Next: Self-calibration for the SIRTF GOODS Legacy Project
Up: Calibration
Previous: Generalized Self-Calibration for Space VLBI Image Reconstruction
Table of Contents -
Subject Index -
Author Index -
Search -
PS reprint -
PDF reprint
Goscha, D., Mehringer, D. M., Plante, R. L., & Sarma, A. 2003, in ASP Conf. Ser., Vol. 295 Astronomical Data Analysis Software and Systems XII, eds. H. E. Payne, R. I. Jedrzejewski, & R. N.
Hook (San Francisco: ASP), 195
Calibration of BIMA Data in AIPS++
Daniel Goscha, David M. Mehringer, Raymond L. Plante
National Center for Supercomputing Applications
Anuj Sarma
University of Illinois
Abstract:
We summarize the general approach adopted for the calibration of millimeter
interferometer
data from the BIMA telescope using AIPS++ and illustrate the use of the
relevant software tools. In particular, we will discuss flagging, phase
calibration, flux calibration, and polarization calibration, and we will show
how we take advantage of the unique capabilities of AIPS++ to meet the special
needs of BIMA data. We will show how BIMA calibration tools can be used to
hide some of the complexity of the processes while still allowing access to
specialized variations if desired. We will illustrate how these tools are
pipelined together for end-to-end processing both within the BIMA Image
Pipeline and on the user's desktop. Finally, we will present a comparison of
data calibrated in MIRIAD and AIPS++.
We present the results of a comparison of calibrated millimeter data from the
Berkeley-Illinois-Maryland Association (BIMA) Array using both AIPS++ and
MIRIAD. In addition,
we discuss the unique calibration capabilities of AIPS++ in calibrating BIMA
data both on the user's desktop and in an end-to-end (e2e) pipeline.
In particular we present:
- A brief discussion of calibration of BIMA data using AIPS++;
- A qualitative comparison of data calibrated and cleaned in AIPS++
and MIRIAD;
- A quantitative comparison of the RMS and dynamic range of data
calibrated and cleaned in AIPS++ and MIRIAD.
AIPS++ allows for the concealment of some of the complexity of calibrating BIMA
data through the use of custom tools. The
bimacalibrater tool in AIPS++ is such a tool. bimacalibrater
contains several functions needed in the calibration process, many of which are
friendly wrappers around functions of the AIPS++ calibrater tool. These
wrappers hide parameters not normally needed in the calibration of BIMA data and provide
more suitable defaults for other parameters. The bimacalibrater functions
hide much of the complexity of the calibration process while still allowing
a high degree of customization for varied data.
One of the important aspects of the calibration process is the ability to view
the antenna based gain solutions, flag bad data in the solution, and fit the
solutions. Gain solutions are written to
a gain table that can be accessed by the AIPS++ table tool,
allowing for a high level of accessibility to the data. Once this
has been done, an interactive user can use the plotcal function of the
bimacalibrater tool to examine the gain table. If any bad data were
noted after examining the gain table, a user could simply flag the bad data
using the autoflag tool, or, interactively using the msplot tool
(both part of AIPS++). In addition, it is also possible to flag and fit gain
table solutions using the gainpolyfitter tool.
The calibration process consists of three primary steps:
- Filling;
- Flagging/Editing;
- Calibration.
All of these steps can be carried out interactively on the user's desktop using
the AIPS++ GUI, interactively using the Glish (the scripting language front end to AIPS++) command line interface, or in
an automated fashion using custom Glish scripts. The BIMA Image Pipeline
currently employs the bimacalibrater tool to do automated calibration of
BIMA data.
In order to assess the robustness of calibration of BIMA data within AIPS++,
several comparisons were made between data calibrated with and cleaned within
AIPS++ and MIRIAD. Great care was taken at each step of the calibration
process to ensure we were comparing ``apples to apples'' - data that were
flagged in one data set were flagged in the other, the same clean algorithms
were used in both cases, gain solution fits were both two point interpolations,
etc.
The following comparisons were carried out:
- Peak fluxes of calibrated versions of the calibrator were compared;
- Antenna based gain solutions were compared;
- Contour maps with the contours chosen to highlight the noise levels
were made for qualitative comparison;
- The data were cleaned using the same number of iteration and noise
and dynamic range calculations were made;
- The data were cleaned to the same peak residual level and the number
of iterations needed to reach that level was noted and subsequent
image noise and dynamic range measurements were made.
As a check on the flux density calibration, images of the calibrator 1733-130
were made. The specified flux density of this source during calibration was
2.8 Jy. The flux densities of the dirty maps were compared.
The MIRIAD image produced a peak flux density of 2.66 Jy, while
the AIPS++ data yielded a flux density of 2.81 Jy. In this
particular case, AIPS++ did a better job in reproducing the correct flux density during
calibration.
The next step in the calibration comparison was to examine the gain solutions
produced in the two packages. There was no noteworthy difference in the gain
solutions other than the fact that gain solution amplitudes in MIRIAD are the reciprocals of their AIPS++ counterparts.
A qualitative comparison of the images after a 1000 iteration clean (using the
Clark clean algorithm) was then performed. The same data were calibrated and
cleaned in AIPS++ and MIRIAD and then imaged. Figure 1 shows the results. In
both cases, contour levels were chosen to highlight the background noise levels
so qualitative comparisons between the calibrations could be seen more clearly.
The contour levels are the same for both images.
Figure 1:
SGRB2N calibrated and cleaned in MIRIAD and in AIPS++.
|
Using the imstat command in MIRIAD and the image analysis tool
in AIPS++, the RMS noise level for each cleaned image
was measured. In both cases, the same off-source
region was used. The results of this comparison are summarized in Table 1.
Lastly the data were cleaned to a maximum residual cutoff of 0.115 Jy/beam and
a similar comparison of background noise done. It should be noted that AIPS++
cleaned to that level faster than MIRIAD - 1437 iterations in AIPS++ and
4451 iterations in MIRIAD. The results are summarized in Table 2.
Comparison of calibration of BIMA Array data in MIRIAD and in AIPS++ has been carried out. We found no significant
difference in the gain solutions and images made from calibrated data from either package.
Specifically we found the following:
- The AIPS++ bimacalibrater tool and associated functions
provide a useful interactive GUI, command line, or pipeline solution
for calibrating BIMA data;
- In comparing antenna based gain solutions, it was found that
gain solution amplitudes in MIRIAD are the recipricals of their AIPS++
counterparts;
- In comparing flux densities of the phase-calibrator from maps made
using each package, AIPS++ did a better job reproducing the correct density;
- When the target data were calibrated and cleaned using the same
number of iterations, the dynamic range of the image is roughly 2% higher
in the AIPS++ image;
- When the target data were calibrated and cleaned to the same intensity
cutoff whilst AIPS++ required fewer iterations, the dynamic range of the
MIRIAD image was 4% higher than the AIPS++ image.
© Copyright 2003 Astronomical Society of the Pacific, 390 Ashton Avenue, San Francisco, California 94112, USA
Next: Self-calibration for the SIRTF GOODS Legacy Project
Up: Calibration
Previous: Generalized Self-Calibration for Space VLBI Image Reconstruction
Table of Contents -
Subject Index -
Author Index -
Search -
PS reprint -
PDF reprint