# Nov 2003

The Mar03 gain curve fits vs za (new curve)
The mar03 gain curve fit to az and za (just for comparison)
The error made by using the old gain curve mar03 thru 02nov03

The lband wide receiver was replaced in feb03. The cal values were measured on 6mar03. Data was taken from march through sep03 using the interim correlator. Since the correlator can only measure 4 frequency bands simultaneously, two different frequency groups were used :(1175,1300,1375,1415) and (1415,1550,1610,1666). Gain curves were then computed fitting for gain as a function of za (fits using azimuth terms did not improve the fit enough to make a difference).  The fitting function (with za in degrees) is:

g(za)=c0 +c1*za + c2*(za-14)^2 +c3*(za-14)^3
where the ^2 and ^3 terms are only applied for za ge 14 degrees.
There were a total of 2482 measurements over all frequencies. 12 of these were rejected do to bad measurements.
• Fig 1 shows the az,za distribution for the data. The top plot is the az,za positions for thefrequency group (1175,1300,1375,1415) while the bottom plot has the positions for the (1415,1550,1610,1666) group.
• Fig 2-3 are the individual fits for each frequency. The data is in red, the black lines are the fits. The fit values are listed on each plot. The rms of the fit is listed at: (sigma(y-yfit)) for each plot.
• Fig 4-5 are the fit residuals in K/Jy for each frequency.
 freq # of points fit rms K/Jy 1175 357 .43 1300 354 .47 1375 356 .46 1415 633 .45 1550 255 .40 1610 254 .48 1665 261 .41

1415Mhz has twice the measurements for the fits since it was included in both frequency groups. The gain increases from 1175 to 1415 Mhz (this could be real or just an error in the cal measurements). Looking at the lbw system performance, you can see that the SEFD is also  increasing as you go to higher frequency. Most of the SEFD increase is from Tsys decreasing but some may be from the cal.

Note: The fits using za and az terms are available for comparison. They are not the curves being used.

#### Errors made by using the wrong gain curve

The gain curve for data taken after mar03 was installed on 02nov03. People who have analyzed this data between mar03 and 02nov03 have been using the old gain curve of nov02. Any data analyzed after 02nov03 will use the correct gain curve for the data taken mar03 to 2nov03. To help people in deciding whether it is worth their while to reanalyze their data, the error that people have been making in estimating the lbw gain is plotted versus frequency. The curve is evaluated at az=270,za=10.
• The top figure plots the two gain curves vs frequency for az=270,za=10. Black is mar03 while red is Nov02.
• The bottom figure is the ratio of the two gain curves vs frequency. This is the error that people have been making by using the incorrect gain curve.  Since the nov02 gain curve was larger, users have been overestimating the telescope gain by up to 8%.

The idl routines gainget() or corhgainget() will now return the lbw gain for data after 01mar03 using these fits. The coefficients can be found in the ascii file  data/gain.datR5 (this is provided in the AO idl distribution for correlator routines). You can also find a copy of it at AO in /home/phil/idl/data/gain.datR5.
processing: x101/lb/Mar03/dogainfit.pro
` home_~phil`