Absolute Radiometric Calibration and Corrections
Absolute Radiometric Calibration and Corrections
The radiometric calibration of the SUMER spectrograph was performed in the laboratory using a secondary transfer standard traceable to BESSY I (Hollandt et al., 1996, Appl. Optics 35, 5125).
The calibration was refined in flight with the help of star data and line ratio methods (Wilhelm et al., 1997, Appl. Optics 36, 6416; Schühle et al., 2000, Appl. Optics 39, 418).
The stabilities of both detectors were monitored during the operational phase. Results of the first 15 months demonstrate that there was no inflight degradation experienced (Schühle et al., 1998, Appl. Optics 37, 2646). This monitoring has continued until the present time and no degradation was detected before the loss of attitude of SOHO in June 1998. Thereafter, a loss of responsivity of 43% (on average) has occurred (Schühle et al., 2000, A & A 354, L71; Wilhelm et al., 2000, Metrologia, in press [*]). To use the programme for periods before this loss set the keyword /before, otherwise set the keyword /after. For safety reasons there is no default value set, and one of the keywords has to be used.
Based on these results, calibration curves were derived in both orders and for both photocathodes. They have been made available through the SUMER software tree. A comparison with SOLSTICE (Woods et al. 1998, Solar Phys. 177, 133) indicated a discrepancy (albeit within the uncertainty margins) near C IV (at 1548 and 1550 Å) late in 1998 (Wilhelm et al., 1999, Adv. Space Res., 24, 229). Part of this discrepancy could be attributed to a mathematical smoothing error, which was corrected on 04 December 1998.
Following this correction, the star data were re-evaluated with the aim of obtaining a finer wavelength resolution, and the curve fitting procedures were critically assessed. The resulting SUMER calibration for detector A is in good agreement with SOLSTICE from 1200 Å to 1560 Å, where the instruments have overlapping wavelength ranges. SUMER is about 10% low in this range, if we accept SOLSTICE to be exact (Wilhelm et al., 1999, A & A, 352, 321). The SUMER calibration has also been compared with CDS (Pauluhn et al., 1999, Appl. Optics, 38, 7035).
The revised responsivity curves are now available for the KBr photocathode and the bare MCP, respectively. Although the example of the responsivity curves in Fig.1 are shown versus wavelength in units of nm, RADIOMETRY.PRO works in Angstroem (Å). The deduction of the calibration curves follows the scheme outlined in Wilhelm et al. (1997), but with some modifications as follows:
Uncertainties are estimated to be 15% from 540 to 1205 Å, 30% above 1250 Å on KBr. Below 540 Å only very few lines can be observed, and the data need further confirmation. At this stage the uncertainty is estimated to 50% (1 sigma). The KBr data for detector A in this range need further study.
A comparison of the responsivity ratios S_new/S_old of detector A demonstrates that only at very short wavelength near 500 Å for the KBr photocathode, is there a change greater than the uncertainty levels announced for the previous curves. It results from the revised interpolation concept and it remains to be seen whether continued work with and on the SUMER calibration can establish which solution is to be preferred. In this spirit, the radiometry programme has been amended to accept another keyword "Epoch_x" with the following definition:
Default: Epoch_6 (recent calibration)
Epoch_5 (27 Apr 1999 - 08 Oct 1999)
Epoch_4 (10 Feb 1999 - 27 Apr 1999)
Epoch_3 (30 Jan 1999 - 09 Feb 1999)
Epoch_2 (06 Aug 1997 - 29 Jan 1999)
Epoch_1 (13 May 1997 - 05 Aug 1997)
Epoch_0 (20 Nov 1996 - 12 May 1997)
This will allow users to study in detail the changes which occurred.
It is important to note that detector A has always been in the same calibration status (at least before the attitude loss last year). What has changed is our best estimate on the calibration. So, in principle, the most recent calibration should be applied (no epoch keyword needed). However, since this estimate changed, the epoch keyword provides the users with the means to compare the recent calibration with the one used at a certain time in the past. Which epoch has to be set thus will be determined by the time of calculating the fluxes.
The responsivity curves of the detector B have a slightly different history. The detector B has only been used on two occasions (in February and August 1996) mainly for calibration monitoring purposes, before it was decided to use the detector B continuously for scientific observations starting on 24 Sep 1996. As for the detector A, the operating high voltage had to be adjusted for the continuous gain loss that occurred due to the charge extraction from the channel plates. For the detector A the high voltage was always raised such that the pulse height distribution was well above the discriminator thresholds. After half a year of operation of the detector A, a comparison performed with detector B confirmed that both were still at the same responsivity level as during the laboratory calibration. With the intention to increase the detector B life time by reducing the scrubbing of the channel plates due to high charge extraction, it was decided to operate detector B at a lower gain level. (A reduced high voltage by 200 V was possible, since the noise was so low that the discriminator thresholds could be lowered as well). This led to a pulse height distribution always close to the thresholds. Consequently, a loss of gain resulted immediately to loss of counts (pulses falling below the thresholds) and thus, a gradual responsivity loss until the high voltage was raised again.
A comparison was made to determine the difference that this makes to the instrument responsivity. According to an analysis done by J. Hollandt, the lower gain level setting leads to a 13% decrease in responsivity independent of wavelength. The variation due to the gain loss between voltage adjustments is estimated to be always less than -15%. However, since the gain loss may be not uniform across the detector, it cannot be accounted for precisely, and it thus leads to a higher overall uncertainty in the calibration.
The epoch keyword also works for detector B. Until May 1997 no calibration was available for detector B in the SUMER software tree. When using the laboratory calibration data for radiometric calibration, the results obtained should be increased by 13% plus an average value for the gain loss. To account for both effects we decided to implement a 17% correction to the responsivity curves and increase the relative uncertainty level of the calibration to 20%. This is included in the refined restore files used by the radiometry routine. However, until 05 Aug 1997 no correction was implemented.
As for detector A, the responsivity curves have been established in the laboratory calibration, and the refinement has been done in a similar way during the flight using star observations and line ratio measurements (Schühle et al., 2000).
Above 1236 Å the star observation data have been adapted to the laboratory curve of the KBr photocathode part up to 1467 Å. The uncertainty of these measurements are estimated at 30%. The KBr/bare ratio which has been measured during flight was used to transfer these values to the bare curve. In contrast to detector A, for detector B we could take advantage of a larger range of wavelengths where first and second order overlap. By observing lines in first and second order on KBr and on bare photocathodes we could tie together the four responsivity curves in a consistent way in the wavelength range between 676 Å and 721 Å. The KBr/bare ratio could also be used to improve the interpolation between laboratory calibration points at 769 Å and 889 Å. Finally, line ratio measurements have been used to extrapolate the second order curves on the short wavelength side down to 465 Å. However, the uncertainty below 540 Å remains as high as 50%. In the main part of the spectral range, from 540 Å to 1236 Å, we feel that, with the arguments mentioned above, the uncertainty is below 20%.
The responsivity curves of detectors A and B in first and second order for both photocathodes have been and will be updated as required.
A programme for converting SUMER count rates (in count s¯¹ px¯¹) into physical units (such as W sr¯¹ m¯² ů¹) is also provided here: an IDL routine that you can download to your directory by simply clicking on its name RADIOMETRY.PRO (30 KB). The programme contains comments explaining what is needed as input and how it can be called within IDL.
The programme requires the IDL (XDR) save files with the calibration curves. They are provided in this directory of the SOHO/SUMER software tree at NASA/GSFC.
If programme and files are in different directories, you have to use
commands like the following to set the environment (or "logical") RADIO
IDL> setenv,'RADIO=disk:[user.directory]' in VMS, or
IDL> setenv,'RADIO=/user/name/directory' in UNIX
where disk, user, name, directory refer to the path where the files
can be found on your system.
Please note that it is not necessary to transfer these calibration curves while working on the SUMER machines at GSFC or at MPAe, because the restore files already exist in the SUMER software tree that can be used there. If you want to run the most recent calibration, only the files *_4_99.RST are required for detector A and *_5_99.RST for detector B.
It follows a short history of RADIOMETRY.PRO:
First version 20 Nov 1996, KW, WC Programme can now accept arrays of rates. 10 Dec 1996, TAK Full Sun scheme modified and editing 08 Jan 1997, KW Keyword 'arcsec' added 13 Jan 1997, KW Input of actual wavelength 31 Jan 1997, KW Wavelength and rate same dimensions 11 Feb 1997, KW All slit dimensions with uncertainties given 29 Apr 1997, KW Uncertainty of telescope area (for information) 29 Apr 1997, KW Detector B added 12 May 1997, KW Extrapolation detector A in 2nd order 11 Jun 1997, KW Gain correction detector B (preliminary data) 06 Aug 1997, US, KW Refined detector B calibration data now available 11 Dec 1997, US Corrected dispersion for detector B 09 Feb 1998, IED,KW 2nd correction of dispersion for detector B 11 Aug 1998, KW 3rd correction of dispersion for detector B 01 Sep 1998, KW The programme is now using the flight S/W concept 26 Nov 1998, KW The mean spectral pixel size of detector A set to 0.0266 mm Procedures: Magnification, GRATING_B_511, LAMBDA_N and PARAM_CORR separately established Warning issued if KBr or det_a used as default 01 Feb 1999, KW Keyword: Epoch (Access to old responsivity curves) 07 Apr 1999, KW New responsivity evaluation for detector A 27 Apr 1999, KW New responsivity evaluation for detector B 27 Apr 1999, US Corrected responsivity for detector B 07 May 1999, US Editing 19 May 1999,KW Responsivity for detector A without Ne VII ratio: Ne VII (465) identification considered unreliable 08 October 1999,KW Change in responsivity after SOHO attitude loss 28 Dec. 1999,KW,US
Several keywords can be used to control the function and its input and output parameters. They are:
Input keywords:
Bare - Data on bare part of the detector
(For Lyman alpha observations on the attenuator, it should noted that the
attenuation is nominally a factor of 10 and that behind the attenuator
there is the bare MCP.)
KBr - Default
Px - Default
Line
Sun_line
arcsec
det_a - (Detector A) Default
det_b - (Detector B)
Epoch_x - Default: Epoch_6 (recent calibration)
before -
after - loss of attitude (no default set)
Output keywords:
Watts - Default
Photons
Input parameters:
1. slit slit number from image header
(widths and lengths
in arcseconds)
1: width of 4.122 +- 0.5%
2: width of 0.986 +- 1.6%
3 - 5: width of 0.993 +- 1.6%
6 - 8: width of 0.278 +- 4.5%
1: length of 299.2 +- 0.3%
2: length of 299.2 +- 0.3%
3 - 5: length of 119.6 +- 0.5%
6 - 8: length of 119.6 +- 0.5%
(Area of primary mirror (9 * 13 cm^2 +- 0.2%)
2. wavelength in Angstroem
(either array of same dimension as rate or single value)
3. order of diffraction your judgement! (3rd order
calibration not yet available.)
4. count_rate either
(either array of same dimension as wavelength or
single value)
rate_spec in count/(s*px_spat*px_spec)
(Default)
rate_line in count/(s*px_spat*line)
(Set keyword LINE)
rate_sun (full Sun) in count/(s*line)
(Set keyword SUN_LINE)
(if you want to have full Sun values, sum count rates over the full Sun.
The slit selection has no consequence in this case. Any dummy parameter
will do.)
rate_mean (average) in count/(s*arcsec^2*line)
(Set keyword ARCSEC)
Output:
Result either in
(Default) in W/(m^2 sr A)
in W/(m^2 sr line)
or (PHOTONS keyword set) in photon/(s m^2 sr A)
in photon/(s m^2 sr line)
or (if SUN_LINE keyword set)
Result either in
(Default) in W/(m^2 line)
(PHOTONS keyword set) in photon/(s m^2 line)
The IDL routine LOCAL_GAIN_CORR.PRO corrects for local gain depression in detectors A and B. It should be applied after deadtime correction (Wilhelm et al., 1997, Solar Phys., 170, 75).
The programmes are documented. They have to be run either in the same directory or in an environment called 'CONVOLUTION'. - Building the look-up tables takes a little patience. Please note that it is not necessary to generate these look-up tables while working on the SUMER machines at GSFC or at MPAe, because the tables already exist in the SUMER software tree that can be used there.
![[Home Page]](../pictures/sum_icon.gif){kind=icon}
SUMER Home Page