 |
NICMOS ETC - Performance and Testing Some usefull links:
|
 Cycle 7/7n CGI NICMOS ETC (Atem) Development etc (Cycle 11 ->) Usefull testing information: Regression Info APT information Unit Conversion tool Standard Star Magnitudes |
CDBS Spectra (or calspec)
Kurucz Models note: this apparently does not lead to the full listed file, just the wavelengths listed for stis. If you look in synphot at the spectra file that it calls there seems to be full wavelength coverage.You can look at the full files here: Kurucz . NON-APT Related ISR Documents: Science Verification of the CGI ETC: ISR (pdf) | ISR (ps) CGI ETC Software and Installation: NICMOS-ISR-2000-005.pdf CGI ETC Algorithms and User Interface: NICMOS-ISR-00-001.pdf NICMOS Thermal background analysisNICMOS-ISR-2003-007.pdf  |
Abstract
A robust test of the newest version of the NICMOS Exposure Time Calculator (ETC) was conducted in order to assess its performance against actual NICMOS datasets in predicting an accurate Signal - to - Noise ratio for observation planning. All available signal regimes were explored and the results show good agreement with the ETC for all cases.
Introduction
The NICMOS ETC was completely redesigned in Spring 2000 to support the NCS-era of NICMOS observations. It also introduced new algorithms to correct the estimate of the thermal background contributions from the telescope. Further information on the redesign of the ETC and the new thermal library can be found in Sivaramakrishnan et al, 2000. At the time of its release the code had been checked for accuracy against the original ETC, but no testing was completed on its agreement with archived NICMOS datasets. Therefore, this project was designed to confirm that the observed signal-to-noise ratio (hereafter referred to as SNR) was in good agreement with the predicted one and that the performance of the ETC was consistent over the available temperature range.
Data and Analysis
All data used for this analysis were gathered from the HST archive and represent a broad range of available NICMOS data, spanning the initial operating temperature of the instrument (61.5 K - 62.5 K). The basic regimes which were tested are as follows:-
High SNR observations: these datasets test the Poisson noise limited regime.
- Low SNR observations, where faint objects were observed just above the background, achieving a typical signal to noise below 20. These datasets test the background and instrument limited regimes.
The above scenarios were tested for point sources and extended objects in each of the NICMOS cameras, and when available, in several different filters. Unfortunately data about known faint stars was unavailable in the database, therefore no testing was done in this part of the PSF regime. The SNR for each image was calculated as:
where:
- S is the total signal for the source in electrons / pixel / second
- Sigma is the rms noise in the background in electrons / pixel / second
- t is the total exposure time for the observation
- Sigma is the rms noise in the background in electrons / pixel / second
Each dataset was evaluated separately and the respective properties for each object were entered into the ETC web form to calculate the predicted SNR.
Point Sources
The ETC returns only the SNR for the central pixel of the observed PSF. To achieve this, the total flux value calculated by SYNPHOT is multiplied by the pixfrac - the fraction of total energy that falls on the central pixel of the PSF. This assumes that the image is centered on that pixel. In practice, it's possible to measure the signal in the peak pixel using the imexam routine in the IRAF package. Imexam fits a gaussian profile to the pixels surrounding the center of the star and then returns the peak value for that fit. This is preferable over using imstat to obtain the highest pixel value around the center of the star because of individual pixel sensitivities and detector cosmetics. The stars were carefully selected to avoid known bad pixels and grot (Sosey et al, 1999). No preference was given to any quadrant in any of the cameras, object locations were spread across the entire array.
Each selected dataset was fully processed through the NICMOS pipeline and associated algorithms when necessary (such as pedsky ), using the most current reference files, which in some cases included synthetic darks and flatfields (to account for changes in detector temperature). Care was taken in images affected by many cosmic ray hits and other anomalies to get a realistic measurement of the background noise. No extinction values were calculated for any of the point sources.
The NICMOS standard stars were chosen for testing bright, high SNR PSFs. The requisite Kurucz spectra was chosen for the solar analogs, and a black body spectrum at the appropriate temperature was used to simulate the White Dwarf. The Kurucz spectra are stored in the CDBS archive and cover the wavelength range from 1000Å - 10 microns. They are specified by effective temperature, metalicity and gravity and called through the SYNPHOT routine icat which then interpolates between the spectra in the database which bracket the specified values. Table 1 lists information about each of the standard stars. Table 2 details which stellar datasets were chosen and their resulting measured and predicted SNR values.
Table 1. Information for each of the NICMOS standard stars used
|
Star Name |
Type |
Mag |
Other Info. |
|---|---|---|---|
|
P330E |
Solar Analog |
12.01 (J) |
G2V-Kurucz |
|
G191-B2B |
White Dwarf |
12.6 (J) |
T eff ~ 61300 |
|
P177D |
Solar Analog |
12.47 (J) |
G2V-Kurucz |
Location of the object on the detector is not a provided option, as the ETC does not attempt to model flat field variance, it merely uses the average DQE at the pivot wavelength of the source. Since some of the standard star observations were well dithered they also provided a decent check on how much the variation in pixel sensitivity affects the SNR. Figure 1 shows the ratio of predicted and measured SNR measurements vs. wavelength for P330E. The large range in the Camera 2, F110W measurements illustrates the effects of varying pixel sensitivity across the array. Data from proposal 7693, the pupil transfer function, were used for these calculations. It is emphasized again that these are measurements on the peak flux from the central pixel, not from a full aperture. Aperture photometry in NIC 2 has been proven at the 2% level (Sosey et al, in preparation). The observed spread in SNR is too large for intrapixel sensitivity to be the main culprit. Camera 2 has 75 mas pixels which critically sample the PSF at all wavelengths. The same spread can be seen in cameras 1 and 3 to a similar extent. Figure 2 shows the ratio of predicted and measured SNR for camera 3, P177D, for the F110W and F222M filters. In the case of Camera 3 observations intrapixel sensitivity plays a larger role since the PSF is undersampled at all wavelengths, with flux variations of up to 30% for individual images of a single star observation. For more information on intrapixel sensitivity see Storrs et al, 1999. The average measured SNR for all the stellar datasets taken in a particular camera and filter combination was used for comparison against the ETC and match the predicted values from the ETC extremely well. Table 2 shows for each of the stellar datasets which were examined
Figure 1: SNR measurements for P330E in all cameras
Figure 2: SNR Measurements for P177D in camera 3
Table 2 : Results for each of the stellar datasets which were examined
The largest disagreement between the SNR values predicted by the ETC and those measured from the data can be found in camera 3. The measurement of P330E in the F110W filter only agreed to 47%. Other images of P177D and G191B2B in the same filter agreed quite well with the ETC estimates. Closer examination of the P330E image itself reveals it to be about twice as noisy as the others (See Figure 3). This dataset was taken 30 minutes after a 23 minute long SAA (South Atlantic Anomaly) passage. Other images which executed in the same proposal (7152) also show increased noise that improves as the image start time is increasingly farther from the SAA exit time. This is a good example of how cosmic ray persistence can have adverse affects on the SNR of NICMOS observations. For more information on NICMOS and the SAA see Najita, et al. 1998.
Figure 3: NIC3 comparison of P177d and P330e images at the same stretch
Extended Sources
For extended sources, the ETC returns the SNR for a pixel which is fully illuminated by the observed source. It accepts information about the source as either the surface flux from the galaxy in Jy per arcsec 2 and the central wavelength, or the magnitude in Vega mags per arcsec 2 . Only elliptical galaxies were chosen from the database since they have the most well defined source spectrum (SED) in the ETC (the elliptical galaxy spectrum was provided by M. Rieke and is unpublished). Each of the datasets were processed through the NICMOS calibration pipeline and associated routines when necessary (such as pedsky ). The surface flux from each object was measured in a circular radius and an estimate for the background noise was calculated through the IRAF routine iterstat . Incomplete color information about each of the sources makes it hard to calculate a reasonable extinction, therefore, none was used to correct the elliptical SED.
Table 3 is an example dataset of extended object measurements. The data is taken from proposal 7895, a snapshot survey of field galaxies. Both objects were compact and each was measured using an aperture of 6 pixels. Even in this faint SNR regime, the ETC does a fairly good job of predicting the actual SNR for both the individual datasets and the average combined measurements. The compiled results for each of the chosen extragalactic datasets can be found in Table 4.
Table 3: A closer look at proposal 7895, all data are from Camera 2, F160W
Table 4: Results for each of the galaxy datasets that were chosen
Conclusions and Changes to the ETC Web Interface
As a general estimate of exposure time, the ETC predicts a very accurate SNR for bright point sources. Except for some of the medium and narrow band filters, all of the SNR predictions were within 10% of the measured values. Even for the case of varying sensitivity across the detector, the average SNR of the measurements was well estimated. The results from proposal 7693 show that the local SNR can vary by a large amount, so users should avoid the lowest DQE areas when possible. Although no faint stellar sources were tested, the results from this report suggest that the ETC would also do a good job in predicting accurate SNR's for faint stellar sources.
In general, the ETC did a good job of predicting SNR for both faint and bright extended objects. The predicted and measured values are generally within 20% of each other and in some cases better than 10%. Low SNR observations can suffer greatly from background noise induced by thermal emission from the telescope optics (affecting the longer wavelength filters) and the typical noise characteristics of the detectors. In these cases it is hard to accurately predict the final observation characteristics. The results from the faint galaxy measurements of proposal 7280 show how difficult it is to predict and measure SNRs around 1. They also show that the ETC did a fairly good job of calculating an accurate SNR. Note that the large percentage difference between the predicted SNR and measured SNR is a result of the small numbers that are involved. The ETC also appeared to report slightly better SNR values for the faint blue galaxies that were measured in F160W. Users observing such objects may wish to slightly overestimate their exposure times in order to get the desired SNR.
The following items in the operation of the NICMOS ETC graphical user interface have been changed:
-
Under the HST standard star spectra that are listed, only the standard stars that have somewhat well defined infrared spectra in CDBS (i.e. are populated past 0.9 microns by more than a few points) were saved. All other spectra have been removed from the options box. However, it should be noted that only the NICMOS standard stars (P33-0E and G191B2B) have spectra which have been obtained using both ground and space based data as reference. Other spectra contain only general extrapolations into the infrared from the visible.
-
The "Real Object" templates should include a text description of their spectral classification in the option box, for easy reference. They cannot be extended to include this because it would interfere with the scripting of the ETC. This information may be found in Appendix A3 of NICMOS-ISR-00-0001 and there is already a link in the ETC GUI directly to that ISR on the NICMOS website.
-
The option to use Grism filters for NIC3 has been removed since they are not supported by the ETC.
- New aperture corrections were computed as part of work being done to improve the NICMOS photometric calibration. The old values have been updated to agree with the new aperture corrections (results of the improved photometric calibration will be presented in a future ISR).
- The option to have more than 1 number of reads (nread) has been removed. The default value is now 1 and no longer a user option.
NICMOS users in cycle 11 and beyond should expect to use this version of the ETC to aid in calculating exposure times for their programs. The datasets tested for this analysis spanned the entire range of available temperatures. After NCS is installed NICMOS is expected to operate at the higher temperature of 78K, this is now the default temperature in the ETC web interface. The increased temperature will decrease the large range of DQE values that were seen during the initial lifetime of the instrument. Users should still avoid the low DQE regions when designing their observations, but the difference between the lowest and highest DQE regions will be smaller. Work is in progress to integrate all of the HST instrument exposure time calculators into a more uniform, observatory wide ETC as part of the Astronomy Proposal Tools package, but the basic functionality of the ETC should not change.
References
Burstein, D., Heiles, C., ApJ, 225 :40-55, 1978
Calzetti, D., Kinney, A., Storchi-Bergmann, ApJ, 429:582, 1994
Rieke, G. H., Lebofsky, M. J., ApJ, 288 :618-621, 1985
Najita, J., Dickinson, M., Holfeltz, S., NICMOS ISR-98-001, 1998
Seaton, M. J., MNRAS, 187 , p.75, 1979
Sivaramakrishnan, A., Holfeltz, S., Sosey, M., Simon, B., Robberto, M. NICMOS ISR-00-0001, May 2000
Sosey, M., Bergeron, L. E., NICMOS ISR-99-008, 1999
Storrs, A.,R. Hook, M. Stiavelli, C. Hanley, W. Freudling NICMOS ISR-99-005, 1999
Zombeck, M. V. 1990, Handbook of Astronomy and Astrophysics, Second Edition: Cambridge University Press)