Another difference is that IRAC2b does not have a shutter, nor does it shuffle charge around like a CCD, so the concept of a bias frame is meaningless.
The steps to reduce IRAC2b data differ slightly between the two read out methods.
For DCR one:
For NDR one:
Not all these steps may be necessary; for NDR only step 3 is essential and
for DCR steps 1 and 3 are essential.
back to top
The dark frames for the two read out methods differ substantially. Dark frames should be taken at least once per night.
A dark taken with the DCR method has a mean value of -1000 ADU, shows vertical stripes and becomes more negative with increasing exposure time. If you subtract a dark from your data and you are using DCR, it is essential that the dark has the same exposure time as the science frame. The dark current for DCR is not a linear function of time. Note that the camera suffers a slight light leak, so observers should only take dark frames during the night or during the day with the dome dark. The perfect time to take darks is before the afternoon flats.
A dark taken with the NDR method, on the
shows positive counts which increase with exposure time, has considerable
structure and is made up of three components. The first and by far
dominant component is the heat from the output amplifiers, one in each
corner, and depends only
on the number of reads. It is independent of the integration time. The
number of reads increase from a minimum of 2 for integrations of 0.5
seconds to a maximum of 20 for integrations of 6.25 seconds. For longer
integrations the number of reads remains at 20. The second component is the
actual dark current of the array and is less than 0.6 ADU/sec. The
third component is a light leak. The light leak means that observers should
only take dark frames during the night or during the day with the dome dark.
Furthermore, due to a slight memory effect it is not recommended to take
DCR darks after an exposure that has been exposed to a bright source (this
includes the K-band sky). The perfect time to take darks is before the
afternoon flats. If you subtract a NDR dark from your data it must
have exactly the same exposure as your science frames if the exposure of
the science frames is shorter than 6.25 seconds. For longer exposures it
is sufficient to subtract a dark with exposures greater than or equal to
For example it is acceptable to subtract a 10 second dark from an
exposure lasting 160 seconds, but it is not acceptable to subtract a
10 second dark from a 5 second exposure, nor is it acceptable to divide
the 10 second dark by 2 and then subtract it from the 5 second exposure.
back to top
IRAC2b, is a non-linear device. Up to 10,000 ADU the non-linearity is less than 2%. Beyond this flux, the number of pixels that become highly non-linear increases dramatically.
In general, the linearity correction is not required, especially if you keep to counts below 10,000 ADU.
For NDR data the linearity correction is applied to all data. For DCR the linearity correction is applied to dark subtracted data only.
The average non-linearity of the array can be represented by the following formula. The constants in the formula differ for the two read out methods.
Two words of caution. The non-linearity may depend on flux rate and
wavelength. Furthermore, the equation is the average over a small
region of the array. It may not be applicable to individual pixels.
back to top
Although there are four commonly used lenses available for IRAC2b, flats for lenses LA, LB and LC should be taken with lens LC only. Images with lens LA and LB have light concentrations which do not accurately reflect the sensitivity of the array. Flats for lens LD can be taken with either lens LC or lens LD. We recommend you try both and see which of the two you prefer.
In principle it is best to match both the exposure time and the flux level in the flats with what you will get in the observations. However, in practice this is either time consuming or not possible, particularly with flats created by imaging the dome. Flats should be taken at least once during your run. Taking flats with IRAC2 is very easy, so you might consider taking flats every night.
There are three commonly used methods to create flat fields for your data. By far the most common method is to take images of the dome flat-field screen with the flat-field lamp on and off. The recommended sequence of exposures is with the flat-field lamp first off, then two exposures with the flat field lamp on and finally a single exposure with the flat-field lamp off. The flat field is the difference between the "on" and "off" exposures. The usual detector settings are a DIT of 1 second and a NDIT of 30. This method has the advantage of subtracting the dark automatically as well as any contribution from the telescope which can be substantial at K. The disadvantage is that the flat field does not represent the large scale sensitivity of the array. This is corrected with an image which is called illumination correction - more on this later.
Flats can also be created from the twilight sky (sky flats) or from the observations themselves (super flats). In both cases it is necessary to subtract a dark from the flat first. These flats have the advantage that they better represent the large scale sensitivity of the array. The disadvantage of these flats is that the signal from the telescope, which may be non-uniform, is not subtracted. For super flats, there is the additional advantage that the exposure time and flux level matches that of the observation; however, this method is not well suited for the narrow band filters. Furthermore, observations done in lens LA or LB cannot use this method. This method is generally used for deep K-band observations with lens LC.
Firstly, the flat may vary because the illumination of the flat field screen has changed. This can be caused by a change in the position of the flat field lamp. In the quotient of two flats taken with the flat field lamp in differnet positions, the changee appears as a herringbone pattern or as a slope. The peak to peak difference in this image generally ranges from less than 1% for J to less than 3% for K.
Secondly, the differences can occur if the lens wheel changes position very slightly. In the quotient image this usually appears as pattern of arcs near the centre of the image. The effect is less than 1\%.
Thirdly, there can be variations in the sensitivity over certain regions of the array, in particular (x,y)=(200,190), (x,y)=(230,90) and (x,y)=(5,130). In the quotient of the two flats taken two nights apart this usually appears as excess noise in the regions affected. The effect is less than 1\%.
And fourthly, there is a small triangular shaped region of the array near pixel (1,256) which is vignetted. In the quotient image this usually appears as either a region excessively high or excessively low counts. This region cannot be accurately flatfielded. We recommend you either flag this region as bad, or replace the pixel values in this region with a constant.
Over a period of months where the camera has been taken off and then put
back on the telescope and where the detector has been temperature cycled,
variations in the flat-field can occur for all the reasons outlined above,
especially the third. As an example, the RMS variations of the quotient
of two flats taken two months apart is shown for J, H, K and K' flats in
the following table.
|Filter||RMS of the Quotient Image|
J Flat ffjn980320.fits.gz (220 kb)
H Flat ffhn980320.fits.gz (220 kb)
K Flat ffkn980320.fits.gz (220 kb)
K' Flat ffkpn980320.fits.gz (220 kb)
back to top
The flat fields generated by methods outlined above do not represent the true large scale sensitivity of the array. The large scale sensitivity can be corrected by scanning a bright star, usually a standard, over the array, fitting a surface (3rd order in x and y without cross terms is usually sufficient) to the normalised fluxes (not magnitudes), and then multiplying this surface into the flat field. For lens LA and LB, 9 positions should be sufficient, for lens LC and LD, 16 positions are required. In the following table we list the residual error in instrumental magnitude of the scanned star before and after the correction. The table is complete for lens LC, partially complete for lens LB. There is no data for lenses LA and LB.
All flat fields were taken in lens LC.
|Lens/Filter||Dome Flat uncorrected||Dome Flat Corrected||Sky Flat uncorrected||Sky Flat corrected|
If a consistent zero point (ie better than 5%) over the entire field of view is important for your work, then an illumination correction is essential.
The illumination correction depends slightly on telescope position. The second column in the following table shows the residuals in the magnitude of a bright star scanned over sixteen positions for five different telescope positions. The third column shows the residuals after a 3rd order surface was fitted for each individual telescope position. The final column shows the residual if the fit at Zenith is used to correct for all telescope positions.
Thus one can expect the zero-point to be consistent to better than 2.5% over
the array for reasonable telescope positions.
|Lens/Filter||Telescope position||Before Fit||Residuals of the Fit||After ZD Fit|
|LC/K'||10 degrees S. of Zenith||0.047||0.015||0.017|
|LC/K'||43 degrees W. of Zenith||0.036||0.016||0.028|
|LC/K'||27 degrees N. of Zenith||0.040||0.016||0.020|
|LC/K'||40 degrees SE. of Zenith||0.042||0.012||0.022|
|LC/J||43 degrees W. of Zenith||0.060||0.021||0.028|
|LC/J||27 degrees N. of Zenith||0.064||0.017||0.020|
|LC/J||40 degrees SE. of Zenith||0.069||0.013||0.016|
The illumination corrections also change between runs. The RMS variations in the quotient images can be as high as 3%, although it is not entirely clear how much of this variation can be attributed to errors in the surface fit. In general the largest variations are for K and K' and occur in the corners of the array.
J Correction illumj970316.fits.gz (200 kb)
H Correction illumh970316.fits.gz (200 kb)
K Correction illumk970316.fits.gz (200 kb)
K' Correction illumkp970316.fits.gz (200 kb)
back to top
Sky subtraction can vary from the simple, subtracting one frame from another, to the complex, creating a running sky from a large number of frames. Rather than going into details of the reduction process itself, it is better to pass on a few suggestions The number of sky frames one observes, the pattern of observations and the frequency one switches between object and sky positions is crucial for good sky subtraction. I will describe two common methods.
There are two points to keep in mind. First, the more sky positions the better, and second, one must interleave the observations of the sky with observations of the programme object. Here is one example. Consider a single field we wish to observe. Keeping the above two points in mind we can construct a series of exposures which takes sky frames that are East and West (or North and South) of the object field. One can also construct a sequence of exposures where the skies are East, West, North and South of the object. After aligning the telescope with the object we first take a sky frame to the East of the object (we could have also started to the West), second we observe the object, third we observe the object again (a small offset can be inserted between the second and third exposures), fourth we take a sky frame to the West of the object, fifth we take another sky frame near this same place (an offset between the forth and fifth exposures is essential) , sixth we observe the object, seventh we observe the object again (an offset between the sixth and seventh exposures can also be performed) and finally we take a sky frame East of the object and near our first exposure. If the return to origin flag is set, the telescope will move back to the object. Thus one will have four exposures centred on the object and four exposures centred on the skies that are east and west of the object.
More complicated schemes can be implemented. For example, each sky exposure in the above sequence can be divided into two separate sky exposures, with each exposed for half the time of the original and with a small offset in between. Thus one would have 8 exposures on the sky and 12 exposures in total. With this pattern one still spends the same amount of time on the sky as on the object. This pattern is useful if the field is crowded.
The detector in IRAC2b contains a lot of bad pixels, approximately 700. In
general the number
of bad pixels is flux and wavelength dependent. There are more bad pixels
at low flux levels. The list
of bad pixels is defined by deviding a high
flux exposure with a low flux exposure and selecting all pixels that
deviate from the local background by more than 10%.
Fabry Perot Data
back to top
The Fabry Perot unit sits outside the cool environs of the camera. As the FP sits in the air, all vacuum wavelengths must be multiplied by 1.000273. The throughput varies between 70 and 80% over the 2.0 to 2.5 micron range and the finesse is greater than 50 for most of this range. It has a wavelength resolution of 0.0016 microns.
Obtaining and calibrating data taken with the Fabry-Perot requires some care. Flat fields are required for all wavelength settings and a bright reflection arc near the Eastern edge of the array (only seen in lens LC images) makes the array difficult to flat field in this region.
For estimating line fluxes it is possible to achieve accuracies of 5% with six wavelength settings, four near the line and two well separated from the line centre and on either side. Note that separate skies are required for each wavelength setting. Using images at different wavelengths as skies will lead to less accurate results.
Obtaining velocities requires even greater care. The wavelength of maximum
transmission varies across the array. If accuracies better that
200 km/s over a large part of the array are required, then this wavelength
shift must be corrected. This contour plot shows the wavelength shift in
terms of velocity. It is also available as fits file. The area of
missing data is caused by the reflection arc.
back to top
The IR window between 1 and 2.5 microns contains several large absorption troughs that are primarily due to water vapour in the atmosphere. As the amount of water vapour varies so will the amount of absorption. Unfortunately, the edges of some IR filters, particularly J and K, are defined by these absorption features. Thus, when the column density of water vapour is variable, good photometry can be difficult to achieve. On good nights (generally when the humidity is low) it has been possible to achieve better than 1% photometry; however, on most nights this should be considered as the lower limit.
To get the best photometry you should choose standard stars that are as close as possible to your programme objects and you should observe them before and after you observe your programme objects. During the night you should observe at least three different standards and you should observe at least one standard every two hours. On nights where the humidity (which is a rough measure of the water column vapour) is varying considerably (let's say by 40\% in one hour) you will need to observe standards more frequently.
Generally, a standard star is observed twice in different parts of the array. Thus one observation acts as the sky of the other.
It is important that you take care where the standard star lies on the array. Use the PPS display to find areas that are clear of bad pixels. Furthermore, check what other objects are in the field. Furthermore, the offset between exposures should not be such that other objects in the field interfere with flux of the standard when one frame is subtracted from another. Alternatively, one can take several exposures each with the standard star in different location. Observers commonly use a sequence of five exposures. One in the centre of the array and one in the centre of each of the quadrants.
Note that most standards were observed with single channel photometers with a very wide aperture. Thus close, companion stars were probably included.
There are several standard star lists at the telescope. Many of the standards are far too bright for IRAC2b. It is important not to let the value of any pixel to exceed 10000 ADU. If this limit is exceeded, try defocussing the telescope. As a guide a ninth magnitude star imaged under 0.8" seeing will reach this count in 1 second for lens LC.
The following ZPs are for DCR and lens LC. The ZPs before the aliminisation took place were taken by a variety of observers and reduced with a variety of techniques. They should be used as a rough indication of the ZPs at those times.
|14th December, 1997||DCR||ffjn971214.fits.gz (220 kb)|
|7th November, 1997||DCR||ffjn971107.fits.gz (220 kb)|
|7th July, 1997||DCR||ffjn970707.fits.gz (220 kb)|
|16th March, 1997||DCR||ffjn970316.fits.gz (220 kb)|
|16th January, 1997||DCR||ffjn970116.fits.gz (220 kb)|
|14th December, 1997||DCR||ffhn971214.fits.gz (220 kb)|
|7th November, 1997||DCR||ffhn971107.fits.gz (220 kb)|
|7th July, 1997||DCR||ffhn970707.fits.gz (220 kb)|
|16th March, 1997||DCR||ffhn970316.fits.gz (220 kb)|
|16th January, 1997||DCR||ffhn970116.fits.gz (220 kb)|
|14th December, 1997||DCR||ffkn971214.fits.gz (220 kb)|
|7th November, 1997||DCR||ffkn971107.fits.gz (220 kb)|
|7th July, 1997||DCR||ffkn970707.fits.gz (220 kb)|
|16th March, 1997||DCR||ffkn970316.fits.gz (220 kb)|
|16th January, 1997||DCR||ffkn970116.fits.gz (220 kb)|
|14th December, 1997||DCR||ffkpn971214.fits.gz (220 kb)|
|7th November, 1997||DCR||ffkpn971107.fits.gz (220 kb)|
|7th July, 1997||DCR||ffkpn970707.fits.gz (220 kb)|
|16th March, 1997||DCR||ffkpn970316.fits.gz (220 kb)|
|16th January, 1997||DCR||ffkpn970116.fits.gz (220 kb)|
|16th March, 1997||illumj970316.fits.gz (200 kb)|
|16th January, 1997||illumj970116.fits.gz (200 kb)|
|16th March, 1997||illumh970316.fits.gz (200 kb)|
|16th January, 1997||illumh970116.fits.gz (200 kb)|
|16th March, 1997||illumk970316.fits.gz (200 kb)|
|16th January, 1997||illumk970116.fits.gz (200 kb)|
|16th March, 1997||illumkp970316.fits.gz (200 kb)|
|16th January, 1997||illumkp970116.fits.gz (200 kb)|