- Chop Tails
- Chop Frequency
- Radiative Offset and Nod Dwell Time
- Symmetric vs Asymmetric Chopping
- Frame Times and detector well depth
- Readout Mode
- New default readout mode
- Detector Features
- Slow Guiding Accuracy
Since September 2010, an important effort has been put into getting GTC ready for commissioning and scientific operation of CanariCam. Pre-commissioning tests to gauge the operability of the telescope with CanariCam have been taking place almost every month. Finally, CanariCam commissioning was performed between June 19th and 24th. This run was followed by two more runs, one in July (29th and 30th) and one in August (4th and 5th). Commissioning observations and analysis have been performed in a combined effort of UF and GTC teams.
During these three periods it was possible to commission the imaging mode as well as the 10-micron low resolution spectroscopic mode. Useful data were also taken in the polarimetric and 20-micron low resolution spectroscopic modes. No data have been taken yet in coronograpic or high resolution spectroscopic mode.
In all, commissioning data analysis shows that GTC is ready for the operation with CanariCam in imaging mode (10 and 20 microns) and in low resolution spectroscopy at 10 micron, although there are some limitations that will be described below. Preliminary results from low resolution spectroscopy at 20 micron and polarimetry at 10 micron (note that polarimetry at 20 micron is not available in CanariCam) show that these modes are in good health, but more work needs to be done before they can be offered widely for common use by the GTC community.
Further commissioning tests have been performed on December 6th to 11th and 17th to 19th, 2011 and on January 5th to 8th, 2012. During these commissioning periods more data were taken in the polarimetry mode and low resolution spectroscopy at 20 micron. Also the first high resolution spectra at 10 micron were obtained. Data analysis for the polarimetry and low resolution spectroscopy is currently being performed. Given the amount of data gathered and initial analysis results, we are in the position of offering low resolution spectroscopy at 20 micron for semester 2012B.
In what follows, we show some of the commissioning results in imaging mode, which can be useful for proposal preparation and observation planning. Commissioning results regarding low resolution spectroscopy will be posted soon.
Sensitivity and its dependency on PWV
The following table shows the sensitivity in some key filters, since they are also used in T-ReCS at Gemini and therefore they can be used for comparison. Measured sensitivities are also compared with the values estimated with the CanariCam ITC.
The column Sensitivity shows the sensitivities measured using the star HD38944 (Q1 filter) and the star HD70272 (all 10 micron filters), on March, 3rd, 2012. The method to calculate the sensitivity was the following. Aperture photometry was performed on each image with increasing aperture radius. Two aperture radii were taken, one containing the full flux of the star (Rfull) and the other one giving the maximum SNR (Ropt). A template spectrum of the star and the transmission curve of each filter was used to estimate the theoretical flux of the star in each band in Jy (column Flux). For each filter, the theoretical flux in Jy was divided by the number of ADU in the Rfull aperture and by the on-source integration time to obtain the conversion factor Jy/ADU/s. The measured SNR in each filter within the Ropt aperture was scaled to an on-source time of 30 min. Finally, the conversion factor (Jy/ADU/s) was used to estimate the flux of a source yielding a SNR of 5 in 30 min in the Ropt aperture, which is our nominal definition of sensitivity (column Sensitivity). This definition is the same as the one used in Gemini, which allows us straight comparison of our values with the ones published in their T-ReCS web site (see the original T-ReCS values here).
Column PWV in the table shows the amount of water vapor in the atmosphere in terms of millimeters of Precipitable Water Vapor (PWV) as measured by the IAC real-time PWV monitor during each observation.
The last column of the table shows the sensitivity as calculated with the CanariCam ITC. The ITC seems to yield slightly higher sensitivity values than the real CanariCam data. Nevertheless, when estimating the time required to perform an observation during proposal preparation, the ITC should be used .
To estimate the impact that the atmospheric water vapor has in the mid-IR observations, the following table shows the sensitivities obtained on June 5th, 2012, using the standard star HD96833, in the same set of filters as in March 3rd, 2012. Each column from the table below represents the same quantity as the previous table. Data from June 5th were taken with a PWV which was a factor of 2 higher than in March, 3rd.
A clear degradation in the sensitivity can be noted by the increase the amount of water vapor in the atmosphere. It can also be noted that there are differences in the FWHM of the PSF when comparing data for the same filter in both nights, since the seeing conditions were not exactly the same in both nights. To eliminate any dependency of the sensitivity on the image quality, it is more appropriate to estimate a new set of sensitivities using the aperture that contains the total flux, rather than the optimal-radius aperture.
The following table shows the sensitivities obtained for the same set of data as in the previous two tables, but using the full-flux aperture (columns 3 and 6). For completeness and to facilitate the visualization of the data, the sensitivities using the optimal aperture and the PWV for both nights are reproduced again in this table. One can see inmediately how the sensitivity values are higher when using the aperture containing the full flux, whose radius is always larger than the radius of the aperture that maximizes the SNR. This degradation in the sensitivity is expected, since by increasing the aperture radius, we are also increasing the contribution of the background to the SNR, and therefore the sensitivity worsens.
|March, 3rd, 2012||June, 5th, 2012|
|Ropt (mJy)||Rfull (mJy)||(mm)||Ropt (mJy)||Rfull (mJy)||(mm)||ratio (Rfull)|
The last column in the table shows the ratio of sensitivities obtained using the full-flux aperture from June 5th (PWV ~ 10 mm) and March 3rd (PWV ~ 5-6 mm). With some exception, one can see that in general, when increasing the PWV by a factor of 2, the sensitivity values for the Si filters are degraded by roughly a factor of 2 . However, the same increment in PWV has a stronger impact in the Q1 filter, whose sensitivity is degraded by a factor 3.4 . These results are in accordance with the expectations that variations in the PWV have a stronger in the 20 micron window than in the 10 micron window (see the plot on atmospheric transmission as a function of PWV in Section Mid-IR Observing ).
The following table shows the sensitivity (5-s in 30 minutes on-source) for all CanariCam imaging filters. All data correspond to images taken on March 3rd, 2012 using the bright mid-IR standard HD70272. The sensitivity values reported here were obtained using the optimal radius aperture, i.e. the aperture that maximizes the SNR for each image. In some filters two measurements corresponding to two different images taken during the same night are given to show the type of uncertainties that one can expect from sensitivity measurements.
|Filter||FWHM (“)||Ropt (“)||Rfull (“)||FlxJy (Jy)||Sens (mJy)||PWV (mm)|
Currently, GTC can only do slow guiding, i.e. guiding corrections are sent to the telescope axes, which occurs at a typical frequency of ~0.2 Hz. However, to be able to reach the diffraction limit of the telescope with CanariCam, it is necessary to implement the functionality of fast guiding. This consists of commanding the telescope secondary mirror at high frequency (>50Hz), to correct for tip/tilt motions on the image. Until fast guiding is implemented, it will be difficult to reach the GTC diffraction limit with CanariCam, particularly at short wavelengths. Hence, CanariCam images are currently seeing-limited. Still, it is possible to see diffraction rings in cases where the seeing is good (≤1" in the optical).
The following figure shows the measured EER curves in several filters, compared with the corresponding diffraction limit EER80 at each filter's wavelength. These data were taken on June 25th, in good seeing conditions (optical seeing of 0.9").
It is clear that the measured EER80 (red vertical line) is worse than the diffraction-limited EER80 (blue vertical line – based on the optical model of CanariCam+GTC). In contrast, the following figure shows the same analysis during a night with bad seeing conditions (1.3" in the optical):
In this case the measeured EER80 is always >1", being more than twice the theoretical EER80.
The following figure illustrates the diffraction rings in the filters Si5 (11.6 microns) and Q1 (17.65 micron).
Observations were done on June 22nd, corresponding to the mid-IR standard HD140573, which has a flux of 30.7Jy and 13.1Jy in the Si5 and Q1 filters respectively (45.3Jy in the N filter). The FOV shown in the images is 7x7 arcsec. The FWHM of the PSF in each image is 0.36" (Si5) and 0.46" (Q1), which corresponds to 1.29 and 1.07 times the diffraction FWHM, respectively. Two diffraction rings can be seen in the Si5 image and only one weak ring can be seen in the Q1 image.
Previous images were reduced using the IRAF/GEMINI package, where individual savesets (see CanariCam User Manual for a description of the reduction tools and CanariCam data structure) were stacked to produce the shown images. An alternative to improve slightly the image quality is to reduce the data by registering and aligning each individual saveset (i.e. shift and add). This reduction technique mitigates slightly the lack of fast guiding. Note the round shape of the PSF core in the following figure (FWHM = 0.31", FOV also 7x7 arcsec) with respect to the left panel in the previous figure.
It is worth noting that sensitivities obtained using the shift-and-add method for data reduction are improved by a factor of 2 with respect to the sensitivities based on images reduced with the stacking method.
However, the reduction method consisting on shifting and adding individual savesets only works well with point sources, sufficiently bright for a good centroid determination. Otherwise, the resulting image could be worse than by simply stacking all savesets. This is, for instance, the case of the Q1 data shown in the previous example, where registering and aligning savesets degrades the image quality. This statement applies to the shift-and-add capability available within the GEMINI/IRAF software, based on the cross-correlation technique to calculate the offsets. However, other shifting and adding techniques, which were not explored during commissioning (e.g. use of the brightest pixel within the PSF), may work for certain science programs even for faint sources.
On-chip vs Off-chip Chopping and Nodding
For point sources or for very compact sources (up to ~5" diameter), it is possible to perform chopping and nodding on the chip, so that the negative images (from the off-source chop beam) are also seen in the detector image.
During the early stages of operation of CanariCam, only chopping parallel to nodding will be used. In this case, the positive image will have double number of counts than the negative image. The on-source time, whose value is in the image header keywords EXPTIME and OBJTIME, corresponds to the positive image.
Users can opt for using the negative images to increase the SNR in their data. This can be done by folding the two negative images of the target into the positive image, during the data reduction. This is shown in the following figure, where the usual accumulated image produced by CanariCam (left panel), has been folded into one single image (right panel).
A series of images taken on June 25th were folded during data reduction to compare the SNR measured on the positive image and the SNR measured on the folded image. Images correspond to the same standard star, HD186791, and were taken with a chop and nod throw of 10 arcsec at a PA of 45º with respect to the North (vertical detector axis). The following table shows the SNR calculated in both, the unfolded and folded images.
|Filter||FWHM /FWHM_diff||Flux||Noise||SNR||Flux folded||Noise folded||SNR folded||SNR_fold /SNR|
The SNR was calculated by performing aperture photometry with a radius of 35 pixels centered on the positive image of the star. The table illustrates how the signal and the noise increase in the folded image. However, the increase in noise is less than the increase in signal, and as a result the SNR is ~1.4–1.5 times better in the folded image. This is approximately what one would expect, since the exposure time associated to the folded image is a factor of 2 longer than the exposure time on the unfolded image.
It is also important to note that this test was made under conditions such that the FWHM of the PSF was between 1.1 and 1.5 times worse than the diffraction-limited FWHM. Under this conditions there is no degradation in the image quality by folding in the negative images, which in the case of asymmetric chopping (used here) correspond to the aberrated chopping beam. In strict diffraction-limited conditions, there would be a degradation in the image quality by folding in the negative beams. Even with such degradation, may still be worth to fold in the off source due to the gain in SNR.
Hence, we recommend to choose chop and nod throws such that the negative images fall inside the CanariCam detector, whenever the scientific target is compact (up to ~5" diameter).
The plate scale was measured using the double star WDSC 20467+1607. The average plate scale in the 10-micron window, measured using four filters (N-10.36, ArII-8.99, Si2-8.7 and Si5-11.6), is 0.0798 ± 0.0002 arcsec/pix. This plate scale ensures a FOV of 25.6" x 19.2" for the full detector. Laboratory measurements of the plate scale at 20 micron using a pinhole mask indicate values entirely consistent with the on-sky 10-micron measurement.
Distortion in imaging mode
Measurements of the distortion across the FOV in imaging mode were obtained during the instrument acceptance testing performed between January and February, 2007, at the University of Florida. To measure the distortion, a uniform grid of pinholes was observed against a background source in all CanariCam filters. The theoretical position of each spot image was compared with the actual measurement on CanariCam images. The distortion in X an Y for all filters is shown in the following table.
The analysis was performed using the GEOMAP task within IRAF. The input parameters were the measured positions of the pinholes on the detector, and their theoretical position on the sky. The transformation between both coordinate systems included distortion terms. The minimum and maximum of the residuals from the fit are the minimum and maximum distortion on each axis (columns 3 to 6 on the previous table). For every filter, the center-to-border distortion on each axis (columns 7 and 8 on the previous table) was estimated as the difference between the maximum and these measurements show that distortion is sub-pixel in the X axis and ~1.5 pixels in the Y axis.
This section shows some of the images that were taken during CanariCam commissioning (and pre-commissioning) to proof the current capability of the instrument and telescope.
NGC 7469 is an infrared bright Seyfert galaxy that forms part of a pair of interacting galaxies. It is located at a distance of 40 Mpc from Earth and has an infrared luminosity of 11.6 L_sun (within the 3" diameter nuclear region). High spatial resolution Mid-IR observations of the nuclear region within these type of active galaxies is fundamental to determine the nature of the source that heats the dust producing its strong infrared emission. If dust is heated by a central source, mid-IR sub-arcsecond observations would show a compact point source. However, if the dust is heated by hot stars in starburst regions, the morphology of the nuclear region would be extended.
Next figure shows two mid-IR images of NGC 7469 taken with CanariCam at 8.7 and 11.3 micron, respectively, during pre-commissioning, on May 29th, 2011. Both images show that dust emission is produced not only in the nuclear region but also in an extended area of 3–4 arcsec (~0.5kpc) diameter around the center.
Images are accumulated signal images produced automatically by the CanariCam software out of the raw data cubes. A smoothing top-hat function of 3 pixels was applied to increase the SNR and to be able to see the nuclear structure. Due to this spatial smoothing, the FWHM of the core is degraded with respect to the original data to a spatial resolution of 0.30", as inferred from a point source taken immediately before the observation of NGC 7469. The central core appears elongated at a PA~200 deg.
For comparison, next figure shows an image of NGC 7469 taken with LWS@KECK (Soifer et al. 2003). This image has a similar exposure time as the CanariCam image, but was taken at a slightly longer wavelength of 12.5 micron. The image was processed using a deconvolution algorithm, which improved the spatial resolution from its original 0.26" to ~0.08".
It is worth noting that several of the features that are seen in the deconvolved LWS image are also seen in the 11.3 micron image taken with CanariCam, which has no special processing but a simple smoothing of the data.
Technical data for the image of NGC 7469 taken with LWS@Keck-I (2003):
- Chopping: 5Hz, 15" throw
- Nodding: 15" throw, parallel
- Filter: 12.5 micron
- Total on-source time: 324 sec
- FWHM: 0.27", deconvolution (0.08")
Technical data for the images of NGC 7469 taken with CanariCam@GTC (2011, January 19):
- Chopping: 2.9Hz, 15" throw, N-S
- Nodding: 15" throw, parallel, nod dwell time 33 sec
- Filters: Si2-8.77 and PAH2-11.3
- Total on-source time: 318 sec per filter
- FWHM: 0.30" @ 11.6 micron (measured on one of the components of the double star WDSC 20467+1607)
This is a massive star forming region located at a distance ~450 pc, in the Orion nebula. The region exhibits several compact IR peaks and it is the center of an extensive outflow. As high-mass stars usually form in clusters, it is likely that several proto-stars and massive Young Stellar Objects (YSOs) are embedded in the region. One of the puzzles to be solved in these kind of environment is which of the IR blobs actually contain a proto-star or a massive YSO and which ones are blobs of dust and gas heated by external sources. High spatial resolution mid-IR observations are fundamental solve this puzzle.
The image below shows a comparison between data taken with CanariCam at GTC in January 19th, 2011 and LWS at Keck in 2002. LWS@Keck-I image is from Shuping et al. (2004). CanariCam@GTC image was reduced with the MIDIR tasks from the GEMINI/IRAF package. Individual subtracted savesets were registered to a common position defined by the peak of the BN source. Intensity levels in the three color channels are in logarithmic scale to show the faintest details of the extended emission.
Both images have labels in some the blobs that are probably composed of dust heated by the massive stars that are being forming in the region. By comparison between radio data (which probes the densest environment) and these mid-IR data its is possible to know that the main energy sources that light this region are BN and IRS 'n'. These two mid-IR peaks are genuine proto-stars/YSO's while the other peaks are heated externally.
Technical data for the image of BN/KL taken with LWS@Keck-I (2002, Nov 16):
- Chopping: 2Hz, 30" throw, E-W
- Nodding: 30" throw, parallel
- Filter: 12.5 micron
- Mosaic of 30 images, 10.2"x10.2" per image
- Total on-source time: 828 sec
- FWHM: 0.38"
Technical data for the image of BN/KL taken with CanariCam@GTC (2011, Jan 19):
- Chopping: 2.9Hz, 30" throw, E-W
- Nodding: 30" throw, parallel, nod dwell time 33 sec
- Filters: Q1-17.65(red), PAH2-11.3(green), ArIII-8.99 (blue)
- One single image per filter 25"x19"
- Total on-source time: 159 sec per filter
- FWHM: 0.34" @ 11.3 micron (measured on the standard HD95578)
It is worth noting that most of the features shown in the LWS image are also seen in the CanariCam three-color image. It is also important to mention that a mosaic of 30 images with LWS is required to map nearly the same area covered by one single shoot with CanariCam. Finally, note that CanariCam data are nearly as deep as LWS data but using only a 20% of the LWS on-source time.
NCG 7027 is one of the youngest known planetary nebulae. Mid-IR imaging of these type of objects is useful to trace the distribution of dust heated by the intense radiation from the central core of the red giant that is loosing its outer atmospheric layers. Dust particles can be formed in the outer regions of this expanding shell of gas. Furthermore, mid-IR imaging with narrow-band filters, specifically tuned to detect forbidden line emission, can be used to map the distribution of the gas excited by the core's strong ultraviolet radiation.
The following figure shows a three-color composite of filters Q4 (20.5 micron), Si5 (11.6 micron) and Si2 (8.7 micron) of NGC7027.
Images where reduced using the standard GEMINI/IRAF package, where registering option was selected to align individual savesets. The image shows how the cooler dust, which is seen in the Q4 filter (red channel) appears to be slightly further away from the central core that hotter dust, which is seen by the 8.7 and 11.6 micron filters.
The next figure shows an image taken during MICHELLE commissioning at Gemini, in 2003 in the 7.9 and 18.5 micron filters. The on-source time in this case was about half of the on-source time in the CanariCam images. The same ring-like and filamentary distribution of the emission is seen with both instruments, which gives an indication of the good health of CanariCam performance.
We also took two images of NGC7027 with CanariCam in the narrow filters tuned to the [NeII] line at 12.8 micron and adjacent continuum emission (13.1 micron). The following figure shows the spatial distribution of the pure [NeII] emission, since it was built by subtracting the continuum image from the line image.
The excited gas distribution is very similar to the dust distribution, with an important decrease in this emission towards the Northwest and Southeast, where the ring-like structure of the nebula is disturbed by the presence of a bipolar outflow known to originate in the central star.
Technical data for the image of NGC7027 taken with MICHELLE@Gemini (2003, June - July):
- Chopping: 15" throw, E-W
- Nodding: 15" throw, parallel
- Filters: 7.9 and 18.5 micron
- Total on-source time: 30 sec per image
- FWHM: 0.25"
Technical data for the images of NGC7027 with CanariCam@GTC (2011, June 20):
- Chopping: 2.0 Hz, 15" throw, PA=45 deg
- Nodding: 15" throw, parallel, nod dwell time 45 sec
- Filters for the three color image: Q4-20.5 (red), Si5-11.6 (green), Si2-8.7 (blue)
- Filter for the line image: NeII-12.8 minus NeII_ref2-13.1
- Total on-source time: 122 sec Q4-20.5, 73 sec Si5-11.6, 69 sec Si2-8.7, and 122 sec NeII-12.8 and NeII_ref2-13.1
- FWHM: 0.34" @ 11.3 micron (measured on the standard HD95578)
This section summarizes some of the most relevant results from the commissioning of the low resolution spectroscopy (10 and 20 micron) mode in CanariCam. The data used to generate these results were obtained during several observing runs that took place between June 2011 and August 2012.
To understand the limitations of the spectroscopy sensitivity calculations, we must first show which equations and approximations have been used. CanariCam standard star spectra were reduced and extracted using the GEMINI-IRAF MIDIR package. Each extracted spectrum was divided by the corresponding out-of-the-atmosphere Cohen spectrum in order to calculate a conversion factor between ADU/s and physical units (Jy). The SNR in each image was transformed into a SNR per frame time, which in turn was transformed into a SNR for a 30 minutes on-source image yielding the final 5-s detection in 30 minutes.
Each reduced spectrum (using the AVERAGE option in MISTACK) has a signal corresponding to the following saveset time in seconds:
Where FRMTIME is the frame time in milliseconds and SAVCOADD is the number of frames co-added in a saveset. Additionally, each reduced spectrum has a noise corresponding to its OBJTIME, i.e. the total on-source time. It should be noted that only the positive image of the spectrum in the final reduces image is used for this calculation.
The signal and standard deviation of the background in the reduced spectra and per individual frame are given by:
The signal per frame and noise per frame are related with the signal and standard deviation of the background emission with the expressions:
Where A is the area of the aperture used for estimating the source signal, G is the theoretical detector gain (189 e-/ADU in shallow well and 586 e-/ADU in deep well), SAVESETS is the number of savesets in the CanariCam raw data cube (4th dimension), NODSETS is the number of nod pairs A-B or B-A, and NNODS=2.
In the noise equation above, the first term corresponds to the shot noise from the source and the second term corresponds to the rest sources of noise (background noise, dark current and readout noise). Even though the shot noise has not been neglected in the sensitivity calculations, its contribution to the total noise is negligible as it is shown in the following figure, which compares the variance from the shot noise and the variance from the other sources of noise for an observation of the standard star HD31398 performed on January 1st, 2012.
The figure clearly illustrates that the shot noise is less than 3 orders of magnitude smaller than the other sources of noise. This particular behaviour is similar for all bright standards observed up to now.
If we assume that the noise scales as the square root of time, the 5-s sensitivity in 30 minutes can be related with the noise per frame in physical units using the following equation:
The noise per frame in physical is obtained from the noise per frame in ADU using the following equation:
Where CK is the transformation constant between ADU/s and Jansky:
The calculation of the transformation constant (CK) involves the slit factor (SF), the theoretical flux of the standard star from the Cohen template spectra and the extinction factor (K). The slit factor (SF) is the ratio between the total flux of the star (from the acquisition image) and the flux of the star within the slit (from the throughslit image). The extinction factor is used only in the 10 micron spectra and is given by the expression:
The following figure shows the 5-s sensitivity for an integration of 30 minutes on-source as a function of wavelength using four different standard stars observed in three different nights.
The figure shows that the 10 micron sensitivity in spectroscopy mode ranges from ~50 mJy to ~150 mJy depending on the night and the star used. In one of the nights the PWV was ~4 mm but in the other two the nights the PWV was over 8 mm. The airmass was ranging between 1.0 and 1.4 depending on star and night and the slit correction factor was ranging from 1.1 to 1.9. Two different slits were used, 0.52” and 0.36”.
It is important to note that there is an important dependency of the sensitivity on the slit factor. The larger the slit factor the better (smaller) the sensitivity. It is also important to bear in mind that these sensitivity values were obtained using a trace of constant aperture across the whole wavelength range. This means that at bluer wavelengths, where the PSF is narrower, there is a higher contribution from the background within the aperture than at redder wavelengths.
The following figure shows the 5-s sensitivity for an integration of 30 minutes on-source as a function of wavelength using three different standard stars observed in two different nights. Note that in this case we did not consider the first star from the 10-micron data set because there were no data available at 20 micron using the same slit as for 10 micron.
The figure shows that the 20 micron sensitivity in spectroscopy mode ranges from ~300 mJy to ~800 mJy. In this case, there is a stronger dependency of the sensitivity with wavelength than in the 10 micron window, due to the presence of strong water lines in the 20 micron window. The same acquisition image as for the 10 micron spectra was used, and therefore the slit correction factors are the same for the first two stars as for the 10 micron spectra. The slit correction factor is different in both observing windows for the third star because a slit of 0.52” was used for the spectrum while the acquisition image was taken with the 0.36” slit.
The following table shows a comparison between the sensitivity values obtained from the commissioning data, the sensitivity estimates using the CanariCam ITC and the sensitivity values for T-ReCS extracted from the Gemini web site.
|CanariCam Commissioning||CanariCam ITC||T-ReCS web page|
|10 micron||20 micron||10 micron||20 micron||10 micron||20 micron|
It seems that the measured sensitivity at 10 micron with CanariCam is worse than the estimates from the ITC and also worse than the quoted values for T-ReCS sensitivity. There is a good agreement between the estimates from the CanariCam ITC and the measured values in the 20 micron window.
A different way to estimate the system performance is to calculate the throughput of CanariCam and GTC. The throughput is calculated using the formula:
Where the different terms are slit correction factor (SF) the detector gain (G), the extinction correction factor (K) the energy of the photon (E) the area of the telescope pupil (A), the pixel size in wavelength units (δλ), the flux of the standard star in ADU/s and the flux of the star outside the atmosphere in physical units from the Cohen template spectra database.
The next figures illustrates the total throughput of GTC+CanariCam in the 10 micron window and in the 20 micron window.
The best throughput measured at 10 micron is ~14%, while the best throughput measured at 20 micron is ~3%.
The comparison of background emission spectra observed with CanariCam with theoretical sky and telescope spectra can give us an estimate of how much the telescope background emission contributes to the total emissivity. We used a background spectrum observed with CanariCam on January 25th, 2012, corresponding to an observation of the standard star HD31398. The telescope structure during the observations was at a temperature of 5.3C, the airmass was 1.06 and the PWV was 9mm. A theoretical sky spectrum with a PWV of 10 mm and an airmass of 1.0 was selected from a grid of ATRAN atmospheric spectra available on the Gemini web site for Cerro Pachón observatory. The sky spectrum was degraded to the resolution of the CanariCam observations. The telescope emissivity was modelled as a back body at a temperature of 5.3C. The total background emission was assumed to be the sum of the theoretical sky and the black body. A grid of emissivity spectra was obtained where the emissivity of the telescope was varied between 0% and 40%. The following figure shows the observed background spectrum (thick black line) and the grid theoretical spectra corresponding to different telescope emissivities.
It is important to bear in mind that an offset has been given to the observed background spectrum to avoid negative values due to detector bias. Such offset has been selected to normalize approximately the observed spectrum at the ~7.9 micron sky emission feature for 0% telescope emissivity, i.e. we are assuming that the telescope emissivity does not contribute at wavelengths below 8 micron.
The previous figure shows that the mean level of the observed background is clearly higher in the 10.2-12.5 micron region than in the 8-9 micron region, where there is a clear increase in emissivity between 8 and 9 micron. It should be noted that behaviour shown in this particular spectrum is detected in all 10 micron spectra observed with CanariCam. The comparison between the theoretical grid and the observed spectrum indicates that the telescope emissivity ranges between 10% and 25%. A single-temperature black body plus sky emission cannot explain the behaviour of the background emission in CanariCam observations. Telescope emissivity is still under investigation.
Measured Spectral Resolution
On the night of January 25th, 2012, sky spectra were taken with CanariCam using different slits. Once the spectra were wavelength calibrated, we could determine the spectral dispersion for each of the low resolution gratings by measuring the pixel and wavelength value on the band edges. The dispersion for the LowRes-10 grating is 0.018 micron/pix and the dispersion for the LowRes-20 grating is 0.029 micron/pix.
The spectral resolution was obtained by measuring the FWHM of some sky emission lines. The resolution for each line is shown in the following table.
|Image||Slit||λ||Δλ||R = λ / Δλ|
The table shows that the spectral resolution varies from ~50 with the 1.04” slit to ~120-130 with the 0.17” slit. In some cases a narrower slit does not yield a higher resolution probably due to intrinsic variability in the sky lines widths. It can also be noted that the measured highest resolution is lower than highest expected resolution (R=175), which probably has to do with the sky lines being resolved with the narrowest slit width. A better determination of the spectral resolution can be obtained by observing emission-line objects with unresolved forbidden emission lines. This is the case of the planetary nebula NGC7027, which was observed with CanariCam using the LowRes-10 grating and 0.17” slit (See Section on Verification Spectra, yielding a measured spectral resolution of 160 at 10.51 micron, fully consistent with the expected performance of the instrument.
The following table shows the sky emission lines at 10 and 20 microns that can be used for wavelength calibration.
|Wavelength (A)||Wavelength (μm)||Line ID|
The line maps for the LowRes-10 and LowRes-20 gratings, can be seen in the following plots.
The 20 micron window is highly populated with emission lines that can be used for wavelength calibration even with the widest slit (1.04”). However, wavelength calibration using sky lines in the 10 micron window can be difficult, particularly if the 0.52” or 1.04” slit are used.
CanariCam has a Polystyrene plate that can be inserted in the optical path just before the entrance window. Polystyrene emission lines be used for wavelength calibration of 10 micron spectra. This can be particularly useful when observations are performed with the widest slits. Grantecan will deliver an additional Polystyrene calibration spectrum together with all 10 micron spectra. The Polystyrene spectrum will be taken just before the target spectrum. The set of Polystyrene lines is listed in the following table.
|Wavelength (A)||Wavelength (μm)||Line ID|
The following figure shows line map for identifying the Polystyrene emission lines.
A linear transformation between the pixel and the wavelength scale is recommended when performing the wavelength calibration of the observed spectra.
On July 4th, 2012 the planetary nebula NGC7027 was observed with CanariCam in spectroscopy mode with the LowRes-10 grating and the 0.17” slit. The nebula has a ring-shaped emission region (See Section on verification images) and the slit was oriented perpendicular to the narrowest dimension of the ring, where the emission lobes are brighter. The spectrum was reduced using two different methods. On the one hand, the GEMINI/IRAF package was used following the instructions described in the CanariCam User's Manual. On the other hand, the RedCan pipeline was used (Gonzalez-Martín et al. in prep.). The spectra were calibrated in flux using the standard star HD213310, which was observed immediately after the planetary nebula observation. The FWHM of the PSF in the standard star was 0.47” in the Si5-11.6 filter.
CanariCam spectra of NGC7027 extracted using both reduction methods are shown in the following figure and compared with an archive spectrum of the same object from ISO-SWS.
An arbitrary flux scaling factor was applied to the CanariCam spectra for comparison with the ISO-SWS spectrum. The SWS aperture encloses the whole planetary nebula while the CanariCam slit (0.17” width) only includes a small part of the nebula, this is why, without scaling, the spectrum taken with CanariCam and reduced with GEMINI/IRAF is a factor of 10 fainter that the spectrum from ISO-SWS, while the spectrum reduced with RedCan is a factor of 3 fainter than the ISO-SWS spectrum. The discrepancy in the overall flux between the CanariCam spectra using both reduction methods is due to the fact of having used different extraction apertures. All of the emission features that appear in the ISO-SWS spectrum can also be detected in the CanariCam spectra, even the faint and broad PAH emission at 8.6 micron, which is a proof of the excellent performance of the spectroscopy mode of CanariCam@GTC. The fact that the forbidden emission lines appear wider with CanariCam than with ISO-SWS is due to the difference in spectral resolution, which is R~160 with CanariCam and R~1200-2500 with ISO-SWS.
The differences in the relative strength of the emission lines observed with CanariCam and with ISO-SWS actually show the strong advantage of using mid-IR diffraction-limited spectroscopy from the ground, allowing to map independently different regions of the nebula, which may have different ionization temperatures and chemical composition.
In the period between June 1st and 6th, 2012, we were performing a series of on-sky commissioning tests for the CanariCam imaging polarimetry observing mode. In this section we show the most relevant results from this commissioning run. The work presented here would have been impossible without the input from the CanariCam team at the University of Florida, particularly from Enrique López, who was participating in the observing run at the GTC control room, and Chris Packham, Frank Varosi and Charles Telesco, who were helping and giving advise remotely during the whole observing run. Most of the results shown here have been produced by Enrique Lopez and Chris Packham with a suite of IDL scripts that Enrique has programmed himself.
During commissioning of the imaging polarimetry mode, the S1R1_CR and S1R3 readout modes (see Section Readout mode and New default readout mode were also compared. The horizontal noise pattern that appears in the S1R1_CR mode, even though it is at a very low level, produces a strong bias in the calculation of the polarization angle of astronomical sources (see next figure).
Basically, during commissioning we noticed that we were systematically obtaining a polarization angle of ~90º, i.e. along the horizontal detector axis, for all sources. Therefore, S1R3 was selected as the standard readout mode for imaging polarimetry.
To measure the instrumental polarization (CanariCam plus GTC) several unpolarized standards were observed in imaging polarimetry mode with high enough SNR to ensure a polarization uncertainty of <0.3% (in one case it was only possible to achieve a polarization uncertainty <0.7%). To cover the full spectral range in the 10-micron window, measurements were made in all Silicate filters. Data were taken at two different orientations of the field of view to measure any possible dependency of the instrumental polarization on the detector orientation.
The following table summarizes the measured degree and angle of polarization for an instrument position angle (IPA) of 0º, i.e. with the Y detector axis aligned with the celestial North.
The table shows that the degree of polarization is <0.6% in all cases. Even though the polarization angle calculations are completely biased towards 90º due to the pickup noise in the S1R1_CR mode, the calculation of the degree of polarization is not affected by the noise.
The following table shows the measured degree and angle of polarization for an instrument position angle of 90º.
In this case, the degree of polarization is ∼0.2% for all filters and the polarization angle is again biased towards 90º due to the use of the S1R1_CR readout mode.
Summarizing the results shown in the previous two tables, the instrumental polarization is in the range of 0.1% - 0.6%, depending on the filter.
Polarization measurement efficiency
The polarization measurement efficiency is the ratio between the degree of polarization of the input beam and the observed degree of polarization. The efficiency in the polarization measurement was obtained by inserting a wire-grid analyzer in the optical path, just before the entrance window. The wire-grid is a KRS-5 mounted type with a polarization efficiency of 99.7% at 10 micron, which is basically constant in the whole 10-micron window. We observed three different unpolarized standards through the wire-grid in different silicate filters to cover the whole 10 micron window. The polarization in each filter was measured within an aperture that ensured a polarization uncertainty of 0.5%. The following table shows the measured polarization efficiency.
In this case the S1R3 readout mode was used, so that the polarization angle of 65º-66º is real, corresponding to the theoretical polarization angle of the wire-grid. The table shows that the maximum efficiency is obtained for the Si4-10.3 filter and that the efficiency drops towards the edges of the 10 micron band. The main cause for this drop is the fact that the wave-plate is not achromatic, being the retardation exactly half wave only at 10.6 micron.
Imaging polarimetry fo a polarized source
On June 1st, we observed with CanariCam/GTC a well studied polarized source, AFGL2136, using imaging polarimetry in all silicate filters. An on-source time of 73 seconds per filter were used, with a chop/nod throw of 10 arcseconds along the East-West direction, which corresponded with the longest dimension of the polarization mask slots (see next figure).
The polarization properties of AFGL2136 are described in the paper by Smith et al. (2000, MNRAS, 312, 327), shown in the following figure.
The degree and angle of polarization were measured on CanariCam images within an aperture where the SNR ensured an uncertainty of polarization 0.1%. A summary of the polarimetric measurements is shown in the following table, as well as a comparison with the results shown in Smith et al. (2000).
|Filter||FWHM of PSF
Smith et al. (2000)
Since CanariCam data were taken with the S1R1_CR mode, the polarization angle measurements are not reliable, and therefore are not presented in the table. The measured degree of polarization with CanariCam is shown before and after correcting by the efficiency factor shown in the Section on Polarization measurement efficiency. The comparison between the two last columns in the table show that the degree of polarization measurements in AFGL2136 obtained with CanariCam/GTC are completely consistent with the results published by Smith et al. (2002).
Distortions in polarimetry mode
Measurements of the distortion across the FOV in polarimetry mode were obtained during the instrument acceptance testing performed between January and February, 2007, at the University of Florida. The same procedure as in the Section on distortion in imaging mode was used to obtain the distortion terms. The calculation were performed for spots corresponding to the ordinary and extraordinary rays separately. The following table shows the distortion values in all filters for the ordinary ray image. Note that only the 10-micron filters appear in this list since CanariCam does not currently have 20-micron polarimetry capabilities.
|Filter||Wavelength (μm)||Min. X distortion||Max. X distortion||Min. Y distortion||Max. Y distortion||X distortion||Y distortion||X Pol/Imag||Y Pol/Imag|
|Filter||Wavelength (μm)||Min. X distortion||Max. X distortion||Min. Y distortion||Max. Y distortion||X distortion||Y distortion||X Pol/Imag||Y Pol/Imag|
When we take into account all filters in common for imaging and polarimetry, the distortion in the X axis is 1.3±0.1 times larger in the O ray and E ray images than in the imaging mode. In the case of the Y axis, the distortion is 1.19±0.05 and 1.08±0.05 times larger in the O ray and E ray images, respectively, than in the imaging mode.
In mid 2012 an upgrade was made to the M2 control system which reduced the problem of the chop tails. However, there is still some degradation of the PSF at large chop throws, which is expected to occur due to the chopping-induced coma (see this section). Additionally, we have also found that there is some discrepancy between the chopping and nodding distances which can affect the data quality for large throws (see this information). Hence, even if the GTC secondary mirror can operate up to a chop throw of 60 arcsec, when preparing the observations, it is very important to take into consideration the image quality degradation that will occur for large throws.
Chopping performance of the secondary mirror during the first months of CanariCam operaction was limited because the position of the star image in either chop beam varied with time. There was a jitter or oscillation in the image position that occurred predominantly along the chopping direction.
The following figure shows the integrated image of an observation with CanariCam while M2 was chopping at 2Hz with a chop throw of 50" and a chop angle of 45º. Each frame single frame had an exposure time of 23ms and the integrated image corresponds to an elapsed time of ~ 6min (on-source time of 2min), in the PAH2-11.3 filter. The image corresponds to the chopping beam where M2 is aligned with M1 (on-source beam), as seen by CanariCam. The off-source beam is somewhere outside the detector FOV (50" chop throw is larger than the detector size, which is 25"x19") towards the lower right corner of the image.
The result is an image that has a clear elongation along the chopping direction of 1.4arcsec, while the FWHM of the PSF perpendicular to the chopping direction is 0.6arcsec (as a reference, the diffraction limit at 11.3micron is 0.3arcsec).
Chop tails were less noticeable when shorter chop throws were used, since they become embedded within the PSF.
The optimal chopping frequency to observe with CanariCam is the result of a balance between the background correction achieved and the duty cycle of the chopper. The following table shows the 5-sigma sensitivity in 30min on-source for different filters and chop frequencies. The table also shows the chopper duty cycle associated to each frequency.
|Filter||Chop Frequency||Chopper duty cycle||Sensitivity|
The observations were done on May 29th, using the standard star HD140573. The water vapor conditions were poor (PWV ~ 5 mm) but the seeing was good (0.6" – 0.7" in the optical). A chop throw of 10 arcsec was used and 30 seconds on-source in all cases. Note that the chopping frequencies are not integer values because the CC software adjusts the input chop frequency automatically to accommodate an integer number of frames in each chopping beam.
It can be seen how the sensitivity generally improves as the chopping frequency is increased from 2Hz to 3Hz. However, there is not a clear gain in sensitivity when increasing the chopping frequency from 3Hz to 4Hz. On the other hand, the duty cycle is ~15% less efficient at 3Hz than at 2Hz, and 30% less efficient at 4Hz than at 2Hz. Hence, we discard completely chopping at 4Hz. Even though the sensitivity is slightly better chopping at 3Hz than chopping at 2Hz, the duty cycle is clearly less efficient at 3Hz. Therefore, 2Hz was selected as the nominal chopping frequency for observations with CanariCam.
Radiative Offset and Nod Dwell Time
The optimal nod dwell time to minimise the radiative offset can be found as a balance between the radiative offset removal and the nodding duty cycle.
The radiative offset is a noise pattern that remains after applying the chopping technique because the background that is seen by CanariCam is not the same in both positions of the secondary mirror. The radiative offset can be removed by nodding the telescope, once every several seconds, so that the on-source and off-source chopper beams are swapped. Even with nodding, some of the radiative offset my remain in the final images because the pupil that is seen by CanariCam is continuously rotating and hence it is not the same in both nod beams. It is expected that a residual radiative offset would be more prominent at high elevations where the pupil rotates faster than at low elevations.
This was actually measured on June, 19th, by looking at sky regions at different elevations while chopping and nodding. The following figure shows the radiative offset at an elevation of 75 deg using a nod dwell time of 60 seconds (left panels) and 45 seconds (right panels).
The diagonal wavy pattern noise seen in the top left image is the radiative offset residual, which can be seen much clearer in the Fourier space (bottom left panel). The pattern almost disappears (both, in the image and Fourier space) when a nod dwell time of 45 seconds instead of 60 seconds is used. By measuring the radiative offset residual at different elevations, we found out that the residual can be removed reasonably well with a nod dwell time of 45 seconds at all elevations. It would be possible to nod faster, but the gain in offset removal is counterweighted by a lower observing efficiency, which is 91%, 88% and 84% for nod dwell times of 60, 45 and 30 seconds, respectively. It would also possible to nod slower than 45 seconds at low elevations, but this would add unnecessary complexity to the observation. Therefore, the nominal nod dwell time for radiative offset minimization in CanariCam is 45 seconds at all elevations.
Symmetric vs Asymetric Chopping
On May 30th, we tested CanariCam in chop mode using chop throws of 10" and 20", symmetric and asymmetric (with respect to the pointing position), leaving the rest of the observing parameters constant in all images. The standard star HD153210 was observed in the filter Si2-8.7. The seeing was good (0.7"–0.8" in the optical) and the PWV was ~5–6 mm during the whole night.
The following table shows the results of this experiment in terms of FWHM of the PSF in the final accumulated image and in terms of sensitivity (5 sigma detection in 30 minutes on-source). Each row indicates if the chopping was symmetric or asymmetric. Two reduction methods were used, in one case savesets were simply stacked together, while in the other case savesests were registered, shifted and added. Note that each saveset corresponds to an integration time of 2 seconds, in the case of these observations. The values in the table where measured in the on-source beam, which corresponds to the nominal, aligned position of M2 in the case of asymmetric chopping.
The data show that, for a given chop throw, the FWHM of the PSF is better in the case of symmetric chopping than in asymmetric chopping. Also, the sensitivity is clearly better in the case of symmetric chopping. In other words, it appears that the image is less degraded when M2 swings from, say -10" to 10" (symmetric chopping) than when it swings from 0 to 20" (asymmetric chopping), even though the amplitude of the movement is the same.
It is worth noting that in the table above, we also see the same trend as in other analysis (see Image Quality section), that the image quality is improved by registering the savesets during the reduction.
As a conclusion to this test, it seems that symmetric chopping gives better image quality and sensitivity than asymmetric chopping when operating in the engineering mode chopping only (no nodding). However, further tests should be done in chop-nod mode to decide which type of chopping (symmetric or asymmetric) is best for science observations.
Frame times and detector well depth
Before dealing with the optimization of the frame times, it should be clear that frame times are not controlled by CanariCam users. The only time definition the users should be concerned of is the on-source time. Any of the other multiple definitions of time in CanariCam (see here) is setup automatically by the CanariCam control system, depending on the observing conditions (airmass, PWV and temperature).
The frame time is the fundamental time unit in CanariCam observations. All other time parameters (e.g. chop frequency, nod dwell time, on-source time, etc) are optimized by the CanariCam control system to be integer multiples of the frame time. During commissioning, we had to estimate the optimal frame time for each filter. Such optimization is made by looking at background images in stare mode and ensuring that the detector wells are filled to more than 50% of its capacity without saturation.
It is also possible to change the well capacity by choosing either shallow or deep well depth. Normally, filters where the background is very high, require deep mode and short frame times, while very narrow filters, where the background is low, require shallow mode and allow for longer frame times. The following table shows the relationship between well depth, gain and readout noise (assuming a nominal readout noise of 9 ADU/frame). The table also shows the FITS header keyword representing each value.
|Well depth||Gain (e-/ADU)||RON (e-/frame)|
The previous table shows that data will preferably taken in shallow well, since the readout noise will be a factor of 3 smaller than in the deep well. However, in many cases (wide filters and relatively high backgrounds) the use of shallow well is not possible and the observations must be performed in the deep well (high noise) configuration. As with the frame time, it is important to bear in mind that the well depth is not controllable by the user, but it is default to a value pre-defined by optimization during commissioning of the instrument.
The following table shows what we found to be the optimal frame times and well capacity modes for all imaging filters:
|Well depth||Observing mode|
|Si1-7.8||34||deep||imaging and polarimetry|
|Si2-8.7||25||shallow||imaging and polarimetry|
|Si3-9.8||34||deep||imaging and polarimetry|
|Si4-10.3||25||shallow||imaging and polarimetry|
|Si5-11.6||34||deep||imaging and polarimetry|
|Si6-12.5||34||deep||imaging and polarimetry|
|SiC-11.75||17||deep||imaging and polarimetry. Recommended only for low water vapor conditions.|
|N-10.36||17||deep||imaging and polarimetry. Can only be used in very low water vapor conditions (PWV < 3mm)|
|NeII-12.8||34||shallow||imaging and polarimetry|
|NeII_ref2-13.1||34||shallow||imaging and polarimetry|
|SIV-10.5||51||shallow||imaging and polarimetry|
|PAH1-8.6||34||shallow||imaging and polarimetry|
|PAH2-11.3||25||shallow||imaging and polarimetry|
|ArIII-8.99||51||shallow||imaging and polarimetry|
|QW-20.8||17||deep||imaging. Can only be used in very low water vapor conditions (PWV < 3mm).|
Unlike observations in the visible with CCDs, where the gain is typically the same for all filters, in the mid-IR one can have a set of data for an object where the detector gain in a filter will be different to the gain in another filter. The observations of the calibration targets will be made with exactly the same gain as the scientific targets so that data can be properly compared. Nevertheless, CanariCam users are advised to put special care when comparing data taken in different filters to make sure that the comparison is made in physical units. Otherwise, if the comparison is made in ADU, the factor of 3 gain difference between data taken in shallow and deep well must be taken into account.
CanariCam detector array is a Raytheon CRC-774 Si:As IBC, which has 320x240 pixels. Pixels are read vertically in 16 channels, each of them having 20 columns. There are several modes of reading out the detector (S1, S1R3 and S1R1_CR), which are briefly described in here. In Section 1.6 of the CanariCam User Manual there is a comparison between the S1 and S1R3 modes, which favors the S1R3 mode for imaging, particularly with narrow filters, based on lab tests performed during CanariCam acceptance testing.
During on-sky commissioning, the S1R3 mode was compared with the S1R1_CR mode. The S1R1_CR mode was implemented to minimize the impact of the level drop pattern, a type of artifact that appears in images of bright sources when observed with the S1R3 mode (see Section 1.6 of the CanariCam User Manual). The comparison test was performed on August 8th, under poor seeing (> 1.5" in the optical) and PWV (8 mm) conditions. Images were taken in both readout modes with the Si2-8.7, Si5-11.6 and Q1-17.65 filters. The mid-IR standard HD186791 and the planetary nebula NGC7027 were observed to compare the SNR attained in both readout modes for point sources as well as for extended sources. The observations were performed in chop-nod mode, with a chop frequency of 2 Hz and a nod dwell time of 30 seconds. Chopping and nodding were parallel with a PA of 45 degrees to avoid negative images falling within the level drop (see here) below for a description of the artifacts that can appear in CanariCam images) artifacts area. An on-source time of ~70 seconds was used in all images. The analysis described below was performed on the accumulated-signal images produced by CanariCam.
Aperture photometry was performed with an aperture radius of 30 pixel on the images of HR186791 (inner yellow circle in the following figure). The background noise was estimated in two rectangular areas, one outside (green box in the figure) and the other one including (yellow box) the level-drop pattern.
The level drop artifact is clearly seen in the previous images as a repetitive horizontal pattern of opposite sign to the star image. Note that the image contrast has been tuned to show noise features, and therefore the stellar images (positive and negative) appear completely saturated. The level drop artifact is stronger in the S1R3 mode than in the S1R1_CR mode. In the later mode, though, there is an additional correlated noise component that shows up as horizontal stripes all over the image (see here).
In the case of the the planetary nebula NGC7027, aperture photometry was performed in one of the extended features (inner yellow circle in the next image). The background noise was estimated in two rectangular areas, one outside (green box in the figure) and the other one including (yellow box) the level-drop pattern.
The figure above shows how the level-drop noise is more prominent in the S1R3 mode than in the S1R1_CR mode. Also, the channel boundaries are more marked in the case of the S1R3 readout mode.
In all images used for this test, the SNR was obtained by dividing the integrated flux within the circular aperture by the product of the standard deviation of the background in the rectangular boxes multiplied by square root of the area of the circular aperture. The following table summarizes the SNR measured in each filter for each readout mode.
|Object||Filter||Readout Mode||SNR (off level drop)||SNR (on level drop)|
The last two columns in the table show that the difference between the SNR achieved in the S1R1_CR mode and in the S1R3 mode is minor when we measure the noise outside the level-drop pattern area. However, there is a clear decrease in the SNR when the noise is measured in the level-drop feature area. Such a decrease is more acute for bright point sources than for extended sources. The table also shows that the SNR is generally better in the S1R1_CR mode than in the S1R3 mode. Besides, the S1R1_CR mode allows shorter frame times, which is very convenient in situations when the background is high. The S1R1_CR mode also yields more efficient chopping duty cycles. Therefore, at the time of commissioning on June 2011, S1R1_CR was chosen as the favorite readout mode for CanariCam. However, due to the strong horizontal noise pattern which has proven to be a problem to detect faint extended features, the S1R3 mode was subsequently established as the default readout mode for imaging and spectroscopy from March 2013 (see Section New default readout mode).
New default readout mode (March 2013)
S1R3 vs. S1R1_CR readout modes:
At the time of commissioning CanariCam on the sky, the readout mode S1R1_CR was selected as the default readout mode for imaging and spectroscopy. This is a novel readout mode compared with the standard readout mode S1R3 that has been used in other mid-IR instruments using the Raytheon CRC-774 320x240 Si:As IBC array, such as COMICS on Subaru, Michelle and T-ReCS on Gemini and TIMMI2 on the ESO 3.6m telescope. The S1R1_CR mode was originally implemented because it has the advantage of minimizing the effect of the level drop; a set of negative images associated with a bright source that appear horizontally in every of the 16 detector output channels (see section Detector Features and the CanariCam User's Manual). It also has the advantage that it allows for shorter frame times than S1R3, which is useful in high background conditions and a slightly smaller readout noise because the net signal is derived from fewer internal electronic samples. A detailed study of all unwanted effects associated to the Raytheon array can be found in Okamoto et al. 2003 and Sako et al. 2002.
However, the S1R1_CR mode has a horizontal non-Gaussian noise pattern (pickup correlated noise) that is persistent over time (see next two figures).
When computing the noise statistics on the images showing the correlated noise (S1R1_CR mode), it can be clearly seen that the standard deviation along the detector columns is higher than along the detector lines (approximately a factor of 2 on average). However, the standard deviation along lines and columns in the case of the S1R3 mode are the same, since the structure of the noise is rather Gaussian. This pickup noise pattern degrades the detection level of faint objects and extended features. The next figure shows two consecutive observations of the planetary nebula NGC7027 taken with the S1R3 and the S1R1_CR readout modes. Apart from the different readout modes the rest of the observing configuration is the same for both images except for the slight adjustments due to the fact the the frame times are not exactly the same in both modes. The color scale has been stretched between -9 and 9 sigmas of noise in the case of the image in the S1R3 mode and -5 and 5 sigmas in the case of the S1R3 image, where the standard deviation (sigma) chosen in either case is calculated along the detector columns. Note that due to the pickup noise pattern, the faint filament emerging from the main nebula ring at pixel [220, 70] is detected at 5 sigma in the S1R3 image while it is lost within 2 sigma in the S1R1_CR image.
The fact that the pickup noise affects the whole detector surface, tips the balance in favor of using the S1R3 as default readout mode, since the negative effect of the level drop in the S1R3 is very localized (it only affects along the detector X axis).
Sensitivity comparison between readout modes:
We have tested that the change in readout mode for imaging and spectroscopy does not have any negative effect in the sensitivity. The following table shows a comparison between the sensitivity in the S1R1_CR mode and the S1R3 mode in all CanariCam filters.
|S1R1_CR (2012/03/03)||S1R3 (2013/01/29)|
Sensitivities in both cases correspond to 5-sigma detections in 30 minutes on-source. The data for the S1R1_CR can also be found in the section on Sensitivity and its dependency on PWV. The imaging sensitivity values in the S1R3 mode are consistent with previous values obtained with the S1R1_CR mode. In some filters it was not possible to take the measurements in both nights. In some other filters two measurements were taken on the same night. Both values are presented to give an idea of the uncertainties in the sensitivity calculations. The sensitivity is clearly degraded in the Q4-20.5 and Q8-24.5 filters with the S1R3 mode, but this is mainly caused by the highest PWV when the S1R3 images were taken.
On December 31st, 2013, LowRes-10 and LowRes-20 spectra of the standard star HD3712 were taken in both readout modes, S1R1_CR and S1R3. The next two figures show the sensitivity comparison for each of the low resolution gratings, where it can be seen that sensitivities (5-sigma in 30 minutes on-source) are very similar in both readout modes.
Therefore, the change in readout mode from S1R1_CR to S1R3 does not have any negative impact on the CanariCam imaging and spectroscopy sensitivities.
To bear in mind in observations with the S1R3 readout mode:
The change in readout mode is completely transparent to the user when preparing their proposals at the Phase-1 and Phase-2 and when estimating integration times with the CanariCam ITC. All required changes are made internally in the CanariCam control software. The frame times, which are calculated automatically by the control software, will be typically slightly higher in the case of the S1R3 readout mode. One should expect:
- Images and spectra with a more uniform noise pattern than in the case of the S1R1_CR mode.
- A level drop artifact along the detector X axis, at the position of any bright source. The level drop depth is < 1% of the peak emission. In general, if bright sources are going to be observed and the chopping/nodding is on-chip, it is recommendable to use a chopping/nodding along the detector diagonal, to avoid the level drop artifact from the negative and positive images to overlap. The diagonal chopping nodding pattern will have the disadvantage, though, that there will be three stripes of level drop instead of only one. The best chopping/nodding pattern to use will depend on each specific science case.
- The use of longer frame times will make the observations with the wide filters N-10.36 and Qw-20.8 nearly impossible in high background conditions, due to saturation of the detector. Note that this is also the case for the S1R1_CR mode. In general, the N-10.36 and Qw- 20.8 filters cannot be used unless the PWV is below 3 mm.
There are several detector features that must be taken into account when dealing with CanariCam data. Some of the features are always present and some of them only appear in certain conditions.
- 1. Level-drop noise. This is seen as a negative image of a bright source that is repeated horizontally in all 16 channels. Its depth is normally <0.2% of the source peak and only affects the detector rows where the bright source is located. This is why we recommend NOT to use chopping and nodding along the horizontal direction of the detector, particularly for bright sources.
- 2. Detector clipping. This feature has been found to appear when too short frame times are chosen in high background conditions (wide filters and/or high PWV conditions). It appears as a cross-hatched region in the upper left area of the images. During commissioning care was taken to select appropriate frame times for each filter to avoid detector clipping while not saturating the detector. Therefore, this type of feature should normally not appear in scientific data, but it has been include here for completeness.
- 3. Correlated noise. This feature appears as a pattern of horizontal stripes. This noise is characteristic of the S1R1_CR readout mode. The noise in the regions of the detector where this pattern is present is normally 10% higher than the noise in the regions where it is not present. Using appropriate data reduction techniques (cross-correlating savesets before adding them), it is possible to remove this pattern noise.
The following figure illustrates the different type of detector features described above.
Slow Guiding Accuracy
Until fast guiding is available, CanariCam observations must rely on slow guiding (i.e. guide corrections provided by the telescope axes at frequencies as low as 0.2Hz), which is only possible in one of the chopping beams. One of the commissioning tests consisted of measuring the position of a star in the CanariCam detector as a function of time (during 55 min), while chopping, nodding and slow guiding were at work. The star position was measured in both chopping beams and nodding beams, because chop and nod throw were such that the positive and negative images were on chip. The test was performed on June, 22nd using the standard star HD187462 using the filter ArIII-8.99. We used chop and a nod throw of 10 arcsec in the NE-SW direction (PA=45º), a chop frequency of 2 Hz and a time lapse between nods (nod dwell time) of 30 seconds. Guiding was performed at a frequency slower than 0.5Hz, i.e. the guiding images exposure time was 2 seconds.
Centroids were calculated in every saveset difference, which were generated every 1.4 seconds. The following figure shows two typical saveset differences, the left one for Nod A and the right one for Nod B. Each one is the difference between the on-source and the off source chop beams. It can be seen that every saveset difference has a positive and a negative image of the same star because the chopping and nodding were defined on chip.
The following figure shows the offsets with respect to the mean of all centroids within a given nod beam. The upper and lower panels show the X and Y detector coordinates, respectively. Red dots represent the offsets for the positive images in Nod A, while the blue dots represent the offsets for the negative images in Nod A. Red and blue plus symbols represent the offsets for the positive and negative images in Nod B, respectively. The figure also shows the average and standard deviation of all centroids in each type of image (positive or negative) and nod (A or B).
The figure shows that there is a dispersion on the position where the star appears in CanariCam, which is ~0.17" (adding in quadrature both Cartesian components) on average and can be as large as ~0.8" peak-to-peak. As a consequence, the accumulated image generated from these data has a FWHM of 0.44", which is a factor 2 larger than the diffraction limit at 8.99 micron, even though images in individual savesets have a FWHM which is only a factor 1.2 larger than the diffraction limit.
By looking at the average values of the centroids (shown on the right side of the plots), we also note that on average, the position of the positive image in Nod A coincide with the negative image in Nod B, within a fraction of a pixel. This is an indication of the accuracy in the chop and nod throw definition and of the stability of the slow guiding. Nevertheless, the need of fast guiding is clear, not only to correct the jitter between different savesets, but also to ensure that every individual saveset is diffraction-limited.
Last modified: 14 August 2014