PDS_VERSION_ID = PDS3 RECORD_TYPE = FIXED_LENGTH RECORD_BYTES = 80 OBJECT = TEXT PUBLICATION_DATE = 2000-04-16 NOTE = "Brief description of Clementine bistatic radar data processing at Stanford University. Includes naming conventions for software, data, etc." END_OBJECT = TEXT END Bistatic radar data were collected during four Clementine revolutions of the Moon on 1994-04-09 and 1994-04-10 and during three revolutions on dates 1994-04-23 and 1994-04-24. In the first case all three NASA Deep Space Network (DSN) 70-m antennas were used (DSS 14, 43, and 63); in the second case only DSS 43 and 63 were available. Only the single revolution captured by DSS 14 yielded a backscatter anomaly in the original analysis by Nozette et al. (Science, 274, 1495-1498, 1996). A Stanford reanalysis focussed on the DSS 14 data in an attempt to reproduce the anomaly and highlight its characteristics, but no anomaly was found. The steps in the Stanford analysis are summarized below. Data Processing Path ------------------------ PREPMO and File Naming: Data flow at Stanford is summarized in the diagram below. ODR tapes (containing real 8-bit samples at 50000 samples/second) were preprocessed by program PREPMO, which separated header information from data samples and stored the data from the right-circularly polarized (RCP) channel and the left-circularly polarized (LCP) channel in different files as 16-bit integers. The output sample files from PREPMO have names of the form ydddhhmm.sPc where y is the year on which the file begins ddd is the day-of-year on which the file begins hh is the hour on which the file begins mm is the minute on which the file begins s indicates that data were collected at SPC s s=1 for DSS 14 4 for DSS 43 6 for DSS 63 P indicates that the generating program was PREPMO c indicates that the source is Channel c c=1 for S-band RCP 3 for S-band LCP For the ODR tape with data starting at 1994-04-09T18:36:45, the RCP and LCP output files from PREPMO were 40991836.1P1 and 40991836.1P3, respectively. A typical ODR spanned 16 minutes; each PREPMO output file contained on the order of 96 MB (48M samples * 2 bytes + 2048 bytes of header). Quick-Look Diagnostic Programs: PREPPOWER and PREPLOOK are programs which can give quick-look diagnostic views of the data emerging from PREPMO. PREPPOWER finds the mean and variance of the 16-bit samples, plotting the latter after averaging over one second intervals. PREPLOOK computes power spectra, averaging those over one minute intervals before plotting. PREPHIST is a third diagnostic program which plots the histogram of all 48M samples from a single channel. ODR Original data (on tape). 2 channels, | 50000 8-bit samples/sec on each. V -------- | PREPMO | Parse by channel, store in separate files -------- | --------- PREPLOOK (power spectra plots) and V | PREPLOOK| PREPOWER (average sample power vs time *.1P{1,3} -->| and | plots) used to find times at which signal | |PREPPOWER| levels changed. V --------- --------- | PREPFND |<-- FL14S{R,L}01.SPC Filter and decimate (1:1) with separate --------- equalizing filter for each channel. | Filters based on power spectra averaged V over 19:56:47 (71807 sec) to 20:01:51 *.PF{1,3} (72111 sec). | | --------- |--> | FND | Look at low-side window (bins 201-328 of | | (8,201) |-> *.F1{1,3} 1024) to monitor time variability of noise | --------- power (no carrier, no surface echo here) | | --------- |--> | FND | Look at high-side window (bins 701-828 of | | (8,701) |-> *.F4{1,3} 1024) to monitor time variability of noise | --------- power (no carrier, no surface echo here). V ------ Compensate for attenuator changes, | GAIN | spontaneous maser/amplifier gain changes; ------ calibrate RCP and LCP to same absolute | amplitude standard. V *.GN{1,3} | --------- |-->| FND | Look at low-side noise window after gain | | (8,201) |-> *.F2{1,3} correction; carrier moves into window at | --------- end of run. | | --------- |-->| FND | Use high-side window (bins 701-828 of | | (8,701) |-> *.F3{1,3} 1024) for tape-to-tape amplitude | --------- calibration. | V ------- Adjust frequency so that carrier is in | STEER |<--COEFFICIENT_SET_1 the center of the processing bandwidth ------- (at 12500 Hz). | V *.S1{1,3} | V ------- Adjust frequency by the difference | STEER |<--COEFFICIENT_SET_2 between carrier and South Pole echo ------- (South Pole bin is at 12500 Hz). | V *.S2{1,3} | V --------- Compute 1464 16384-point power spectra; | FNDLOOK | save bins 7357-8380 (South Pole is in --------- bin 8193). Input is from ODR covering | 1994-04-09T18:36:45 to 18:52:45. V FNDLOOK.MAT.S2{1,3} PREPFND: Program PREPFND can do several things. First, it converts real samples (stored as 16-bit integers) to double-precision complex samples. Second, it can digitally reduce the bandwidth represented by the samples. Third, it can compensate for filter characteristics in the receiver hardware by 'equalizing' (flattening) the filter response. Output from PREPFND is in a standard 'FND' format which is described in a separate document (file FNDO.TXT, also in this directory). All subsequent programs accept FND format files as input; most also produce output files in FND format. For the Clementine data, there was no bandwidth reduction at the PREPFND stage; but equalizing filters were applied to both channels. The filters were computed from receiver response when only noise was present at the input. Power spectra representations of the filters are included in the archive (files FL14SR01.SPC and FL14SL01.SPC in the CALIB directory); the corresponding voltage spectra with assumed linear phase dependence on frequency were used to normalize the data samples in PREPFND. PREPFND output files have names of the form ydddhhmm.PFc where the characters have the same meanings as above (and 'PF' denotes output from the PREPFND stage). RCP and LCP files starting 18:36:45 have names 40991836.PF1 and 40991836.PF3, respectively. With no bandwidth reduction each 96 MB PREPMO file expanded to 384MB at this stage (each pair of 16-bit real samples became one 128-bit complex sample). FND: FND is a general filter and decimation program. It accepts an input file in FND format and, based on user-suppled parameters, carries out digital bandwidth reduction first by filtering and then by shifting the remaining bandwidth to lower frequencies. In the diagram above, bins 201-328 in a 1024-point spectrum have been extracted to yield files ydddhhmm.F11 and ydddhhmm.F13 in the first case; bins 701-828 have been extracted to yield files ydddhhmm.F41 and ydddhhmm.F43 in the second case. In both, the bandwith has been reduced from the original 25000 Hz by a factor of 8 to 3125 Hz. The objective in the first case was to measure radiothermal noise below the Clementine signal(s); in the second case it was to measure noise on the high side. The 'clear-frequency' noise measurements were important inputs to the next processing stage. GAIN: GAIN adjusts the amplitude of the complex time samples based on a set of polynomial coefficients and times. For Clementine, all gain changes were assumed to be linear in voltage over the time interval specified. The RCP and LCP coefficients are stored in files G099C141.TAB and G099C143.TAB,respectively, in the CALIB directory of this archive. Gain changes could result from a number of factors including attenuator adjustments in the receiving system and slow drift in the masers. Noise levels varied depending on the pointing of the ground antenna and whether calibration sources were connected to the receiver inputs. Measurement of and correction for various gain changes are discussed elsewhere (file DSS14TSY.TXT, also in this directory; and in Simpson and Tyler, J. Geophys. Res., 104, 3845-3862, 1999). Output files from the GAIN stage have file names of the form ydddhhmm.GNc. Output could be used as input to FND (with filters on both the high and low sides of the carrier) to check whether gain corrections had been accomplished successfully. The gain calibration results are stored in file TABLE2.TAB in the CALIB directory of this archive. STEER: STEER adjusts phase of each double precision complex sample to account for deterministic and other Doppler-like effects. Although it would have been possible to compute exact Doppler shifts expected from motion of the spacecraft and ground receiver, the stability of the onboard oscillator was questionable. Instead, the carrier frequency was measured from the data, and COEFFICIENT_SET_1 was used to center the carrier empirically. The oscillator noise and the errors expected from this procedure are discussed in detail in Simpson and Tyler, J. Geophys. Res., 104, 3845-3862, 1999. The coefficients are stored in file STEERCF1.TAB in the CALIB directory of this archive. Output files from first-order 'steering' have names of the form ydddhhmm.S1c. For the data begining at 18:36:45 these are 40991836.S11 and 40991836.S13 for RCP and LCP, respectively. The expected separation in frequency between the carrier and the echo from the lunar South Pole was computed and used to place the South Pole frequency bin at 12500 Hz. The program STEER was run a second time for this step; the input file of coefficients is shown in the diagram above as COEFFICIENT_SET_2. COEFFICIENT_SET_2 is stored under file name DF2SCM.TAB in the CALIB directory of this archive. Output files from second-order steering have names of the form ydddhhmm.S2c. Using these RCP and LCP files (40991836.S21 and 40991836.S23, respectively) it was then possible to measure the echo power from strips on the lunar surface with known frequency offsets from the South Pole. FNDLOOK: FNDLOOK computes power spectra from FND format time samples. In this case 16384-point spectra were computed from the second-order steer output -- one spectrum every 0.65536 sec with frequency resolution of approximately 1.5259 Hz per bin. By extracting bins 7356-8379, the user is left with an array of 1464 1024-point spectra where the South Pole is in bin 838. This frequency range captures the majority of the surface echo and includes over 100 bins on the high side of the Pole (beyond the limb) with only system noise. The 1464 time steps cover a bit less than 960 seconds of data acquistion beginning approximately 18:36:45. The bistatic angle is minimum at the South Pole at 18:46:36.5 (time step 971). The RCP array is stored Geometrical Calculations ------------------------ To locate the South Pole (and other surface points) in frequency, it was necessary to compute the observing geometry using the spacecraft orbit solution derived by the Geodynamics Group at Goddard Space Flight Center. Their solution was converted to the JPL Navigation and Ancillary Information Facility standard 'text' transfer format by NAIF and delivered to Stanford. The diagram and text below summarize this aspect of the calculations. BSRTRAJ computes state vectors (position and velocity) for Earth, Moon, and spacecraft as a function of time; these are listed in a standard format which can be read by other programs. BSRGEOM computes an extensive set of parameters from the BSRTRAJ output including vectors from the center of the Moon to the DSN receiver, the spacecraft, and the target point (South Pole). The angles of incidence and reflection at the specular point, its latitude and longitude, and the sensitivity of these quantities to small changes in radius are also computed. An abbreviated output file (BSRGEOM.TAB) is included in the GEOMETRY directory of this archive; a second version of the output file retaining full numerical precision was used for subsequent calculations. SPK94099.TSP Ephemeris file, based on orbit solution | from GSFC, reformatted by JPL/NAIF. V --------- | BSRTRAJ | Generate state vector file --------- | V BSRTRAJ.OUT | V --------- Calculate bistatic radar geometry. | BSRGEOM |-->BSRGEOM.TAB BSRGEOM.TAB is for archive (readable). --------- BSRGEOM.OUT has full precision, for | additional calculations V BSRGEOM.OUT------ | | | V | -------- Calculates spacecraft Doppler, Doppler | | BSRDOP | for South Pole, and their difference. | -------- Differences can be used to create | | coefficients for second-order steering. | | | V | BSRDOP.OUT-->COEFFICIENT_SET_2 V --------- For an orthographic projection of the | BSRGRID | Moon's southern hemisphere, computes --------- geometrical and scattering parameters | in grids of 700x700 points. V *.IMG BSRDOP uses the geometrical information in BSRGEOM.OUT to calculate Doppler shift of the spacecraft carrier (as observed at the DSN station) and the Doppler shift for one of the other 'points' calculated by BSRGEOM. In this case, the (fixed target) South Pole was used. The Doppler differences were then converted to a table of piecewise-linear polynomial coefficients which could be used to 'steer' the South Pole frequency bin to the center frequency at 12500 Hz. BSRGRID also uses the geometrical information in BSRGEOM.OUT, but calculates geometrical and other values on a grid tangent to the lunar surface at the South Pole. The grid has 700x700 points with 5 km spacing; the South Pole is at the center of the grid +/-2.5 km in (x,y) from each of four grid points. Each grid point is associated with a surface point from the lunar southern hemisphere projected orthographically onto the grid. Values at grid points allow computation of the bistatic radar equation for the surface. Grid values are stored in 700x700 arrays with file names of the form sssssppp.IMG, where sssss is the time in seconds UT from 0h at the receiving station on 1994-04-09 and ppp is defined in the table below. For example, file 67596BET.IMG gives the bistatic angle at 18:46:36 Earth Receive Time (ERT) on 1994-04-09. ppp Parameter Definition for the Surface Point Associated with (x,y) --- ---------------------------------------------------------------- BET bistatic angle (degrees) DAR area (square meters) DBR offset from boresight at the (DSN) receiving antenna (degrees) DPR incremental received power (watts) from the surface element FQZ Doppler frequency relative to the South Pole (Hz) GAM tilt of the surface element (degrees) GRX receiving antenna gain at the surface element (dB) GTX spacecraft transmitting antenna gain at the surface element (dB) RRX distance from the receiving antenna (meters) RTX distance from the spacecraft transmitting antenna (meters) SBR offset from boresight at the spacecraft antenna (degrees) SG0 assumed specific radar cross section for calculating DPR THI incidence angle at the surface element (degrees) THS scattering angle at the surface element (degrees) VAL validity mask (1 = grid point within lunar disk; 0 = outside) DFQ time derivative of FQZ (Hz/sec) Arrays have been calculated and stored at time steps of 100 seconds between 18:14:56 (65696 s) and 19:51:36 (71496 s) in the IMG directories in the archive. Since the DFQ array is found from differencing two FQZ arrays, there is no DFQ array at either 18:14:56 or 19:51:36. DPR values are evaluations of the radar equation using cross sections stored in the SG0 array, which was set to the (unrealistic) value of unity. The VAL array may be used as a mask to screen out grid points which do not correspond to a surface point. Although grids were computed only every 100 seconds, most parameters vary smoothly and linear interpolation between grids provides good accuracy. Sorting ------- Each FQZ grid provides a mapping between surface location and frequency in the southern hemisphere, with the South Pole as the reference (f=0). The mapping varies with time. For times near the South Pole backscatter condition (1994-04-09T18:46:36.5) contours of constant frequency map onto the surface mutally visible to spacecraft and receiver as 'horseshoes'. Each point in the spectra generated by FNDLOOK can be associated with one of these truncated annuli. The objective in the Clementine bistatic experiment was to seek an enhancement in near-backscattered power and/or a change in echo polarization around the time when the bistatic angle at the South Pole passed through zero. Contours of constant bistatic angle (BET) do not align well with the frequency contours, however; establishing an algorithm to sort the data for this dependence is not straightforward. The Stanford reanalysis (Simpson and Tyler, J. Geophys. Res., 104, 3845-3862, 1999) sought time vs. frequency correlations in the data for individual and aggregated frequency bins. None was found. An alternative approach assumes that a backscatter enhancement at an individual point will raise the echo power (or change the polarization) in a single frequency bin regardless of the bistatic angles elsewhere on that Doppler contour. Then, the locus of points for which the bistatic angle equals zero can be used to sort the data. Each of these 'target' points satisfies the condition BET=0 at some time during the experiment; a scattering function can be constructed for each target covering a range of bistatic angles (including 0). By adopting the convention that the bistatic angle is negative prior to BET=0, one can obtain a smoothly varying function which shows both dependence on bistatic angle and on time. The individual functions for each target may then be aggregated to improve signal-to-noise ratio depending on criteria developed elsewhere. The file BETAZER0.TAB in the GEOMETRY directory of this archive gives the location of BET=0 target points at one second intervals. BETAZER1.TAB in the same directory provides the same information, but with coordinates given in degrees rather than radians. Both files list 72 target points, with the last being 6.5 seconds after BET=0 at the South Pole. Sorting then reduces to associating each frequency bin in a spectrum generated by FNDLOOK with the closest target point, finding the actual BET angle at that time by interpolating among grids, and storing the echo power in a new array indexed by BET over the range -5 to +5 degrees in steps of 0.1 degree. The procedure must be followed for all 1024 frequency bins and for both RCP and LCP data. Many echoes are from surface elements where the |BET| < 5 degree condition is not met. But as many as 42 measurements were captured in some bins of some functions. The values, which are noisy since the spectra were not averaged, can be handled in several ways. On a point-by-point basis (for each of the 72 targets and each of the 101 BET values) these include computing: A. average values in each polarization, then ratio of averages B. medians in each polarization, then ratio of medians C. ratio RCP/LCP, then average of ratios D. ratio RCP/LCP, then median of ratios The results of the average, median, and/or ratio calculations are 72 functions giving echo power as a function of bistatic angle. If one chooses, the functions can be aggregated according to the area contributing. By comparing Figure 4 in Nozette et al (1996) with the Doppler contours at 18:46:36.5, one can estimate that target numbers 58-70 approximately correspond to the coverage used in their analysis. The scattering functions in RATIOFNS.TAB in the SORT directory are the result of processing the data according to (A) and (D) above, aggregated over those 13 targets.