Summary of Stardust NAVCAM Data Calibration Procedure

The Stardust NAVCAM data calibration procedure takes the EDR (raw) Data Number (DN) images and converts them to radiance (or just to data calibrated DN if the exposure duration is zero).

The procedure comprises several steps summarised in the table below. The procedure uses information from the EDR image label as well as from ancillary sources such as calibration files and model parameters.

See also Figure 1 below.

The procedure also appends quality, signal-to-noise and uncertainty maps as PDS OBJECTS to the data-calibrated output image in the output file.

Refer to [KLAASENETAL2013] for details.

Step Mnemonic Description Calibration info or file(s) (see data set directory CALIB/) Uncertainty

MASK Flag any pixels that were outside WINDOWs and/or bad and/or missing. The flags are stored as bits in the QULMAP_IMAGE OBJECT. Flagged pixels are not processed any further in the data calibration process.
Pixels outside windows were never downlinked.
Bad pixels are identified in bad pixel calibration file.
Missing pixels will have a raw value of zero
CALIB/NCBADP.LBL - list of bad pixels -

SATU Flag saturated pixels based on raw Data Number (DN) value (255 if compressed; 4095 if uncompressed). Also flag pixels that are above or to the right of the saturated pixels as potentially pixels into which excess signal has bled. The flags are stored as bits in the QUALITY_MAP OBJECT. - -

DCMP For uncompressed images do nothing. For compressed images, invert on-board 12- to 8-bit NAVCAM compression lookup table: replace 8-bit raw DN with center of corresponding 12-bit bin from lookup table. Save the bin size for each pixel for NOISe step (below).
The instrument calibration is based on 12-bit DNs from the raw data.
The detector measures data as 12-bit DNs, but the instrument firmware may be commanded to compress those data, by representing multiple contiguous 12-bit values (composing a "bin") as a single 8-bit value, before on-board storage and telemetering of the 8-bit values to Earth.
Note that this represents a loss of information in DN resolution, mitigated by a non-linear, near-square root correlation between measured 12-bit DN magnitude and bin size, which is why the bin size is carried forward to the NOISe step.
CALIB/NC_COMPR.LBL - compression lookup table in on-board NAVCAM ROM (Read-Only Memory).
This table provides the mapping, performed on-board by the NAVCAM firmware, between measured 12-bit DN and 8-bit DN stored in the raw file.
-

BIAS Calculate bias and subtract from DNs. Use the first of the following methods which succeeds.
Use resistant mean last three columns of BaseLine Stabilization (BLS) pixels from overscan pixels saved in BLSIMG_IMAGE OBJECT in EDR (raw) data file for current image. This method only works if current image is a full-frame image; this method fails if current image is a WINDOWed image because BLS pixels are not stored or downlinked for windowed images.
Interpolate between two full-frame images taken within two days before and after current image. Use temperature model correction to compensate for any temperature difference between full frame images and current image. If two such images do not exist, or if BLS pixels are otherwise missing.
Extrapolate using time model from the last time the NAVCAM heater was turned off. Time model assumes a nominal temperature; make same temperature correction as for previous method.

Temperature bias model: -3.5 DN/degC
For temperature, use value of FOCAL_PLANE_TEMPERATURE keyword from PDS label, or of FOPLTEMP keyword from FITS header.

Time bias model: 20.435 * LN(heater off, d) + 427.53DN
Time bias model assumes nominal temperature of 240.795K
Table of NAVCAM events: CALIB/NAVCAM_ACTIVITY_LOG.LBL
Heater on/off events are called ANNEAL events in that log.
Time bias model assumes minimum time of 0.1d, maximum time of 100d.

Method 1 - 0DN
Method 2 - 10DN
Method 3 - 30DN or 50DN for heater off time less than or greater than, respectively, 2 days.

NOIS Calculate sum-squared NOISe from BIAS-subtracted DNs. Noise will be divided into BIAS- and DARK-subtracted DNs to make Signal-to-Noise Ratio (SNR) map, SNRMAP_IMAGE OBJECT. NOISe units are DN-squared.
Quantization noise = Quant*Quant/12
- Quant = 1 for uncompressed images.
- Quant = bin size for compressed images (cf. DCMP step above).
Shot noise = DN / 25
Read noise = 3.2 * 3.2 = 10.24
Coherent Noise and Fixed Pattern Noise (FPN) are neither estimated for, nor included in, in the SNR map. Refer to [KLAASENETAL2013] for further, if limited, discussion of coherent noise and FPN.
-

DARK Calculate dark current for and subtract from current image DNs. Dark current builds up continuously from end of previous CCD read and through EXPOSURE_DURATION of current image. Get time (MET) of previous NAVCAM read event from table of NAVCAM events. Dark current rate model is K * t^λ, DN/s; t is time since previous read in seconds.
Table of NAVCAM events: CALIB/NAVCAM_ACTIVITY_LOG.LBL
K, λ after 2009-01-01:
3.057E-13, 1.065E-01
K, λ after launch:
4.411E-11, 8.879E-02
Choose latest date for K, λ.
Twice the dark model estimate.

BDFX Bias- and Dark-Fix. Examine median of result of previous step to see if median is less than zero. If it is, add some DN back to the image to make that median zero. When calculating this median, exclude the N brightest pixels where the area of N pixels is equivalent to a circle of a given radius at the range of the comet. N will be zero for all but the encounter flyby images. Nominal radius to use: 3.5km (9P/Tempel 1); is big enought to include some coma. -

SNRM Calculate SNR Map for SNRMAP_IMAGE OBJECT. Divide DNs from BDFX step by result of NOISe step. - -

FLAT Divide by flat-field image to remove pixel-to-pixel variation. CALIB/NAVCAMNORMPRLSLOPE_TESTSL1.LBL - flat-field file -

RATE Divide DN from output of FLAT step by per-pixel exposure duration to get DN rate, DN/ms. This step is not performed if current image is a bias frame i.e. if the nominal exposure time is zero. Nominal exposure duration is in keyword EXPOSURE_DURATION, in ms, to which a per-pixel offset is added to get the actual exposure duration for each pixel. The per-pixel exposure offset, in ms, is a polynomial function of image line number, L, and shutter blade polarity: forward=FWD; backward=BCK; unknown=UNK. Shutter blade polarity is set to FWD at NAVCAM power-on, and toggles between FWD and BCK for every non-bias frame. See DOCUMENT/NAVCAM_SHUTTER_POLARITY.LBL for more details.
The shutter polarity is implicit in the PDS lable keyword SHUTTER_OFFSET_FILE_NAME for calibrated (RDR) data; the value of the keyword will be either NC_FWD_SHUTTER.LBL or NC_BCK_SHUTTER.LBL for FWD and BCK polarity, respectively. This keyword will be absent in raw (EDR) data; it will also be absent if the shutter polarity either is BIAS (zero-exposure) or is unknown.
Table of NAVCAM events for determining shutter blade polarity: CALIB/NAVCAM_ACTIVITY_LOG.LBL
The polynomials below model the per-row exposure offset; the resulting value units are ms.
L is pixel row offset from the bottom (first) row in an image; the range of L is 0 to 1023, inclusive.
The polynomials are evaluated for each row and time period in the CALIB/ products with PDS labels CALIB/NC_FWD_SHUTTER.LBL and CALIB/NC_BCK_SHUTTER.LBL.

FWD polynomial after 2010-08-01:
+ 1.524E-17 * L⁶
- 5.294E-14 * L⁵
+ 7.216E-11 * L⁴
- 4.828E-08 * L³
+ 1.678E-05 * L²
- 2.275E-03 * L¹
- 4.521E-01

BCK polynomial after 2010-08-01:
+ 6.556E-13 * L⁴
- 1.885E-09 * L³
+ 2.271E-06 * L²
- 1.289E-03 * L¹
+ 1.667E+00

FWD polynomial after 1998-04-01:
+ 1.525e-17 * L⁶
- 5.294e-14 * L⁵
+ 7.216e-11 * L⁴
- 4.828e-08 * L³
+ 1.683e-05 * L²
- 2.308e-03 * L¹
- 5.732e-01

BCK polynomial after 2010-08-01:
+ 6.556e-13 * L⁴
- 1.885e-09 * L³
+ 2.124e-06 * L²
- 1.073e-03 * L¹
+ 1.590e+00
If shutter blade polarity is unknown, the per-pixel exposure offset is zero.
0.1ms in pixel exposure duration

ABSC Convert from the DN rate (DN/ms) output of the previous step to radiance. Also save the I/F coefficient.
After 2011-02-11:
Radiance=2.01E-9 (W/(cm^2*nm*sr))/(DN/ms);
I/F=4.05E-5 (I/F at 1AU)/(DN/ms).
After 1998-04-01:
Radiance=1.93E-9 (W/(cm^2*nm*sr))/(DN/ms);
I/F=3.89E-5 (I/F at 1AU)/(DN/ms).
If the scan mirror angle is less than 17deg, then 100%, else zero.

Step Mnemonic	Description	Calibration info or file(s) (see data set directory CALIB/)	Uncertainty
MASK	Flag any pixels that were outside WINDOWs and/or bad and/or missing. The flags are stored as bits in the QULMAP_IMAGE OBJECT. Flagged pixels are not processed any further in the data calibration process. Pixels outside windows were never downlinked. Bad pixels are identified in bad pixel calibration file. Missing pixels will have a raw value of zero	CALIB/NCBADP.LBL - list of bad pixels	-
SATU	Flag saturated pixels based on raw Data Number (DN) value (255 if compressed; 4095 if uncompressed). Also flag pixels that are above or to the right of the saturated pixels as potentially pixels into which excess signal has bled. The flags are stored as bits in the QUALITY_MAP OBJECT.	-	-
DCMP	For uncompressed images do nothing. For compressed images, invert on-board 12- to 8-bit NAVCAM compression lookup table: replace 8-bit raw DN with center of corresponding 12-bit bin from lookup table. Save the bin size for each pixel for NOISe step (below). The instrument calibration is based on 12-bit DNs from the raw data. The detector measures data as 12-bit DNs, but the instrument firmware may be commanded to compress those data, by representing multiple contiguous 12-bit values (composing a "bin") as a single 8-bit value, before on-board storage and telemetering of the 8-bit values to Earth. Note that this represents a loss of information in DN resolution, mitigated by a non-linear, near-square root correlation between measured 12-bit DN magnitude and bin size, which is why the bin size is carried forward to the NOISe step.	CALIB/NC_COMPR.LBL - compression lookup table in on-board NAVCAM ROM (Read-Only Memory). This table provides the mapping, performed on-board by the NAVCAM firmware, between measured 12-bit DN and 8-bit DN stored in the raw file.	-
BIAS	Calculate bias and subtract from DNs. Use the first of the following methods which succeeds. Use resistant mean last three columns of BaseLine Stabilization (BLS) pixels from overscan pixels saved in BLSIMG_IMAGE OBJECT in EDR (raw) data file for current image. This method only works if current image is a full-frame image; this method fails if current image is a WINDOWed image because BLS pixels are not stored or downlinked for windowed images. Interpolate between two full-frame images taken within two days before and after current image. Use temperature model correction to compensate for any temperature difference between full frame images and current image. If two such images do not exist, or if BLS pixels are otherwise missing. Extrapolate using time model from the last time the NAVCAM heater was turned off. Time model assumes a nominal temperature; make same temperature correction as for previous method.	Temperature bias model: -3.5 DN/degC For temperature, use value of FOCAL_PLANE_TEMPERATURE keyword from PDS label, or of FOPLTEMP keyword from FITS header. Time bias model: 20.435 * LN(heater off, d) + 427.53DN Time bias model assumes nominal temperature of 240.795K Table of NAVCAM events: CALIB/NAVCAM_ACTIVITY_LOG.LBL Heater on/off events are called ANNEAL events in that log. Time bias model assumes minimum time of 0.1d, maximum time of 100d.	Method 1 - 0DN Method 2 - 10DN Method 3 - 30DN or 50DN for heater off time less than or greater than, respectively, 2 days.
NOIS	Calculate sum-squared NOISe from BIAS-subtracted DNs. Noise will be divided into BIAS- and DARK-subtracted DNs to make Signal-to-Noise Ratio (SNR) map, SNRMAP_IMAGE OBJECT. NOISe units are DN-squared.	Quantization noise = QuantQuant/12 - Quant = 1 for uncompressed images. - Quant = bin size for compressed images (cf. DCMP step above). Shot noise = DN / 25 Read noise = 3.2 3.2 = 10.24 Coherent Noise and Fixed Pattern Noise (FPN) are neither estimated for, nor included in, in the SNR map. Refer to [KLAASENETAL2013] for further, if limited, discussion of coherent noise and FPN.	-
DARK	Calculate dark current for and subtract from current image DNs. Dark current builds up continuously from end of previous CCD read and through EXPOSURE_DURATION of current image. Get time (MET) of previous NAVCAM read event from table of NAVCAM events. Dark current rate model is K * t^λ, DN/s; t is time since previous read in seconds.	Table of NAVCAM events: CALIB/NAVCAM_ACTIVITY_LOG.LBL K, λ after 2009-01-01: 3.057E-13, 1.065E-01 K, λ after launch: 4.411E-11, 8.879E-02 Choose latest date for K, λ.	Twice the dark model estimate.
BDFX	Bias- and Dark-Fix. Examine median of result of previous step to see if median is less than zero. If it is, add some DN back to the image to make that median zero. When calculating this median, exclude the N brightest pixels where the area of N pixels is equivalent to a circle of a given radius at the range of the comet. N will be zero for all but the encounter flyby images.	Nominal radius to use: 3.5km (9P/Tempel 1); is big enought to include some coma.	-
SNRM	Calculate SNR Map for SNRMAP_IMAGE OBJECT. Divide DNs from BDFX step by result of NOISe step.	-	-
FLAT	Divide by flat-field image to remove pixel-to-pixel variation.	CALIB/NAVCAMNORMPRLSLOPE_TESTSL1.LBL - flat-field file	-
RATE	Divide DN from output of FLAT step by per-pixel exposure duration to get DN rate, DN/ms. This step is not performed if current image is a bias frame i.e. if the nominal exposure time is zero. Nominal exposure duration is in keyword EXPOSURE_DURATION, in ms, to which a per-pixel offset is added to get the actual exposure duration for each pixel. The per-pixel exposure offset, in ms, is a polynomial function of image line number, L, and shutter blade polarity: forward=FWD; backward=BCK; unknown=UNK. Shutter blade polarity is set to FWD at NAVCAM power-on, and toggles between FWD and BCK for every non-bias frame. See DOCUMENT/NAVCAM_SHUTTER_POLARITY.LBL for more details.	The shutter polarity is implicit in the PDS lable keyword SHUTTER_OFFSET_FILE_NAME for calibrated (RDR) data; the value of the keyword will be either NC_FWD_SHUTTER.LBL or NC_BCK_SHUTTER.LBL for FWD and BCK polarity, respectively. This keyword will be absent in raw (EDR) data; it will also be absent if the shutter polarity either is BIAS (zero-exposure) or is unknown. Table of NAVCAM events for determining shutter blade polarity: CALIB/NAVCAM_ACTIVITY_LOG.LBL The polynomials below model the per-row exposure offset; the resulting value units are ms. L is pixel row offset from the bottom (first) row in an image; the range of L is 0 to 1023, inclusive. The polynomials are evaluated for each row and time period in the CALIB/ products with PDS labels CALIB/NC_FWD_SHUTTER.LBL and CALIB/NC_BCK_SHUTTER.LBL. FWD polynomial after 2010-08-01: + 1.524E-17 * L⁶ - 5.294E-14 * L⁵ + 7.216E-11 * L⁴ - 4.828E-08 * L³ + 1.678E-05 * L² - 2.275E-03 * L¹ - 4.521E-01 BCK polynomial after 2010-08-01: + 6.556E-13 * L⁴ - 1.885E-09 * L³ + 2.271E-06 * L² - 1.289E-03 * L¹ + 1.667E+00 FWD polynomial after 1998-04-01: + 1.525e-17 * L⁶ - 5.294e-14 * L⁵ + 7.216e-11 * L⁴ - 4.828e-08 * L³ + 1.683e-05 * L² - 2.308e-03 * L¹ - 5.732e-01 BCK polynomial after 2010-08-01: + 6.556e-13 * L⁴ - 1.885e-09 * L³ + 2.124e-06 * L² - 1.073e-03 * L¹ + 1.590e+00 If shutter blade polarity is unknown, the per-pixel exposure offset is zero.	0.1ms in pixel exposure duration
ABSC	Convert from the DN rate (DN/ms) output of the previous step to radiance. Also save the I/F coefficient.	After 2011-02-11: Radiance=2.01E-9 (W/(cm^2nmsr))/(DN/ms); I/F=4.05E-5 (I/F at 1AU)/(DN/ms). After 1998-04-01: Radiance=1.93E-9 (W/(cm^2nmsr))/(DN/ms); I/F=3.89E-5 (I/F at 1AU)/(DN/ms).	If the scan mirror angle is less than 17deg, then 100%, else zero.

Figure 1: NAVCAM data calibration processing logic flow from [KLAASENETAL2013]. (file nc_cal_overview_fig1.png)