UVIS calibration update
Greg Holsclaw, Bill McClintock
June 18, 2012
Braunschweig, Germany
Extended source vs point source
• An extended
source appears
to exhibit flat
field effects
• The total signal
from a point
source, as a
function of
position on the
detector, does
not
The total signal in the two curves is equivalent, but the red line
has more data points.
Local mislocation of counts?
Say a photoevent
located in this pixel is
counted by the pixel
above
• Hypothesis:
photoevents that
occur within the
geometric area
of an adjacent
spatial pixel are
erroneously
counted
Local mislocation of counts?
80μm
100μ
m
140μ
m
60μm
120μ
m
• The effect of these
mislocations is a
change in the
effective width and
position of spatial
pixels.
• The flat-field
variation is caused
not by changes in
QE, but by changes
in effective area.
Spica slow scans
• Objective
– Characterize the geometric response (“effective height”) of
each pixel in the spatial dimension
– Develop a detector model and calibration methodology
that separates geometric effects from sensitivity effects
• Observations
– Standard calibration scans slew the star at a rate of
0.020mrad/sec (0.9rows per 45sec integration period)
– Slow scans use a 5x slower slew rate of 0.004mrad/sec
(0.18 rows per 45sec integration period)
– Because this takes longer, coverage of the complete
detector was broken into four separate observations
spanning ~1 year
Coverage
Date: 2011-329
Rows: 51 - 61
Date: 2011-329
Rows: 37 - 53
Date: 2012-071
Rows: 18 - 34
Rows of
‘height’,
vertical,
spatial
Date: 2012-124
Rows: 3 - 18
Columns of ‘width’, horizontal, spectral
Pixel output vs time/position
-500
-1000
time (seconds)
0
•
1000
500
1.0
pixel [900,25]
normalized signal
0.8
0.6
Pixel output vs
time as the star
image is
scanned across
•
0.4
0.2
-5
-4
-3
-2
-1
0
rows
1
2
3
4
5
This time series
represents the
convolution of
the star spectrum
image with the
pixel response
A gaussian
function can be
fit to this curve to
arrive at a time at
which the image
crossed the pixel
and the “effective
height” of
response
Pixel output vs time/position
-500
-1000
time (seconds)
0
1000
500
1.0
• FUV pixel
height is
systematically
larger at
shorter
wavelengths
pixel [100,25]
pixel [900,25]
normalized signal
0.8
0.6
Pixel output vs
time as the star
image is
scanned across
0.4
0.2
-5
-4
-3
-2
-1
0
rows
1
2
3
4
5
Effective pixel height map
EUV
FUV
• This shows the convolution of the vertical image with the vertical pixel response
• Interpretation:
• At long wavelengths, the image height is smaller than a pixel
• At short wavelengths, the image height is larger than a pixel
• Though this describes the vertical image size, it is likely the spectral image width is
similar
• Thus, the spectral resolution should be worse at the lower left of the detector
• Impact on spectral modeling?
Rows are of systematically different
heights
• Individual
rows can
exhibit
systematically
larger
effective row
heights than
adjacent rows
1.6
row 54
row height (rows)
1.4
1.2
1.0
0.8
row 53
0.6
0.4
0
200
600
400
column
800
1000
Variability in pixel height is repeatable
effective row height for row 50
1.6
observation 1
observation 2
row height
1.4
1.2
1.0
0.8
0.6
0
200
400
600
800
1000
800
1000
column
height minus smooth by 21
difference
0.4
0.2
0.0
-0.2
-0.4
0
200
400
600
• Row 50 was
covered by
both
observation
1 and 2
• Structure in
the effective
height
calculation
appears
similar
Variability in pixel height is repeatable
• Plot of effective pixel
height for each pixel in
row 50 against that
derived from an
overlapping observation
– After smoothing by 21.
That is, a plot of the
two vectors in the
lower plot of the
previous slide against
each other.
• This illustrates the high
correlation between the
two, suggesting real
variations in effective
height between pixels
Pixel heights for one column
1.4
column 800
effective height (rows)
effective height (rows)
1.4
1.2
1.0
0.8
0.6
0
10
20
30
40
row number
EUV
50
60
column 800
1.2
1.0
0.8
0.6
0
10
20
30
40
row number
FUV
50
60
signal (counts/second)
Evil pixel response
-1000
-500
500
25
20
1000
good [783,25]
evil [783,45]
15
10
5
0
-5
-4
-3
-1000
normalized signal
time (seconds)
0
-2
-500
-1
0
rows
1
time (seconds)
0
2
3
500
4
5
1000
0.14
0.12
0.10
0.08
0.06
0.04
0.02
-5
-4
-3
-2
-1
0
rows
1
2
3
4
5
• Evil pixels
exhibit
similar
effective
heights as
good
pixels
Burned pixel response
400
600
200
time (seconds)
-200
0
normal pixel [900,20]
burned pixel [900,31]
count rate
15
10
5
0
-3
1
0
row
time (seconds)
-200
0
200
400
3
2
-1
-2
600
count rate (normalized)
-600
-400
-600
-400
0.15
0.10
0.05
0.00
-3
-2
-1
0
row
1
2
3
• Effective
height of
burned pixels
seem
comparable
to normal
pixels
Time image crossed each pixel in a row
time at which image crossed row 50
150
0.6
100
0.4
50
0.2
0
0.0
-50
-0.2
-100
-0.4
-150
-0.6
1000
0
200
400
600
column number
800
rows
time (seconds)
-0.00063 rows per column
• Spectral tilt
of 0.64 rows
across the
FUV
detector
width
Time image crossed each pixel in a row
0.6
0.4
observation 1
observation 2
50
0.2
0
0.0
-50
-0.2
-100
-150
800
-0.4
-0.6
1000
800
1000
0
200
400
600
column number
time minus linear trendline
difference
100
50
0
-50
-100
0
200
400
600
rows
time (seconds)
time at which image crossed row 50
150
100
• Pixel to pixel
variation
appears to be
reproducible
• Pixels are
systematically
offset from a
nominal
position
Time image crossed each pixel in a row
• Plot of the time the image
crossed each pixel in row
50 for one observation
against that derived from
another observation
– After subtracting linear
trend. That is, a plot of
the two vectors from the
lower plot of the
previous slide against
each other
• This illustrates the high
correlation between the
two, suggesting real
displacement in center of
response
Applying an effective height flat-field corrector
Standard calibration,
raw data summed in
time
Divided by row height
Multiplied by Steffl preburn corrector
row 20 height
2.0
row height
1.5
1.0
0.5
0.0
0
200
400
600
column
800
1000
Conclusions
• Individual pixel elements appear to exhibit
unique effective heights and positions
– Based on reproducibility between observations
• Evil pixels and burned pixels have similar
effective heights as good pixels
• A row-to-row corrector using the average row
height improves the data, but not to the
extent as the Steffl flat-field corrector
Calibration observations
• Steffl calibrations once/year, gamma Vel
– Mike Evans says ok
• Standard calibration once/year, alpha Vir
– Mike Evans says ok
• Desired: one inter-comparison effort, standard
calibration of both stars within a month
– Ask Mike Evans if this is possible