Transcript Lecture 1

Lecture 17 – AIPS and other animals
• Some thoughts about the design of big
astronomical software packages.
• Packages to analyse radio interferometer
data.
• AIPS description.
• High-energy astronomy.
NASSP Masters 5003F - Computational Astronomy - 2009
Astronomy data reduction packages
• Software which implements (often
complex) algorithms to calibrate and
process (sometimes very large) volumes
of data from astronomical instruments.
Journal of Very Good
Research
DATA
Science
NASSP Masters 5003F - Computational Astronomy - 2009
Design criteria (my opinions):
• User interface is crucial – but difficult.
– Principles of good design have been worked
out – but you cannot ‘re-educate’ users. You
have to cater to their prejudices.
• Trustability
– Again users and designers may have different
ideas about when code deserves to be trusted!
• Who should write the code – professional
software engineers, or astronomers?
– Professional software engineers give a good
product – robust, maintainable, ‘trustable’ in
their eyes – but it may not be what the users
want. There is also a high incidence of manic
zealotry among SEs.
NASSP Masters 5003F - Computational Astronomy - 2009
Design criteria (my opinions):
– Astronomers have intimate knowledge of what
they want – but are not usually professional
coders – although since coding is ‘easy to
learn but hard to master’, they may not be
enough aware of their deficiencies in this
respect.
• Solution: code should be written by
professionals, but under close
supervision/consultation with astronomers.
• Code should not be ‘owned’ by a single
person.
• Portable installation.
NASSP Masters 5003F - Computational Astronomy - 2009
Design criteria (my opinions):
Application
programs or ‘tasks’
Common code in
well-documented
libraries
User-contributed
applications.
A well-documented Applications
Programmer Interface (API)
Data in a well-documented file format.
NASSP Masters 5003F - Computational Astronomy - 2009
Design criteria (my opinions):
• How finely to divide up the work between
tasks?
– People seem to have different opinions. In
some packages there are giant tasks which
seem to allow you to do almost anything by
judicious choice among a zoo of parameters.
– On the whole though I prefer tasks which
each do something fairly simple.
• Then join them up in a script if more complicated
single-command processing is necessary.
NASSP Masters 5003F - Computational Astronomy - 2009
Interferometry reduction packages:
• Astronomical Image Processing System (AIPS)
– Started 1978; 1.4 M lines of code; hosted by NRAO.
Fortran 77 but highly ‘tweaked’ for speed.
• Miriad
– 80s? Designed to process ATCA data (Australia).
• CASA (née AIPS++)
– Started about 2004 (building upon the AIPS++ suite
which was started in the 90s but abandoned in
controversial circumstances). Designed for ALMA and
eVLA. C++, but python scripting interface.
• MeqTrees
– First release 2007. Hosted by ASTRON
(Netherlands). Implements more rigorous
interferometer theory.
NASSP Masters 5003F - Computational Astronomy - 2009
AIPS
• http://www.aips.nrao.edu/
• AIPS is showing its age in many ways but
it is still the ‘default package’ for most
interferometry data reduction. A huge
number of man-years of experience on
AIPS has been accumulated and
astronomers have come to trust it deeply.
• For the casual user, its drawbacks are:
– There is a very large zoo of tasks. Almost
certainly there is one which will do your
required job: the problem is finding it.
– Many tasks have a very large zoo of
parameters.
NASSP Masters 5003F - Computational Astronomy - 2009
AIPS
– The parameter interface is rudimentary. As is
usual with even moderately complicated
systems of parameters, some parameters
control whether others are read or not. In
AIPS, these logical relations are often
obscure. It is therefore difficult for the casual
user to be confident that all needed
parameters have been set, and no angst has
been spent on setting unnecessary ones.
– AIPS has its own shell, its own private file
system, and in general AIPS forces you to
play by its rules. You cannot easily get into
your data and fiddle with it.
NASSP Masters 5003F - Computational Astronomy - 2009
AIPS
• By far the best way to learn AIPS is by experience
processing real data, under the supervision of an
expert user.
• Alas... there are no expert users at UCT.
• This is a pity, because here in SA we will shortly
have a ground-breaking interferometer in our own
backyard - MeerKAT. Opportunities will be there for
the taking!
– But this is also a good opportunity to move away from
AIPS to something a bit more modern.
• Why learn anything about AIPS then?
– Because it is the only package I have had any experience
with!
– And because there are many similarities across
packages.
NASSP Masters 5003F - Computational Astronomy - 2009
High-Energy Astronomy
• Means x-rays and gamma rays.
• It’s convenient at this end of the spectrum
to concentrate on the particle part of the
quantum wave-particle duality.
– So we usually talk about the energy of the
photons which make up the radiation, rather
than their wavelength or frequency.
– A convenient energy unit is the electron volt
(eV).
– Confusingly, x-ray fluxes are often cited in ergs.
• 1 eV ~ 1.6 ×10-19 joules ~ 1.6 ×10-12 ergs.
– From E=hν, 1 eV ~ 2.42 ×1014 Hz.
NASSP Masters 5003F - Computational Astronomy - 2009
High-Energy Astronomy
• X-rays: roughly speaking, from ~100 to ~105 eV.
– We speak of hard (= high energy) vs soft x-rays.
• Gamma: anything higher.
• Physical sources: as the name ‘high energy’
implies, very energetic events tend to generate xrays and gamma rays.
– Thermal radiation if the temperature is >106 K – eg:
• the sun’s corona ( soft x-rays)
• Black Hole accretion disk ( hard x-rays)
– Nuclear fusion ( soft x-rays)
– Matter falling into a gravity well.
• Supernova ( hard x-rays, gammas)
• GRB (?) ( gamma rays)
• It is interesting that radio and x-ray images often
follow similar brightness distributions.
– Because hot plasma  relativistic synchrotron
emission.
NASSP Masters 5003F - Computational Astronomy - 2009
X-ray Observatories
• You need to be in space, because the
atmosphere efficiently absorbs x-rays.
• Most of the currently interesting results are
coming from:
– Chandra (good images, so-so spectra)
– XMM-Newton (so-so images, good spectra)
– SWIFT (looks mostly at GRB afterglows)
although there are several more.
• How do you make an image from x-rays?
Don’t they go through everything? So how
can you make a reflector?
NASSP Masters 5003F - Computational Astronomy - 2009
X-ray Observatories - mirrors
• Wolter grazing-incidence mirrors:
– reflectivity of any EM radiation gets higher for a
low angle of incidence.
Double reflection
Can nest
many shells
of differing
diameter.
Paraboloid
Hyperboloid
Behaves like a thin lens set here.
NASSP Masters 5003F - Computational Astronomy - 2009
X-ray Observatories - detectors
• Charge Coupled Devices (CCDs) are used, just
as for optical. The workings are slightly different
though.
e-
e- e-
e- e-
e-
e- e-e- e-
e- e- e-e- e-
e- e- e-e-e- e-
...then read out.
At optical wavelengths:
At x-ray wavelengths:
e-
e- e-e- -e--e- -e- e-eee- e
e
...then read out.
Time
NASSP Masters 5003F - Computational Astronomy - 2009
X-ray Observatories - detectors
• Basic substance of the CCD is silicon – but
‘doped’ with impurities which alter its electronic
structure.
• There are several sorts, eg:
– Metal Oxide Semiconductor (MOS)
– pn
• The CCD surface is divided into an array of
pixels.
• A photon striking the material ejects some
electrons which sit around waiting to be
harvested.
– The number of ejected electrons is proportional to the
energy of the photon.
NASSP Masters 5003F - Computational Astronomy - 2009
X-ray Observatories - detectors
• CCD operation is in frame cycles.
Long accumulation
time
Short readout time
NASSP Masters 5003F - Computational Astronomy - 2009
X-ray Observatories - detectors
• CCDs have to be read out sequentially
(slow).
Rows
Columns
Digitizer
(ADC)
Out
Readout row
NASSP Masters 5003F - Computational Astronomy - 2009
•
X-ray Observatories - detectors
The readout operation consists of the
following steps:
– For each CCD row, starting with that nearest
the readout row, move the charges into the
next lowest row.
– In the readout row, starting at the pixel nearest
the output, shift the charges into the next lowest
pixel.
– Convert the analog charge quantity (it is an
integer number of electrons, but such a large
integer that we can ignore quantum ‘graininess’)
to a digital number.
– This is done in an Analog to Digital Converter (ADC).
– Send that number to the outside world.
NASSP Masters 5003F - Computational Astronomy - 2009
X-ray Observatories - detectors
• For x-ray detection, we want to arrange the
frame duration so that we expect no more than 1
x-ray per pixel per frame.
• Why? Because if we can be pretty sure that all
the charge per pixel per frame comes from a
single x-ray, we can determine the energy of the
x-ray.  x-ray spectroscopy.
• XMM example:
– Spatial resolution is ~1 arcsec (fractional ~10-3).
– Spectral resolution is ~100 eV (fractional ~10-2).
– Time resolution is ~1 second (fractional ~10-5).
NASSP Masters 5003F - Computational Astronomy - 2009