 |
WFPC2 Associations
Frequently Asked Questions |
What is an Association ?
-
An association is a logical group of exposures.
The deepest of such exposures is called the leader
of the association.
NOTE: The association identifier (asn_id) is chosen after the leader.
Why bothering building associations ?
For an archive user it is not easy, nor meaningful,
to browse through more than 100,000 WFPC2 individual exposures:
- certain exposures, meant to be one single long exposure, are split
for cosmic rays considerations (CR-SPLIT)
- certain exposures are dithered (POS-TARG),
both for removing detector defects (e.g. bad columns and hot pixels),
and for allowing the construction of enhanced resolution products.
CR-SPLIT and POS-TARG have to do with the Data Flow System of the
HST observatory, not with the scientific exploitation of the data.
The associations provide a way to isolate the archive user
from those technical details.
There would be little to gain in constructing associations
if accurate offsets among the various members of the association
weren't provided.
Once the shifts are given,
- an automatic pipeline can build the mosaic
product of the association
alleviating the archive users from tedious work.
When does an exposure belong to an association ?
An exposure is associated to a given leader
if
- they belong to the same programme (same proposal id)
- they are observed with the same filter (or combination of filters)
- their sky projected distance is not greater than 10 arcsec (100 WF4 pixels)
- their position angle doesn't differ by more than 0.03 degrees
These criteria do not prevent an exposure from belonging to two associations.
To avoid this, when an exposure could be associated to two (or more) leaders,
the closest one is chosen.
With these criteria about 88% of the scientific observations of WFPC2 end up
grouped into associations.
Which pointing information is used to group exposures into associations ?
To identify which observations should be grouped together,
coarse pointing information is sufficient.
In this case, the WCS (World Coordinate System) coordinates have been used.
How are the shifts computed ?
There are three possible methods:
- Shifts are measured via cross-correlation
The cross-correlation method,
is supposed to offer the most reliable offsets.
It actually measures the shifts onto the images providing also the error.
To ensure the correctness of the measurements two things are taken into
consideration.
- The provided error
- Since the shifts
of each individual WFPC2 chip are computed separately,
a voting system is introduced to make sure that the intra-chip
shifts agrees.
- Shifts are computed using the jitter information (Observatory Monitoring System)
For each observation, RA, DEC, ROLL measurements
(averages and standard deviations) are extracted, along with
some of the telemetry keywords; the comparison of RA and DEC
with those of the association leader provides the
jitter offsets.
The quality of the jitter is also evaluated,
see the
"jitter flag" and the jitter table.
These offsets were used in the previous associations
(associations of type "A").
- Shifts are computed using the WCS information (science headers)
Each association member's WCS is compared with the WCS of the
association leader
(defined to be the deepest first observation in the group).
Of the three methods, preference is given to the cross-correlation algorithm.
In those cases where the cross-correlation method fails
(not enough signal, typically in blue filters),
the jitter information is used.
In case the jitter information is missing or provides non-reliable offsets,
the World Coordinate System (WCS) information is used.
Can higher level products be built ?
Associations trace back the observing strategy (CR-SPLIT, POS-TARG) adopted by
a PI.
Such information is offered via our web interface,
changing the way our Archive Community is used to browse
the HST Mission Log.
Even better,
the knowledge of what actually happened during the observations
is available.
Having measured the offsets among the members of associations,
the archive users are offered an
automatic pipeline to generate higher level products:
| Processing steps |
| I |
On-The-Fly Re-calibration (OTF)
OTF
is applied to each individual exposure |
| II |
Association File
an association
file, which describes the associations and its members, is generated
from the DB |
| III |
Mosaic: Cosmic Ray Rejection and Co-Addition
By knowing the shifts among the exposures in the association
(from the association file) an IRAF procedure is able to co-add
the on-the-fly calibrated exposures. Cosmic ray removal is also taken care of.
Software packages:
|
All the steps are user selectable.
Why yet another type of associations ?
Those of you, already familiar with the first version of WFPC2
associations (type A), might want to know what pushed CADC and ST-ECF
to embark onto this project (type B).
The answer is not provided in this FAQ.
Please refer to the Introduction.
Idiosyncrasies & Warnings
The user shall be aware of some idiosyncrasies:
Associations might change with time
At different times the methods used to compute the offsets within
a given association may vary.
It is a physiological thing, since WFPC2 is a living instrument.
The WCS coordinates are available as soon as the observations are carried out.
The same is true for the jitter files, but a delay is introduced by the fact
that the jitter files need to be processed; typically, ST-ECF receives
the jitter files one/two months after they are produced at STScI.
Instead, the cross-correlation pipeline cannot run until the members
of the association are publically available; that is after the proprietary
period (typically 1 year) expires.
At different times an association might have a different number
of members.
As of Oct 2001, almost 60% of all association members have offsets measured
with cross-correlation; the offsets of the remaining members are computed
using the pointing information (Jitter 22%, WCS 20%)
WCS in the dataset FITS header might be unreliable
Before April 1996, the WCS information
in the fits header didn't reflect the dithering strategy.
That is, all dithered images had the same WCS
(the WCS of the first observation carried out in that visit).
Danger!
The users should carefully check whether their selected associations
were observed before April 1996 AND the shifts were derived by
WCS information;
if so, it might be that a POS-TARG observing mode is
wrongly believed to be a simple CR-SPLIT.
Still, a real CR-SPLIT will be properly handled.
|
The shifts within a given associations might not be homogeneous
About 12% of the associations have members whose offsets are computed
with different methods.
This might happen for several reasons:
- some members might still be in their proprietary period (see above)
- the cross-correlation process might have failed for some exposures
(typically the faintest ones) and succeeded with others (the deepest ones)
- the jitter files might be missing, or the observations might not have
the correct "FINELOCK" status
- etc.
Danger!
The users should be aware that the offsets might
not be compatible when computed with different methods.
In those cases (12%) use the provided stacked images and shifts
with due caution.
|
- Comparing offsets obtained with different methods
(jitter and cross-correlation),
we have discovered a zero point drift
of the jitter pointing information.
Danger!
The users should be aware that the jitter offsets might
be wrong up to ~0.05 arcsec (up to 0.5 WF4 pixels) when the offsets
are not negligible.
Yet, the jitter information is reliable when it
indicates that two exposures
have a relative offset of less than 0.1/0.2 pixels.
|
Comments to Alberto Micol (Firstname.Lastname@eso.org)
|