memo21.tar

memo21.tex0000644000175000017500000000614713706551774011053 0ustar  njjnjj\documentclass[12pt]{article}
\textwidth 160mm
\textheight 245mm
\oddsidemargin -3mm
\evensidemargin -3mm
\parskip 3mm
\parindent 0mm
\topmargin -20mm
\pagestyle{empty}
\usepackage{graphicx}
\graphicspath{{.}{./figures/}}
\usepackage{color}
\usepackage[utf8]{inputenc}
\usepackage{hyperref}
\usepackage{subcaption}
\usepackage{amssymb}
\usepackage{gensymb}
\newcommand{\memohead}[3]{\clearpage\begin{centering}\mbox{}\hrulefill \vskip 50mm \mbox{}{\Huge \bf LBWG memo #1}\\ \mbox{}\vskip 40mm {\Huge \bf #2}\mbox{}\vskip 40mm{\Large #3}\\\end{centering}\vfill\hrulefill\clearpage}
\usepackage{listings}
\usepackage{rotating}
\begin{document}

\memohead{21}{Clock-TEC separation parameter test}{Atvars Nikolajevs, Kaspars Prūsis, 2019.04.24}
During LOFAR Long-Baseline Busy Week in Dublin (15.04. - 18.04.), we been working on finding optimal Clock-TEC separation parameters.
We tested both types (gaincal and ddecal) by changing some parameters, mostly solint (number of time slots on which a solution is assumed to be constant) and nchan (number of channels on which a solution is assumed to be constant).
Results (.h5 files and plots) are stored on CEP3:
\begin{verbatim}
    /data019/scratch/nikolajevs/
\end{verbatim}

In our opinion the best results we acquired are using {\tt gaincal\_tec\_solint5\_nchan10}. Full parset file:
\begin{verbatim}
    msin=L1327+5504.ms
    msin.datacolumn=DATA
    msout=.
    steps=[gaincal]
    gaincal.type=gaincal
    gaincal.caltype=tec
    gaincal.parmdb=gaincal_tec_solint5_nchan10.h5
    gaincal.solint=5
    gaincal.nchan=10
    gaincal.sourcedb=L1327+5504.ms/sky
\end{verbatim}

By using ddecal type, resulting plots with slightly different parameters significantly differed from each other - there was no common pattern. We did not observe a difference between using caltype tec and tecandphase.

List of parameters we tested (other parameters are default):
\begin{verbatim}
    ddecal_tec_solint3_nchan10_propagate_all_approx_tec
    ddecal_tec_solint3_nchan10_propagate_all
    ddecal_tec_solint5_nchan10
    ddecal_tec_solint5_nchan10_propagate_all
    gaincal_tec_solint2_nchan10
    gaincal_tec_solint2_nchan20
    gaincal_tec_solint3_nchan10
    gaincal_tec_solint3_nchan15
    gaincal_tec_solint4_nchan10
    gaincal_tec_solint5_nchan8
    gaincal_tec_solint5_nchan10
    gaincal_tec_solint5_nchan12
    gaincal_tecandphase_solint5_nchan10
\end{verbatim}

{\noindent \bf Background}. The LOFAR stations have their own clocks which are not perfectly synchronized with the single clock that is used for all the core stations. This causes a strong phase delay across the frequency band for the remote-to-remote and core-to-remote baselines. The large clock offsets need to be corrected before calibration of the directional part.
There is a method called "Clock-TEC separation" to derive the clock values from an observation of a bright calibrator source. Once the clock values are determined the target field data for this effect can be corrected with BBS. Removing the time varying clock is equivalent to applying a frequency and time dependent phase correction to each station.
(van Weeren et al. 2016)
\end{document}