Long-baseline intensity interferometry: data transmission and correlation Erez Ribak Department of Physics Technion Israel Institute of Technology Workshop on Hanbury Brown & Twiss interferometry Nice, 13 May 2014
Participants Graduate (MSc) students Coral Moreno Yaron Shulamy Faculty S. G. Lipson Pini Gurfil Technical support Dr Dan Spektor Hovik Agalarian Communications lab, EE dept. Distributed Space Systems Lab 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 2
Contents Why intensity interferometry Why revival New possibilities Previous experiments Correlation at a distance Transmitting the intensities Future directions Summary 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 3
Amplitude interferometry pioneers Stephan, inside the aperture, 1870 0.8 m Michelson, outside the aperture, 1920 6 m (HBT, 1965-72) Labeyrie, outside the telescope, 1975 13 m 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 4
Why intensity interferometry Science developments Quantum physics Astrophysical knowledge Technology developments Fast detectors Radio technology Light collectors Fast electronics Correlators Coaxes Human intervention not necessary (cf. Michelson s eye) 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 5
Why revival Science developments Astrophysical knowledge Photonics Other fields of physics: particles, biophysics Technology developments Detectors: faster and redder Electronics: faster and digital Optics: fibres Space 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 6
New possibilities Science developments Astrophysical knowledge photonics Technology developments Detectors: faster and redder Electronics: faster and digital Optics: fibres Space New opportunities Ĉerenkov arrays Antarctica 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 7
Previous experiments (Technion) I Theoretical studies Ofir and Ribak: higher-order correlations b ν = electronic bandwidth α = quantum efficiency Σ = system efficiency m = optical channels b ν 1GHz, α 0.8, Σ 0.8, m=1 NSII performance: 0 m star, 1 hr, SNR=27 b ν = 100MHz, α = 0.2, Σ = 0.2, m = 1 4.99 4.91 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 8 dishes
Previous experiments (Technion) II Theoretical studies Ofir and Ribak: higher-order correlations Klein, Guellman, Lipson: space source k x All wavelengths possible Formation flight Satellites orbits Keeping constant baselines Optimal fuel consumption Control laws k x k y k y source k x source k y 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 9
Previous experiments (Technion) III Theoretical studies Ofir and Ribak: higher-order correlations Klein, Guellman, Lipson: space intensity interferometry Laboratory studies Spektor, Lipson, Ribak Blue LEDs metres from Fresnel lenses Fast photomultipliers, lock-in amp. 2000 s correlation electronics: AD8302 LED spectrum 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 10 no signal with signal
Correlation at a distance Asher Space Research Institute Physics and Aeronautics Distributed Space Systems Laboratory ERC support Air table, vehicle location Dark room 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 11
Scheme LED Pinhole Light Source Satellite 3 Telescope Detector Transmitter Satellite 2 Telescope Detector Transmitter Space Satellite 1 Telescope Detector Transmitter Receiver A\D Converter Correlator Earth PC 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 12
Components I Three receivers 0.95 GHz bandwidth @ 3.1, 4.2 & 5.9 GHz Photomultiplier Light collector Tilt mechanism Preamplifier + transmitter Rotation propellers Antenna Common antenna 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 13
Components II Analogue-to-digital converters Up to 5 giga-samples per second (GSPS) Virtex-6 FPGA Delay Correlation Integration Transfer to host PC 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 14
Dark-room experiment Blue LED + pin-hole Beam-splitter (non-polarising) Two channels 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 15
Dark-room set-up Battery 1 Attenuator LPF Electric al Circuit 1 TIA BS Pinhole PMT 1 ADC Corr PC Attenuator LPF PMT 2 Electric al Circuit 2 TIA Battery 2 Light Source Some experiments: switch A/D and correlator by fast scope and memory 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 16
Typical input Three channels (red inactive) LED: 415 nm. Power: 2.8 W. Pin-hole: 15 µm. H = 78 cm. D = 0 cm. 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 17
Electronic delay Correlation of 15MHz modulated LED Cables Same length Cable A is longer by 2.5m Cable B is longer by 2.5m 4 x 106 Correlation for 15MHz Modulated LED N correlation = 1, N samples = 2 19 Correlation for 15MHz Modulated LED N correlation = 1, N samples = 2 19 Correlation for 15MHz Modulated LED N correlation = 1, N samples = 2 19 3 X: 2 Y: 2.521e+06 3 x 10 6 Cable A is longer by 2.5m than Cable B X: 18 Y: 2.454e+06 3 x 10 6 Cable B is longer by 2.5m than Cable A X: -14 Y: 2.452e+06 2 2 2 γ [ mv 2 ] 1 0 γ [ mv 2 ] 1 0 γ [ mv 2 ] 1 0-1 -1-1 -2-2 -2-30 -20-10 0 10 20 30 τ [ sample ] -30-20 -10 0 10 20 30 τ [ sample ] -30-20 -10 0 10 20 30 τ [ sample ] 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 18
Electronic delay Correlation of low power LED for different baselines Correlations for different baselines Correlation for different baselines LED 15µm, h = 150cm, P 2mW Maximum Correlation for different baselines Maximum correlation for different baselines LED 15µm, h = 150cm, P 2mW average( corr(i 1, I 2 ) ) [volt 2 10-12 ] 20 15 10 5 0 N correlations = 10, N samples = 20M, f = 1.25GHz d = 0 cm d = 0.5 cm d = 1 cm d = 1.5 cm d = 2 cm d = 2.5 cm d = 3 cm maximum correlation [volt 2 10-12 ] 18 16 14 12 10 8 6 N correlations = 10, N samples = 20M, f = 1.25GHz experiment theory normalized to experiment -5-400 -300-200 -100 0 100 200 300 400 τ [nano-sec] 4 0 0.5 1 1.5 2 2.5 3 baseline [cm] 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 19
Digital band-width Short integration times Effective continuous sample rates Our setup: 3 500,000 samples / 0.8 s = 1.875 Msamples / s HBT: 24 MHz 2 = 48 Msamples / s HBT integration time: 1.5 h For the same number of samples we need to integrate for 38 h Solution: clip measurements, use 1 bit correlation (not 10 bit) Also: remove all mobile phone signals (use µ-metal shield) 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 20
Transmitting the intensities Growing baselines, on the ground and in space Coaxial transmission difficult or impossible Fibre optics for stellar light transmission not likely Limited space-bandwidth product: low efficiency Delay still has to be performed electronically or mechanically Can we compress the detected currents? We first check the method of Compressed Sensing 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 21
Compressed sensing principle Original time trace Requires dense sampling After using the proper filter (e.g. low-pass, wavelets) Allows sparser sampling 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 22
Compressed sensing: matrix notation K N = N x D α Sparse and Redundant Representation Modeling of Signals Theory and Applications By Michael Elad 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 23
Compressed sensing f x x y G x ( ) = 2 + ( ) f ( x) arg min xˆ { } 0 G x = λα λ Lx ( ) = 0 for x = Dα 0 2 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 24
The correlation properties N 1 ym nl n n l m n [ ] = ( [] )( [ + ] ) N l = 1 1 2 [ ] αδ [ ] y m = m m0 + β xm [ ] ( ) x[ m] Normal 0,1 τ = = T c 2 2 α n γd, β n N N =10 13 12 14 12 13 3 10 ym [ ] m 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 25
Taking a simple case [ ] αδ [ ] y m = m m0 + β xm [ ] 0 { [ ]} { αδ } α Yk [ ] = ym = [ m m] = e π [ k ] 0 [ k ] β 0 i2 km / N α ym [ ] m 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 26
A simple case: Fourier domain [ ] αδ [ ] y m = m m0 + β xm [ ] 0 { [ ]} { αδ } α Yk [ ] = ym = [ m m] = e π [ k ] α α α 0 [ k ] β 0 i2 km / N N 1 1 Y [ k ] = n1 [ l+ m ] n n2 [ l ] n n ˆ 1 [ k ] n n ˆ 2 [ k ] n N = l= 1 N ( )( ) ( )( ) Re(Y[k]) 0 Im(Y[k]) 0 0 k α α α 0 Re(Y[k]) α 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 27
Poisson noise in Fourier domain N f k x m e x k ix k N i2 πκ ( kmn ) / [ ] β [ ] = β 1[ ] + 2[ ] m= 1 2 ( ) β = 0 β = 0.001α β = 0.01α β = 0.1α Correlation (circle radius) buried in Poisson noise Dropping frequencies drops points, not noise 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 28
Other compression methods n <<1 { = 2, = 2, ( )} = 1, = 2, ( ) p y x nt { } p y x nt 1 ( ) e = nt2 << 1 Sparse photons, sparse photoelectrons 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 29
Compression efficiency ( n) b 2 n log2 2 comp ( n ) = = = 2 n n log2 n k 1/ n ( ) δω 10 T 10 η 9 09. 14 A 30m 2 Various compression schemes, here showing run-length No gain for brighter objects, m < 8 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 30
Future directions System improvement steps Rewrite correlator to 1 bit to improve flow Add third channel, test closure Test on moving platforms Test other compression options 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 31
Research directions Use electronic analog correlation (still faster) Use photonic correlation (e.g. HBT in OCT) Nonlinear optics Requires very narrow beams, optical delay lines 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 32
Summary We built a lab system to test space HBT Integration and testing proceeding Checked the options of compressing data Compressed sensing depends on reduced band-width Requires widest band possible Other compression methods useful at low flux 13 May 2014, Nice, France Workshop on Hanbury Brown & Twiss interferometry 33