Ast w392G

1
Ast w392G

• Lecture 3 Calibrations Part I
• Bias, Flat Fields, Darks, Fringe
• Frames, and Focus

2
Preliminary Processing

• There are two types of instrumental signature to
  remove
• Additive
• Bias Level
• Bias Structure
• Dark Counts
• Multiplicative
• Q.E. variations on all scales
• Data processing, especially flat-fielding, is as
  much art as science.

Constant of counts added independent of the
brightness of the source(s).
Constant fractional effect
3
Bias Correction

• Bias level and any gradient along columns is
  taken out via overscan subtraction.
• Bias structure is taken out by subtracting a
  zero-level frame.
• In IRAF ccdproc takes care of both.

4
Overscan

• After reading out the real pixels in a row, CCD
  electronics continue to read out virtual pixels
  and record the bias level and read noise of the
  amplifiers. These virtual pixels are called the
  overscan region.

Amp A/D
Active area of CCD
Row or line
Column
overscan region
5
Multi-amplifier overscan
The overscan regions for multiple amps are often
placed together at one end of each row during
post-readout processing.
6
E.g., Four-amplifier overscan
Amp 1
Amp 2
Amp 3
Amp 4
7
Overscan subtraction

• First you need to identify the relevant columns.
• In IRAF use the format
• x1,x2y1,y2 e.g., 10451054,11024
• Often want to avoid first couple columns due to
  noise (typically from capacitors in circuitry)

Sky level
Overscan1
Oscan2
In IRAF, plots like this are made using implot
8
Colbias

• The overscan subtraction is accomplished by
  fitting a smooth function to the average of
  several columns in the overscan region.
• The value of the fitted curve is subtracted from
  each row. This accounts for a mean bias level and
  any gradient in along columns.

9
Smooth fit is subtracted from each row
Average several
10
Applying a bias (zero) frame

• Most modern ground-based CCDs have little
  significant structure in bias frames after
  subtracting the overscan
• If structure is significant
• Average several (10) bias frames
• Subtract from program frames (can use ccdproc)
  before further processing

11
Trimming an image

• Trimming is used to remove overscan and other
  non-useful segments of an image
• All images must be trimmed to same image segments
  as flat fields
• Trim after subtracting overscan
• Use ccdproc, which can overscan-subtract and trim
  at the same time.

12
Why Flat Fields?

• Each pixel has unique QE you want to normalize
  to a single value.
• Each chip has large-scale QE fluctuations due to
• Dust on Filter
• Vignetting
• Uneven coatings and thinning
• Other issues
• You want to take this out, too!

13
Flat Fielding

• If you illuminate the CCD uniformly, then
  normalize the mean to 1, this image could be
  divided into every frame.
• Want high counts, I.e. 106 e- gives (in theory)
  0.1 flat-fielding
• For direct imaging, usually use a combination of
• Dome Flats
• Twilight Flats
• Night Sky Flats
• For spectroscopy, use dome flats or internal
  flats (will talk about much later in class)

14
Dome Flats

• Put some quartz (hot, continuum source) lamps on
  the telescope and illuminate a white screen or
  spot on the dome.
• These often dont work very well for two reasons
• The lamps are always too cool (red)
• The dome is not even close to infinity and
  usually illuminates the primary differently than
  the sky
• But, you can collect a lot of photons during the
  day

15
Twilight Flats

• These often work pretty well
• The Sun is pretty hot, the scattering surface
  illuminates the telescope just like the dark
  night sky
• Doesnt use dark time
• Sky gets dark quickly often hard to get enough
  images with enough counts in enough filters
  before sky is too dark.

16
Night-sky Flats

• These tend to work very well. They match the sky
  illumination perfectly
• They sometimes require useful dark time
• If your objects are small and not concentrated
  (like distant galaxies or open star clusters),
  combine your actual data.
• They sometimes contain fringes

17
Stars and Galaxies

• For twilight and night sky flats you have a
  problem they contain stars and galaxies.
• Move the telescope (dither) between exposures
• Make a non-registered stack of the frames in each
  filter.
• Stars are present even when you cant see
  them!!!!
• Median (or better yet minmax rejection) in the
  frame combining eliminates stars and galaxies in
  the combined flat.

18
Tips for taking twilight flats

• Make sure you are several degrees out of the
  Galactic plane
• Remember the sky is getting darker (or brighter)
  youll need to lengthen or shorten each exposure
  (usually 1.5x or 2x the previous exposure time)
• What order? If doing UBVRI in twilight, start
  with U and B, then order not so important (though
  it might be best to do I next). Reverse in dawn.
• Check your sky levels and avoid saturation!
• Dont forget you need flats for each filter!

19
Minmax rejection
Average/median combine frames with minmax
rejection reject at least 2 highest values.
NOTE! Must normalize frames to common mean or
mode before combining!
20
Combining Frames

• In IRAF, imcombine is the task to combine frames.
• combineaverage (or median)
• rejectminmax
• scalemode
• nlow0
• nhigh2

21
Normalizing Flat Fields

• Each of your flat fiield images likely has a mean
  of 10,000-30,000 counts.
• When applying flat fields, goals are
• Remove QE variations
• Preserve flux noise statistics
• Need to normalize flat to mode of 1. Three
  choices
• Use imarith to divide flat to make a normalized
  flat
• Use imdiv with rescale set to numerator
• Use ccdproc with flat field correction

22
Flat fielding tricks

• Dome flats (high counts, bad illumination) are
  good for correcting pixel-to-pixel variations.
• Night sky flats (low counts, excellent
  illumination) are good for correcting large-scale
  variations.
• Spatially smooth (or fit low-order surface to)
  both combined dome and combined dark sky.
• Remove dome low-spatial-frequency pattern
  Dome?smoothed Domepixel-to-pixel variations
• Smoothed night sky is low-frequency response.
• Best of both worlds is (Dome / smoothed Dome) x
  smoothed night sky
• NOTE Smoothing does poorly at amplifier
  junctions fit each piece separately

23
I-band
PFCam flat fields
Dust on filter
Note differences with color. This means that
objects with different spectra will be
flat-fielded slightly incorrectly.
V-band
Rings due to non uniform thinning
U-band
Two amplifer readout
24
(No Transcript)
25
Flat-field tests

• Divide flat field into component images first
  look for strange structures (stars/galaxies
  okay). Reject individual odd frames and re-build
  flat-field.
• Take cuts through your flat-fielded frames and
  make sure the sky is flat (check corners). IRAF
  implot
• In blank areas, make sure the pixel-to-pixel
  variations are consistent with shot noise from
  the sky level. IRAF imexam and the m key.

26
One caveat of flat fields

• If pixel scale varies significantly (such as
  in many mosaic cameras), flat-fielding hurts
  photometry
• You are forcing sky counts to be the same
• Larger pixel scale more sky counts ? more
  object counts
• Forcing flat sky artificially dims real objects
• Cannot ignore! (Mosaic reduction routines often
  correct for this, but make sure!)

27
Dark frames

• If dark current is significant, take several dark
  frames
• Expose as long as longest exposure without
  opening shutter
• Take shorter darks and see if dark current is
  linear with exposure time
• If so, take several long dark frames, average
  them.
• Scale by science frame exposure times and
  subtract from science frames
• Make sure dark frame is stable over time
• If not, need to take dark frames during night.

28
Fringe Frames

• At long wavelengths or with narrow-band filters,
  a fringe pattern appears
• Amplitude 1
• Interference pattern from internal reflections
• Due to emission lines in night sky
• Intensity can vary from frame to frame
• Is an additive signature
• Does not usually appear in flat fields

29
Removing Fringing

• minmax combine flat-fielded night sky images to
  make an image of sky without sources
• Heavily smooth image to create image without
  fringes
• Subtract the two to make a fringe map
• Scale fringe map to each individual image and
  subtract (ccdproc can do this)

30
Telescope Focus

• Whether you focus yourself or the telescope
  operator does it for you, you need to always be
  checking radial profiles.

Plot intensity values for pixels that intersect
circles
31
Poor focus (scatter is lack of roundness and
probably astigmatism)

• Using the r command in IRAFs imexam, you can
  produce radial profiles for any object in a
  frame. .snap eps in the graphics window will
  output an .eps file.

Excellent focus. Very round (and dangerously
sharp) image
32
Galaxy
Saturated star
Well-focused star
33
Way out of focus (donut).
34
Focus frames

• Set focus, start exposure.
• Pause exposure, move telescope, change focus
• Repeat
• Make a double telescope move on the last focus
  value, then read out

35
Splitting Exposures

• How long to expose? Once in the sky-limited
  regime, the S/N only depends on the total
  exposure time. There is only the CCD readout time
  penalty to be paid by splitting long exposures
  into multiple shorter exposure.
• Why do shorter exposures?
• Cosmic ray rejection
• Increase dynamic range
• in-field dithering along slit or on the sky
  can help with flat fielding
• Reduce risk