RA-2 LEVEL 1B PROCESSOR VERIFICATION

Transcription

1 RA-2 LEVEL 1B PROCESSOR VERIFICATION M. Roca (1), M.P. Milagro (2), R. Scharroo (3), S. Laxon (4), D. Calabrese (5), B. Greco (6) and S. Baker (7) (1) ESA/ESTEC, Keplerlaan 1, 2200AG Noordwijk, (The Netherlands), (2) ESA/ESRIN, Via Galileo Galilei, I-00044, Frascati, (Italy), (3) DEOS/Delft University, (Nederland), (4) CPOM/UCL Gower St., London WC1E 6BT, (UK), (5) Alenia Aerospazio, Via (Italy), (6) ESA/ESRIN, Via Galileo Galilei, I-00044, Frascati, (Italy), (7) MSSL/UCL, Dorking RH5 6NT(UK), ABSTRACT The L1b processor is designed to process raw, level0, data from the RA-2 instrument to apply engineering and other corrections. From the Level 0 data (which are basically telemetry data) the Level 1b processor is applied and output the Level 1b product. In doing this auxiliary data is used. These algorithms and the parameters used in this process shall be verified and possibly tuned. Later the output of the Level 1b will be used to produce the Level 2, which is the product commonly used by the scientific community. The investigations, modifications applied to the algorithms as results of the investigations and the final results are described in this paper. 1. INTRODUCTION The EnviSat commissioning phase, and in particular applied to the altimeter, has the following objectives: verification of the instrument behaviour and optimisation of the on-board parameters [1]; verification of the Level 1b processor; verification of the Level 2 processor [2]; geophysical validation [2]; absolute calibration of instrumental parameter as range and backscatter [3]; relative calibration to other altimeters of all the geophysical parameters [4]. In the following sections, the activities performed in the verification of the algorithms within the Level 1b processor are described. The main outputs of the Level 1b product are: window delay, from which the range will be obtained; datation, time tagged of the measurement that has to correspond to the time in which the window delay is given; sigma-0 calibration factor, which is later used in the Level 2 in order to obtain the final Sigma-0. The verification of the algorithms used in the Level 1b processor in order to produce these 3 outputs, has been performed during the calibration phase of EnviSat. The objective of the Level 1b verification is to ensure that the algorithms (more precisely the equations) used in this processor are correct. Later, tuning of some of the parameters involved in these equations may occur. This is the case of some of the ground measured parameters that are later calibrated in-flight by means of different absolute calibrations (range and sigma-0) and the relative one. 2. THE LEVEL 1B PRODUCT Main characteristics of the Level 1b product: Contains 1 full orbit.

2 Conversion of satellite time to UTC, and time tag computation. Instrumental corrections applied: - IF filter: corrects power distortions of echo waveforms. - Point Target Response (PTR): corrects time delay for internal path. - Corrections for drifts of reference timing source (USO). Waveform alignment with the corresponding estimated on-board parameters used in the computation of the window delay and the attenuation. Window delay computed (but no retracking of waveforms performed -a responsibility of the Level 2). Total attenuation (AGC) computed. Waveforms are included, and corrected by the IF filter. Level 1b output formatting. The key processing algorithms required for the generation of the L1b product starting from the RA2 Level 0 product are described in the following sub-sections. All of them require verification, however, some are more likely to have errors than others, or affect more directly to the 3 main outputs of the Level 1b: window delay, from which the range will be obtained; datation, time tagged of the measurement that has to correspond to the time in which the window delay is given; sigma-0 calibration factor, which is later used in the Level 2 in order to obtain the final Sigma-0. The method used to verify the Level 1b is mainly through analysis of these 3 outputs and comparing them with the expected values. When problems are detected, the equations that produce these outputs are analysed.the description of the analysis and the correction of the errors encountered will be described in Section 3 (for window delay and datation) and Section 4 (for Sigma-0). The purpose of this activity is Level 1b verification, which aims to ensure that all the equations included in the Level 1b processor are correct. The accuracy of some of the terms of these equations might not be precise enough at this time. Their accuracy is supposed to be verified and improved by means of other activities as the Absolute Calibration for both range and sigma-0, and the Cross-calibration Level 0 data decoding Each field of each Source Packet contained in the raw data stream (Level 0) is extracted and converted to engineering units according to its binary format and the value and units of the least significant bit Orbit Data generation This function provides the calculation of latitude/longitude coordinates for each processed data block in the source packet and the required spacecraft orbit parameters. Data are retrieved through the EnviSat Orbit computation library that uses data from the Orbit State Vectors files. In order to call this library the time stamp (or datation) in UTC is required. This is calculated as described in Section 2.3. However, a further correction to the UTC to compensate for the propagation delay can only be performed after the first estimation of the satellite height. This correction is performed in this section Conversion of satellite time to UTC The UTC datation for each data block of a source packet shall be computed and included in the output data records. A data record is in fact associated to each data block (18 Hz data). UTC time shall be derived through correlation of the RA-2 ICU clock counter datation OBDH (43 bits) with the satellite on-board reference binary time which is in turn correlated to an UTC reference time. The UTC datation is computed in the UTC Transport Format (MJD 2000) which consists of three 32 bit signed integer values: the first one representing the number of days between the current day and 1 January :00 hours; the second one representing the number of seconds on the current day since 00:00 hours; the third one representing the number of microseconds elapsed since the last second.

3 The datation OBDH derived from the ICU counter and stored in the Level 0 is relevant to the first Tx pulse of the first Data Block of the current Source Packet analysed. However the datation word shall be the time in which the averaged waveform was at the surface. These corrections are taken into account in the computation of the Level 1b datation [5] Instrumental corrections applied Three instrumental correction are applied. they are briefly described below. More detailed information can be found in [6]. Point Target Response (PTR) data processing: PTR data measurements are processed every Source Packet (SP) in order to retrieve the relevant correction factors required to compensate for the variability of instrument errors. The computed calibration factors, if valid, are then used to correct the measurement data of AGC, for the retrieval of the sigma-0, and window delay, for the retrieval of range. IF shape correction: This function implements the correction of the waveform samples for the transfer function shape distortions due to the receiver noise. The main contribution to this distortion comes from the IF (intermediate frequency) filter, reason why is usually called IF correction. The correction consists mainly of multiplying each Ku/S waveform sample by the corresponding sample of the effective correction mask. The mask has been properly modified to account for the amount of fine shift applied to the waveforms in the on board processing. Corrections for drifts of reference timing source (USO) In addition to the satellite clock, the altimeter is provided with and Ultra Stable Oscillator (USO) clock. This clock is used in the computation of the window delay. As with any clock, an USO is likely to drift (variation of its frequency or period) throughout the mission life time. Therefore, changes to this frequency needs to be monitored. This is done in the IECF and an Auxiliary file is generated, which is used by the Level 1b processor with the new value of its clock period Waveform/Parameters alignment An alignment process to correct for the time lug which exists in the RA-2 on-board processor between the accumulation of echoes and their subsequent processing, is applied. The waveform samples (both for Ku and S bands) extracted from the j-th data block of the i-th source packet shall be associated to the estimated parameters (e.g. Rx_dist_c, Rx_dist_f, AGC_c1, AGC_c2, AGC_f, AGC_Dist, AGC_Rate, htl_dist, htl_rate) for the (j-1)-th data block of the i-th packet if j 1. In case j=1, then they are associated to the last data block (the 20th) of the i-1 source packet. The result of this processing is written to the L1b record corresponding to the j-th data block of the i-th SP (or to the last DB in the (i-1)-th SP) Window time reference extraction This function evaluates the total window delay time of the processed waveforms (Ku and S) from the Rx delay coarse and fine components computed by the on-board tracker. It also corrects it for instrument errors and Doppler shift. The four steps are the following: computation of the on board Rx delay (separately for Ku and S) from the Rx coarse and fine components extracted from the DB s of the Source Packet, and converted to time units using the Tx/Rx clock period; this measurement is further compensated by the height rate and again converted to time units using the Tx/Rx clock period; the time delay measurements are corrected by instrument induced errors using both ground (from the Characterisation file) and in-flight (from PTR) calibration parameters; the Doppler effect caused by the satellite radial velocity is also compensated for.

4 2.7. Automatic Gain Control calibration The on-board AGC measurement is corrected by the characterisation corrections extracted from the Characterization Auxiliary File to provide corrected AGC values for Ku and S data. Any gain variation in the radar Tx/Rx chain with respect to the nominal condition characterised on the ground is compensated using in-flight calibration data derived from PTR measurements. The power scaling factor required to evaluate the sigma_0 in the level 2 processor is additionally evaluated Level 1b data formatting The output is formatted as a RA-2 Level 1b product [7]. 3. WINDOW DELAY (RANGE) AND DATATION VERIFICATION 3.1. Introduction To derive the height of the mean surface, the radar altimeter needs to measure the range. This range is actually derived from the measurement of the time between the transmission of the pulse and its reception. It is also necessary to know when this measurement is obtained in order to locate the height measurement in the surface, in other words the time when this pulse was on the surface. Since the altimeter is a very accurate instrument the importance of the time accuracy of the measurement is very high, not only on the range but also locating this range, on the Earth s surface. This procedure is called datation. Since the altimeter performs measurements of the range for every 100 pulse average, we will consider this averaged waveform as a single one in order to build a single range value consistent with a single datation value. The averaged waveform should, therefore, be considered as a single waveform located in the middle of the 100 pulses averaged. Now we can define Datation as: The exact time when the averaged waveform is reflected from the surface, and hence, when the middle of the 100 averaged waveform is at the surface. An important remark to be kept in mind during the complete process of the calculation is that RANGE and DATATION have to be consistent. A number of factors have to be taken into account. They are: Location of the mid point of the averaged waveform. Time when the value of the on-board datation (or OBDH) is actually generated by the hardware. Propagation time delay. Time when the value of the range is taken from the on-board hardware. At the same time, when we calculate the range, or in the case of the Level 1b the window delay, we also have to take into account the following points: at which point during the100 pulse average, the information of range (in time), Rx_dist coarse and fine, are written in the SP (only one of the 100 window positions are contained in the SP); in case it is not the centre of the averaged waveform, propagate the time stamp to this centre using the range rate, also given in the SP. Details of this calculations are reported in [5], and details of the instrument behaviour and when and how these parameters are written in the SP, reported in [8] Results of the data analysis An analysis of ESRIN FDMAR data from 23 August, computing the sea level differences, was performed. The sea level differences were computed as the satellite altitude minus the altimeter range and minus the corrections, both due to atmosphere and sea state. Details of this computation are listed below:

5 corrections = dry tropospheric correction + MWR wet tropospheric correction + dual frequency ionospheric correction + SSB + inverse barometric correction + solid earth tide + pole tide + geocentric tide height + ocean tide + loading tide + MSS orbit changed with DEOS DORIS/SLR orbits MSS replaced with an in-house version of GSFC00.1 dry tropospheric correction. set to -2.3 m wet tropospheric correction. set to m ocean/load tide from GOT00.2 The results are shown in Fig timing bias = 78.3 ms, range bias = m Sea level anomaly (m) ascending passes descending passes Orbital altitude rate (m/s) Fig. 1. Sea level differences. The X-axis is the orbital altitude rate (computed from the DEOS orbits); the Y-axis is the sea height anomaly (applying all corrections). The interpolated line through the measurements in this graph gives the offset (intercept with the Y-axis) which becomes an estimate of the range bias. The slope of this line is a measure of the time tag or datation bias. Conclusions: 1 Offset indicates an estimated range bias of ~10.5 m (Envisat ranges too long by this amount) 2 Slope of linear trend suggests timing bias of ~78 ms (RA-2 time stamps late by this amount) 3.3. Range Investigation The way the time window delay is computed has been described in Section 2.6. The way this window delay shall be computed in the Level 1b processor is (Eq.1):

6 W Delay( middle, 49.5) = W Delay( 17) + range rate ( ) PRI ε g ε PTR + doppler (1) where W-Delay(17) is the position of the window (translated into window delay) for the 17th pulse in the 100 average, which is the only information contained in the SP. The middle position of the averaged waveforms corresponds to the equivalent pulse 49.5; ε g is the time delay corrections of instrument induced errors using ground (from the Characterisation file) calibration parameters. This correction is actually composed by the sum of 4 instrument internal delays. It is depicted in Fig.2; ε PTR is the time delay corrections of instrument induced errors using in-flight (from PTR) calibration parameters. It is depicted in Fig.2; and doppler is the doppler correction due to the altitude range rate ε g ε PTR ε PTR Fig. 2.Time delay corrections due to instrument internal delays. The instrument delays, ε g and ε PTR, are delays artificially added to the measurement of the total time delay. The measurement reference for the time delay computation is the antenna flange. However, those two delays are accounted in the time measurement because it is computed at the receiver. Therefore, that shall be subtracted. The old processor was actually adding these terms, inducing a range bias. Further more, one of the contributions to the ground calibration factor (ε g ) had also the wrong sign. The sign has been corrected and the numerical value adjusted accordingly. The Level 1b reference processor immediately implemented these modifications, tested and verified them. The operational processor (IPF) has also been modified and will be integrated by mid November, from version number 4.53 on. Better accuracy of the range bias shall now be provided by the absolute and cross range calibrations Datation investigation The way the datation is computed has been described in Section 2.6. The equation, in the Level 1b processor that computes this datation is (Eq.2): UTC DB = ( SBT 0 UTC 0 ) + UTC H c (2) where

7 SBT 0 -> UTC 0 represents the translation of the satellite binary tie (or ODBH in the SP) into UTC; UTC are all the factor included to compensate by the location of the mid point of the averaged waveform; the time when the value of the on-board datation (or OBDH) is actually generated by the hardware; and the time when the value of the range is taken from the on-board hardware. And H/c is the propagation delay. A full description of the parameters can be found in [5]. They are also depicted in Fig.3. UTC range_rate SBTo UTCo T R Fig. 3.Graphic showing all the contributor of the datation equation. Another contributor to the datation is the hidden in the window delay computation (Eq.1). Any factor in the window delay computation that depends on the range rate will appear as a change in datation. In other words, any error in the factor of the range rate in the window delay computation will appear as a datation error.

8 Further, during the waveform accumulation on-board, the estimated parameters are written in that data block. However the averaged waveform is not available until after the complete data block has been written. Therefore, that waveform is reported in the following data block. However, as explained in Section 2.5, this waveform should be associated to the previous data block. This achieved by re-aligning the waveforms and associated parameters within the L1b processor. If this is not performed correctly a datation error will result due to a misalignment of the retracked range and the time stamp. Three contributors to the datation error were found: 1 The units of the instrument range_rate were wrongly decoded from the Level 0, giving a datation error of 17.2 miliseconds. 2 The re-alignment between waveform and estimated parameters was performed incorrectly, resulting in a datation error equivalent to a complete data block, 55.7 milliseconds. 3 The correction for the propagation delay in order to evaluate the time when the pulse was on the surface (see eq3), had the wrong sign, resulting in a datation error of approximately (since depends on the satellite altitude) of 2.6 miliseconds. These modifications were immediately implemented into the Level 1b reference processor, tested and verified. The operational processor (IPF) has also been modified and will be integrated by mid November, from version number 4.53 on Verification, conclusions and final results Data used for the verification are from 1st September. The ERS-2 OPR1 data were available as well. Just as the Envisat data they are merged into RADS, which basically means that all the data are repaired as mentioned in [9], plus the latest tide models, etc., just as for Envisat. The ERS-2 data are corrected for a -1.3 ms timing bias. The orbits used are DEOS fast-delivery SLR-only orbits for Envisat and fast-delivery SLR/Altimetry-based orbits for ERS-2. All meteo corrections were discarded, which means the same constant values were used for both Envisat and ERS-2. Assuming the wet and dry troposphere, and ionosphere delays not to change significantly over 30 minutes, this is actually a better way of analysing the difference between the altimeters than using their respective measured MWR wet tropospheric correction, or a mix of dual-frequency and modelled ionospheric correction. The result of co-linear sea height difference is plotted in Fig.4. As a conclusion we can say: 1 The datation bias has disappeared. The estimate is actually as low as milliseconds. More data will be analysed in the future in order to estimate this difference to a better accuracy. 2 After the thorough review of the equations, a relative bias between Envisat and ERS-2 of approximately 47 centimetres (Envisat measuring SHORT) remains, which is supposed to be due to the inaccuracy of the ground measurements. This is the reason for a range calibration. More data will be analysed in order to retrieve this difference to a better accuracy. However, the final evaluation of the RA-2 absolute range bias will be performed by various range calibration activities. 4. SIGMA-0 VERIFICATION 4.1. Introduction As mentioned in Section 2, one of the main outputs of the Level 1b processor is the sigma-0 calibration factor, which is later used in the Level 2 in order to obtain the final Sigma-0. From the radar equation we can derive the Sigma-0 as given in Eq.3:

9 2 Envisat Cycle 10, Passes (18 October 2002) 1.5 Sea level anomaly (m) ascending passes descending passes timing bias = ms, range bias = -472 mm Orbital altitude rate (m/s) Fig. 4. Co-linear sea height difference between RA-2 and the ERS-2 altimeter. The horizontal scale is altitude rate. Any slope of the fit betrays a relative timing error, the offset a relative bias. Only one out of two points is plotted. Blue pulses for ascending tracks, red crosses for descending tracks. Sigma 0 P r 64π log P 2 10 H 3 H = + + log log( τ S t R c ) 2G A cλ ku (3) where P r /P t is the ratio between the received and transmitted power; λ is the radar signal wavelength; H is the altitude above the reference ellipsoid; (1+H/R) the correction for the Earth sphericity; τ is the radar pulse length; S c is the chirp slope (that depends on the current bandwidth); and G A is the antenna gain. The sigma-0 calibration factor groups all factors that do not depend on the surface characteristics (or in other words, the waveform power level), such that in the level 2 processor this factor can be used directly by multiplying it (in linear power units) by the maximum power, derived from the maximum of the return waveform. In particular the sigma-0 scaling factor depends on the transmitted power; the geometry; the attenuation applied to the waveform, AGC; and the total gain of the

10 altimeter, G Tx/Rx, (antenna, G A ; electronics, G e ; and digital gain, G SPSA ); but not on the received power retrieved from the waveform, P echo. The total gain of the altimeter can in turn, be referred to a reference such as the PTR Results of data analysis The first check of Sigma-0 retrieved from the RA-2 Level 2 product was performed with data from 20 to 31 March on the basis of global ocean statistics. These results were compared with the ERS altimeter retrieved Sigma-0. A first estimation of the difference between the two was obtained. A second check, more precise, was performed using collocated RA-2 data with ERS altimeter data, from 5th May. This comparison showed a difference between the two retrieved Sigma-0 of db, being the RA-2 one higher. The shape of the Sigma-0 and AGC histograms in Ku-band over sea were found to be very similar to those of the ERS-2 altimeter, hence the plausibility of the RA-2 sigma-0 measurement is ascertained, making possible to align the RA-2 sigma-0 measurements to ERS-2 by the introduction of a simple bias. The histograms of sigma-0 and AGC at Ku band of RA-2 and the equivalent histograms of ERS (ERS-2 takes data for March 2001, as there were no concurrent data those days), are shown in Fig.5 and Fig.6. RA-2 Sigma-0 histogram ERS-2 Sigma-0 histogram sigma-0 [db] sigma-0 [db] Fig. 5. Sigma-0 histograms for RA-2 and ERS-2 altimeter. The vertical axes is the percentage of occurrences and the horizontal axes is the value of the Sigma-0 in db s. It can be observed that the two histograms have the same behaviour, however there is an offset between the two. This constant offset between the two measured sigma-0 is not necessary a bias in RA-2 (no altimeter has calibrated sigma- 0 in absolute terms, but they are all referred to GeoSat sigma-0 measurement). However, given the magnitude of the value, it is a clear indication of an error in the computation of sigma-0 in the Level 1b, and therefore the algorithm was reviewed Sigma-0 bias investigation In order to investigate the source of the error in the sigma-0 scaling factor the attention was focused in the term P r /P t, since it was the most likely source of error. As mentioned in Section 4.1, this term depends on several factors. The AGC was

11 RA-2 AGC histogram ERS AGC histogram AGC [db] AGC [db] Fig. 6. AGC histograms for RA-2 and ERS-2 altimeter. The vertical axes is the percentage of occurrences and the horizontal the AGC in db s. The two histograms have the same behaviour. The values do not have to be the same because they depend upon several instrument parameters as antenna gain, setting of the reference power value, etc. demonstrated to be correct by comparing the histogram to the corresponding ERS histogram. All the other factors were investigated and re-computed by an independent method. As a result of the investigation several problems were found. They are: 1 Error in the computation of the digital gain of the instrument: G SPSA. 2 Error in the decoding of the P PTR from the Level 0. The energy of the PTR was considered instead of the power as for all other parameters. 3 Error in the ground calibration values used in the Level 1b processor and given in the Characterisation auxiliary file. The other terms of the scaling factor equation were also investigated. Another miscomputation was found: 4 Error in the correction by the Earth sphericity. The correction shall be applied only once and not to the power of 3. The 4th term of Eq.3 was written as: H 30log H R (4) and shall be rewritten as: H 30log( H) + 10 log R (5) These 4 errors lead to a wrong evaluation of Sigma-0 of about 8-9 db. These modifications were immediately implemented into the Level 1b reference processor, tested and verified. The operational processor (IPF) has also been modified and will be integrated by mid November, from version number 4.53 on.

12 4.4. Verification, conclusions and final results The results of the two independent methods, the one used in the Level 1b processor and the new one used for comparison, we computed. The comparisons are shown in Fig.7 and Fig.8 for Ku and S -bands respectively. We can say that the two independent methods give extremely consistent results. Sigma-0 scaling factor in Level 1b K_cal_Ku [db] 20.5 Series record Fig. 7.Sigma-0 scaling factor for Ku-band. Comparison between the two independent methods. Scaling factor in the Level 1b for S-band Sigma-0 determination K_cal_S [db] ecord Fig. 8.Sigma-0 scaling factor for S-band. Comparison between the two independent methods.

13 In order to be able to use the Witter and Chelton wind models [10], an offset of approximately 4.7 db shall be applied to the RA-2 retrieved sigma-0. The residual error in the sigma0 values will be calibrated. This will be done as a result of the sigma-0 absolute calibration activities by means of a radiometric transponder [3] and the sigma-0 passive calibration by using natural targets [11]. 5. REFERENCES 1. Roca M., and Laxon S., EnviSat RA-2: Tracking Performance and Resolution Selection Logic, EnviSat Calibration Workshop Proceedings, The Netherlands, Milagro M.P. and Benveniste J., RA-2 Level 2 algorithm verification, EnviSat Calibration Review Proceedings, The Netherlands, Roca M., et al., RA-2 Absolute Range and Sigma-0 Calibration and In-Flight Verification, ERS-EnviSat Symposium, Göteborg, October Benveniste J., et al., RA-2/MWR Cross-Calibration and Validation Plan, PO-PL-ESR-RA-0005, iss 1.1, 14 February Roca M., Datation in RA-2 Level 1b Product, ESA Technical Note, PO-TN-ESA-GS-00588, Issue 2/A, February Celani, C., et al., Instruments Corrections Applied on RA-2 Level-1B Products, EnviSat Calibration Review Proceedings, The Netherlands, EnviSat RA-2 Product Handbook, available at the EnviSat web site. 8. Roca M., The RA-2 On-Board Tracker and its Autonomous Adaptable Resolution, ESA Technical Note, PO- TN-ESA-RA-1316, 1.a, 13 February R. Scharroo, et al., A recipe for upgrading ERS altimeter data, Proceedings of the ERS-ENVISAT Symposium, Gothenburg, Sweden, October 2000, Eur. Space Agency Spec. Publ., ESA SP-461, 10 pp., Witter D. L. and Chelton D. B., A GeoSat altimeter wind-speed algorithm and a method for altimeter windspeed algorithm development, JGR, Vol. 96 (C5), , B. Greco, et al., A novel approach for absolute backscatter calibration of spaceborne altimeters, Proceedings IGARSS 2000, IEEE 2000 International, pp vol.5, 2000.