[Top] [Prev] [Next] [Bottom]
36.2 Calibration Steps Explained
Each calibration step is described in this section along with the corresponding calibration switch and various reference files required by that step. A flowchart of the process is illustrated in Figure 36.1.
Data Quality Initialization (DQI_CORR)
This step applies data quality initialization using the reference file dqihfile, which contains a data quality flag for each diode. Each data quality flag in the .q0h file (Table 37.10) is compared with the corresponding flag in the data quality initialization file (.r5h), and the most severe flag is kept. Quality flags are not additive and are never decreased. The most severe data quality flags are written to the output file (.cqh). Table 37.5 defines these flags (by order of decreasing -severity).
Conversion to Count Rates (EXP_CORR)
This step converts the input data to count rates by dividing by the exposure times. The exposure time is computed for each bin as:
Diode Response Correction (DIO_CORR)
The diodes within the Digicons of the GHRS do not have identical sensitivities. This calibration step divides the observed count value (or count rate) by the diode's response (near unity) to correct for diode nonuniformity using the diode response file diohfile. When COMB addition is used, a smooth diode response array is computed using a weighted average of diode responses. Data with a diode response value less than the minimum diode response value set in the ccr3 table are set to 0.0.
Paired Pulse Correction (PPC_CORR)
This step corrects count rate data for the finite response time of the detector electronics. The measured deadtime is 10.2 µsec, determined for detector D1 and assumed to apply to D2 as well. What this means is that the correction is only 1% at an input rate of 2,000 counts per second, which occurs only for very bright stars. The values used are stored in table ccg2.
Photocathode Mapping (MAP_CORR)
This step computes where the spectrum was located on the detector's photocathode. This mapping of the location is performed in LINE and SAMPLE space. LINE position is perpendicular to dispersion, and SAMPLE is parallel to the dispersion. This calculation is used by the following calibration steps: ADC_CORR, VIG_CORR, PHC_CORR, and MER_CORR. If MAP_CORR is omitted, then those steps will also be omitted, in spite of their settings.
The position of the individual substep bins are mapped into photocathode space using the following:
Doppler Compensation (DOP_CORR)
This step corrects for Doppler compensation when removing photocathode nonuniformities (i.e., when PHC_CORR is set to PERFORM) or vignetting effects (VIG_CORR set to PERFORM). Do not confuse this with the on-board Doppler compensation indicated in the science header by the value of the DOPPLER (DPOZER and DOPMAG) keywords. The on-board Doppler compensation is needed to avoid blurring of the spectrum because of apparent shifts introduced by the spacecraft motion. The on-board compensation involves moving the spectrum on the photocathode in the opposite sense to the effect caused by the spacecraft. That means that the spectrum is not at a fixed location over the course of the exposure, and this step attempts to correct for that fact.
This DOP_CORR step computes the percentage of time spent at each Doppler offset in the original observation. These are computed by dividing the observation into time segments and computing the deflection offset for each segment. The SHP packet time is used as the start of the readout and the packet time of the first science packet is used as the ending time of the readout. This step is not applied in the pipeline because GHRS observations are interruptible. We may know the start time of the observation and that it was interrupted, but we have no way of knowing when the interruption began, so we don't know where we are in Doppler space.
Photocathode Nonuniformity Removal (PHC_CORR)
This step removes the photocathode granularity using a reference file that has a granularity map, phchfile, when MAP_CORR is also performed. The GHRS pipeline does not use this feature because obtaining the flatfield exposures for all possible grating positions was impractical. Therefore the phchfile used by the pipeline is a dummy file that is populated with ones and zeroes. FP-SPLITs were generally used for high signal-to-noise work. However, post-COSTAR G140L flatfields are available; they are especially useful for data having S/N > 30, see GHRS ISR 076 for more information.
This map is intended to have a granularity vector for multiple line positions. At each line position, the granularity is tabulated with a constant starting sample for all lines and a constant delta sample. To compute the response for a given line and sample, bilinear interpolation is used within the reference file. If Doppler compensation is specified (DOP_CORR = `PERFORM'), the response is smoothed by a weighting function describing the motion of the data samples along the photocathode.
Vignetting Removal (VIG_CORR)
The response of the detector to light of a given wavelength depends on the grating used, but there is also a dependence on position on the photocathode. The first effect is the overall sensitivity, and the second is called vignetting. This vignetting resembles optical vignetting in that it is not present at the center of the detector but causes lower counts (by a few percent) near the edges; however, the cause is in the detector, not in the optics. The vignetting was determined by observing standard stars at enough grating settings so that there was redundancy at any one wavelength, allowing sensitivity and vignetting to be disentangled.
The VIG_CORR routine removes the vignetting and low frequency photocathode response using a reference file that has a vignetting map, vighfile, when MAP_CORR is also performed. The vignetting map has a vector for multiple line positions and carrousel positions. At each line position the vignetting response is tabulated with a constant starting sample for all lines and a constant delta sample. To compute the response for a given line and sample, tri-linear interpolation is used within the reference file over carrousel position, line position and sample position. If doppler compensation is specified (DOP_CORR = `PERFORM'), the response is smoothed by a weighting function describing the motion of the data samples along the photocathode.
Merging Substep Bins (MER_CORR)
This routine merges the spectral data when MAP_CORR is also performed. If MER_CORR is omitted, then the background correction (BCK_CORR) will also be omitted. Unmerged output data are then just a copy of the input data. BINID# keyword value are used by calhrs to determine how many groups of data are to be merged. Any STEPPATT that accumulates two or more bins of spectra as listed in Table 35.3 will generally require merging. To illustrate the merging, consider input data having values Dbin.diode for bin number bin and diode number diode. The data would look like the following:
bin 1 D1.1 D1.2 D1.3 D1.4 ...
The position of the data points in the two-dimensional data array mapped into a one-dimensional data array are . This routine maps the data into the output array for half-stepped data as:
bin 2 D2.1 D2.2 D2.3 D2.4 ...
bin 7 D7.1 D7.2 D7.3 D7.4 ...
D1.1 D2.1 D1.2 D2.2 D1.3 D2.3 ...
And for quarter-stepped data as:
D1.1 D2.1 D3.1 D4.1 D1.2 D2.2 D3.2 D4.2 D1.3 ...
Determine Wavelengths (ADC_CORR, GWC_CORR)
This step converts the sample positions on the photocathode to wavelengths by applying the dispersion constants using tables ccr5, ccr6, ccr7, and ccrc; these contain spectral order, dispersion, and thermal constants. ADC_CORR computes spectral orders and wavelengths, when MAP_CORR is also performed. For the first order gratings, the spectral order is set to 1. For the echelle gratings, the spectral order is computed from the following formula:
- NINT is the nearest integer.
- A, B, C are in table ccr5.
- a, b, and d are in table ccr5.
- carpos is the carrousel position.
- ydef is the Y-deflection adjusted for the proper aperture (LSA: 128 added to it, SC1: 128 subtracted from it).
The wavelengths are computed by solving the dispersion relation for wavelength using Newton's iterative method. The dispersion relation is described by the following equation:
- m is the spectral order.
- l is the wavelength.
- a0,a1,... are the dispersion coefficients.
- s is the sample position.
The dispersion constants are calculated in one of two ways. If the switch GWC_CORR is set to PERFORM, then the dispersion coefficients are calculated from the ccrc table's set of global coefficients which define a function based on carrousel position. If ADC_CORR is PERFORM but GWC_CORR is set to OMIT, then the dispersion coefficients are read from the ccr6 table, which contains the dispersion coefficients for a few carrousel positions. Therefore, when GWC_CORR is OMIT, interpolation is performed between two sets of coefficients bracketing the required position, if that particular position is not in the ccr6 table. If ADC_CORR is omitted, then many other switches relying on it are also omitted, namely: IAC_CORR, VAC_CORR, HEL_CORR, FLX_CORR, ECH_CORR, BCK_CORR, and GWC_CORR. The subject of wavelength calibrations and their improvement is treated in the next chapter.
Background Removal (MDF_CORR, MNF_CORR, PLY_CORR, BCK_CORR, BMD_CORR)
This step removes the background, or dark counts, from the observed flux. The switch BCK_CORR determines whether or not background removal is done. The other switches, MDF_CORR, MNF_CORR, PLY_CORR, and BMD_CORR determine how the background is smoothed before subtraction. If BCK_CORR is omitted, then MDF_CORR, MNF_CORR, PLY_CORR, and BMD_CORR are all omitted, too. If the dispersion constants are not applied (ADC_CORR is omitted), then no background subtraction is done.
If MDF_CORR is set to PERFORM, then a median filter is applied to the background. This size of the filter box is found in table ccr3 in columns SKY_MDFWIDTH and INT_MDFWIDTH. This switch is not normally applied in RSDP: it is provided as a recalibration option.
If MNF_CORR is set to PERFORM, then a mean filter is applied to the background. The size of the filter box is found in table ccr3 in the columns SKY_MNFWIDTH and INT_MNFWIDTH. This switch is not normally applied in RSDP: it is provided as a recalibration option.
If PLY_CORR is set to PERFORM, then a polynomial is fit to the background and the function is subtracted. The order of the polynomial is found in table ccr3 in columns SKY_ORDER and INT_ORDER. Currently the order is set to 0 but can be modified in your own copy of the ccr3 table as a recalibration option.
It is possible to have all three filter options set, in which case, they are all performed in the order given above. However, if any of the above are set to PERFORM and BMD_CORR is set to PERFORM, then the background model correction is omitted.
There are three ways to calculate the background counts: the first involving resampling by linear interpolation and the other two internally measuring the background.
If the science data are composed of multiple substep bins, the sky background will be resampled by linearly interpolating adjacent smoothed background data values. The background will then be scaled by:
Internal Measurement Methods
The two other methods involve internal measures of the background. Both methods use the same formula for determining the background vector:
- The background can be measured from inter-order spectra. The background is measured by the science diodes by observing the photocathode above and below the science data. Ui is set to the upper background spectrum and Li is set to the lower background spectrum.
- The background can be measured from the corner diodes. There can be up to six substep bins sampling both the upper and lower background diodes. The background for each corner diode is the average of all measurements for that particular corner diode:
Background Diodes Used by Substepping
- The background can be calculated using a model. There are good reasons for doing this. Sometimes the number of counts accumulated in the background diodes is very small because the exposures are fairly short. When this occurs, the counting statistics limit the accuracy of the background correction. The model, on the other hand, is based on the cumulative experience of the GHRS in orbit. There are also reasons to avoid the model. In particular, if the observations span a range of orbital conditions, then the model's estimate of background may be significantly wrong. This is especially a problem in the region around the South Atlantic Anomaly. Also, for longer observations, the measured background may be well-measured anyway.
BMD_CORR is provided as a recalibration option; it controls the application of the background count rate model available in calhrs v.1.3.11 (March 1997). When performed, the background model correction will calculate and subtract a model background based on location of the telescope in its orbit instead of using the background bins obtained with an observation. Because results from the background count rate model are not reliable in or near the SAA, the SAA Model 7 contour (defined in SAAHFILE) is used to issue warning messages per readout of an observation. The ccre table contains mean derived background count rates as a function of geomagnetic latitude and longitude. The model is actually the blue FOS model derived by Rosenblatt, et al. (1992) multiplied by a GHRS-detector-specific scaling factor. Header keywords in the table contain the multiplicative scaling factors required to scale the results from the model to the appropriate GHRS detector. See GHRS ISR 084.
Apply the Incident Angle Correction (IAC_CORR)
All GHRS wavelengths are referenced to the SSA. After the dispersion constants are applied (ADC_CORR is performed), this routine adjusts the zero-point of the wavelength array for the effects on wavelength of the difference in incidence angle of apertures LSA, SC1, and SC2 from the SSA. New post-COSTAR LSA offsets were determined in 1997 and are now available (GHRS ISR 080).
The ccr8 table is searched for the correct grating, spectral order, aperture, and carrousel position to obtain two coefficients, A and B. Interpolation of the coefficients (in carrousel position) is used if an exact match is not found. These coefficients are then used to compute an offset using the following formula:
Echelle Blaze Correction (ECH_CORR)
This step is appropriate only for observations made with an echelle grating. The large-scale change of response with wavelength is characterized as sensitivity, while vignetting accounts for minor detector effects. An echelle grating, however, produces a spectrum that drops as one goes away from the blaze peak; this is the blaze function, sometimes called the ripple function.
Tables ccr9 and ccra contain echelle blaze constants. This step performs the echelle ripple removal (if the data were taken with one of the echelle gratings) after the dispersion constants are applied (ADC_CORR is performed) by dividing the flux by the following echelle ripple function:
Absolute Flux Conversion (FLX_CORR)
This step converts the input flux to absolute flux units, after the dispersion constants are applied (ADC_CORR is performed), by dividing the input count rate by a sensitivity stored in the abshfile (sensitivities) and the nethfile (wavelengths for the sensitivities) files. Quadratic interpolation is used within the sensitivity file to compute sensitivities for the input wavelengths.
The nature and quality of the flux calibrations is treated in the next chapter.
Heliocentric Correction (HEL_CORR)
This step converts wavelengths to a heliocentric system after the dispersion constants are applied (ADC_CORR is performed). This step corrects for the Earth's motion around the Sun and its rotation, and modifies the wavelengths appropriately. The wavelength correction is computed as follows:
Vacuum Correction (VAC_CORR)
This step converts vacuum wavelengths to air wavelengths above 2000 Å, after the dispersion constants are applied (ADC_CORR is performed). This correction is not applied in the calibration pipeline. GHRS data are routinely calibrated to vacuum wavelengths. This step is provided as a recalibration option. The following formula is used:
[Top] [Prev] [Next] [Bottom]
Copyright © 1997, Association of Universities for Research in Astronomy. All rights
Last updated: 01/14/98 15:51:00