Questions for experts
(Deferred from earlier sessions.)
Q. CASA improvement request: a way to keep track of antennas that might be contributing to shadowing, but aren't known by CASA because they weren't in the observing array.
- Brainstorming solutions: keep track of ALL antennas, including the ones that don't have data; however, this can confuse later CASA tasks. All antennas need to be kept track of in the online system. This could be approached as part of the horizon problem and the shadowing flags determined before the data reaches the user; this would not allow the user to choose a shadowing diameter to apply.
Q. Interp = linear vs interp = nearest. Does it behave correctly when there is only one solution? (i.e. if only one bandpass solution in time, are the results the same for interp='linear' and interp='nearest'?) What if you are linearly interpolating, but have one region that requires extrapolation (i.e., not bracketed by calibration observations)?
- A: Applycal will NOT extrapolate. For times that aren't bracketed, it will switch to the behavior of "nearest". For only one calibration solution, it will behave as "nearest".
Q. In fluxscale -- how is the SNR calculated? Peak flux / rms in some area - where and how is the rms calculated?
- A: Fluxscale calculates the gain between the flux calibrator and the gain calibrators, for each baseline. Rms determined by averaging baselines (trying various methods). Averages the baseline results in a way that leads to inconsistencies across spectral windows. George is working on making improvements to this (stay tuned).
Q. What is the behavior of applycal with the parameter calwt=True? Are the weights continually modified in a way that the user cannot recover from if they want to backtrack?
- A: Anytime calibration is applied, the scale of data is changed; weights (inverse variance) should be modified accordingly. So calwt=True is in theory the right thing to do. Only works in practice if the input values of the weights were true! (In ALMA, the data from the correlator is normalized such that starting with uniform weights are probably correct.) Note that re-applying the Tsys calibration returns the original Tsys weights. CASA improvement request: carry both the weights and the sigma column when splitting, to keep track of both the weights column and the original sigma column (can be used to revert to the original weights).
Q. How do we account for the fact that some bandpass calibrators have (non-zero) spectral indices?
- A: CASA improvement request: Eventually, should fluxscale report spectral index. Then can go back and calculate new bandpass solution, after running fluxscale. This is not currently available in fluxscale, because of the above-noted issue that the results are not always continuous across spectral indices.
(Other questions:)
Q. "gainfield" and "field" parameters in applycal are not set to the same value for the WVR corrections (but they are for the Tsys correction), as recommended in the
AntennaeBand7? guide. How important is it to do this?
Q. Can we get weather plotting directly into CASA? (Todd has written weather plotting for ALMA in
AnalysisUtils. See the
ALMACalQuestions page.)
--
AmyKimball - 2011-09-08