7 Imaging - Procedures & Database Products

 

This chapter deals with these elements:

Flat field:

An image of a uniformly illuminated (``flat'') field. (In this section, ``flat field'' refers specifically to imaging mode.) Flat field images will be assumed to be normalized to unity at a specified ``normalization point'' (which in practice will be a cosmetically-clean region on the fiducial CCD). Flat fields are used to remove the signature of the CCDs (``pixel-to-pixel variations'') and the system throughput variations (``vignetting''; others?) across the field of view. Flat fields are identified by their source of illumination -- twilight flats of the twilight sky, dome flats of the dome illuminated by special lamps, and internal flats of the hatch illuminated by lamps within DEIMOS. In imaging mode, a flat field is alwyas associated with a particular filter and is useful only with that filter.

Fringe frame:

A flat-fielded, normalized image of the dark sky (with the continuum component subtracted), showing a summation of interference fringes caused by strong night-sky emission lines. The stability and/or utility of fringe frames is yet to be established.

Relative flux calibration:

Adjusting counts across an image to the same count-vs-flux ratio. This is traditionally performed by division by a flat field image, which has been normalized to unity at a specified ``normalization point.''

Absolute flux calibration:

Conversion of count rates into physical flux units. This consists of two constants for each filter, a zeropoint ( ) and an extinction coefficient ( ). The defining relation is:

where x is the airmass at the time of observation. , sometimes refered to as total throughput, contains all the information about system efficiency or throughput for the normalization point of the flat-field image.

Standard magnitudes:

Magnitudes on a widely accepted system such as Johnson UBV or the SDSS sytem. Conversion from instrumental to standard magnitudes traditionally takes the form

The coefficients , and can be deduced from observations of standard stars, or ``synthesized'' using measured transmission curves and published spectral energy distributions for a wide range of spectral types.

Relative Astrometry:

Determination of the relative positions, in arcseconds on the sky, of positions in an image. This can be expressed as a distortion mapping between the detector and the gnomonic projection of the refracted sky (corresponding to reverse transforms of D,A,C in the RSCS ICS mapping of Chapter 2). We assume independfent polynomial fits in and :

Absolute Astrometry:

Celestial coordinates tied to a fundamental system, and the procedures for determining such coordinates.

7.1 Background and Issues

Flux Calibration:

[To be written]

Defringing:

[To be written]

Astrometry: There are three levels of astrometry, all using astrometric mapping (CCDCSi ICS FKCS) that will be available from the database. The first level uses telescope telescope pointing information from the image header -- expected relative accuracy should be good to [ 1 arcsec] ignoring differential refraction and aberration; absolute accuracy is poor [ 10 arcsec]. The second level adds the measurement of one star (or more) of known coordinates -- this improves the absolute accuracy to the level of relative accuracy [ 1 arcsec]. The third level requires three or more astrometric reference stars in the image, and provides a correction to the database astrometric mapping. The precision of this level is the minimum required for slitmask design. It is expected that three reference stars will be adequate to obtain 0.1 arcsec accuracy. [Review all the accuracy estimates. The biggest unknown here is the exact pixel scale, and whether it will vary with and/or focus.]

7.2 Procedures

In the following, ``relevant'' refers to images and constants that are associated with the filter used in the observation.

Relative flux calibration (Flat-fielding): Retrieve the relevant flat-field image from the database; divide.

Defringing: Retrieve relevant fringe frame from database; fit to sky fringes by scaling; subtract.

Absolute flux calibration: Retrieve the relevant constants (zeropoint and extinction coefficient) from the database; calculate the constant of proportionality ( ); place constant in the header for use by analysis applications. [What about gain - should this be included??

7.3 Database Products

A library of standard flat-field frames through each filter will be maintained for each CCD. [In practice, this may be a product of pixel-to-pixel flats and large-scale flats.]

Nominal system throughputs for each filter will be maintained in the database.

Standard extinction values for Mauna Kea will be maintained in database.

Transforms from instrumental to standard photometric systems will be maintained in database.

If useful/stable, a library of fringe frames for the night sky will be maintained for each filter.

Distortion mappings will be maintained in the database.

NOTES -- Delivery: 0=Commissioning, 1=Pipeline.

Required Hardware: - grid-of-holes (GOH) mask for differential distortion map.

7.3.1 To create a flat field image (method 1):

For each filter:

1.

Take 3 or more twilight flat frames with adequate exposure. The telescope should be tracking during this sequence, but offset between exposures.

2.

Reduce each through bias-level correction.

3.

Combine by taking the median level at each pixel (or use some other robust level determinator that rejects cosmic rays). The individual flats should be scaled by their mean/median levels in a cosmetically clean region before combining, to remove differences in sky intensity.

4.

Normalize to unity at the normalization point.

Advantages: Straight-forward; illumination by sky; better ``color balance.'' Disadvantages: Difficult to obtain images with proper exposures (a model of sky counts per filter per sun altitude would help!); narrow time window; stars in field often leave residual features.

In the above, adequate exposure depends on the accuracy desired. Usually ;SPMlt;1% accuracy is the goal, which means electrons level in each image is adequate.

7.3.2 To create a flat field image (method 2):

For each filter:

1.

Take N dome flat frames with adequate exposure.

2.

Reduce each through bias-level correction.

3.

Combine by taking the median level at each pixel (or use some other robust level determinator that rejects cosmic rays). The individual flats should be scaled by their mean/median levels in a cosmetically clean region before combining, to remove fluctuations in lamp intensity.

4.

Remove large-scale variations by dividing the combined image by a heavily smoothed version of itself -- this leaves a normalized image of pixel-to-pixel variations (``pixel-to-pixel flat'').

5.

Obtain one (or more) twilight flat(s). (Combine multiple images by first scaling to same level, then taking median value at each pixel.) Smooth and/or spatially-median filter or fit by a surface to retain only large-scale variations (``large scale flat'').

6.

Multiply the large scale flat by the pixel-to-pixel flat.

7.

Normalize to unity at the normalization point.

Advantages: Dome flats easy to acquire; pixel-to-pixel variations can be monitored and combined with large scale flats to update the database more often. Disadvantages: Color balance not as good as with sky.

As above, a goal of ;SPMlt;1% accuracy requires (electron level) .

7.3.3 To create a fringe frame:

For each filter:

1.

Take 3 or more deep frames in darktime with significant sky counts. The telescope should be tracking during this sequence, but offset between exposures. The field should contain as few stars as possible.

2.

Reduce each through relative flux calibration.

3.

Combine by taking the median level at each pixel (or use some other robust level determinator that rejects stars and cosmic rays). The individual frames should be scaled (or offset?) by their mean/median sky levels.

4.

Smooth on scales of a quarter the fringe wavelength.

5.

Normalize to unity.

7.3.4 To measure system throughput:

For each filter:

1.

Take exposures of N photometric standard stars of varying spectral type; process through relative flux calibration.

2.

Determine the count rate for each standard star. Adjust for expected atmospheric extinction. (Adjust for gain.)

3.

[?? The method of defining the instrumental system has not been investigated yet.]

7.3.5 To create a distortion map:

1.

Take one or more images of an astrometric field; process through relative flux calibration.

1a.

Take one or more images of the grid-of-holes mask before/after to allow for future differential updates (dome or twilight sky illumination).

2.

Measure the pixel positions of stars with known celestial coordinates.

3.

Fit the pixel coordinates to the gnomonic projections of refracted sky coordinates; fit the same sky coordinates to the pixel coordinates. [Functional form not yet known, but probably low-order polynomial will be fine.]

7.3.6 To create a differential update to distortion map:

1.

Take one or more images of the grid-of-holes mask (dome or twilight sky illumination).

2.

Measure the pixel positions of holes in mask; compare with previous grid-of-holes measurements.

3.

Fit the differences in pixel location; this becomes the differential update to be used in conjuction with the distortion map.

7.4 Other Products

Instrument-control scripts and instructions to create the database products listed above will be required.

7.5 Outstanding Issues

What are the standard DEIMOS filters? Do instrumental-standard conversions exist for filters of these specific design? Should the zeropoints follow Johnson, AB, or some other magnitude system?

What are the extinction values for MK? (See Keck Web pages)

Do we want a more elaborate scheme for characterizing the color dependence of pixel-to-pixel variations? This could be done spectroscopically.

There is no discussion of fringing here. This awaits analysis of fringing data.

Is distortion mapping the same for all filters, modulo scale change? This can be checked with grid-of-holes comparisons in different filters.

[The role of gain is not fully developed in absolute flux cal discussions.]

We need to discuss the effect of spatial distortion (different pixel scales throughout the image) on flux calibration. If the distortion is mapable (it should be!), there is no problem correcting the data for this.

How wide is the time window for twilight flats? How many filters can we expect to do in each window?