PDS_VERSION_ID = PDS3 RECORD_TYPE = STREAM OBJECT = TEXT INTERCHANGE_FORMAT = ASCII PUBLICATION_DATE = 2019-06-01 NOTE = "Description for Venus Climate Orbiter (VCO, also known as PLANET-C and AKATSUKI) IR1 data calibration." END_OBJECT = TEXT END IR1 data calibration ==================== 1. Definition of region A, B, C, D ---------------------------------- The following descriptions are in the full-size image coordinate, i.e., after vertical flipping of the raw image data. A: lower-left quadrant, [ 1, 512]x[ 1, 512] B: lower-right quadrant, [ 513, 1024]x[ 1, 512] C: upper-left quadrant, [ 1, 512]x[ 513, 1024] D: upper-right quadrant, [ 513, 1024]x[ 513, 1024] (1,1024) (512,1024) (513,1024) (1024,1024) +-----------------++-----------------+ | || | | || | | || | | C || D | | || | | || | | (512,513)||(513,513) | (1,513) +-----------------++-----------------+ (1024,513) (1,512) +-----------------++-----------------+ (1024,512) | (512,512)||(513,512) | | || | | || | | A || B | | || | | || | | || | Y +-----------------++-----------------x ^ (1,1) (512,1) (513,1) (1024,1) | +---> X 2. Smear noise reduction ------------------------ We estimated the non-uniform offset noise, smear in raw image, assuming that it is caused by weak photo-sensitivity of vertical transfer device in the sensor unit and its level is constant in each vertical line (column) in each quadrant. For example, the offset noise level in the i-th column in region A, N_A(i), is calculated as follows, 512 N_A(i) = C_A * Sum s_in(i, j) j=1 where C_A is the smear noise coefficient for region A, s_in(i, j) is the input signal of the pixel (i, j). The output signal s_out(i, j) is expressed as s_in(i, j) + N_A(i). Thus, the input signal (after subtraction of smear noise) is expressed as 512 Sum s_out(i, j) j=1 s_in(i, j) = s_out(i, j) - C_A * ------------------- (1 + 512 * C_A) If s_out(i, j) has dead pixel flag value, saturated pixel flag value, or missing value, processing is different in level 2b (l2b) and in level 2c (l2c). In level 2b (l2b), if s_out(i, j) has a special flag value, all the value s_in(i, *) within the quadrant are filled with the missing value specified by value of P_MPIXV keyword. In level 2c (l2c), if s_out(i, j) has a special flag value, the value of s_out(i, j) for the calculation of amount of smear will be estimated from the values at the pixels around j in the column i, instead of using s_out(i, j) as is. If both side of the unavailable pixel(s) are available and exist within the same quadrant, value of s_out(i, j) will be estimated by linear interpolation using values of s_out(i, j0) and s_out(i, j1) at both side of the pixels, where j0 < j < j1. If only value of one side pixel is available (i.e., within the same quadrant), unavailable pixel value for the calculation will be filled by the available pixel value s_out(i, jj), where jj < j or j < jj. Note that if s_out(i, j) has a special flag value at pixel, the resultant image also has a special flag value at the pixel. The estimated values are used only for the estimate of the amount of smear. The smear noise coefficients for A, B, C, and D region are listed in Table 1. Note that the values for the dayside filter with exposure time of 7.833 s are different from those for the nightside filters with exposure time of 30.833 s. Table 1. Smear correction coefficients v0.1 C_A C_B C_C C_D ---------- ---------- ---------- ---------- Dayside 0.0017274 0.0017215 0.0017316 0.0017838 Nightside 0.00066193 0.00066193 0.00071513 0.00071513 NOTE: We can make dayside cloud distribution clearer, but until now, we cannot explain why smear noise level in raw image is not uniform in the column. The version of smear correction coefficients is recorded in the FITS header keyword, I1_SCVER as string. The values of C_A, C_B, C_C, and C_D are also recorded in the FITS header keywords, I1_SCF00, I1_SCF10, I1_SCF01, and I1_SCF11, respectively. 3. Flat field correction ------------------------ When we take a long exposure image of dayside of Venus near the periapsis, fine structure is smoothed because of the high ground (cloud top) velocity. We used 4 images ir1_20160303_213458_09d_l1b_v10.fit, ir1_20160813_200607_09d_l1b_v10.fit, ir1_20160916_003208_09d_l1b_v10.fit, ir1_20160916_012708_09d_l1b_v10.fit for flat field correction. The average of four images after smear correction was used. NOTE: The four images are away from "identical". The variation is ~25% at maximum. The sensitivity variation within the several pixels can be corrected by this method, but sensitivity variation in the several tens to hundreds pixels is possibly incorrect. The flat field correction is done by dividing data by the flat field except for the pixel with special flag value. The filename of the flat field used for the correction is recorded in the FITS header keyword, I1_FLAT. 4. Boundary correction ---------------------- Even after flat field correction, discontinuity at the boundaries of A, B, C, and D regions can be seen in several images. This shows our methods above is imperfect. This boundary discontinuity may be caused by the difference of property (e.g., gain) among 4 readout devices and incompleteness of smear noise reduction. The sensitivity correction for four regions should be done to reduce the discontinuity at the boundaries. For dayside images, the following sensitivity correction is applied. The correction coefficient of A-B boundary, R_AB, is expressed as R_AB = ( 1.5*S(512) - 0.5*S(511) ) / ( 1.5*S(513) - 0.5*S(514) ) where 512 S(i) = Sum s(i, j). j=1 In the right hand side of the above equation, the numerator is linearly extrapolated value at i=512.5 using values at i=511 and i=512. The denominator is linearly extrapolated value at i=512.5 using values at i=513 and i=514. If there are any special value in s(511, j), s(512, j), s(513, j), and s(514, j), these values are not added to the sums, S(511), S(512), S(513), and S(514). To avoid using values at deep space, threshold value v_thd is introduced. If values s(511, j), s(512, j), s(513, j), and s(514, j) after smear noise reduction are greater than v_thd, these values are added to the sums. If not, these values are not added to the sums. Currently, the threshold value v_thd is set to 200 in counts. In the case that R_AB is less than 0.5 or more than 2.0, calculation of sensitivity correction factor is identified as failed, and the value of R_AB is set to 1. If one of the value s(511, j), s(512, j), s(513, j), and s(514, j) has been calculated using estimated amount of smear, these values are not added to the sums for keeping sensitivity corrected data reliable. This case is only appeared in L2c. The other sensitivity factors, R_CD, R_AC, and R_BD are also estimated in the same way. There are four sensitivity correction factors, R_AB, R_CD, R_AC, and R_BD. However, only three factors are needed in this algorithm. The application strategy of sensitivity factors is as follows: - If all sensitivity factors are available, using the sensitivity factor associated with the darkest boundary is avoided. The left three sensitivity factors are used with a unique possible combination. - If a sensitivity factor for a boundary cannot be calculated, using the sensitivity factor associated with the boundary is avoided. The left three sensitivity factors are used with a unique possible combination. - If more than one sensitivity factors are unavailable, the only available sensitivity factors are used as good as possible for sensitivity correction. Which sensitivity factors are applied are recorded in the FITS header keywords, I1_QC_0X (left two quadrants), I1_QC_1X (right two quadrants), I1_QC_X0 (lower two quadrants), and I1_QC_X1 (upper two quadrants). For nightside images, no sensitivity correction is applied at now. The values of sensitivity correction coefficient for each quadrant, 1.0 (lower-left), R_AB (lower-right), R' (upper-left), and R'*R_CD (upper-right), are also recorded in the FITS header keywords, I1_QCF00, I1_QCF10, I1_QCF01, and I1_QCF11, respectively. 5. Sensitivity calibration -------------------------- Sensitivity calibration was performed using the image of epsilon, lambda, delta, tau, sigma, and zeta Sgr, obtained in February and September in 2016. We estimated the irradiance at the wavelengths of 0.90, 0.97, and 1.01 um from the visual magnitudes and the temperatures with black body assumption (Table 2), and calculated sensitivity coefficients in mW/cm**2/um/(ADU/s). Table 2. List of stars for calibration star m_v T (K) I_0.9n* I_0.97* I_1.01* ----------- ----- ----- ------- ------- ------- Sgr epsilon 1.85 9960 180.4 143.7 126.8 lambda 2.82 4770 199.4 176.4 164.0 delta 2.668 4192 304.4 277.7 262.2 tau 3.31 4459 146.5 131.6 123.3 sigma 2.067 20370 100.4 76.8 66.4 zeta 2.585 8799 102.4 82.5 73.2 Note: * in 10**-12 mW/cm**2/um. m_v is the apparent visual magnitude. The observations of stars with the 0.90-um nightside, the 0.97-um, and the 1.01-um filters were performed three times on February 24, and five times on September 9. Some stars in some images were not used because S/N is low (Table 3). Table 3. List of data used for calibration 2016-02-24 2016-09-09 ---------- ------------------ star filter 14 16 18 09 12 15 18 21 ----------- ------ -- -- -- -- -- -- -- -- Sgr epsilon 0.9n x x x x x x x x 0.97 x x x x x x x x 1.01 x x x x x lambda 0.9n x x x x x x x x 0.97 x x x x x x x x 1.01 x x x x x x delta 0.9n x x x x x x x x 0.97 x x x x 1.01 x x x x x tau 0.9n x x x x x x 0.97 x x x x x x x x 1.01 x x sigma 0.9n x x x x x x 0.97 x x x x x x x x 1.01 x x x zeta 0.9n x x x x x x x x 0.97 x x x x x x x x 1.01 x x x The sensitivity coefficients for the three filters were estimated by weighted mean considering shot noise. The one for the 0.90-um dayside filter was estimated considering the difference of spectral transmittance from the 0.90-um nightside filter. As a result, the sensitivity coefficients for the four filters were estimated as follows: filter sensitivity coefficients --------------- ------------------------------------- 0.90 um dayside 61.7 +/- 4.7 uW/cm2/um/sr/(ADU/s) 0.90 um nightside 75.6 +/- 5.8 nW/cm2/um/sr/(ADU/s) 0.97 um 608 +/- 70 nW/cm2/um/sr/(ADU/s) 1.01 um 1.35 +/- 0.91 uW/cm2/um/sr/(ADU/s) Note: The S/N of star observation result with the 1.01-um filter is apparently low. We may miss something in error estimation, and probably the reliability of the value especially for 1.01 um. The conversion method from ADU counts into radiance is briefly recorded in the FITS header keyword I1_C2F as string. The coefficients used to convert ADU counts into radiance are also recorded in the FITS header keywords, I1_C2FK1 and I1_C2FK0.