Water Vapour Radiometers

Path length changes through the atmosphere cause significant errors for astronomical radio interferometry at high frequencies (e.g. 100 GHz). The atmosphere can give rise to a typical differential excess path length of 1mm for antennas separated by 1km, and has fluctuation time scales of greater than 10 seconds. However, by sensing the water vapour emission, the excess path can be deduced and corrected.

Following encouraging results from a prototype system trialled on the ATCA in the late 1990s (see arXiv:astro-ph/0701016), a LIEF grant was obtained to outfit all ATCA antennas with four-channel Water Vapour Radiometers (WVRs). This UNSW-led project has resulted in the development of WVRs that are capable of determining signal path length fluctuations by measuring small temperature fluctuations in the atmosphere using the 22.2 GHz water vapour line. By measuring the line of sight variations of the water vapour, the induced path excess - and thus the phase change - can be estimated and corrections can then be applied during data reduction, reducing decorrelation of the source signal.

A paper describing the new Water Vapour Radiometers has been published in Publications of the Astronomical Society of Australia (PASA), vol. 30, article e035 (2013): "Water Vapour Radiometers for the Australia Telescope Compact Array" ) by B.T. Indermuehle, M.G. Burton, and J. Crofts.

Papers that make use of the WVRs in their data reduction are requested to include in their Acknowledgements section: "The Water Vapour Radiometers were funded through an Australian Research Council Linkage Infrastructure Equipment and Facilities (LIEF) program grant awarded to the University of New South Wales, with the support of the University of Sydney, Swinburne University of Technology, James Cook University, and CSIRO Australia Telescope National Facility."

Using the Water Vapour Radiometers during observations

There are a number of things that observers should be aware of when planning to use the WVRs during their observations.

You should ensure that the calibration of the WVRs is enabled before you begin observing. Calibration involves periodically moving an absorber in front of the WVR feed horn on each antenna. This is done via automatic paddle scans, similar to how the 3mm system works. To ensure that paddle scans are made during your observations, give the command:

enable wvr

in caobs before you begin.

While calibration is enabled, caobs will put the paddle in front of the WVR when either of two conditions have been satisfied:

The first cycle on-source during a scan that has been marked with the C CalCode.
The first cycle on-source after 20 minutes since the previous paddle scan.

The paddle scan lasts approximately 40 seconds.

When observing in the 3mm or 15mm bands, paddle calibration of the WVR will not affect the observations. The WVR paddle briefly passes in front of the 7mm feedhorn as it moves between its "in" and "out" positions. When this happens, the temperature seen by the 7mm system increases very noticably, and so these scans are automatically flagged by the on-line system. These cycles are not replaced though, so if you ask, for example, for a 2 minute dwell on a source marked with CalCode C, each antenna will be flagged out in two cycles of the twelve that are spent on-source (assuming 10 second cycle times). But since the paddles are operated independently on each antenna, it is observed that often there are four cycles that have at least one antenna flagged. If you require a minimum integration time on each visit to your calibrator in the 7mm band while using the WVR, we recommend that you ask for four cycles more than the minimum you require.

Water Vapour Radiometer data reduction pipeline

This section describes in detail the steps required during data reduction to prepare and use the WVR corrections. Some of the scripts described here need direct network access to the MoniCA server in Narrabri, so are best run within that network. Alternatively contact Narrabri staff for instructions on how to tunnel MoniCA via SSH.

Preparation

Begin as normal for mm-band data reduction, by loading the raw data using the atlod task in Miriad. You do not need any special options at this step.
If you are at 3mm, or if you wish to apply baseline length corrections, use atfix at this point, as normal.
You can now generate an index file using uvindex; this will be required if you will use the "traditional" method to determine the WVR corrections. To do this:

uvindex vis=uvfile log=logfile.uvindex.txt

Here, uvfile is the output of atlod, and you may change the name logfile.uvindex.txt to something more descriptive if you wish.
Split the data into separate frequency bands using uvsplit, by using options=nosource. This will make applying the WVR corrections simpler later in the process.

You may now proceed to determine and apply the WVR corrections before you begin the usual calibration of your data. There are two methods to achieve this: the "traditional" method, which is more involved, but may give you a better understanding of what is going on, and the "easy" method, which does almost all of the work in a single command.

Computing the corrections - the traditional way

The scripts mentioned here are found on kaputar:/DATA/KAPUTAR_1/operations/WVR and they will run as long as the following dependencies are met:

Perl with the following CPAN modules:
- Astro::Time
- Math::BigInt
- Time::HiRes
- IO::Socket::INET
IDL v5 or greater, with the following non-standard procedures (available as a .tar.gz here):
Miriad version newer than Dec 2012

Having met all these prerequisites, follow these steps here.

First you need to know the exact date and time range, in UTC, for which you require corrections.
Run the get_monica_wvr.pl script using the start and end dates as parameters:

./get_monica_wvr.pl "2012-05-23 07:00:00" "2012-05-23 21:00:00"

Don't forget the quotation marks around the date/times.

This step will produce one file for each antenna, named ca01 - ca06, containing the following data:
- day fraction
- 16.5 GHz channel voltage
- 18.9 GHz channel voltage
- 22.3 GHz channel voltage
- 25.5 GHz channel voltage
- Total power channel voltage
- T1 in °C
- T1a in °C
- T2 in °C
- T3 in °C
- T4 in °C
- T5 in °C
- RF TEC Current in A
- Shield TEC current in A
- +10 V reference in V
- -10 V reference in V
Set up your IDL environment to allow easy access to the prerequisite non-standard procedures listed above. We suggest copying them into a directory listed in your IDL_PATH environment variable.

If you are running on kaputar, ensure that you have enabled IDL in your .login.packages file in your home directory. Then make a directory idl in your home directory and copy all the prerequisite procedures into it. Then add the line

setenv IDL_PATH ".:+${HOME}/idl:+${IDL_DIR}/lib"

in your .cshrc file in your home directory.

In the directory that you are doing the data reduction, copy the files wvr_phase.pro, wvr_wavelet.pro and cal_data from the /data/KAPUTAR_1/operations/WVR directory.
If you wish to analyse the behaviour of the raw data with a wavelet decomposition, you can do so now by starting IDL and running the wvr_wavelet.pro script. For that to succeed, open the wvr_wavelet.pro script and adjust the parameters so the appropriate input files and signal channels are loaded and analysed. It is pre-configured to produce sensible results, so you'll typically want to simply select which data to plot. This can be set around line 70 in the code. Look for the "Adjust data below" and "..above.." comments in the code.
Now we process the uvindex file of your observations you created earlier. This file is needed so the WVR correction routine can provide a corrections table for the correct time stamps. Copy process_uvindex.pl from /data/KAPUTAR_1/operations/WVR, and run it:

./process_uvindex.pl logfile.uvindex.txt phase-cal target-source frequency

The parameters are:
- logfile.uvindex.txt: The file you generated with uvindex.
- phase-cal: The name of the phase calibrator you used.
- target-source: The name of the target source you observed.
- frequency: The frequency you observed at, in GHz, that you want the path delay calculations to work for. (Tweak alert: If you're getting unsatisfactory results, you may want to consider increasing the amplitude of the corrections by increasing or reducing the frequency here.
All parameters are required. The output is written into the file run_wvr.pro in the current directory. If you have multiple phase calibrators and/or target sources and/or frequencies, you will have to repeat this stage and subsequent processing for each combination.
Manually adjust the wvr_phase.pro file if you want to selectively ignore short baseline corrections. Simply open that file, and fill in the baseline lengths array based on the configuration the array was in at the time the observations were made, and then specify the lower cutoff:

baselines = [ '21', '31', '32', '41', '42', '43', '51', '52', '53', '54', '61', '62', '63', '64', '65' ] ; manual step: If you want to ignore certain baselines, set the blcutoffbelow. You must also fill in ; the bllengths array. This will be automated at some point. bllengths = [ 367, 1163, 796, 2449, 2082, 1286, 2526, 2158, 1362, 77, 5878, 5510, 4714, 3429, 3352 ] blcutoff = 0 ; Baselines shorter than this are set to 0.

Make sure you put the baseline lengths in the correct order (same order as the baselines array!)
Start IDL and run run_wvr.pro.

If you have the current directory in your IDL_PATH, you can do this by giving the following command at the IDL prompt:

IDL> run_wvr

NOTE: This deletes any previously generated atwvr_input* files.

NOTE 2: If the script dies because it can't find the atwvr_input* files, simply comment the FILE_DELETE instructions at the top of the run_wvr.pro script.

Depending on how much data you're working on, you can now go for a coffee. On a reasonably fast computer, this step takes on order of 15 minutes for a 24 hour period's worth of WVR data. This step also creates a WVR phase plot for every baseline according to the naming convention: WVR_Phase_Xs_T.ps, where X = 5 or 10 depending on whether it's the 5 or 10 second integration version, and T = the starting time in day fraction of the respective data segment.
The generated atwvr_input* files need their dates adjusted to the date of the observation. This is simply achieved by running atwvr_input.pl (from the /data/KAPUTAR_1/operations/WVR directory). The resulting files are atwvr_input_10s and atwvr_input_5s (note: just dropping the .txt extension).

Computing the corrections - the easy way

The scripts mentioned here are found on kaputar:/n/ste616/atcacode/general_reduction/wvrcode and they will run as long as the following dependencies are met:

Perl with the following CPAN modules:
- Astro::Time
- PDL
- PDL::GSL::INTERP
- PDL::Graphics::PGPLOT
- PGPLOT
And the ATNF::MoniCA module from the open-monica project.
Miriad version newer than Dec 2012

Having met all these prerequisites, follow these steps here.

Run the wvr_ops.pl script (found in the directory /n/ste616/atcacode/general_reduction/wvrcode), using the file you created in atlod as the argument:

wvr_ops.pl uvfile

Running the script with the "-h" option will make the script output a comprehensive description of the available options.

By default, this script will output the corrections for every 5 seconds during the observation period. To output the corrections for every 10 seconds, use the option "-c 10".

The script will output an atwvr_input file for each frequency that it finds in the dataset. Each file will have the central frequency as part of the filename.

The computation of the corrections should take only a few minutes so long as you have a fast hard disk and a fast network connection to the MoniCA server. Running this script on kaputar is ideal.

Applying the corrections

Apply the corrections to your data using the atwvr Miriad task and the input files

atwvr_input_10s or atvwr_input_5s, if you used the traditional method.
atwvr_input_Ns.freq, if you used the easy method; here N is the correction time you specified with the "-c" option (or 5 as default), and freq is the centre frequency of the data to apply the corrections to, in MHz.

It is a matter of experimentation which correction cadence works better.

For example, after the uvsplit step, you might have the datasets uvsplit.43000 and uvsplit.45000 in your directory. If you have generated the WVR corrections for the 43 GHz dataset with the traditional method, you can apply the 10 second corrections with the command:

atwvr vis=uvsplit.43000 out=uvsplit.corr.43000 wvrphase=atwvr_input_10s

If you have generated the WVR corrections with the easy method, you can apply the 5 second corrections with the command:

atwvr vis=uvsplit.43000 out=uvsplit.corr.43000 wvrphase=atwvr_input_5s.43000

The atwvr task generates a lot of on-screen output.

You can now use uvsplit to split up the corrected datasets into separate sources and continue the calibration procedure as usual.

Original: BI/PGE (16-oct-2014), Modified: JS (24-oct-2013)