Phase error assessment of MIRASSMOS by means of Redundant Space Calibration.pdf
- 1. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory Phase error assessment of MIRAS/SMOS by means of Redundant Space Calibration Rubén Dávila(1), Francesc Torres(1), Nuria Duffo(1), Ignasi Corbella(1), Miriam Pablos(1) and Manuel Martín-Neira (2) (1) Remote Sensing Laboratory. Universitat Politècnica de Catalunya, Barcelona.SMOS Barcelona Expert Centre (2) European Space Agency (ESA-ESTEC). Noordwijk. The Netherlands 1/20
- 2. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory The Soil Moisture & Ocean Salinity Earth Explorer Mission (ESA) Aperture Synthesis Interferometric Radiometer • MIRAS instrument concept - Y-shaped array (arm length ~ 4.5 m) - 21 dual-pol. L-band antennas / arm - spacing 0.875 λ (~1400 MHz) -no scanning mechanisms, 2D imaging by Fourier synthesis -(u,v) antenna separation in wavelengths 2D images formed by Fourier Synthesis (ideal case). Cross correlation of the signals collected by each antenna pair gives the so- called: Visibility samples V(u,v): Launched November 2009 TB (ξ, η) − Tph 2 V(u, v) =< b1 (t)b (t) >= F * 2 F(ξ, η) (SMOS artist’s view, by EADS-CASA Space Division, Spain) 1−ξ −η 2 2 IGARSS 2011 Vancouver 2
- 3. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory Simplified block diagram of a single baseline MIRAS measures normalized correlations: antenna 1 Mkj antenna 2 antenna planes System temperatures measured by a power detector in each receiver Visibility sample at A TsysAk TsysAj V = M kj the antenna plane kj jφkj A Fringe Wash function at the origin (τ=0): Gkj (0) e • Modulus (≈1) IGARSS 2011 Vancouver • FWF Phase at antenna plane 3
- 4. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory Framework of the activity SMOS is producing images within expected performance. However, there is some degree of image distortion (spatial errors) due to a number of causes. This research activity is devoted to assess the different contributions of spatial errors, with two objectives in mind: • SMOS Improved performance • SMOS follow-on specifications The RSC method is devoted to assess the peformance of phase calibration. For calibration purposes, the phase calibration term (antenna plane) is modeled as: φkj = (φkant − φ jant ) + (φkrec − φ jrec ) + φkj A FWF Antenna phase terms Receiver phases Fringe-wash term IGARSS 2011 Vancouver 4
- 5. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory SMOS phase calibration strategy • Receiver phase drift is calibrated by periodic (2-10 min) correlated noise injection (LO phase track) • Antenna phase term (manufacturing tolerances): Measured on ground • Fringe washing term due to filter response differences (negligible) Antenna Receiver plane φkant plane φkrec Antenna phase test set-up A L receiver " k " η C M kj Noise injection Correlator Switch Front end phase model receiver " j " IGARSS 2011 Vancouver 5
- 6. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory Redundant Space Calibration (RSC) Redundant baselines measure the same visibility using a different pair of antennas Redundant baselines Visibility phase measured by a baseline: φVkj = φk − φj + φscene,kj RSC phase differences are independent of the phase of the scene Baseline phase differences: φVkj − φVji =k − 2φj + φi φ IGARSS 2011 Vancouver 6
- 7. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory RSC system of equations A system of equations can be built using independent RSC equations 0 0 1 −2 1 0 … … 0 0 0 0 1 −2 1 0 … 0 Applied on calibrated … … … … … … … … … visibilities the RSC method retrieves the 0 … … … … 0 1 −2 1 ·φreceivers = φphase differences … residual phase error … −1 1 … −1 1 .. … … −1 … 1 … … 1 −1 … … … … … … … … … … A matrix: 66 x 69 Underdetermined system Receivers vector: 69 x 1 (three unknown phases, rank = 66) Phase differences vector: 66 x 1 Moore-Penrose pseudoinverse matrix 66 equations, 69 unknowns Averaging is required to reduce uncertainty due to thermal noise IGARSS 2011 Vancouver 7
- 8. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory Averaging: visibility measurements must be carefully selected • Low visibility amplitude: produces unwanted variations and jumps • Fast scene changes: phase bias in land-ocean transitions • RFI: interferences that spoils the phase values Land-ocean transition Low visibility amplitude RFI IGARSS 2011 Vancouver 8
- 9. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory RSC: examples of good quality visibility samples Averaging area Averaging area Averaging area Arm A Arm B Arm C Red line: Average snap-shots IGARSS 2011 Vancouver 9
- 10. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory RSC: Impact of undetermination The 3 unknown phases have a physical meaning: Tilt angle Steering angle Pointing error Common path delay Irrelevant IGARSS 2011 Vancouver 10
- 11. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory RSC: Pointing error in the phase retrievals Simulations show that a pointing error yields a linear phase error directly related to the antenna position in the arms. φerror,bslN = a·u bslN + b·v bslN a b TBcalibrated (ξ, η) = TBideal ξ − ,η− 2π 2π a ξps ' = ξps − 2π b ηps ' = ps − η 2π Retrieval error linear in each arm The pointing error can be corrected, if required, using a point source (e.g, an interference at a known position ξps , ηps) IGARSS 2011 Vancouver 11
- 12. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory Assessment on the pointing error in RSC retrievals Simulation: SMOS point source retrieval by the RSC method: random phase error Ideal Phase corrupted Corrected •Image blurring (example, σphases = 25º) • Secondary lobes increase • Small pointing error: the maximum has been displaced. Once the point source is RSC calibrated, image blurring and secondary lobes are corrected. However, the pointing error is not compensated. 12 IGARSS 2011 Vancouver
- 13. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory RSC implementation (i): Good/bad estimations Due to pointing error, the difference between two phase retrievals must be linear. This property is used to discard bad estimations of the RSC phases φretrieved = φIVT,error + φpoint ing error 1 1 φretrieved = φIVT,error + φpoint ing error 2 2 φretrieved − φretrieved = φpoint ing error − φpoint ing error 2 1 2 1 Linear Bad estimations Good estimations IGARSS 2011 Vancouver 13
- 14. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory RSC retrieved phases Final RSC phases retrieved by averaging RSC phases from 38 orbits over the ocean Horizontal Phases Vertical Phases RSC Phase Error dispersion σH =5.97º σV =3.17º • RSC gives a conservative Horizontal Mean Phases Vertical Mean Phases upper bound for SMOS residual phase errors • RSC phase dispersion very much contributed by pointing error IGARSS 2011 Vancouver 14
- 15. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory RSC: phase error impact of pointing error Mean pointing error (H) Horizontal Phases Simulation r SMOS std <r> Horizontal Std σphases (°) Simulation: point source shift for 200 cases with σph=20º. 95% of points within a radius σH =5.97º σV =3.17º r=2mrayleigh centred at the point source real position rH = 0.00066 rV = 0.00037 ∆L H = km 0.76 ∆L V = km 0.43 ΔLH and ΔLV below 2% of SMOS resolution (42 km) IGARSS 2011 Vancouver 15
- 16. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory RSC peformance assesssment: RFI in the Caribbean Sea Interference from a vessel (11/02/2010, 21:23 semi-orbit) IGARSS 2011 Vancouver 16
- 17. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory RSC peformance assesssment: RFI in the Caribbean Sea Horizontal IGARSS 2011 Vancouver 17
- 18. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory RSC peformance assessment: RFI in the Caribbean Sea – Primary to Secondary Lobe Ratio (H): Case Primary to Secondary Lobe Ratio Real Point Source 17,40 dB Corrected Point Source 16,50 dB – Primary to Secondary Lobe Ratio (V): Case Primary to Secondary Lobe Ratio Real Point Source 17,40 dB Corrected Point Source 16,65 dB – The uncorrected RFI presents a main-to-secondary lobe ratio very close to an ideal point source. – The RSC method uncertainty above SMOS phase error accuracy!! 18
- 19. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory RSC implementation: Interference in Cáceres (Spain) Vertical 19
- 20. •Remote Remote Sensing Laboratory •Sensing Universitat Politècnica de Catalunya •Laboratory Conclusions • The RSC method cannot be used to phase calibrate SMOS in a per snap shot basis due to the need for long averaging and filtering • SMOS orbital phase drift requires periodic (2-10 min) correlated noise injection (LO phase track) • The RSC is used to validate the consistency of SMOS phase calibrated visibilities: •RSC phase retrieval accuracy limited by undetermination (pointing error) •SMOS phase errors well below σH=5.97 º and σV=3.17º, probably very close to the σ =1º target •Assessment on point sources (RFI) shows that the impact of SMOS residual phase errors on image distortion is probably negligible IGARSS 2011 Vancouver 20