DRIFT-SCAN TIMING OF ASTEROID OCCULTATIONS
John
Broughton
(Software
updated 2008-05-10)
Occultations
present the opportunity to remotely investigate shape and dimensions of
planetary objects with orders-of-magnitude gain in resolution over direct imaging.
I have in the past observed visually a spectacular Jupiter occultation of
2.6-magnitude Beta SCO and measured brief disappearances of a fifth magnitude
star by ringlets of Saturn but until 2003 I had never observed the more common
variety of occultation by an asteroid.
Following the development of Dave Herald’s Occult software,
the turning point came with the advent of Steve Preston’s updated predictions,
the accuracy of which made viable a CCD imaging and timing technique I had
under consideration a decade earlier.
CCD
Due to their slow image
download, most astronomical CCD cameras cannot record short-term variability on
consecutive frames without missing out on most of the action; hence an
occultation is best recorded on a single frame. One way to do this is to
smoothly trail the star across the field by taking advantage of Earth’s
rotation. Another involving electron shifting during the exposure has the
advantage of providing a wide range of motion for recording anything from a
high-speed image of the diffraction effects prominent at the lunar limb to
potentially a TNO occultation of the faintest stars in the UCAC catalogue. Since few cameras including my own have
operating software supporting the electronic option, a stationary telescope
acquiring trailed images at the sidereal rate is the method discussed here.
With the advantage of noise
reduction, a cooled CCD camera provides a substantial magnitude gain over
low-light video. From a moderately
light polluted location under otherwise favourable circumstances, sidereal-rate
star trails as faint as magnitude 14 can be acquired with a telescope of 25cm
aperture. A single image provides a
convenient record for analysis, producing in almost all cases an unambiguously
positive or negative result. Although cloud induced disappearances can mar an
observation, they equally affect all nearby trails, making them easy to differentiate
from the real thing.
Rigorous timing methods
were devised and first employed for the Lutetia occultation of
COORDINATING CCD AND
VIDEO DRIFT-THROUGH OBSERVATIONS
The telescope is
pre-pointed to a fixed position in the sky where by virtue of Earth’s rotation,
the asteroid will drift through the centre of field at the time of occultation.
Because the trail ends of drift scans are used in the timing calculations, the
exposure must begin and end while the star is within the frame boundary,
therefore it is important to employ accurate and fail-safe methods of
coordinating both telescope and camera, especially when the field of view is
small. Originally developed in 2004 for drift scans, ScanTracker includes two
new pre-pointing modes and a printable chart of naked-eye alignment stars,
making it easy to point all kinds of telescopes whether they are used for CCD
drift-scan, video drift-through or even visual observations. The new modes
enable a one-star alignment on a naked-eye star followed by a single-axis
offset and no star hopping is required. Altazimuth mode is especially useful
for rapid set up of portable telescopes. See help for details.
The camera is preferably
oriented north up to produce left to right drift along rows of the CCD image;
quite straight forward on a polar-aligned equatorial. For an altazimuth SCT,
the east-west line of occultation displayed in ScanTracker will give an idea of
the orientation. Final adjustments can be made with the aid of trial and error
imaging of star trails.
In the drift-scan exposure
section, its length is adjustable from 10 seconds up to a maximum dependant on
field width and declination. An exposure
20 times longer than duration of the occultation is necessary if you want to
cover the region in which an asteroid satellite can reside. Exceeding this can
be disadvantageous since there is loss of limiting magnitude by sky glow the
longer the exposure lasts. A 200 second exposure for instance may lose 2
magnitudes over one of 20 seconds at a moderately light-polluted site. For this
reason it is good practise to shorten the exposure if circumstances are
unfavourable for obtaining a well recorded image such as might be caused by
bright moonlight or twilight, poor seeing, small magnitude drop, low elevation
or combinations thereof. A minimal
exposure should always be employed when there is threat of obscuration by cloud. With reference to an updated prediction, a
rule of thumb I use to find minimum exposure is to add the maximum duration of
the occultation to the uncertainty ellipse diameter plus 15 seconds for some
margin of safety. Another reason to use an exposure shorter than the maximum
allowed for your field-width is to avoid an overlap by a trail of equivalent or
brighter magnitude. Such overlaps contain their own contrast-reducing seeing
variations. GUIDE 8 with UCAC stars enabled is my
preferred planetarium software when checking a target’s surroundings for
interfering stars lying within 8 arc seconds of the same declination and within
the intended exposure length in seconds of RA.
The simulated drift scan at
the bottom of Scantracker displays field-of-view, aperture and estimated
limiting magnitude. Changes made to exposure and aperture influence the trail's
brightness and limiting magnitude figure.
This magnitude represents the faintest star whose occultation is
potentially detectable under favourable moonless conditions. A 2-magnitude loss from the listed limit can
be expected when a full moon is present.
MANUAL TIMING OF THE EXPOSURE
In the Windows operating system, the
time listed in image file headers can easily be off by a second, even shortly
after setting of the computer’s clock and as in my case the exposure length may
not be exactly as commanded, so for timing occultations I disregard this
information. Because computer equipment floods short-wave with static, a
digital timer previously synchronised to WWVH can be visually monitored during
the beginning and end of exposure. If the camera lacks an audible shutter, a
cardboard disk taped to a tennis racket can be used to unblock and block the
telescope aperture (without actually touching the telescope) at the intended
times within an electronic exposure a few seconds longer in duration. If as is
preferable the camera has a mechanical shutter, the times when the shutter is
heard to open and close are written down with the fractional second part estimated.
Either method may be as accurate as 0.2-second, depending on the individual.
RIGOROUS TIMING OF THE MECHANICAL SHUTTER
A
$10 quartz analogue clock provides a low-cost means to derive accurate timing
of the sub-second part, using the ticking sound. The small unit containing the
workings can easily be removed from the clock and placed in contact with a
microphone. Audio tests on mine revealed an hourly drift rate of only .008
second! The battery can be inserted by
trial and error until the ticking is heard synchronized with short-wave UTC.
This is then recorded on one channel of a stereo tape recorder while short-wave
UTC is simultaneously recorded on the other through a second microphone or
line-in. This recording is done prior
to, or after an occultation when computer and camera equipment are not
operating and causing radio interference to the short-wave time signal. The
microphone that recorded from the radio is then fastened in contact with the
CCD camera to record its shutter clicks during the occultation. As in the first
recording, the clock ticks are recorded on the other channel, guaranteeing a
clear time signal at the critical time. Audio software such as GoldWave
is later used to analyse both recordings in deducing the times the shutter
opened and closed relative to UTC, as is shown in the diagram of a 1-second
interval between clock ticks. Such timing measurements done within 1-second
intervals are not prone to errors caused by tape stretch. If the short-wave
time signal is faint and immersed in static, the clearest part of the recording
can be selected, the volume maximised and noise reduction button pressed a few
times until the UTC second marker stands out more clearly. The spike in the
audio plot repeats every second in the same place relative to the clock tick
and in my experience, if the signal can be heard then its position can be
measured. To correct for propagation, 0.01-second for every 3000 km from the
short-wave transmitter should then be added to the shutter timings. The previously mentioned manual timing is
also done to provide the time to the correct second and serve as a backup.
GPS
More conveniently, the short-wave
time signal and clock can be replaced by a GPS-based timer for simultaneous
recording with the shutter clicks, in which case only a single recording is
necessary and propagation does not apply. Suitable devices are the KIWI system,
the GPS
Clock and the newly developed VNG-uc
GPS Time Signal Generator. This
last one is the only fully self-contained timing device capable of recreating
short-wave UTC signals but it is not yet in production.
ELECTRONIC SHUTTERS
Cameras lacking mechanical shutters
give no audible cue we can use to time the exposure, so a different approach is
required. A tie-clip mike can be
attached to the RA drive motor instead of the camera. If sidereal tracking can
be turned on for just a second or two near each end of the exposure, a star
image will show up near each end of the trail to become measurement points
instead of the trail ends themselves. From the audio recording of the time
signal and drive motor noise, it is the time of the end of the first short
tracking period and beginning of the second that represent the star image
centres on the trail. The tracking periods are kept short to minimise any
degradation of timing accuracy by periodic error. I’ll later refer to trails of this type as having star-bump
profiles to distinguish them from normal ones.
IMAGE ANALYSIS
As with any CCD image, the drift
scan should be calibrated with bias, dark and flat field frames to achieve the
highest signal to noise ratio. Contrary
to popular belief, timing resolution of an occultation using a CCD drift image
is not limited by pixel size, except in unlikely events of very short duration.
Trail ends can be measured at the sub-pixel scale by interpolation of the
values contained in adjacent pixels; in effect joining the dots to find the
point where the profile crosses a certain brightness level. For best results,
the whole width of the trail needs to be taken into account by averaging X
values over several rows on the Y axis. In MaxIm
DL this can be achieved automatically using a horizontal box aperture. First though it may be necessary to
precisely level the trail using the edit menu rotate function with the bicubic resample box ticked. Next hold down
the left mouse button to drag a long narrow aperture around the trail starting
at the left edge of the frame, excluding any adjacent trails and as much
background as possible. Then use the view menu line profile function, select
horizontal box and the mean sample option and press the export button to save
it to disk in the form
of a comma-separated-values file.
A normal profile is a stretched
version of a fixed star’s bell-shaped curve and being a time variable image, it
is modified by atmospheric turbulence.
Calculations I did in 2004 showed the end of trail profile to be
stretched 200% compared to that of a fixed star and the point of origin to be
located exactly midway in height between the trail and background levels. Measurement levels derived from the lengths
of many rigorously timed trails in 2003 averaged within 1% of the 50% level.
Even under the influence of diffraction, the measurement level of the occulted
part was calculated to be at or very close to this level. This is different from the measurement level
of high frame-rate video recordings where diffraction theory indicates should
be made at the 25% level.
SCANALYZER
This trail measuring application
includes dynamic vertical scaling, smoothing of scintillation and signal noise,
cancellation of optical distortion effects and calculation of overall timing
accuracy. The LOAD button facilitates loading a profile such as included
example file 050521 Bilkis.csv. The 4
measurements on the profile are always made in left to right order. Clicking
the plot produces an expanded view whose width in pixels can be zoomed in as
far as 32 or all the way out using those arrows that appear either side of the
SMOOTH button. Clicking in this expanded view where the levels change brings
about a sub-pixel X measurement displayed in green. A right click returns the
full profile where the next position can be selected. Once the fourth measurement is made, times for the occultation
are computed and displayed. Measurements can be re-done by use of the BACK
button. Scanalyzer now recognises and
can measure star-bump profiles, in which case ‘Star Bump’ automatically
replaces the terms ‘Trail Start’ and ‘Trail End’ after the first position is
measured.
Since in practice the moving star is
a distribution of light on the CCD over 2 or more pixels, sudden changes caused
by an occultation always have a slope and are less abrupt than the rapid
variations due to scintillation and signal noise. The SMOOTH button applies a
gaussian filter to suppress this high-frequency interference and improve
accuracy without biasing the measurements.
Smoothing should be applied at least once and can be done in either view
up to 3 times before the original state is restored. Lights below the button indicate the level of smoothing and
timing results are updated automatically.
When loading a profile file,
Scanalyzer also loads a txt file of the same name if present in the same
directory. Previously edited in Notepad by the user, this file holds 2 lines
for trail end shutter timing followed by 1 line for their accuracy, followed by
4 lines of astrometric data used in averting a distortion-induced warp in
timing. If this file is absent, timings
can be manually entered into the white boxes or alternately the file can be
present but omitting the 4 lines of astrometry. Lacking astrometry,
computations are based on a simple extrapolation of image coordinates to time,
which scarcely matters in the case of telescopes with negligible distortion
such as unmodified SCTs. On the other hand a focal reduced SCT will have a
little distortion while Newtonians and Cassegrains suffer considerably more.
The astrometric data can be measured from any image taken with the same optical
configuration. After calibration to
celestial coordinates using Astrometrica,
4 positions are measured near the X and Y image coordinates equivalent to trail
start (alternately star-bump), disappearance, reappearance and trail end
(alternately star-bump). A Ctrl-click
operation enables measurement within a pixel or 2 of where intended and the X
parameter is displayed to 2 decimal places in the PSF-Fit box. This figure can
be manually incorporated into a contiguous object name of the form X468.55 when
saving a position. X can be followed by anything up to 4096 but there must be
no spaces. The 4 lines of data are then
copied from Astrometrica’s MPCReport.txt and added to the timings as in the
example file 050521 Bilkis.txt. These positions show up as red dots in the
Scanalyzer profile plot to indicate astrometric data is loaded and verify their
locations align approximately with the 4 measurement positions. Relative scales
between these astrometric positions are derived and applied as corrections to
the timings.
Faintly recorded occultations having
depth comparable to the amplitude of seeing irregularities and random noise
might produce what is obviously a bad measurement. A hump or hollow may
interfere where a level is being averaged or could cause the profile slope to
level out temporarily just where X is being interpolated. In such cases the
keyboard Ctrl-up and Ctrl-down arrow keys enable modification of the profile at
the position of the cursor. This is best done after returning the profile to an
unfiltered and unmeasured state via the SMOOTH button. After adjustments to the
problem area, the profile is resmoothed and remeasured. Keep in mind that such modifications
disappear if the profile is again restored to its original state.
Timing accuracy is derived from the
sum of the audio recording measurement uncertainty and profile measurement
uncertainties for both occultation and trail end. The latter calculations involve simulated drops in light in an
unocculted part of the profile made at 40 consecutive pixel locations before
being averaged. A magenta coloured line represents that region. The Bilkis
drift-scan observation made with a 0.5-m aperture seems to indicate that for
larger apertures, the best events measured in Scanalyzer will approach
0.01-second accuracy once the audio measurement uncertainty is effectively
zeroed with the aid of direct GPS-based timings.
FREEWARE
These are benign applications that
only access the txt and csv file types mentioned above and nothing is written
to disk. Included are icons for Scantracker and Scanalyzer you can assign when
you place shortcuts on your desktop alongside Events.txt. There are now 5 example profiles you can
load into Scanalyzer. This article
doubles as help for the software and is revised as the applications are
improved and as the need arises to add detail when observers request
clarification. If assistance is
required, I can be contacted at skyrock at bigpond dot com and I’m interested
to learn of successful observations that ensue. On the principle they are
provided freely with the user accepting full responsibility for any undesirable
consequences of their use then the software is available here for download.
ACKNOWLEDGEMENTS
The camera used by the author was
acquired following a Planetary Society Shoemaker NEO research grant, largely
for the purpose of NEO survey and tracking in the southern sky. My thanks to Keith Gelling for calculating
the effect of diffraction. Special
thanks are owed to Graham Blow of the RASNZ Occultation Section for his
unqualified acceptance, encouragement and promotion of drift-scan observations
since my first report.
LINKS IOTA Euraster
RASNZ Occult Updated Predictions Guide
8 MaxIm DL
Astrometrica GoldWave Vesta Occultation