- Introduction
- Getting Started
- Calibrating SMA Data
- Getting Help
IntroductionWelcome to Eric Keto's MIR Website.
History of MIR
Millimeter Interferometer Reduction (MIR) is an IDL based software package, originally developed by Nick Scoville at Caltech for use with the OVRO millimeter array. While Nick was working at OVRO on MIR (late 1990's), at the SMA, Eric Keto and Ken Young were developing the SMA's offline and online software that handles the SMA data. Eric Keto designed the SMA raw data format and calibration procedure following MIR and adapted many of the MIR programs for the SMA. In the end, MIR was never used at OVRO, principally because of the merger of OVRO and BIMA, and the SMA inherited responsibility for the MIR package. At the SMA, Chunhua (Charlie) Qi has been responsible for maintaining MIR. Over the years, he has contributed many useful modifications.
The examples and the programs The MIR cookbook, https://www.cfa.harvard.edu/~cqi/mircook.html, describes the full capabilities of MIR. The web pages here provide examples illustrating basic calibration with an emphasis on before-and-after graphics showing the effect of each of the calibration steps. The description in the first example includes all the equations used in the SMA flux calibration. This is useful for understanding both the basic procedure as well as the finer points. The flux calibration shown in these examples uses flux calibration programs that I modified and wrote to provide graphics illustrating the flux calibration procedure. While the mathematics of flux calibration is the same in these programs as in those described in the MIR cookbook, the programs and the procedure are slightly different. Also my flux calibration programs use new models for the submillimeter fluxes of the asteroids, planets, and their satellites. These new models are those from ALMA's CASA calibration package. While the differences between the old and new models are small in the 230 - 350 GHz range where the SMA operates, the new models provide consistency with ALMA and consistency going forward as calibration of the SMA data with the CASA package becomes more common. One can access my flux calibration programs by including the directory containing these programs first in order in the IDL path or by copying the programs into the current working directory. Details are provided in the web page accessed from the tab, Calibrating SMA Data, Section, 1. Preliminaries.
|
Getting Started
Since MIR only calibrates the SMA data, once you are finished the calibration you will need to convert the data from its native SMA/MIR format into a format readable by one of the packages used for interferometric imaging such as AIPS, MIRIAD, or CASA.
There are two conversion programs in MIR.
The program idl2miriad writes a MIRIAD style UV data set. This program requires a shared object library (libidlsmamiriad.so) built specifically for MIR from compiled MIRIAD programs. The location of this shared object library is indicated by the IDL variable e.idl_sav. If you are using our CF or RGLINUX computers, all this is already managed.
Alternatively one can image in AIPS by writing an AIPS style UV-FITS file using the IDL program write_fits or its interactive equivalent mir_fits. In this case no shared object library is needed.
CASA can read AIPS UV-FITS file. We have not yet tested a procedure within CASA for imaging UV data that has been calibrated in MIR.
The sample calibration on this website uses data from June 3, 2010 (100603_02:56:50), which is non-proprietary and therefore directly downloadable.
Calibrating SMA Data
A visual calibration procedure developed by Eric Keto for the MIR software.
"Calibrating SMA Data" (pdf version)
First draft: 2013
Updated: November 2015
I developed the data analysis flow described in this example so that the astronomer can see each calibration procedure in the graphics of the MIR package.
The sample calibration on this website uses data from June 3, 2010 (100603_02:56:50), which is non-proprietary and therefore directly downloadable.
eketo@cfa.harvard.edu
1.1 Bandpass, Gain, and Flux Calibration
There are 3
steps to calibrate SMA data. First, the bandpass calibration corrects for
variations as a function of frequency across the bandpass. Second, the gain
calibration corrects for variations in time over the night's observation. Third,
the flux calibration sets the absolute flux level.
All the
calibrations, bandpass, gain, and flux, compare observations of a calibrator
against a model. Whatever corrections are needed to bring the observations of
the calibrator in line with the model are then applied to the observations of
the science target. The assumption is that the same corrections needed for the
calibrator are the same needed to improve the observations of the science
target.
1.2 The calibration steps in order
What is the correct order for
the 3 steps in the calibration process, bandpass, gain, and flux?
The gain and flux calibration use the continuum flux of the respective calibrators. The SMA does not measure a broad-band continuum of the correlated power but constructs one by averaging all the spectral line channels together. Therefore, the bandpass calibration which removes variations in phase and amplitude with frequency should be done first to improve the signal-to-noise of the averaging.
The gain calibration corrects for variations over time in the response of the instrument including effects of changing sky conditions that affect the measured amplitudes and phases but does not set an absolute value for the flux. The second step of the flux calibration, comparison with planets, sets this absolute value by multiplying all the observed amplitudes by the ratio of the modeled and observed amplitudes of planets or their satellites. Mathematically it does not matter whether we multiply the data by this ratio before or after the gain calibration. I will describe a procedure with the gain calibration done before the flux calibration. This allows us to work with observations of the gain calibrator that are phase-calibrated for better signal-to-noise and for consistency in their estimated fluxes. The motivation for this order should become clearer as the procedure is described.
1.3 Improved version of the flux calibration
The procedure described here uses versions of the calibration programs that I modified to use models for the brightness of the planets taken from ALMA's CASA software. These models include corrections for spectral line contamination in Neptune and Titan. The modified flux calibration program also plots the observed and modeled visibilities and prints out a detailed list for each baseline so that the astronomer can assess the reliability of the calibration. In the latest version, I used programs taken from the standard MIR distribution in October 2015. The modified programs are compatible to the MIR distribution at least of that date.
The
modified programs are:
flux_cal.pro
flux_primary.pro
gain.pro
flux_measure.pro
In addition
rad.pro is required to read the planet models.
Of course,
the planet models themselves are required.
Here is a
modified setup file that will include the modified programs and the planet
models and run the other programs in the standard distribution of MIR. This
setup file works on our CF computers.
#XDXDIR is the directory with the improved flux calibration programs
setenv XDXDIR /sma/SMAusers/keto/mir/idl/new
#
# Specify the directory with the models of the planets.
setenv PLANET_MODELS_DIR /sma/SMAusers/keto/mir/idl/new/planetmodels
#
#RDXDIR is the directory with all the rest of the MIR programs
setenv RDXDIR /sma/mir
#
setenv IDL_STARTUP $RDXDIR/idl/idl_startup
#
#These lines are normal idl setup lines except that $RDXDIR/idl has
#been added to the front of the path
#
#Include the XDXDIR directory in the IDL_PATH ahead of the RDXDIR.
#
setenv IDL_PATH \+$XDXDIR/:\+$RDXDIR/idl/pro/:\+$IDL_DIR/lib
setenv IDL_DEVICE X
setenv CAMPUS_LOGIN sma
Source the
setup file and start IDL as usual.
On our CF
computers:
source /sma/SMAusers/keto/mir/setup
idl
On our
RGLINUX computers
source /czdata/keto/mir/setup
idl
Read the
data from the current working directory
readdata
With no
arguments, select shows the sources, etc., in the data set
select
Shows
the data has these sources
saturn
mars
1058+015
PG1119+120
1310+323
PG1404+226
You can
usually figure out the observing strategy from this plot. (This is plo_var not plot_var.) Select the continuum band for
plotting.
select, /p, /re, band='c1'
result = plo_var('dhrs','el',frames_per_page=1)
Figure 2‑1 Elevation vs time. The
purpose of this plot is to understand how the observations were made or the
observing strategy. The plot shows two science targets (green) PG1119 and
PG1404, each with its own gain calibrator, 1058+0 and 1310+3, and mars which is
used for both flux and bandpass calibration.
Mars is
observed first and is both the bandpass and flux calibrator.
1058+015
is the gain calibrator for PG1119+120.
1310+323
is the gain calibrator for PG1404+226.
1058+015
is usually about 0.5 Jy
1310+323
about the same
Use mars
for passband
There is
no data for Saturn.
Plot the
system temperature vs time. This uses plot_var instead of plo_var.
plot_var, x='int', frames=9
Figure 2‑2 This plot shows that
antenna 5 has problems because of the sudden drop in system temperature around
integration 400. Needs some
flagging.
Flag
using the integration number to select the data.
select, /p, /re,ant=5
select,int=[355,537]
flag,/flag
select, /p, /re,ant=5
select,int=[649,697]
flag,/flag
select, /p, /re,ant=5
select,int=[775,810],source='PG1119+120'
flag,/flag
Plot the
system temperature again to check on the flagging.
select, /p, /re, band='c1'
plot_var, x='int', frames=16
Figure 2‑3 Plot the data again to
be sure the flagging was done correctly.
Looks OK
now.
We have
finished the initial assessment and flagging of the data.
The next
step is to set the flux levels of the data with the chopper wheel calibration.
3.1 Chopper-wheel calibration
The initial flux calibration
is done with the "chopper-wheel" calibration method invented by Bob Wilson and
Arno Penzias (Penzias and Burrus, 1973, ARAA, 11, 51). This calibration is
determined by the SMA online system before the data is recorded. This method
measures the difference in voltage for a single dish telescope or in correlator
counts for an interferometer between observations of a room temperature (hot)
load inserted in the beam and the cold blank sky. The SMA uses a combination of
the two as will be explained in the mathematics of the chopper-wheel
calibration. In any case, the working assumption is that the emission of the
hot load is that of a black body at the air temperature while the emission of
the sky is also that of a black body at the same temperature but multiplied by
the hopefully small atmospheric opacity. The ratio of the difference in
measured voltages or counts and the difference in the known or modeled emission
of the hot load and the sky is the gain or response of the telescope which is
used to scale the measured voltage or counts of the science target to the true
brightness temperature.
The method corrects for the
variation in instrumental gain with time and for changes in the atmospheric
opacity which may occur as a result of changes in the elevation of the
observation or weather conditions. This is a first step to make the response of
the interferometer uniform over the course of a night.
The scaling determined by the
online system is applied in the offline calibration with the command apply_tsys which is always run at the beginning of the data analysis. There
are no inputs or options. The calibration is done entirely using data provided
by the system. The apply_tsys command also weights the data by the square of
the measured system temperature, Tsys2, to improve the
signal-to-noise of the averaging of the individual integrations observed over
the night.
The apply_tsys command also converts the system temperature measured by the online
system from a double side band (DSB) definition to a single side band (SSB)
definition. These differ by a factor of 2. Receiver temperatures in other SMA
documentation generally refer to the lower temperatures in the DSB definition
so one has to keep this in mind when comparing the system temperatures before
or after apply_tsys to the expected receiver temperatures.
3.2 A simple derivation of the method
The details of the
chopper-wheel calibration are handled by the online system. This derivation of
the mathematics is intended to explain how the method accounts for variations
in gain and atmospheric opacity to better understand the accuracy of the
method.
The power measured from the
blank sky, Psky, which
could be measured in any units can be described as,
where G is a time dependent gain that converts brightness temperature to
whatever units, voltage or correlator counts, the system is measuring. This
represents the response of the entire telescope. The factor B, between 0 and 1, is the fraction of
the beam that is looking at the sky while the factor (1-B) is the fraction of the beam not looking at the sky, for
example, looking at the ground (spill over) or the support legs of the
secondary mirror. The optical depth and brightness temperature of the sky are τ and Tatm, and Tearth is the temperature of the
ground and the antennas. The contribution to the measured power due to receiver
noise and other sources of instrumental noise is PRX which we usually refer to in units of temperature, TRX = PRX/G. All the
factors vary on the timescale of the observation, so I have not written in an
explicit dependence, as for example G(t).
If most of the water vapor,
the major emitter in the submillimeter, is in the lower atmosphere near the
ground, then the atmosphere and the earth may be assumed to be at the same
temperature, T, then,
At a single dish telescope,
the power received when observing an astronomical source is,
However, at the SMA, the
power measured on a source is the correlated power between two antennas. In
this case, there is no contribution from the sky or the receiver because the
noise power is uncorrelated. At the SMA, the gain G2 is different by a factor of 2. The gain is different
because the sky power is measured by a broadband continuum detector which
measures the power before the signal is sent to the correlator. At this point
the bandwidth includes both sidebands of the receiver. The correlator separates
the signals from the two sidebands with each having half the width. Since the
power is integrated over the bandwidth, the change in power from a source
measured through the sideband separating correlator is half the change in power
of the same source measured through the the broad band continuum detector.
with G2 = ½ G.
Define,
for a single dish telescope,
and for the SMA. The single dish
telescope requires a separate observation of the blank sky usually done by
chopping and nodding the antenna. With an additional
measurement of the power received from a source of known temperature we can get
the scaling necessary to determine the true brightness temperature of our
astronomical source. At a single dish telescope this is usually done by
observing through a disk divided into quadrants, two open and two blocked by an
absorber. Every so often the disk is rotated and the power from the sky and the
absorber are measured synchronously. The SMA uses a vane that rotates into the
beam. Measurements of the vane are typically made when the antennas are
slewing. where the factor of 2 applies
to the SMA with G2 = 1/2G but not to a single dish telescope. The SMA online system
actually records the system temperature which is defined as, If you like, you can think of
Tsys as the sky
temperature corrected for atmospheric absorption although this isn't relevant.
The definition is useful because Tsys
is easily measured with the chopper wheel or vane as in the equation above, and
because the brightness temperature of our target source can be found by scaling
the observed power by Tsys, where the factor of 2 applies
again to the SMA and ΔPsrc is our normalized correlator count. The
command apply_tsys absorbs the factor of 2 into the system temperature. So before apply_tsys, Tsys recorded in the data is
the double sideband system temperature measured by the SMA broadband continuum
detectors. After apply_tsys, Tsys
is exactly twice as large. This corresponds to the single sideband system
temperature since a source must be twice as hot to produce the same power in
half the bandwidth. There is one more step to
convert the source brightness temperature to flux in Jy/beam. The factor Ω/Δ2
is the area of the antenna, 9π m2,
so S = 98 Tsrc Jy/beam. However the effective area of the SMA dish
is not exactly that of 6m diameter disk. There are inefficiencies in the
blockage from the secondary mirror and its support legs, beam spillover, and
imperfect reflecting surfaces. The apply_tsys program assumes an efficiency of 75% so that S = 130 Tsrc Jy/beam. The following table from the SMA
Engineering Test Program web page shows that this is a pretty good assumption
for observations around 230 GHz, but the true efficiency is lower by 8.7% in
the 345GHz band. These efficiencies were measured by comparison with the fluxes
expected in observations of planets. Figure 3‑1 Measured aperture
efficiencies. The dates of the last measurement refer to the surface accuracy
and not the efficiency. This chopper wheel
calibration is a reasonably good calibration. It removes variations in gain and
atmospheric opacity over the night and provides an estimate of the brightness
temperature. We don't have an estimate of the accuracy of this calibration that
is measured with the SMA, but from experience with other telescopes, probably
20% or better. With this explanation, you
can assess for yourself the accuracy of the chopper-wheel calibration or at
least have a better appreciation of the sources of uncertainty or error. The Tsys is measured at least as
often as the SMA slews between the gain calibrator and the science target. So
the method corrects for changes in gain and opacity at least this often. In
fact, the sky temperature is measured continuously since the power of the
science target is negligible compared to the power from the sky. As for the
dish efficiency, one would suppose that it is fairly stable in the 230 GHz
although perhaps less so at 345 GHz. Still, the maximum variation in the
efficiencies of the 6 dishes measured at 345 GHz is only 12%. So maybe the accuracy is 10%
or 20%, but certainly not as bad as a factor of 2. The goal of the flux
calibration with planets and their satellites is to improve on this. The
function apply_tsys will
calibrate the data by the chopper wheel method, weight by Tsys2,
and double the system temperatures to the single sideband definition. select, /p, /re apply_tsys Average
the data to recalculate the continuum. uti_avgband mir_save,'mir-save-tsys-applied' That's
it.
An efficiency of 75% is a good estimate for observations
at 230 GHz and 66% at 345 GHz.3.3
Applying the chopper-wheel calibration
The
passband calibration is done in two steps, phase and then amplitude.
4.1 Phase only pass band calibration.
View the
spectrum before calibration.
plot_spectra, /norm, source='mars',
frames_per_page='16'
Figure 4‑1 Phase and amplitude vs channel for the
observation of Mars before the passband calibration. The panels show different
baselines. The phase in the upper and lower sidebands (orange and blue) drift
in opposite directions vs channel. The amplitudes show the shape of the 104 MHz
filters on the individual chunks. The first step is to correct the phase.
Usually the
MIR plotting programs produce several pages of panels or frames showing the
various baselines or spectra. In this tutorial, I am showing just the first
page in most cases.
Passband
calibration. Phase and amplitude separately. Phase first.
select, /p, /re
pass_cal, cal_type='pha', tel_bsl='telescope',
refant=5, smooth=8,frames_per_page='16'
In the
revolving menu type "mars yes" to
select Mars as the passband calibrator. The calibration solution derived from
the Mars data will ultimately be applied to all the other sources in the menu
not selected for use, those marked "NO".
Figure 4‑2 Plot produced by pass_cal program showing
the variation in phase across the individual 104 MHz chunks. The panels are
labeled by "1" for antenna 1, "345" for the receiver, and "sNN" where NN is the
number of the individual 104 MHz chunk of the spectrum. The blue asterisks are the
data and the red line is the fit to the data. The passband calibration consists
of applying the correction determined by the fit to the data. The goal is have
the fit follow the significant variations in the phase across each chunk but
not follow the noise. Change the length of the smoothing over channel to
achieve this.
Since the
smoothed function follows the data reasonably well.
Answer "Yes" to apply the calibration to Mars
itself and the other sources in the menu.
You can
check the pass band calibration by running the calibration function pass_cal
again with the same parameters. The plot shows the calibrated data and
estimates a new smoothed correction. Don't worry about the edges. These overlap
the neighboring filters and won't be used.
Figure 4‑3 Plot of the phase vs
channel after the phase calibration.
The plot
shows the calibration was successful because the phases are reasonably flat
across the filters. Don't worry about edges of the chunks. The chunks are
spaced to overlap each other and the edges are never used in final data.
After
the plotting, be sure to answer "No" when asked whether to apply the pass band
correction otherwise the second correction will just add noise to the data. For
example, the noise in the second correction is easily seen in jagged red
solution for the S13 filter (lower left). Next recalculate the continuum with
the corrected phases.
select, /p, /re
uti_avgband
With the
phases corrected, run the pass band calibration again to correct the amplitudes.
select, /p, /re
pass_cal, cal_type='amp', tel_bsl='telescope',
refant=5, smooth=8,frames_per_page='16'
Figure 4‑4 Plot of the variation in amplitude vs channel number for each of the individual chunks. This plot shows the variation of the amplitude before the amplitude calibration. The blue histogram shows the data and the orange line shows the fit or essentially the correction. The panels are labeled as explained in figure Figure 4‑2 Plot produced by pass_cal program showing the variation in phase across the individual 104 MHz chunks. The panels are labeled by "1" for antenna 1, "345" for the receiver, and "sNN" where NN is the number of the individual 104 MHz chunk of the spectrum. The blue asterisks are the data and the red line is the fit to the data. The passband calibration consists of applying the correction determined by the fit to the data. The goal is have the fit follow the significant variations in the phase across each chunk but not follow the noise. Change the length of the smoothing over channel to achieve this.
The plot
shows the typical 104 MHz filter shape. The smoothing follows the data reasonably
well so apply the calibration by answering "Yes". Then recalculate the continuum. This applies the
solution to Mars and also the other sources in the menu.
select,/re,/p
uti_avgband
Run the
pass band calibration with the same parameters to check the correction. Don't
worry about messages that the fitting failed. Here we are just using the
pass_cal function to plot the data. We won't be applying the correction.
Figure 4‑5 Plot of the amplitude
vs channel after the amplitude calibration.
The
correction looks OK. Again the drop off at the edge of the chunks can be
ignored. Answer "no", do not apply
because we already applied this correction. This time we are running pass_cal
again just to produce this plot and check the result.
Check
the combine phase and amplitude calibration by plotting the spectrum of Mars again.
The phases and amplitudes are now flat except for the spikes from the edges of
the filters. These can be ignored.
plot_spectra, /norm, source='mars',
frames_per_page='16'
Figure 4‑6 Plot of the both the phase and
amplitude vs channel similar to plot Figure 4‑1
except here after the phase and amplitude calibration. Both are flat except for
the spikes on the amplitude at the edges of the chunks.
Check
that the spikes really are from the filter edges by plotting the spectrum again
with the option ntrim=8 that cuts out the 8 channels on either side of the 128
channel
104 MHz
filter band.
plot_spectra, /norm, source='mars',
frames_per_page='16',ntrim=8
Figure 4‑7 Same as figure Figure 4‑6
except using the ntrim=8 option to exclude plotting the channels at the edges
of the chunks. Now the spikes in figure Figure 4‑6
are gone.
This plot also
shows different noise levels on different baselines. The longer baselines have lower
signal-to-noise because Mars is partially resolved, and the signal strength is
less on the longer baselines.
Here is a
plot that shows the inverse correlation of flux with baseline length. The
amplitude (y-axis) is plotted versus projected baseline length (x-axis).
result=plo_var('prbl','ampave',frames_per_page=16)
Figure 4‑8 Plot of the average amplitude vs projected baseline length. This plot shows that the amplitude is strongest on the shortest baselines meaning that this source, Mars, is at least partially resolved by the SMA baselines. This plot is not really useful for the calibration procedure. Included just to illustrate this point.
Baselines
1-6 and 3-6 are the longest and have the lowest amplitudes. Baselines 2-6, 3-8,
and 4-5 are the shortest and have the highest amplitudes. The noise in the
spectrum is inversely correlated with the amplitude. This plot is not part of
the calibration procedure, but is here for illustration.
We use antenna-based solutions selected by the option, tel_bsl='telescope'
Figure 4‑9 Plot of the spectrum of a single baseline after antenna-based pass band calibration showing poor amplitude normalization. The variations are due to the different sensitivities of individual boards in the correlator. Always use baseline-based pass band calibration to avoid this. This data is from an observation of 3C454.3, a particularly bright quasar, and the calibration errors are obvious in this high signal-to-noise spectrum. This observation is a different data set than the one illustrated in this example.
With the pass band calibration finished, this is a good place to save the data.
mir_save, 'passband_calibrated'
Next, run
the gain calibration to remove the variation of phase and amplitude with time.
The
observation was set up with two targets and a different gain calibrator for
each. For each pair, run the phase and amplitude gain calibration separately.
5.1 Gain calibration phase only
This
data set has two science targets and a different gain calibrator for each.
1058+015
is the gain calibrator for PG1119+120
1310+323
is the gain calibrator for PG1404+226
Run the gain_cal procedure on your gain calibrator (quasar)
selecting just the phase and not the amplitude for calibration. Include the
science targets in the selection before running gain_cal. Exclude the flux calibrator. When your gain
calibrator comes up in the rotating (endless loop) menu, it will show a value
for its flux. This was assigned by the procedure apply_tsys. If the fluxes are zero (this can happen if
the calibration procedure was restarted after apply_tsys with mir_restore and mir_save) exit the loop by hitting return a couple of
times and run flux_measure. The flux of the gain calibrator is the reasonable estimate from
the chopper-wheel calibration, but we can ignore it for phase calibration.
Here is
the command to get or update the fluxes if necessary.
flux_measure,/scalar
Scalar average for band c1
Source
Flux
StdDev SNR Nscans On source Mean UT
(Jy)
(Jy)
time (min) (hours)
mars
72.8769
2.0640 35.3 24 11.9 3.02
1058+015
1.3074 0.0056 232.9 268 82.3 6.10
PG1119+120
0.1692 0.0007 254.3 480 237.4 5.93
1310+323
0.9707
0.0046 213.3 71 35.1 10.09
PG1404+226
0.1750 0.0010 175.2 240 118.7 10.11
Start
the gain calibration for the first pair.
select,/re,/p,source=['1058+015', 'PG1119+120']
gain_cal,cal_type='pha',x_var='hours',tel_bsl='telescope',$
refant=4,frames_per_page='16',smooth=2
This
starts the repeating loop for source selection. Enter the name of the gain
calibrator followed by yes as shown in red below. Notice that the program has picked
up the flux of the sources from the last execution of flux_measure although the
flux is not relevant to the phase calibration. Then hit Enter to exit the loop
and start the calibration.
% Compiled module:
GAIN_CAL.
-------------------------------------------
Single Band
Calibration Mode
Main Band Calibration
-------------------------------------------
% Compiled module:
GAIN_INI.
Set which sources will be
used as calibrators and set the calibrator fluxes
The flux should be in Jy at
the frequency of the observation
These are the sources and
their current gain codes
name: 1058+015
gain code: NO flux
(Jy):
1.3074342
name: PG1119+120
gain code: NO flux
(Jy):
0.1694476
Enter source, cal code, and
if cal, flux in Jy, eg: 3C273 YES 3.1
or hit Return if all the
sources are correctly specified
: 1058+015 yes
These are the sources and
their current gain codes
name: 1058+015
gain code: YES flux (Jy): 1.3074342
name: PG1119+120
gain code: NO flux
(Jy):
0.1694476
Enter source, cal code, and
if cal, flux in Jy, eg: 3C273 YES 3.1
Figure 5‑1
Plot of the phase and coherence vs time for observations of the gain
calibrator before the phase calibration. The coherence can be ignored.
Massive
phase jump. This looks like pointing data in the middle of the time range. When
you are done plotting, you will have the option to apply the calibration or
not. I want to flag the pointing data, so exit from gain_cal by entering NO as
shown in red below to not apply the calibration
EXITING FROM THE
INTERACTIVE MODE
% Compiled module:
GAIN_STORE.
% Compiled module:
CAL_STORE.
% Compiled module: DAT_COMB_SEP.
Stacking gain curves to the
ca structure
% Compiled module:
PLO_PLID_REL.
Apply gain solution? [NO
<YES>]: No
NO: nothing done
IDL>
Flag the
data by integration time. Pointing data have less than 10 seconds time on each integration,
so the following filter will select all the pointing data.
result=dat_filter(s_f,' "integ" lt
"10" ')
flag,/flag
select,/re,/p,source=['1058+015', 'PG1119+120']
Also the
solution, the red line in the plot above, is not following the data very well,
for example on antenna 5, lower left corner where the phase shows some
variation between the hours of 3 and 6. So run the amplitude calibration again
with a shorter smoothing time, one half hour to better follow the variations in
phase with time.
gain_cal,cal_type='pha',x_var='hours',tel_bsl='telescope',
$
refant=4,frames_per_page='16',smooth=0.5
Figure 5‑2 Phase vs time, same as figure Figure 5‑1,
but after flagging and shortening the smoothing time. The red line now looks
like a good fit to the data.
Looks
good. Now the red line of the fit follows the wiggles between the hours 3 and 6
in the phase of antenna 5. Exit the plotting mode and then answer Yes to apply
this calibration.
EXITING FROM THE
INTERACTIVE MODE
Stacking gain curves to the
ca structure
Apply gain solution? [NO
<YES>]: yes
YES: apply gains
Applying Gains
% Compiled module:
CAL_APPLY.
% Compiled module:
UTI_INTERP.
IDL>
uti_avgband
Compare
the vector and scalar averages of the fluxes.
flux_measure,/vector
Vector average for band c1
Source
Flux
StdDev SNR Nscans On source Mean UT
(Jy)
(Jy)
time (min) (hours)
1058+015
1.5041 0.0051 294.7 145 71.7 5.95
PG1119+120
0.0011 0.0010 1.1 480 237.4 5.93
flux_measure,/scalar
Scalar average for band c1
Source
Flux
StdDev SNR Nscans On source Mean UT
(Jy)
(Jy)
time (min) (hours)
1058+015
1.5662 0.0051 310.1 145 71.7 5.95
PG1119+120
0.1694 0.0007 251.6 480 237.4 5.93
Now the
scalar and vector averages for 1045+015 are less than 6% different, and we are
done with the phase calibration using the gain calibrator.
5.2 Setting a representative flux value for the gain calibrator
Now we need to set the flux
of the gain calibrator. Run gain_cal again,
this time for amplitude but do not apply the results. We just want to look at
the plot of the measured amplitude with time over the course of the night.
Hopefully, this is flat. If not, and this is unusual, the chopper-wheel
calibration has failed to correct for the variation in gain with time. The gain_cal
procedure will fix this, but it leaves us with the question of what flux we
should use for the quasar in combination with the flux calibration. There are
at least three choices.
The usual case would be to
use the average conditions over the whole night. However, the observation of
the flux calibrator was done at a particular time and elevation. Two
possibilities are to restrict the comparison in the flux calibration step to
either this time or this elevation. Which is the correct choice among the
three? Again, the cases where the chopper-wheel calibration fails badly are not
common, and I do not have an actual example. Therefore I have drawn three
hypothetical cases by hand, shown in the figures below. All three figures plot
the amplitude of the quasar (gain calibrator) versus time.
Figure 5‑3 Case 1: The gain
(response) of the system is stable over time or stable enough that the
chopper-wheel calibration is able to correct any trends or drifts. In this case
the amplitude of the quasar is constant. We can use the average flux over the
night.
Figure 5‑4 Case 2: The gain varies
with elevation so that the flux of the quasar appears to become brighter and
then fainter from rise to set. Hypothetically, the gain and the calibrator flux
could increase with elevation as the atmospheric optical depth decreases as the
path length through the atmosphere decreases. The elevation of course varies
with time as the target rises and sets. In this case, we might select
observations of the quasar at the same elevation as the observation of the flux
calibrator. Here Tcal is the time when the gain calibrator is at the same
elevation as the flux calibrator.
Figure 5‑5 Case 3: The gain varies,
but not obviously with elevation.
We might or might know what causes this, but we could select
observations of the quasar around the time that the flux calibrator was
observed.
At this point, we are not
saying that the true flux of the quasar is the flux selected by one of the methods
above. We are going to correct this flux to the true flux in the flux
calibration steps coming up in a moment. In other words, if the correction in
the flux calibration step is appropriate only at some elevation or some time,
then if we correct the flux that the quasar had at that same elevation or time,
then it will have its true flux after the correction. Because the same
corrections are applied to the target, then the target will also have its true
flux after the corrections.
In any case, whatever flux
you decide to use in the amplitude gain calibration, here is how it is done.
Use the select or dat_filter
commands to select a set of observations of the gain calibrator quasar either
around the elevation of the flux calibrator, or the time it was observed, or
all the observations of the gain calibrator over the whole night. Run the
command flux_measure to get
an estimate of the flux of the quasar with this selection. The numerical value
in Jy/beam is relying for the moment on the chopper-wheel calibration done by apply_tsys. This
command also applied the same calibration to the amplitudes of the target. So
the calibrator and target fluxes are consistent
The command flux_measure has an
option for either the /scalar or /vector average of the amplitudes.The gain_cal program scales the amplitudes so that the scalar average matches the input flux. Thus you must use the /scalar option. This is the default. We are not saying that the true flux of the quasar is given by the scalar average. This is just what the software uses in this calculation.
5.3 Gain calibration for amplitude
Let's see how this works on
our example data set
The
antenna based amplitude calibration does not work. This is simply a bug in MIR.
So we will use baseline based calibration.
gain_cal,cal_type='amp',x_var='hours',tel_bsl='baseline',$
refant=4,frames_per_page='16',smooth=1
Figure 5‑6 Amplitude vs time for different baselines and
sidebands. Plot shows the amplitudes before the amplitude calibration. The red
line is a fit to model the variation in amplitude over time.
These
amplitudes vary a lot. They decline for several hours, then increase after the
observations were interrupted to optimize the pointing. Perhaps the new
pointing offsets successfully centered the primary beams on the gain calibrator,
and the fluxes after the pointing are a better estimate of the true flux of the
quasar. I can't think of a better hypothesis, so use the select command and
flux_measure to average the amplitudes of the quasar from 6.5 hours to the end
of the observations at 9 hours.
Also, I
don't like fit of the smoothed function where it crosses the time when the
pointing data was taken. I don't want to force the quasar amplitudes to follow
the red line here because this would force the data on the science target to
follow the same correction. I think the transition should be a jump from lower
to higher fluxes rather than ramp. I will flag all the data in this transition
period, calibrator and target source.
First,
get out of the gain_cal program. At the prompt, answer "NO", do not apply the
calibration.
Flag the
data from from 6.1 to 6.4. The MIR header variable is dhrs.
select,/re,/p
result = dat_filter(s_f,' "dhrs" gt
"6.1" and "dhrs" lt "6.4" ')
flag,/flag
Run the
selection again to get the average flux of the quasar after 6.5 hours. Remember to use the scalar average which is default.
This select command clears the previous
selection, and the dat_filter command
restricts the time.
select,/re,/p
result = dat_filter(s_f,' "dhrs" gt "6.5"
')
IDL> flux_measure,/vector
Vector average for band c1
Source
Flux
StdDev SNR Nscans On source Mean UT
(Jy)
(Jy)
time (min) (hours)
1058+015
1.7780 0.0062 289.0 56 27.7 7.73
PG1119+120
0.1757 0.0011 163.9 204 100.9 7.60
1310+323
0.9707 0.0046 213.3 71 35.1 10.09
PG1404+226
0.1750 0.0010 175.2 240 118.7 10.11
IDL>
Run the
gain calibration program after first selecting the gain calibrator and the
science target. This time, in the rotating endless-loop menu, type in the value
for the amplitude of the gain calibrator that was just obtained with
flux_measure. This value may already be there on the line for the gain
calibrator. If not, type it in like this. Also shown in red below.
1058+015 YES 1.7780
IDL> select,/re,/p,source=['1058+015',
'PG1119+120']
Here are the data selection
criteria
("source" like "1058+015"
or "source" like
"PG1119+120")
1196286 passed in
filter
IDL>
IDL> gain_cal,cal_type='amp',x_var='hours',tel_bsl='baseline',$
IDL> refant=4,frames_per_page='16',smooth=0.5
-------------------------------------------
Single Band
Calibration Mode
Main Band Calibration
-------------------------------------------
Set which sources will be
used as calibrators and set the calibrator fluxes
The flux should be in Jy at
the frequency of the observation
These are the sources and
their current gain codes
name: 1058+015
gain code: YES flux (Jy): 1.3074342
name: PG1119+120
gain code: NO flux
(Jy):
0.1757000
Enter source, cal code, and
if cal, flux in Jy, eg: 3C273 YES 3.1
or hit Return if all the
sources are correctly specified
: 1058+015 YES 1.7780
These are the sources and
their current gain codes
name: 1058+015
gain code: YES flux (Jy): 1.7780000
name: PG1119+120
gain code: NO flux
(Jy):
0.1757000
Enter source, cal code, and
if cal, flux in Jy, eg: 3C273 YES 3.1
or hit Return if all the
sources are correctly specified
This
will force all the amplitudes of the gain calibrator to this value and apply
the same correction to the data of the science target. Again, as explained in
the section above, we are not saying, yet, that this is the final true value
for the flux of the quasar. It would be if we stay with the calibration set by
the chopper-wheel method. However, we will still run the flux calibration later
for a final correction based on the observation of Mars.
Figure 5‑7 Amplitude vs time same as Figure
5‑6 but after flagging the
data between 6.1 and 6.4 hours.
The fit
of the red line looks OK. No bad data. So, answer Yes, at the prompt to apply the
solution.
Then run
gain_cal again to check the correction. This time we use type='amp,pha' to
check both the phase and amplitude calibration. I use tel_bsl='telescope' (antenna
based) to reduce the number of plots.
Figure 5‑8 Amplitude and phase of the
gain calibrator after calibration
That
finishes the gain calibration for the first calibrator-target pair. Run the
same procedure on the next pair.
5.4 Phase and amplitude on the next calibrator-target pair
select,/re,/p,source=['PG1404+226','1310+323' ]
Need a
flux for the gain cal.
Just for
interest or to make sure the sources have fluxes, you can run flux_measure to
compare the scalar and vector averages.
IDL> flux_measure,/vector
Vector average for band c1
Source
Flux
StdDev SNR Nscans On source Mean UT
(Jy)
(Jy)
time (min) (hours)
1310+323
0.1185 0.0137 8.7 71 35.1 10.09
PG1404+226
0.0021 0.0015 1.4 240 118.7 10.11
IDL> flux_measure,/scalar
Scalar average for band c1
Source
Flux
StdDev SNR Nscans On source Mean UT
(Jy)
(Jy)
time (min) (hours)
1310+323
0.9707 0.0046 213.3 71 35.1 10.09
PG1404+226
0.1750 0.0010 175.2 240 118.7 10.11
Now run
the gain calibration for phase, and enter the calibrator name and yes in the
menu as shown in red below.
gain_cal,cal_type='pha',x_var='hours',tel_bsl='telescope',$
refant=4,frames_per_page='16',smooth=1
-------------------------------------------
Single Band
Calibration Mode
Main Band Calibration
-------------------------------------------
Set
which sources will be used as calibrators and set the calibrator fluxes
The
flux should be in Jy at the frequency of the observation
These
are the sources and their current gain codes
name: 1310+323 gain
code: NO flux (Jy): 0.1184527
name: PG1404+226
gain code: NO flux
(Jy):
0.0021214
Enter
source, cal code, and if cal, flux in Jy, eg: 3C273 YES 3.1
or
hit Return if all the sources are correctly specified
: 1310+323 yes
These
are the sources and their current gain codes
name: 1310+323
gain code: YES flux (Jy): 0.1184527
name: PG1404+226
gain code: NO flux
(Jy):
0.0021214
Enter
source, cal code, and if cal, flux in Jy, eg: 3C273 YES 3.1
or
hit Return if all the sources are correctly specified
Figure 5‑9 Phase vs time same as Figure
5‑1 but for the second gain
calibrator in the data set.
Looks
OK.
Yes, apply to the data. Check the flux
measurements again.
IDL> flux_measure,/vector
Vector average for band c1
Source
Flux
StdDev SNR Nscans On source Mean UT
(Jy)
(Jy)
time (min) (hours)
1310+323
0.8775 0.0054 163.0 71 35.1 10.09
PG1404+226 0.0027 0.0015 1.8 240 118.7 10.11
IDL> flux_measure,/scalar
Scalar average for band c1
Source
Flux
StdDev SNR Nscans On source Mean UT
(Jy)
(Jy)
time
(min) (hours)
1310+323
0.9707 0.0046 213.3 71 35.1 10.09
PG1404+226
0.1750 0.0010 175.2 240 118.7 10.11
IDL>
The vector
flux on the gain calibrator is now much closer to the scalar flux. The science
target has not improved, but this source turns out to be a non-detection. So
there is no correlated flux. Since we just ran flux_measure,/scalar with the scalar average, this is the value stored as the flux of the gain calibrator, and this is what we need for amplitude gain calibration.
Now the baseline-based
gain amplitude calibration for the second pair.
gain_cal,cal_type='amp',x_var='hours',tel_bsl='baseline',$
refant=4,frames_per_page='16',smooth=1
Figure 5‑10 Amplitude vs time, same as
Figure 5‑6, but for the second gain
calibrator in the data set.
The amplitudes for this quasar are constant with time in the
sense of not drifting, so we can use the average amplitude that was just set by
flux_measure. This is the value that was already shown in the menu when we ran
gain_cal. The fit of the red line to the data looks fine. So we can apply these
corrections.
Answer, Yes, to apply the corrections.
Now run
the gain cal again with both amplitude and phase to check the results for the
second pair of target and calibrator.
gain_cal,cal_type='amp,pha',x_var='hours',tel_bsl='telescope',
refant=4,frames_per_page='16',smooth=1
Figure 5‑11 Phase and amplitude vs
time for the second gain calibrator after the phase and amplitude calibration.
Looks, OK. Answer No,
do not apply. This last run of gain_cal was just to produce this plot as a
check.
Now we finished
the gain calibration for both our science targets.
select,/p,/re
mir_save,'gain_calibrated'
The flux calibration program
is easier to run than it is to describe. The concept is perfectly simple. Find
the correction as the ratio of modeled and observed fluxes, but there are
details in the modeling.
6.1 Models of Planetary Brightness temperature
The models for the planets
and their satellites include data on the brightness temperature of the body and
the apparent size from Earth. I use the models for the brightness temperature
from ALMA's CASA software. The apparent size of the planet is computed by the
SMA's guidance software and included among the header variables for each
observation.
The figure below shows
several models for the brightness temperature of Uranus. The red or pink
crosses show a model from the old OVRO MMA which I included when I set up the
SMA's flux calibration programs in 2003. This was later replaced with the model
shown by the blue diamonds, and later with a functional fit shown as the green
dashed line. The ALMA-CASA model is the white line. At 227 GHz, the old MMA
model and its later replacement differ by 12%. The more recent functional fit
and the ALMA model agree quite well within the SMA's operating band of 200
– 400 GHz. This history provides some indication of the uncertainties in
the model of brightness temperature.
Figure 6‑1 Comparison of 4 models
for the brightness temperature of Uranus plotted as temperature versus frequency.
See text above.
The ALMA-CASA models for
Neptune and Titan (shown in the two figures below) include significant
contamination from spectral lines, particularly CO. Uranus as shown in the
figure above is relatively unaffected.
Figure 6‑2
The ALMA-CASA model of the brightness temperature versus frequency for Neptune
Figure 6‑3 The ALMA-CASA model of the brightness temperature versus
frequency for Titan.
The effect of the spectral
lines, principally CO, are included in the models of Neptune and Titan as shown
in the figures below.
Figure 6‑4 Model of the brightness temperature of Neptune as a
function of frequency. The ALMA-CASA model for Neptune is shown by the white
curve with the red crosses indicating the individual frequency-temperature
pairs. My flux calibration software interpolates a set of points inside the
bandwidth (vertical yellow lines) of the observation and integrates to the
average shown by the horizontal green line using the trapezoidal rule.
The next figure shows the
model for Titan for the same observation. Very preliminary results show
agreement deriving the fluxes of quasars using observations of Neptune and
Titan and these two models.
Figure 6‑5 Model of the brightness
temperature of Titan as a function of frequency. The interpolation within the
flux calibration program (shown on the plot above Figure
6‑4 for Neptune) is not shown
here although the flux calibration program corrects for this contamination in
exactly the same way.
The brightness temperature is
converted to flux following the blackbody formula.
flux=
(1.088791135e7)*freq^3*radius^2/(exp(0.047992*freq/temp)-1)
Neptune
versus Titan
Use Uranus for flux
calibration if possible or one of the Galilean moons.
My original implementation of
the flux calibration software (2001) did not have an accurate brightness model
for Neptune. For many years Neptune was not considered a good calibrator for
the SMA. This is no longer the case with the ALMA-CASA models, but the
prejudice still exists. While the broad CO absorption line in Neptune is more
significant relative to the continuum than is the emission line in Titan,
Neptune is considerably stronger, 20 Jy/beam versus 1.5 Jy/beam at 230 GHz.
Assuming that that the ALMA-CASA model of Neptune is adequate, it might be
preferred over Titan because of its stronger flux and therefore better
signal-to-noise.
6.2 Visibilities
The planets are at least
partially resolved by the SMA even in its more compact configurations. The range
of apparent sizes of the three planets useful as flux calibrators is:
Mars 3.5" – 25.0"
Uranus 3.3" – 4.0"
Neptune 2.2" – 2.4"
The comparison of the modeled
and observed fluxes takes this into account. The planets are assumed to be
circular disks with apparent sizes given by the SMA's internal ephemeris, and
their Fourier transforms are computed as 2-dimensional sinc functions. The flux
measured on each of the SMA's baselines is then the product of the flux given
by the blackbody model for the planet and the response of the interferometer on
the spatial (angular) scales of the baselines, often called the visibility
function, which is in this case the sinc function.
For example, suppose we are
observing Mars when the apparent size is 24" in diameter. The 1D Fourier
transform of a rectangle extending ± 12"
is
FT(Π(12")) = Sinc[ π (1/3600 π/180 12") (B/λ))
where
B/λ is the baseline length measured by the wavelength of the observation.
Of course, the system does not observe negative flux, so use the absolute value
of the sinc function.
Figure 6‑6 The visibility function
for Mars with an apparent diameter of 24". The x-axis is baseline length in
units of 1000 wavelengths, kilo-lambda. If the observing frequency is 1 mm, the
units of the baseline length are meters. The dashed line shows the Sinc function,
and the solid line shows the absolute value which is what the interferometer
measures.
6.3 Running flux_cal for flux calibration
The first step of the flux
calibration was the chopper-wheel calibration which is already done. The second
step is the comparison of the observed and modeled fluxes of a planet or
satellite. The first step in this second phase of the flux calibration is to
correct the phases of the observation of the planet.
Use the progam gain_cal as
before except that this time we set the keyword non_point. Use the select
command to select only the flux calibrator 'mars' and run gain_cal like
this. Normally the system will recognize Mars as a calibrator and the gain code
will be set correctly. If not, set it by hand by entering "mars yes" at the prompt. Be sure to spell mars the same
way it appears in the menu. Source names are case sensitive. The measured flux
of the calibrator is not relevant as long as it is not zero. You can use flux_measure to set the flux if necessary.
Be sure not to include
amplitude in the gain_cal because this would force the amplitudes of the
flux calibrator to agree with the flux value that shows in the rotating menu.
This flux value shown in the menu was derived earlier in the chopper-wheel
calibration procedure. So in this case, the ratio of the modeled and observed
amplitudes would be one and the flux calibration would result in no correction.
IDL> select,/p,/re,source='mars'
IDL> gain_cal, cal_type='pha', x_var='int',
tel_bsl='telescope', $
refant=5, frames_per_page='8',
/non_point
-------------------------------------------
Single Band
Calibration Mode
Main Band Calibration
-------------------------------------------
Set which
sources will be used as calibrators and set the calibrator fluxes
The flux
should be in Jy at the frequency of the observation
These are
the sources and their current gain codes
name: mars
gain code: YES flux
(Jy):
72.8769073
Enter
source, cal code, and if cal, flux in Jy, eg: 3C273 YES 3.1
or hit
Return if all the sources are correctly specified:
Figure 6‑7 Phase vs integration in
the observation of Mars. The different panels show averages for different
antennas and sidebands. The flux_cal program will correct the phases to a model
of the visibilities of the planetary disk.
Finish by examining the
plots, and if reasonable, applying the calibration to correct the phases to the
disk model. This will only affect Mars since this was the only source selected
by the select command before running gain_cal.
We will not calibrate the
flux calibrator for amplitude because this would force all the observed
amplitudes to whatever value we set in the menu and the calibration would be
meaningless.
Instead, go ahead and run the
flux_cal program. First we have to include the other sources that should
have the amplitude corrections applied to them. We are going to include the 2
targets and the 2 gain calibrators. We won't use the flux-calibrated gain
calibrators, but it is good to know their fluxes for reference.
IDL> select,/p,/re,source=['mars','1058+015','PG1119+020','1310+323','PG1404+226']
IDL> flux_cal,minvis=0.4
Running
Eric Keto's version of flux_cal with the ALMA-CASA planet models
Current
Table
# Source Flux Cal.
Flux(Jy)
mars
PRIMARY
73.
1058+015
NO
1.7
1310+323
NO
0.88
PG1119+020
NO
0.0021
PG1404+226
NO
0.0027
Enter
source, new flux cal code:
Use p, P,
or 1 as code for primary flux cal
Use s, S,
or 2 as code for secondary flux cal
Use n, N,
NO, or 0 to deselect a calibrator
or hit
Return if all the sources are correctly specified
:
Found
primary cal mars
Running
Eric Keto's version of flux_primary with the ALMA-CASA planet models
Opening
file
/sma/SMAusers/keto/mir/idl/new/planetmodels/Mars_Tb.dat
Using ALMA
model for Mars
number of
data selected for flux cal
96
# USE SOURCE Vis*FLUX SFact SB INT BSL UVDIST VIS TSSB COH EL
1 mars 281.86 1.43 u 24 2-5 20.5 0.62 252 0.29 80.2
1 mars 281.88 1.30 u 23 2-5 20.5 0.62 253 0.29 80.1
1 mars 281.90 1.32 u 22 2-5 20.5 0.62 254 0.29 80.0
1 mars 281.92 1.45 u 21 2-5 20.5 0.62 257 0.29 79.9
1 mars 281.94 1.30 u 20 2-5 20.5 0.62 257 0.29 79.8
1 mars 281.96 1.32 u 19 2-5 20.5 0.62 262 0.29 79.8
1 mars 281.98 1.33 u 18 2-5 20.5 0.62 254 0.29 79.7
1 mars 282.00 1.33 u 17 2-5 20.5 0.62 253 0.29 79.6
1 mars 282.02 1.31 u 16 2-5 20.5 0.62 258 0.29 79.5
---
/ lots of similar lines /---
1 mars 290.37 2.23 l 16 4-5 18.4 0.68 265 0.36 79.5
1 mars 290.38 1.42 l 17 4-5 18.4 0.68 260 0.35 79.6
1 mars 290.38 1.39 l 18 4-5 18.4 0.68 262 0.36 79.7
1 mars 290.39 1.38 l 19 4-5 18.4 0.68 270 0.35 79.8
1 mars 290.40 1.37 l 20 4-5 18.4 0.68 266 0.36 79.8
1 mars 290.41 1.40 l 21 4-5 18.4 0.69 265 0.35 79.9
1 mars 290.42 1.37 l 22 4-5 18.4 0.69 263 0.35 80.0
1 mars 290.43 1.35 l 23 4-5 18.4 0.69 261 0.35 80.1
1 mars 290.44 1.53 l 24 4-5 18.4 0.69 260 0.36 80.2
# USE SOURCE Vis*FLUX SFact SB INT BSL UVDIST VIS TSYS COH EL
Maximum
measured coherence for mars is 0.969351
Maximum
measured visibility for mars is 0.685077
Minimum
allowed coherence
0.00000
Minimum
allowed visibility
0.400000
Flux scaling factor for each sideband:
l 1.33E+00 u 1.48E+00
Stacking
gain curves to the ca structure
Ready to
apply scaling factors. Default is NO.
Apply Flux
Calibration? [NO <YES>]:
Here Mars is a primary flux
calibrator. Primary means that the program will model the flux of Mars by
calculating its brightness temperature and its visibility function. If we had
no primary flux calibrator, we could use a secondary flux calibrator. In this
case, the program would assume that the flux is given by value set in the menu
and that the flux is a point source. The latter means that the visibility
function is uniquely 1. The SMA has no database of secondary flux calibrators
so this option is generally not useful.
The flux_cal program
produces a table, a plot, and calculates the flux correction to be applied to
each sideband. The table is written to the current working directory as flux_cal_visibility_data.txt. Part of the table is written to the screen.
Explanation of the columns:
1 |
USE |
This
indicates whether this integration is uses in the calculation of the flux
correction. Only the integrations-baselines selected are printed to the
screen so all the numbers are 1. The file flux_cal_visibility_data.txt has all the integrations and baselines.
Those not selected have a 0 in this column. |
2 |
SOURCE |
In
principle, the program allows for multiple flux calibrators. A lot of SMA
data sets have only one flux calibrator, so this option has not been well
tested. |
3 |
Vis*FLUX |
This
is the visibility function times the total flux and the ordinate axis on the
plot. |
4 |
SFact |
This
is the correction factor for this integration-baseline which is the ratio of
the modeled to observed flux. |
5 |
SB |
Sideband |
6 |
INT |
The
number or index of the integration. Might be useful for flagging. |
7 |
BSL |
The
antenna pair. |
8 |
VIS |
The
visibility function from 0 to 1. The default minimum for an
integration-baseline pair to be included in the flux calibration is 0.15.
This can be changed with the argument minvis=.
In this example, the minimum visibility is set to 0.4. |
9 |
TSSB |
The
system temperature. |
10 |
COH |
The
coherence of the data. This function was defined at OVRO as the ratio of the
vector averaged flux over the scalar averaged flux averaged over time. The
SMA data generally does not have enough time samples for a meaningful
statistic so the definition was changed to the ratio averaged over channel or
frequency. This new statistic has been described as a "not entirely useless measure
of data quality". Accordingly, an integration-baseline pair is included in
the average if its coherence is "not entirely zero". In other words, the default value for the minimum
value of the coherence is zero. This can be changed with the option mincoh=. |
11 |
EL |
The
elevation of the antennas at the time of the integration. |
So one can look at the table
on the screen or in the text file flux_cal_visibility_data.txt to see if the results are acceptable.
All this data is summarized
in a plot produced by the flux_cal program shown in the figure below.
The red diamonds are the
model visibilities for the projected baseline lengths of the SMA observation.
These are calculated as above. There are two sets of red diamonds, one set for
each sideband. Because the average wavelength of each sideband is slightly
different, so are the projected baselines measured in cycles (kilo-lambda). The
model amplitudes are also slightly different because of the conversion from the
model brightness temperature of the planet to flux density in Janskys. The blue
and yellow asterisks show the amplitudes of all the visibilities measured by
the SMA. The blue asterisks indicate the visibilities included in the
calculation of the flux correction. These correspond to the 1 in column 1 of
the table. The yellow asterisks are the visibilities not selected. These have a
0 in the first column of the data file written to disk. These are not written
to the screen. The model (red diamonds) is not a fit to the data. It shows the
expected flux. The flux calibration consists of finding the constant (one for
each sideband) required to shift the observed visibilities up or down to best
fit the model.
Figure 6‑8 Modeled (red diamonds) and
observed (blue and yellow asterisks) visibilities for the flux calibrator Mars.
The blue visibilities are those included in the calculation of flux correction.
The minimum value for inclusion is 0.4.
In this example, the minimum
value of the visibilities included in the calculation is 0.4. This is based on
the modeled visibility not the observed. This data is problematic because the
highest visibility data with presumably the best signal-to-noise is
inexplicably low. The suggested corrections for the two sidebands, 1.33 and
1.48 are at odds with the chopper-wheel calibration. The correction factor is
unexpectedly large. A large correction could result from bad pointing on the
flux calibrator which might or might not indicate bad pointing in the rest of
the observations. However, here The data in yellow, those not selected to
calculate the correction, suggest that the correction based on the short
baselines alone (blue points – high visibility) is too large.
Let's try the default value
for the minimum visibility of 0.15
IDL> flux_cal
Running
Eric Keto's version of flux_cal with the ALMA-CASA planet models
Current
Table
# Source Flux Cal.
Flux(Jy)
mars
PRIMARY
73.
Enter
source, new flux cal code:
Use p, P,
or 1 as code for primary flux cal
Use s, S,
or 2 as code for secondary flux cal
Use n, N,
NO, or 0 to deselect a calibrator
or hit
Return if all the sources are correctly specified
:
Found
primary cal mars
Running
Eric Keto's version of flux_primary with the ALMA-CASA planet models
Opening
file
/sma/SMAusers/keto/mir/idl/new/planetmodels/Mars_Tb.dat
Using ALMA
model for Mars
number of
data selected for flux cal
336
# USE SOURCE Vis*FLUX SFact SB INT BSL UVDIST VIS TSSB COH EL
1 mars 83.02 0.69 u 1 1-3 34.7 0.18 359 0.97 78.0
1 mars 83.06 0.70 u 2 1-3 34.7 0.18 358 0.97 78.1
1 mars 83.10 0.69 u 3 1-3 34.7 0.18 358 0.97 78.2
1 mars 83.13 0.69 u 4 1-3 34.7 0.18 350 0.97 78.3
1 mars 83.17 0.65 u 5 1-3 34.7 0.18 352 0.97 78.4
1 mars 83.20 0.65 u 6 1-3 34.6 0.18 350 0.97 78.5
1 mars 83.24 0.70 u 7 1-3 34.6 0.18 348 0.97 78.6
1 mars 83.28 0.76 u 8 1-3 34.6 0.18 347 0.97 78.7
---
/ ---
1 mars 290.38 1.42 l 17 4-5 18.4 0.68 260 0.35 79.6
1 mars 290.38 1.39 l 18 4-5 18.4 0.68 262 0.36 79.7
1 mars 290.39 1.38 l 19 4-5 18.4 0.68 270 0.35 79.8
1 mars 290.40 1.37 l 20 4-5 18.4 0.68 266 0.36 79.8
1 mars 290.41 1.40 l 21 4-5 18.4 0.69 265 0.35 79.9
1 mars 290.42 1.37 l 22 4-5 18.4 0.69 263 0.35 80.0
1 mars 290.43 1.35 l 23 4-5 18.4 0.69 261 0.35 80.1
1 mars 290.44 1.53 l 24 4-5 18.4 0.69 260 0.36 80.2
# USE SOURCE Vis*FLUX SFact SB INT BSL UVDIST VIS TSYS COH EL
Maximum
measured coherence for mars is 0.969351
Maximum
measured visibility for mars is 0.685077
Minimum
allowed coherence
0.00000
Minimum
allowed visibility
0.150000
Flux scaling factor for each sideband:
l 1.01E+00 u 1.08E+00
Stacking
gain curves to the ca structure
Ready to
apply scaling factors. Default is NO.
Apply Flux
Calibration? [NO <YES>]:
And the plot is below.
Figure 6‑9 Modeled and observed
visibilities of the flux calibrator Mars. The minimum acceptable visibility is
0.15.
Now the suggested correction
are 1% and 8% for the lower and upper sidebands. What should you do? Apply the
correction or stick with the chopper-wheel calibration? I'm not sure. I would
probably nudge them up as suggested. Answering yes to the prompt applies the
correction and produces your final average fluxes.
Ready to
apply scaling factors. Default is NO.
Apply Flux
Calibration? [NO <YES>]: y
YES: apply
flux cal
Re-calculating
continuum fluxes
Averaging
all filtered spectral bands to regenerate continuum ...
Done!
Calibrated
fluxes:
Scalar
average for band c1
Source
Flux
StdDev SNR Nscans On source Mean UT
(Jy)
(Jy)
time (min) (hours)
mars
76.0275
2.1492 35.4 24 11.9 3.02
1058+015
1.8011 0.0035 516.9 133 65.8 5.93
PG1119+120
0.1988 0.0008 246.4 480 237.4 5.93
1310+323
0.9157 0.0030 301.1 71 35.1 10.09
PG1404+226
0.1688 0.0010 173.9 240 118.7 10.11
The default minimum for the
value of the visibility function, 0.15, rejects all the points beyond the first
null. There is nothing wrong in principle in including some of them. Of course, you cannot apply flux
calibration corrections a second time after already having done this once. (You
could reload the data and the IDL environment from the last mir_save and start the flux calibration over.) As with all
the calibration programs, you can run the program again to make a plot just for
illustration.
IDL> flux_cal,minvis=0.10
Running
Eric Keto's version of flux_cal with the ALMA-CASA planet models
Current
Table
# Source Flux Cal.
Flux(Jy)
mars
PRIMARY
73.
Enter
source, new flux cal code:
Use p, P,
or 1 as code for primary flux cal
Use s, S,
or 2 as code for secondary flux cal
Use n, N,
NO, or 0 to deselect a calibrator
or hit
Return if all the sources are correctly specified
:
Found
primary cal mars
Running
Eric Keto's version of flux_primary with the ALMA-CASA planet models
Opening
file
/sma/SMAusers/keto/mir/idl/new/planetmodels/Mars_Tb.dat
Using ALMA
model for Mars
number of
data selected for flux cal
624
# USE SOURCE Vis*FLUX SFact SB INT BSL UVDIST VIS TSSB COH EL
1 mars 47.79 0.83 l 1 1-8 62.8 -0.11 352 0.93 78.0
1 mars 47.80 0.83 l 2 1-8 62.8 -0.11 351 0.93 78.1
1 mars 47.81 0.96 l 3 1-8 62.8 -0.11 350 0.93 78.2
1 mars 47.81 0.92 l 4 1-8 62.8 -0.11 341 0.93 78.3
1 mars 47.82 0.80 l 5 1-8 62.8 -0.11 342 0.93 78.4
1 mars 47.83 0.79 l 6 1-8 62.8 -0.11 342 0.93 78.5
1 mars 47.83 0.85 l 7 1-8 62.8 -0.11 340 0.93 78.6
1 mars 47.84 0.91 l 8 1-8 62.8 -0.11 341 0.93 78.7
1 mars 47.85 0.83 l 9 1-8 62.8 -0.11 344 0.93 78.8
---
/ ----
1 mars 290.37 2.23 l 16 4-5 18.4 0.68 265 0.36 79.5
1 mars 290.38 1.42 l 17 4-5 18.4 0.68 260 0.35 79.6
1 mars 290.38 1.39 l 18 4-5 18.4 0.68 262 0.36 79.7
1 mars 290.39 1.38 l 19 4-5 18.4 0.68 270 0.35 79.8
1 mars 290.40 1.37 l 20 4-5 18.4 0.68 266 0.36 79.8
1 mars 290.41 1.40 l 21 4-5 18.4 0.69 265 0.35 79.9
1 mars 290.42 1.37 l 22 4-5 18.4 0.69 263 0.35 80.0
1 mars 290.43 1.35 l 23 4-5 18.4 0.69 261 0.35 80.1
1 mars 290.44 1.53 l 24 4-5 18.4 0.69 260 0.36 80.2
# USE SOURCE Vis*FLUX SFact SB INT BSL UVDIST VIS TSSB COH EL
Maximum
measured coherence for mars is 0.969351
Maximum
measured visibility for mars is 0.685077
Minimum
allowed coherence
0.00000
Minimum
allowed visibility
0.100000
Flux scaling factor for each sideband:
l 1.09E+00 u 1.15E+00
Stacking
gain curves to the ca structure
Ready to
apply scaling factors. Default is NO.
Apply Flux
Calibration? [NO <YES>]:
The
suggested corrections are now 9% and 15%. I would not use these because some of
the visibilities included in the ratio are not much above the noise. The noise
level can be estimated from the plot. Each vertical set of visibilities
represents all the integrations on one projected baseline length. So the
vertical spread shows the noise. Taking ratios of two numbers whose amplitude
is within the noise will not produce a reliable result.
Figure 6‑10 Observed and modeled
visibilities. Points selected to calculate the flux correction (blue) are those
with modeled visibilities above 0.10. Some of these look to be at the noise
level or below.
Finished the flux cal.
The accuracy of the flux calibration obviously depends on the accuracy of the observations of the flux calibrator. If the chopper-wheel calibration is as good as 20%, then the correction based on the flux calibrator needs a higher accuracy than this to make any improvement.
The flux calibration in this data is particularly problematic, one reason it was chosen as an example. Here are some plots of observations and models of Mars and Neptune from a different data set observed on 2016-06-02.
The example
here of Mars shows that there can be uncertainties or problems beyond the
signal-to-noise of the observations of the flux calibrator. Mars is a very
bright source. In this data set, some of the antennas may have different
effective efficiencies leading to individual baselines with higher and lower
fluxes than expected. This might be due to pointing since the observation of
Mars was made before the pointing was corrected in the middle of the night. It
may not be possible to identify the cause just from the data set with only one
observation of one flux calibrator.
The flux calibration in this data is particularly problematic, one reason it was chosen as an example. Here are some plots of observations and models of Mars and Neptune from a different data set observed on 2016-06-02.
Figure 7‑1 Observation and model of Mars from a different data set, 2014-06-02 where the flux calibration is working better. The X and Y axes are projected baseline length and visibility. The next figure shows Neptune from this same data set.
Figure 7‑2 Observation and model of Neptune from the data of 2014-06-02.
With two flux calibrators in the same data set, we can compare to assess the accuracy. The observations of Mars and Neptune were made at the beginning and end of the night of 2014-06-02. The scaling factors for the upper and lower sidebands suggested by Mars are 0.73 and 0.79 and by Neptune 0.84 and 0.88. You can enter more than one flux calibrator in the source selection, for example,
Figure 7‑3 Mars and Neptune, 2014-06-02, on the same plot when both flux calibrators are selected prior to running flux_cal. Neptune is just above the X-axis on this plot because the scale is set by much brighter Mars. Of course, you can run the flux_cal program again after applying the corrections to see the results. The model plotted before the flux calibration is not a fit to the data. It is what the data should look like. After the flux calibration, the data are scaled to best match the model.
Figure 7‑4 Mars, 2014-06-02, after the flux calibration. The data are now scaled to match the model and the same correction is applied to all the other sources selected. Figure 7‑5 Neptune, 2014-06-02, after the flux calibration.
select,/p,/re,source=['Neptune', '1733-130', 'G0.489+0.010', '1744-312', 'Mars']
and the flux_cal program will average the corrections and apply them to all the sources in the selection. In this case the suggested corrections are 0.82 and 0.86. The plot produced by flux_cal auto-scales for the brightest visibility as shown below. Mars is 86 Jy, and Neptune at 11 Jy is barely visible above the X-axis.
PG1119+120
was tracked. Doppler correct
PG1404+226 according to:
select, /p, /re
uti_doppler_fix, reference='PG1119+120',
source='PG1404+226'
Final save of MIR
calibration
mir_save,'fully_calibrated'
select,/p,/re,source='1058+015'
plot_var, x='prbl', y='ampave', frame_var='sb',
color='blcd'
Looks
like a point source, equal ampl. with baseline. Not surprising since it was self-calibrated in gain_cal . This test would be more meaningful if the observations had been
set up with a second (check) quasar.
Plot the
spectrum of the target.
select,/p,/re,source='PG1119+120'
plot_spectra, x_var='fsky', frame_var='sb',
color='band', ntrim=8
This
would be more meaningful if there were strong lines in the source. If there
were lines, the sidebands could be selected individually and plotted for more
detail.
.compile idl2miriad
.compile idl2miriad
select,/re,/p
idl2miriad, dir='100603_025520_PG1119+120_usb',
source='PG1119+120', sideband='u'
idl2miriad, dir='100603_025520_PG1119+120_lsb',
source='PG1119+120', sideband='l'
idl2miriad, dir='100603_025520_PG1404+226_usb',
source='PG1404+226', sideband='u'
idl2miriad, dir='100603_025520_PG1404+226_lsb',
source='PG1404+226', sideband='l'
Getting Help
Eric Keto's MIR Distribution
E-mail us at eketo@cfa.harvard.edu and put the following in the subject line:
- "data" if related to computer access, MIR, etc.
- "website" if related to website viewing.
Official MIR Distribution
To view the MIR cookbook, visit https://www.cfa.harvard.edu/~cqi/mircook.html
Updates
Section | Sub-section | Title | Author | Date |
---|---|---|---|---|
Calibrating SMA Data | N/A | NEW RELEASE | Eric Keto | 11/15/2015 |
Calibrating SMA Data | Gain Calibration | Phase Wrapping (pdf file) | Mark Thomas Graham | 09/26/2014 |