A description of the gain curves.
Sep04
Links:
When are new gain curves made
Where are the gain curves located and
links to the files
The format of the gain curve files
The fit types used
Code to evaluate the fits
The gain curves provide the point source gain of
the telescope in Kelvins/Jy. They are fit to the average of polA and polB
(Stokes I/2) and are a function of azimuth and zenith angle. The
data for the fits is taken with the heiles
calibration scans (spider scans).
The gain curves are used to compute the flux of objects
measured by users. This method relies on the cal deflection to transfer
the flux of the known calibrators to the flux of the unknown source observed
by the user. We say that the gain curves are in Kelvins/Jansky since we
think we know what the cal temperature is in Kelvins. A better way would
would have been to say that the gain curves are in calUnits/Jy. It doesn't
really matter if we have the correct cal Temperature. We only need to guarantee
that the cal temperature does not change from the time the gain curves
were made , to when the user used applied the gain curve to their source.
When are new gain curves
made. (top)
New gain curves are computed when the gain of the telescope changes
(e.g.: moving a horn, reshimming the elevation rails, resetting the primary/secondary/tertiary
surface, etc..) or when the cal values change (e.g.: new diodes or some
connection from the diode to the dewar changes). Spider scans are taken
starting at the change until enough data is acquired to do the fit (this
may take several months). When the new gain curve is installed it contains
two dates:

The starting date for this gain curve. Data taken after this date should
use this curve.

The installation data of the gain curve. Data taken between 1. and 2. should
be reanalyzed using the new curve.
Where are the gain
curves located. (top)
The gain curve information is kept in ascii disc files
(one file per receiver). The filename format is gain.datRn where n is the
receiver number (1327, 2=430, 3610, 5lbw, 7sbw, 8sbh, 9cb, 10=cbh,
11=xb, 12=sbn). So the gain curve file for lband wide (lbw) is gain.datR5.
These files are located in the idl directory /pkg/rsi/local/libao/phil/data
at the observatory. Anyone who downloads the ao idl distribution will have
these files as part of the distribution (under ./data). The links below
point at the gain curve files:
The format of the gain curve
files. (top)
The first column of each line tells how this line should
be interpreted:

; the line is a comment

! yyyy ddd This is the start of a gain curve set of data. yyyy ddd
is the start date for the gain curve. yyyy is a 4 digit year (2004)
and ddd is the day number of the year counting from 1.

all other lines are gain curve data. These lines are white space separated
columns containing:

col 1 the frequency in Mhz for this measurement

col 2 The type of fit used for this gain curve (see below for the codes).

col 3 numcoefs. The coef's for the gain curve fit

colLast3 the sigma for the fit in (K/Jy).

colLast2 the polUsed for the fit (this is always I for stokes I).

col last 2. the cal values used (pola,polB) for the computation of the
gain curve at this frequency.
There is a set of column labels above each data set (as a comment line).
The fit types used: (top)
There are currently 6 types of fits that can be used.
They are identified by the 2nd (type) column in the ascii file.

type 1: y=c0 + c1*za + c2*(za14)^2 + c3*(za14)^3 + azterms1
. See Note1.

type 2: y=c0 + c1*(za10) + c2(za10)^2+c3(za10)^3 + azterms1

type 3: y=c0 +c1*(y)+ c2*(2*y*y1.) + c3*(4*y^33*y) + azterms1
where y=(za10)

type 4: y=c0 +c1*za +c2*(za14)^2 +c3*(za14)^3 (same
as 1 but no az terms). See Note1

type 5: y=c0 + c1*(za10) + c2(za10)^2+c3*(za10)^3 with
azterms2

type 6: y=c0 + c1*(za10) + c2(za10)^2+c3(za10)^3 (same
as 2 but no az terms)
The az terms contain:

azterms1: c4*cos(az) +c5*sin(az) + c6*cos(2az) + c7*sin(2az)
+ c8*cos(3az) +c9*sin(3az)

azterms2: c4*cos(az) +c5*sin(az) + c6*cos(3az) + c7*sin(3az)
+ c8*sin(za)*cos(3az)+c9*sin(za)sin(3az)
Note 1: The (za14) terms are only included when the za is greater
than or equal to 14 degrees za. Below za=14 the fit is linear.
Code to evaluate the fits
(top)
There are idl routine that evaluate the gain curves.
Given az, za, rcvr, frequency, and date they will find the correct file,
input the coef's for the correct date range, evaluate the function interpolating
to the requested frequency. The routines will not extrapolate. If the requested
frequency is outside the measured frequencies, then the value at the last
frequency measured is returned. The routines are:

istat=gainget(az,za,freq,rcvrNum,gainval,date=date).
In the generic idl routine section.

istat=corhgainget(hdr,gainval,date=date,az=az,za=za).
In the correlator idl section. The header is an interim correlator header.
It takes the receiver number, az, za, date and frequency from the hdr.
This will also work for wapp data (although alfa may not work..).
home_phil