Intro
The warm s,xband receiver has
electronic gain variations (mainly related to temperature).
2d crosses on continuum sources are used to measure the system
performance (tsys,gain,sefd,beam width, pointing error).
2d Gauss fits are made to the data. The fit includes:
- Tsys (constant term)
- src amplitude
- beamwidths in az, el directions
- pointing error offsets in az and el
- linear fit to elevation change.
If the gain changes during the 2 or 3 minute cross, then the fit
has a harder time fitting the baseline (using the single Tsys
value and the ramp in elevation).
To improve the fitting (especially for weak sources):
- A linear fit in az and el were done and the subtracted
from the original data (prior to the 2d fit).
- In this case the Tsys term was later added as the
average of the midpoints of the two linear fits.
This page shows the improvements in fitting using the linear
baseline removal. The a4003 xband crosses on 3C286 (5.2 Jy)
were used for the comparison.
Movies were also made of the 2d crosses with the fits over
plotted:
- Top frame: azimith strip
- bottom frame: elevation strip
- white: no linear baseline removal before 2dfit
- green: linear baseline removal before 2d fit.
- You can see the variation in the baseline with the white
data.
- The large spikes in the 8789,8931 data comes from a 4
second chirped radar (more
info).
processing: x101/230303/mkmovie.pro
The data:
- 46 crosses on 3C286 were used for the test:
- 3C286 has a flux around 5.2 Jy at xband
- the sefd for xband (at higher el) is around 3700Jy)
- The srcdeflection was 5.2/3700 =.14% of Tsys.
- data taken 03-17aug22
- 7 172 MHz bands were used with center freq of:
- 8219. 8363. 8505. 8647. 8789. 8931. 9073 MHz.
- The az and el strips were 60Amin long (about 6 beams)
- mascrossinp() was used to input the data and compute the
total power
- 6% on each edge of a band was exluded
- the the rms/mean by channel, channels with larger rms
were also excluded (rfi).
- mascrossfit2d() was used to do the 2d fitting. The fitting
was done twice:
- the first set did not include linear baseline removal of
the az and el strips
- the second set included the linear fits to the az, and
el strips.
The fits:
The
plots compare the two fitting
procedures (.ps) (
.pdf)
- Page 1 fit sigmas
- Each color is a different frequency band
- freqBands 8786 and 8930 occasionally had strong
interference
- at times the interference was strong enough to
saturate the entire band (more
info).
- Top: fit sigmas with no linear baseline removal
- units are deg K
- the dashed horizontal line was used to exclude any
fits above it
- 2ndframe: fit sigmas vs cross number for fits with
linear baseline.
- 3rd,bottom frames: histograms of the fit sigmas
- 3rd: histogram of fits with no baseline removal
- bottom: histogram of fit sigmas with baseline removal.
- Page 2: azimuth, elevation beamwidth fits
- the dotted horizontal lines are clipping levels to
exclude fits with bad beamwidths
- < 5amin, > 15amin.. the beamwidth is about 10
amin.
- the #clipped tell how many fits were excluded in each
freq band.]
- The larger number of excluded fits in the 8788, 8930
bands was caused by the rfi.
- top,2nd frames: az,el beamwidths without baseline
removal
- 3rd,bottom: az, el beamwidths with baseline removal.
- Page 3: source deflection by cross and freq
- The vertical scale is deg K.
- Top: src deflection no baseline removal
- the points * at 0 are freqbands where the fit didn't
converge.
- 2nd: src deflection using 2d fits with baseline removal.
- 3rd,bottom frames: source deflections including
only good fits
- a good fit:
- converges
- has beamwidths between 5 and 15 amin
- have fit sigma's below .05 DegK
- Page 4: Fraction of good fits, average source
deflection, and sigma bye freq band
- this only included "good" fits (converged, ok
beamwidth (5-15 Amin), ok fit sigma (<.05K)
- Top fraction of fits that were good .
- black: no baseline removal. abot 40% are usable
- red: with baseline removal: about 90% are usable.
- 2nd: average source deflection.
- black: no baseline removal
- red with baseline removal
- They have give similar results (although the baseline
removed points have twice as many good fits).
- bottom: average fit sigma... about 12Millikelvin
- Page 5: source flux
- Using the 12meter xband gain
curve the deg K was converted to Jy.
- the average flux for the good fits was computed.
- Top: flux vs frequency
- Black: no baselines removed
- Red: baselines removed
- red dashed line: linear fit to flux vs freq
- blue: flux from chris salters' catatlog fits.
- 2nd: sigmas for avg flux
- black: no baseline moved, red: baselines removed.
- the average sigma across the 1ghz is about .45 Jy
- 3rd: sigmaFlux/flux
- the sigma is about 6% of the flux
- bottom:
- ration of measured to catalog flux
- the linear fit to the flux vs freq was divided by the
catalog flux
- The measured flux was about 2.5% higher.
SUMMARY:
- 46 crosses on 3C286 were done during aug22.
- 7 freq bands of 172MHz each were used
- 2d gauss fits to the crosses were done
- no baseline removed
- linear baselines removed prior to 2d fit.
- Removing baselines improved the number of good fits.
- no baseline removal: 40% of crosses were usable
- with baseline removal:90% of crosses usable.
- the average flux be freq had a sigma of about .45 Jy
- deltaflux/flux was about 6%
- the flux with baseline removal was 2.5 % higher than the
catalog values.