωt o u−1( t ) e − αt ↔ 1/ ( jω + α ) 6.661 Fall 2001 Slide 12 ω∆2πf ∞ A( f )e j2πft df −∞ a( t ) = ∫ ωo ≡ 2πfo C1 Transforms: Even/Odd Functions ae ( t ) ↔ R e {A (f )} where: so: 0 6.661 Fall 2001 Slide 13 EVEN ao (t) ∆ [a( t ) − a(− t )] / 2 = ao (− t ) ODD a(t) = ae (t) + ao (t) a( t ) e.g. ao ( t ) ae ( t ) ∆ [a( t ) + a(− t )] / 2 = ae (− t ) t = ae ( t ) 0 t + ao ( t ) 0 t ↔ j Im{A( f )} C2 Transforms: Operators and Gaussians a1( t ) • a2 ( t ) ↔ A1( f ) ∗ A 2 ( f ) a1( t ) ∗ a2 ( t ) ↔ A1( f ) • A 2 ( f ) v(t) σ 0 6.661 Fall 2001 Slide 14 A −( t / σ )2 / 2 −( σω)2 / 2 v( t ) ∆ e ↔ Ae t σ 2π All Gaussians ↓ ↓ A 2 −( τ / σ 2 )2 2 2 −( σω)2 e ↔ A e 2σ π C3 Linear Systems x(t) h(t) y(t) Characterized by: h(t) = " Impulse Response, " where ∞ x( τ)h( t − τ)dτ ← " superposition integral" −∞ y(t) ∆ x(t) ∗ h( t ) ∆ ∫ Test : If x(t) = δ(t), then y(t) = h(t) If h(t) = δ(t), then y(t) = x(t) Note: A ∗ (B + C) = (A ∗ B) + (A ∗ C) A ∗B =B∗A A ∗ (B ∗ C) = (A ∗ B) ∗ C 6.661 Fall 2001 Slide 15 “Distributive” “Commutative” “Associative” C4 Periodic Signals ∞ 2 T 2 Although ∫ v ( t )dt = ∞, ∫ v ( t )dt < ∞ where Period ∆ T o −∞ v(t) –T m =1 m=0 T 0 Fourier Series: 1 T/2 v( t )e − jm( 2π / T )t dt Vm ∆ ∫ T −T / 2 ω = 2πf o v(t) = Infinite energy, finite power 2T 3T where m = 0, ± 1, ± 2,... o ∞ jm 2πfo t ∆ 1/T ) ( V e f ∑ m o m = −∞ 6.661 Fall 2001 Slide 16 E1 Autocorrelation, Power Spectrum Autocorrelation Function: ∞ 1 T/2 2 ∗ v( t )v (t − τ )dt = ∑ V m e jm2πfo τ R( τ) ∆ ∫ T −T / 2 m = −∞ Power Spectrum: Φm ∆ V m 6.661 Fall 2001 Slide 17 2 1 T/2 = ∫ R( τ)e − jm 2πfo τ dt T −T / 2 E2 Compact Notation ↔ Vm ↓ 2 ↔ V m ∆ Φm ↔ Φ(f) v( t ) ↓ R( τ) Typical dimensions: [volts] ↓ [ volts2 ] [volts] ↔ ↓ ↔ [ volts2 ] In 1-ohm resistor: [ watts] 6.661 Fall 2001 Slide 18 ↔ [W ] E3 Transforms of Impulse Trains a − 2T a/T v( t ) 0 2T t 0 v( t ) (a / T ) 2T τ −3/T Vm 0 R( τ) ah area = a2 / T e.g. 1/ h τ Area = a is impulse value 6.661 Fall 2001 Slide 19 0 2 a 2 / T R( τ) − 2T −3/T Vm 2/T 2 f → Φ( f ) 2/T f a2h2 ⎛ 1 ⎞ •⎜ ⎟ T ⎝h⎠ = R(0) = a2h / T τ − 1 / h 0 1/ h Let h → ∞ so area a2 / T becomes impulse value E4 Receiver Noise Processes Receivers, Antennas, and Signals Professor David H. Staelin 6.661 Fall 2001 Slide 20 Random Signals Random signals generally have finite power, infinite energy, and are unpredictable Example: [WH ] z −1 Φ( f ) 0 f MAX f Since we have infinite information for infinite time and a finite frequency band, then “V(f)” is not an analytic function and our approach must be different. New definitions are required. 6.661 Fall 2001 Slide 21 G1 Expected Value of x Finite or infinite ensemble of xi (t) ∞ E[x(t )] ∆ ∑ xi (t) p{xi (t )} → ∫ x p(x ) dx −∞ i t1 xi ε " ensemble" t2 xi ( t ) ∑ p( x i ) ∆ 1 t i p[xi (t )] t x j (t) A “random signal” is drawn from some ensemble [ ] Autocorrelation Function : φv (t1, t 2 ) ∆ E v (t1)v ∗ (t 2 ) 6.661 Fall 2001 Slide 22 G2 Stationarity v(t) is “wide-sense stationary” if: φ v (t1, t 2 ) = φ v (t1 + ∆, t 2 + ∆ ) = φ(τ ) where τ ∆ t 2 − t1 for all t1, t 2 , ∆ v(t) is “strict-sense stationary” if: E[g{v (t1), v (t 2 ),..., v (t n )}] = E[g{v (t1 + ∆ ), v (t 2 + ∆ ),..., v (tn + ∆ )}] for any function g v(t) is “Ergodic” if: v(t) is wide-sense stationary and 1 T ∗ v(t) v φv ( τ ) = lim ( t − τ ) dt, ∫ T − T →∞ 2T i.e., ensemble average = time average transitions Otherwise v(t) is occur only “Non-stationary” – e.g.: 0 1 0 1 1 0 0 1 t at clock ticks (time-origin sensitive) 0 1 2 3 4 5 6 6.661 Fall 2001 Slide 23 G3 Transform Diagram: Random Signals v(t) ↔ (?) ↔ ↓ Φ(f) ↓ φv(τ) Typical Sets of Units [V] ↔ ↓ [ V2 ] ↔ (?) [V] ↓ ↓ [ V2/Hz ] [W] Power to 1-ohm resistor 6.661 Fall 2001 Slide 24 ↔ (?) ↓ ↔ [W/Hz] Power spectral density G4 Power Spectral Density 2⎤ ⎡1 T − j2πft Φ( f ) = lim E ⎢ v( t )e dt ⎥ ∫ T − T → ∞ ⎣ 2T ⎦ Why use E[ ] if v(t) is ergodic? Because lim σ2T ( f ) ≠ 0! where σ2T ( f ) ∆ E[Φ T ( f ) − Φ( f )]2 Spectral resolution increases with T, becoming infinite as T → ∞ e.g. Infinite information in finite t bandwidth unless ensemble is averaged Power Spectral Density Computation: For a single ergodic waveform, take ensemble average over successive intervals of width 2T. Use T adequate to yield desired or meaningful spectral resolution. 6.661 Fall 2001 Slide 25 G5