WILBUR ASI Calibration Procedure epk/tst
Purpose:
The purpose of the calibration exercise is to: 1. take image data and
perform data reduction to allow users of the ASI data to convert from
raw Data Numbers (DN) to physical units (Rayleighs), and 2., to provide
CSA with arrays of numbers they use in their preparation of the ASI
quick-look data. The required data and formats have been established
over time, and are now the same from year-to-year.
Output Data Products, their nomenclature, and the problems:
1. The "R" values. These are the ASI instrumental responsivity; a single
number, produced for each filter position used. The units of "R" are:
dn/R/sec. The "R" values are submitted in the annual Calibration Report.
Over the period 1986-1995, these data were produced for all four ASI
filters: 557.7,427.8,[737.0 or 607.5], and 630.0nm. However, since the
ASI does not routinely use (and has not since 1989), the 4278 filter
position, no "R" value has been produced for it since 1995.
2. The "P" array. This is the flat-field correction for ASI. It includes
two corrections: the general geometric response of the ASI optical
system, and small-scale imperfections of both the optical system and
CCD. This array is delivered to CSA as an ASCII file of 4096 lines of
16 floating-point numbers, normalized to the peak value, preceded by a
120-byte ASCII header containing a description of the calibration source
data.
Unfortunately, the correction for optical system geometric response
variation does not "work" correctly. That is, when applied to real,
full-FOV "sky" data, it over-corrects for brightnesses near the horizon
(Ze > 75 deg). This effect is not noticeable for cropped data sent in
real-time to CSA, sine the cutoff Ze is about 70 degrees.
3. The "Q" array. This array contains the reference calibration source
image. The array is used by CSA to correct changes in the calibration
over the operating season. The basic idea is that the ASI calibration
source intensity is carefully characterized during the annual calibration
exercise. This data is then taken to be the reference calibration data.
During the operating season, data from the calibration source are
included in the real-time data production. These real-time data are
compared to the reference "Q" data, and the physical units of ASI
responsivity are corrected to Q(field)/Q(lab).
The "Q" array is delivered to CSA as an ASCII file of 4096 rows of 16
floating-point numbers, normalized to the peak value, preceded by a
single value representing the peak value.
Unfortunately, there really is no justification for this exercise,
since the basis is unsound. That is, that the calibration source is
more stable than other parts of the system, and that changes in DN
from the calibration source represent changes in the ASI responsivity.
The ground work required to establish this has simply never been done.
4. Alignment. These data are:
1. The position of the optic axis on the CCD.
2. the effective azimuth of the central column of the CCD array,
3. the Ze of the optic axis (tilt).
Additionally, we would like to identify the azimuth of the tilt, but
since the tilt is usually very small, this quantity cannot be
determined with any certainty.
The numbers are delivered in the annual Calibration Report.
During normal real-time operation, ASI takes images of the sky through
a background (non-auroral) filter (607.5 nm). These have come to be
called "star" images. Data from two clear nights are used, selected
somewhat after the ASI's re-installation in the Fall.
Stars in the ASI images are identified, and their positions on the
ASI images are compared with the known positions of the stars.
Coordinate transformations are performed to obtain the required numbers.
Input Data and data-taking methods
1. For the "R" values, two pieces of information are required: the
absolute response of the ASI to a source of known brightness, and the
effective bandwidth of the filter/camera system.
The first data are taken be setting up our standard brightness source,
LBS-A3, in front of the ASI. This is done by setting up ASI on its side,
complete with weather dome installed, and positioning the LBS about
50 cm from the dome, in the zenith of the ASI.
Data are taken by exposing 10 images at each of several LBS aperture-
wheel settings. Additionally, using the 557.7 nm position, a series
of 10 images is taken for all 10 LBS settings, and a check on linearity.
The linearity data are used for an internal (U of C) consistency check,
and are not published (no requirement).
Images are exposed using the same exposure techniques used by ASI when
in the field: 16 images of 0.104 sec each, summed.
The bandwidth data are taken by setting up the ASI as above, but
changing the target source to a piece of white, translucent paper,
illuminated by a beam of collimated light produced by the Acton
monochromator. The target paper is turned 45 degrees so it can be seen
by the ASI. Five images of the target are taken for each of a number
of wavelength points (typically 30-60). The monochromator is operated
manually via its computer interface.
Dark frames (same exposure, shutter closed), are taken before and
after each data-taking operation.
2. Data for the "P" array are taken from two sources. The geometric response
is measured in the field. We have a special jig at Gillam which holds
the LBS perpendicular to the ASI. The LBS is swung through all Ze as
image data are taken. Ten images are taken at each of about 30 Ze
positions.
Data for the small-scale corrections is taken from the "Q" array.
3. Data for the "Q" array are taken by exposing 100 images of the
calibration source, and 100 dark frames. Traditionally, these data
are taken both in the lab, and in the field. The field data are used
for the delivered "Q" array.
4. data for the alignment information are taken automatically during
ASI real-time operations. Using the ASI summary image plots, two or
three clear nights are identified sometime after ASI re-installation
in the Fall. The "star" data are requested from CSA, who send them
on DAT or 8mm tapes.
Data Reduction
1. "R" values.
The data sets contain a set of five or ten dark frames before and after
each data-taking operation. A program, "make_avg_seq", is used to average
these sets of images together, to produce a single dark frame. This dark
frame is subtracted from ALL image frames, before anything else is
calculated.
Spectral Response
The spectral response of each filter must be measured. The data set
containing the monochromator scans is visually inspected. The location
of the source is found, and a small (typically 3x3 pixel) area in the
source is identified. The data set contains five images taken at each
wavelength. The mean and standard deviation are calculated for the
3x3 target, over five images. This is one wavelength point. The scans
are done such that we begin well above the cutoff wavelength, and
go well below. The resulting data are placed into an ASCII file in two-
column tabular form, and fed to a program called FWHM_new. This program
calculates the centre wavelength, FWHM, and effective bandwidth of the
data. This is done for each filter, yielding BW(eff).
ASI Responsivity
The data set containing the LBS images is identified, and the location
of the LBS within the images is identified. The approximate centre of
the LBS array is identified. A "C" program, "calc" is used to calculate
the mean and standard deviation over a set of ten images, in a sub-array
(typically 7x7 pixel) of the LBS image. This is repeated while walking
the nominal centre of the LBS image in row and column, until a maximum
is found. Using this as the image centre, the data set is fed to
a "C" program called "qcalc", which calculates the mean and standard
deviation over the sub-array, but for the entire data set. This produces
a pair of numbers for each LBS setting, for each filter. The numbers are
the mean and mean standard deviation for a sample pixel. These values
are the ASI raw signal response, "S", and the uncertainty. The "S" value
from the (typically 3) LBS brightness settings are averaged together
to produce a final "S" value. The uncertainty from the brightest LBS
(lowest uncertainty) is used as the uncertainty in the final "R" value.
The "R" values are calculated from the following:
R = (S*K)/(Texp*Blbs*Albs*BWeff) = DN/R/sec, where:
Text = image exposure time in seconds;
Blbs = Spectral brightness (R/A) of LBS at wavelength of filter;
Albs = Attenuation factor for LBS pinhole wheel;
BWeff = filter effective bandwidth in A.
K = ratio of filter response at emission line to peak filter
response. K=1.0 for non-emission-line wavelengths. This is
measured visually by plotting the measured filter response
curve and "seeing" where the emission line falls.
These values are calculated and presented in the Calibration Report.
2. The "P" and "Q" arrays.
The data set containing the LBS Ze scan images for the "P" array
are processed by a "C" program called "superpos" which calculates a
superposition of all of the LBS images over the entire data set. This
results in an averaged, combination image of all of the Ze points.
This image is used as input to a very empirical process of determining
the general shape of the ASI response as a function of zenith angle.
The LBS zenith scan goes from horizon-to-horizon, through the optic
axis, but generally not along a row or column. Using IDL, the endpoints
of the scan are deduced from the superposition image, and then a profile
along the scan is extracted, averaging several parallel strips together to
improve SNR.
A curve fit is performed along the extracted scan, and the coefficients
thereof are used as inputs to input to an IDL program called
"p_and_q.pro" (unfortunately now missing). This program takes in the
average calibration source image, its dark reference image, and does all
the processing necessary for the "P" and "Q" arrays, and writes them out
as properly-formatted files.
3. Alignment
The "star" images arrive in the form of a long, continuous stream of
ASI data in the CSA's "beta" format. Using some perusal tool, the images
are examined for cases where the sky view was clear, and several stars
are visible in the images. Optimally, the stars are close to the
cropping horizon, and are widely spaced in azimuth.
The selected image(s) are extracted from the "beta" data set, and tool
such as IDL is used to extract the position of the stars (in pixel row
and columns terms).
Using an ephemeris, the stars are identified, and the expected
topocentric zenith angle and azimuth for each star is entered into a
table, along with the observed CCD position noted above. This table
is input to a program called "starfit". This program assumes a position
for the optic axis in row/column, and sequentially moved the ASI star
positions around in az, ze, and az(ze), minimizing the RMS error
between the observed star positions and ephemeris-predicted positions.
The program is re-run numerous times for different optic axis positions,
and the results with the minimum rms error are taken to be correct.
The ASI alignment numbers are then Azimuth of CCD column east of
North-South meridian, and the magnitude of the tilt angle. These are
reported in the Calibration Report.
TS Trondsen 2000