Transcript calibration frames - Ottawa Centre, RASC

TECHNIQUES FOR IMAGING VERY FAINT OBJECTS
NO MAGIC, JUST DO EVERYTHING RIGHT
(or at least try to!)
by Paul Boltwood
ITS November 8, 2002 Salem Oregon
This talk is on the conference CDR
Copyright (C) 1996-2002 Paul Boltwood
Slightly updated January 15, 2006
1
SKY & TELESCOPE DEEP FIELD CONTEST
• one of the reasons I was invited here was that I won this contest
with an image that reached mag. 24.1
• see Bradley Schaefer’s articles S&T May 1998 and May 1999
• I should not have won because I had:
• suburban skies
• a CCD with 40% peak quantum efficiency
• 16" aperture
• done a crude reduction
• many amateurs surpass me in all of these departments
• as soon as I heard that I won, I re-reduced the image carefully and
reached mag. 24.5 – but too late for S&T
• in this talk I will tell you how I did this
2
PHILOSOPHY BEHIND THE TECHNIQUES
• when I started in 1989, I wanted to do some publishable science
and I ended up doing photometry of blazars, ~100,000 images
• having my data accepted by scientists was my motivating force
and my strategy was:
“try to do everything right”
• most professionals in North America assume that an amateur is
incompetent until proven otherwise
• when I design something, I try to learn the theory behind my
equipment & techniques and be very thorough
3
WHY IS SO MUCH HOMEMADE?
• during this talk you will see that most of my equipment and
software is homemade. Everything in this talk that does not have
a brand name, I have designed, and either built or had built.
• I like designing and building this stuff, and I know how to do it
• I started before much of today’s equipment was available
• to keep costs down, I used my labour rather than money
• for the scientific work especially, I needed to know that
everything works properly and in a known way. Equipment
manufactured for amateurs is deficient in this regard:
• optical performance data is not available, esp. non-visual
• software source code is not available
• mechanical performance is usually inadequate
• database handling is not provided
• wrong approach taken for scientific work
4
WHY DO I HAVE AN OBSERVATORY?
•
•
•
•
allows use of superior, non-portable, equipment
avoids losing alignment and collimation
has electrical power
observing in the Frozen North near Ottawa, Canada makes an
insulated heated equipment room especially desirable:
• makes long exposures feasible
• allows use of long winter nights even at -35°C
• protects computers and electronics
• stores books and accessories
• allows money earning activities while waiting for exposures
5
BOLTWOOD OBSERVATORY (1)
In The Far Backyard At Home
6
BOLTWOOD OBSERVATORY (2)
Roof Rolled Off To The South
7
BOLTWOOD OBSERVATORY (3)
Telescope Room, Old Homemade Camera
8
BOLTWOOD OBSERVATORY (4)
• telescope room is cool in the
sun because double skinned
and ventilated, including
between skins. Thermal design
by Rob Dick
• biggest fault is no dome.
Cannot observe if windy
• walls are black because
original use was visual with a
refractor. Roll-off roof ceiling
is white to provide light while
working on equipment
Telescope Room, AP7p Camera
• convenient access more
important than dark skies for
my work so I located at home
9
MOUNT
• Byers Series 2 (sort of), with
modifications:
• backlash removal
• stiffened DEC drive
(15* better than Byers)
• stiffened mounting plate
• overloaded with 180 lb
optical tube assembly
• tracking error in 2 minute
exposure is imperceptible
• periodic error correction not
used or needed
• no guiding either
• stepping motors limited to
15 times siderial rate
10
OPTICAL TUBE ASSEMBLY (1)
•
•
•
•
•
•
focus flap motors, no longer used
rotating secondary, 4 ports
eyepiece focuser
S.S. angle stiffeners epoxied on
11*80 finder
square wooden tube:
• allows home carpentry
• supports extensive baffling
• corners allow tube currents to
escape away from light path
• easy to mount items on
• corner spaces are useful
• allowed a demountable
11
primary cell
OPTICAL TUBE ASSEMBLY (2)
collimation adj. using a rod end
Primary Cell – Designed And Built By Max Stuart
12
OPTICAL TUBE ASSEMBLY (3)
holes for rod ends
Primary Mirror Support
13
OPTICAL TUBE ASSEMBLY (4)
Spider Assembly And Baffling
14
BAFFLING
• no non-imaging light should get to the focal plane or intermediate
optical surfaces
• optics often have light-scattering dust on them
• baffles at right angles to the light are the primary defence. Black
paint is a secondary defence
• unbaffled focusers, extension tubes, cameras are major sources of
flare
• you should be able to work near the moon as long as it is not
actually shining on your primary
• you should be able to turn on the observatory lights with little
impact while CCD observing
15
OPTICS
• 16" f/4.7 primary on Astro Sitall zero expansion substrate by Peter
Ceravolo
• 2 times better than diffraction limit
• zero expansion substrate replaced a previous pyrex primary for a
several times reduction in focus drift
• 3.1" Newtonian secondary by Galaxy, RMS .016 wave
• Televue 2" Big Barlow normally used
• Televue Paracorr used for wide field
• Televue 5X Powermate intended for narrow field but not fully
tested
• UBVRI filters for photometry by Omega, RGB filters for
astrophotography by Custom Scientific
16
OPTICS PROBLEMS
• amateur optics don't cover CCD spectral extremes
• refractive optics suffer from chromatic aberrations, and filters may
pass unexpected UV or IR. Do not use photographic filters
• wish I had a Ritchey-Chretien Cassegrain matched to camera
• before buying your 4096*4096 CCD, check your optical
aberrations in the corners, and alignment and focus capability.
This stuff is hard enough at 512*512
• before using multicoated optics, check what happens at the
spectral extremes
• design such that all optical surfaces are far from focus. Then dust
will not affect your flat fields very much
• black anodizing is clear in the near IR, so paint it
• I use Krylon Ultra Flat Black spray
17
FOCUSER & CAMERA ASSEMBLY
• anti-backlash
• umbilical
• focuser frame
• focus sled
• optics tube
• collet
• AP7p camera
• filter wheel
• fan for 3°C
cooler CCD
• tarp on ceiling
18
FOCUSER
• 5.5 inch range to
allow for
magnifying optics
• stepping motor
and limit switches
• linear bearings
• the camera is
always at same
rotation on sky
19
EXPLODED CAMERA ASSEMBLY
magnifiers with
buttress thread
baffling
filter
disk
motor
filter wheel
Apogee AP7p
Collet
extra filter holder
camera clamps
20
FILTER WHEEL
•
•
•
•
•
•
•
•
•
10 positions for 1" filters
microstepped stepping motor
wheel is carefully balanced
2 optical switches for positioning wheel:
• one with 1 hole, other with 10 holes
• switches are checked to be sure that the
correct filter is in place at each move
AP7p not fastened in the normal way –
clamps are used
AP7p penetrates into face of housing
filter is very close to housing
permits the use of 1" filters – 2" is normal
note the baffle to correct shiny ring in the
camera
21
CCD CAMERA
• old homemade camera with Thomson
TH7883 chip in vacuum, liquid
cooled, chip at -72°C, low thermal
noise and very few hot pixels
• has little blue sensitivity, 40% peak
quantum efficiency, excellent flatness,
good temperature control, 382*574
pixels. Used for S&T Deep Field
• current camera Apogee AP7p with
SiTe 502A chip
• has blue sensitivity and 80% peak
quantum efficiency, more thermal
noise and hot pixels, poor
temperature control, 512*508 pixels
22
CALIBRATION FLAT FIELD LIGHT SOURCE
• purpose is to flood telescope entrance
aperture to simulate a sky background:
• puts diffuser across full aperture and
in contact. Need exactly the same
light path as for sky background
• diffuser needs to be evenly lit
• spectra to match night sky (not
likely!)
• four 35W tungsten halogen lamps with
blue-green and IR cut filters on each
• regulated power required
• inside covered with crumpled Al foil
• clips onto front of OTA, 40"*40"*11.5"
• 4% variation across diffuser - best I
23
could do in an 11.5" thick package
BOLTWOOD OBSERVATORY (5)
Equipment Room
24
ELECTRONICS (1)
•
•
•
•
•
power and switch panel
rack of microprocessor boards
flat fielder power supply
modular
use Microchip PIC
microcontrollers, one per
function
• complex stepping motor
control firmware in 4 of them
• all time critical functions are
done in the microcontrollers
• almost electronics all are in
equipment room to ease
repairs and reduce thermal
wear & tear
25
ELECTRONICS (2)
• control of telescope fans & heaters,
button box sound, temp. readouts
• RA & DEC stepping motor drives,
button box
• filter wheel & focuser motor drives
• serial backplane controller with parallel
port interface to PC
26
ELECTRONICS (3)
Rack Backplane For Microcontroller Boards
27
COMPUTER AND SOFTWARE SUMMARY
• observatory computer is a Pentium II 333 MHz, 256 MB RAM,
60 GB hard drive
• 17" 1600*1200 display
• software:
• Windows 2000 because it allows simultaneous star mapping,
data base, observing, reduction, and non-astronomical work
• telescope control
• image data base
• reduction
• MaxIm
• star mapping:
• Guide 8.0 with A2.0 catalog to mag 20 - 526,280,881 stars
and a million galaxies
• DSS available at house over web
28
TELESCOPE CONTROL SOFTWARE
• VB and C++ software
• position & exposure info. usually
from “.aux” files or script
• uses MaxIm for camera driver, autofocus, mount jogger, image display
• corrects for refraction and flexure for
position and focus
• corrects for temperature for focus
• optionally runs scripts in my own
language
• images are rotated to have north at
the top immediately upon readout
• often not attended for 90 minutes
• normal exposure is 2 minutes with
many frames merged in reduction 29
DATA BASE
•
•
•
•
has index to 120,000 images, all FITS with extensive headers
allows semi-automatic data reduction
observations are grouped by “group files” to aid reduction
observing and reduction software updates data base automatically
• user interface package:
• searches data base in several ways
• uses MaxIm to display images
• allows display and editing of FITS headers (singly or in bulk)
• allows display and editing of “group files”
• VB and C++ software
30
DATA BASE USER INTERFACE
31
CALIBRATION FRAMES
• I do 16 frames each of bias, deferred charge and flat fields which
are then averaged to reduce noise
• deferred charge frame was used for my old homemade TH7883
camera, not yet for AP7p:
• was several times more important than bias frame
• added to image to compensate for trapped electrons
• 10+ hours of 1000 sec. dark frames for each CCD temperature:
• calibrated and summed to form thermal frame
• prorated by exposure time when used – assumes good camera
• quality is dependant upon camera temperature regulation
• normal image reduction procedures are used to create master
calibration frames from raw calibration frames
• bad pixels in master calibration frames are marked and not used in
image reduction
32
REDUCTION SOFTWARE (1)
• produces master calibration, astrophoto and photometric
reductions
• runs scripts
• all pixel computations are 32 bit floating point, 16 bits inadequate
• reduced images have a large dynamic range due to merging
• C++ software
33
REDUCTION SOFTWARE (2)
• automatic, driven by:
• "group" file listing image files and FITS headers
• control file giving star, sky, and exclusion zone locations
• merges images interpolating between pixels to correct for
translation, scale, and rotation
• cosmic rays removed from calibration and image frames during:
• overscan averaging
• raw frame calibration
• frame merging
• bad pixels (due to many sources) are marked and avoided – merge
is corrected for these missing pixels
• variance frames are maintained (primarily for photometry)
34
REDUCTION PROCEDURE SUMMARY
• manually delete bad images from the group file
• reduction software then:
• calibrates each raw image pixel by pixel
• measures each calibrated image
• merges calibrated images based on those measurements into
buffers. Each image is multiplied by a weight that is larger for
better images
• completes the merge
• optionally flatten the image to correct errors in flat field
calibration
• optional MaxIm deconvolution and other fiddling
• will use S&T Deep Field image as the example in what follows
35
CALIBRATE EACH RAW IMAGE
S&T Deep Field Raw Frame
• scan lines have 32 overscan
pixels beyond the real pixels
• average these along the line
• subtract overscan from each
pixel to remove certain camera
problems
• apply bias, deferred charge,
thermal, and flat master
calibration frames
• for each pixel:
cal = (raw-overscanbias+defchg(prorated_therm))/flat
• bad pixels are marked and not
used
36
MEASURE EACH CALIBRATED IMAGE
• using pattern matching, locate the "key" star
• measure:
• sky background
• centers for each of 2 "locating" stars
• shape of "locating" stars
• reject any calibrated image where "location" star elongation is too
large
• estimate the variance of a faint object in this image:
est_var = sky_var * exp(key_mag - key_min_mag) * (fwhm*fwhm)
is proportional to it
• weight for frame when merging is 1/est_var
• compute translation, scale, and rotation required for registration
• compute weighted sky estimate and add to sum
37
MERGE EACH CALIBRATED IMAGE
• merge this calibrated image into 12 merging buffers
• 12 needed to handle variance, weight, and cosmic ray removal
later
• bad pixels are skipped
• subpixel merge into the image buffers using bilinear interpolation
• subtract sky from each pixel because it is so dominant
• weight each pixel by 1/(est_var)
• add weight to the sum_of_weights buffer for each pixel
• for each pixel location, remember max and next_to_max value
seen in buffers for cosmic ray removal later
38
COMPLETE MERGE
S&T Deep Field Merged
• remove cosmic rays. For a pixel:
• use merged image value
excluding max and next_to_max
values
• statistically decide whether max
is a cosmic ray, ditto
next_to_max
• if cosmic ray, remove from sum
and sum_of_weights buffers
• compute for frame: sky =
weighted_sky_sum /
sum_of_sky_weights
• compute for each pixel: pixel =
(weighted_sum / sum_of_weights) +
sky
39
FLATTENING THE IMAGE
• why isn't it flat now?
• I did flat field every raw frame but there is an extreme
sensitivity to flat field errors due to light pollution
• range of image is 1.444e6 to 1.459e6 photons/pixel, just 1%
• image is 99% light pollution
• sky flatness failure is 6000 photons/pixel, just .4%, but that is
40% of the total image range
• flattening method:
• separate sky from stars and other objects
• examine sky near each final image pixel to get a sky estimate
for that pixel
• for each pixel subtract off local sky estimate, add on overall
sky average
• only works with small objects, and fails with nebulosity
40
S&T DEEP FIELD FINISHED IMAGE
Mag. 24.5 Objects Have SNR Of 3 According To Bradley Schaefer 41
FAINT TARGET URBAN CCD PROBLEM
• flattening the sky background is the #1 problem
• any spectrally independent lack of flatness in the CCD chip does
not matter as long as your master calibration frame is good (this
does require care)
• what matters is any spectrally dependant lack of flatness
• spectral variation of each pixel’s sensitivity creates the problem
because sky background, target, and the flat field light source all
have a different spectral content
• back illuminated chips are especially bad due to interference
effects
42
CCD FLATNESS COMPARISON
• Thomson CSF TH7883 and SiTe 502A were compared
• used master flat calibration frames for B, V, R, and I photometric
filters where the average pixel value was 1.
• measured standard deviations of flat differences between filters:
TH7883
SiTe 502A
B-V
0.0144
V-R 0.0027
0.0052
R-I 0.0058
0.0168
• the TH7883 should be able, ultimately, to go substantially deeper,
but with much longer exposures.
43
SPECIAL URBAN REQUIREMENTS
• high flat field quality
• proper baffling
• perhaps filtering for light pollution. I have not tried this except to
use a photometric I filter to darken sky
• my skies are suburban at <19.3 mag./sq.arcsec (dark is 22) so my
advice may not be entirely appropriate
• the ability with the CCD to subtract the sky makes the big urban
difference in comparison with film or eyesight
• unfortunately, due to higher sky noise and flat fielding failure, an
urban site still cannot match a dark site
44
ASTROPHOTO SUGGESTIONS
• when doing astrophotos, avoid the hackneyed objects, or at least
do some different view of them
• pick suitable targets:
• that fit the chip well
• are overhead at the middle of the observing session
• do long exposures
• get in close
45
EXAMPLE OF CLOSE IN - CORE OF M31 (1)
• usual amateur picture covers 3
degrees and the center 15
arcmin is burned out
• this has .47 arcsec pixels,
3.0*4.5 arcmin
• 7" refractor, homemade CCD
camera, no filter
• 54 min. exposure
• some interesting dark patches
on the left
46
EXAMPLE OF CLOSE IN - CORE OF M31 (2)
• same image, different stretch
• center of the reduced image
was not saturated and this
shows that M31 has a very
bright point-like center - not
evident in most photos
47
EXAMPLE OF CLOSE IN - CORE OF M31 (3)
• same image
• unsharp masked in MaxIm
(but the official way failed)
• mask made using low pass
FFT filter with 2.5% cutoff
and 100% weight
• image - mask + 10000 using
pixel math
• 10000 added because MaxIm
does not understand negative
numbers
48
EXAMPLE OF CLOSE IN - CORE OF M31 (4)
• deconvolved using MaxIm
• in this image the background
is not the sky – it is the star
clouds of M31
49
EXAMPLE OF A DIFFERENT VIEW - NGC 206 IN M31
• I R V photometric filtered
images rendered as R G B
• faintest star visible limited by
confusion - more resolution is
needed to do better, not more
exposure
• pixel size 1.09 arcsec
• exposures: I 348 min.
R 562 min.
V 434 min.
• 16" Newtonian, homemade
CCD camera
• weighted merge was tuned to
enhance sharpness
50
ALIGNMENT FOR GERMAN EQUATORIALS (1)
• I use Project Pluto's Guide
for map of the polar region
• mark refracted pole location
• do rough alignment some
other way
• have RA drive on
• use low CCD magnification
• aim telescope at pole with
telescope on one side of pier
• start 1 min. exposure
• after 30 sec. pull
counterweight shaft around
180 degrees slowly in 30
sec.
51
ALIGNMENT FOR GERMAN EQUATORIALS (2)
Shim OTA on saddle and
adjust DEC to center the half
circles. Set DEC circle to 90°
Center of half circles is where the
polar axis is pointing. Adjust
mount elevation and azimuth until
correct. Sorry – these pictures are
52
not mates
ALIGNMENT FOR GERMAN EQUATORIALS (3)
Doing Precision Adjustment Of Elevation And Azimuth
53
OPTICAL ALIGNMENT TOOL
Homemade HeNe Laser Collimator
• fits into vane holder in
place of secondary to
first align the spider
• fits into eyepiece
focusers 1.25" & 2"
• fits into collet on front
of filter wheel
• HeNe for 1/2 sized spot
• aluminized center dot on
primary with clear
donut around it instead
of a gummed
reinforcement. Gum
will streak when
washing mirror with
54
alcohol
TECHNIQUES FOR IMAGING VERY FAINT OBJECTS
Paul Boltwood,
[email protected]
1655 Stittsville Main St., Stittsville, Ont.,
Canada K2S 1N6
(613) 836-6462
More at ottawa.rasc.ca under the Astronomy button.
A more technical talk on the S&T Deep Field image given at
Starfest 2000 is on the CDR and the web site. The techniques in it
are somewhat different.
55
56