Intro:

    A cross dipole feed was designed by jim breakall at 327 MHz and was placed on the uphill side of carriage house 1. It was to be used to look at geostationary satellites.

Jim's feed design:

• Notes:
•  in some of the reports jim  uses the "height of the antenna". He uses the height above the reflector rather than the distance from the center of curvature.
• The radius of curvature is  870 feet. The paraxial surface is at 435 Feed.
  
• When he says "max gain at height of 421.6 feet", this is a distance (870-421.6)-448.4 feet from the focal point. It is : (448.4-435) = 13.4 feet below the paraxial surface (closer to the dish).
  

• 20190211: Optimized crossed dipole disk with .25 in Cyl elements (.pdf)
  
• This is the feed response it does not include the dish.
• 20190211: Crossed dipole disk feed .25 in Cyl elements Optimized (.pdf)
• shows optimization runs. last page shows far field pattern of the feed.
• 20190212: Results for the 327 MHz Crosssed-Dipole Disk Optimized feed (.pdf)
• This report includes the dish in the modeling. The plots are:
  
• gain vs height (he measures from bottom of dish)
• gain vs frequency. (With theta=za = 22deg, and optimum height).
• vswr vs freq
• gain  za=22,phi=90  vs theta   This looks like a cut along za direction
  
• gain  za=22,phi=90   vs phi . this looks like an azimuth cut along the azimuth direction.
  
• To get the great circle beamwidth you need to multiply the null distance by sin(22deg)=.374.
  
• 20190213: Gain vs Tilt angle of feed. 30,10,5 deg down then 5,10,30 deg up. (.pdf)
  
• This was to check how sensitive the gain was if we didn't have the feed pointing exactly along the radius.
  

The receiver.

• The fiber receiver had very strong birdies at 15-20 MHz.
  
• We had to put a 300 MHz hipass filter after the fiber receiver to get rid of it.
• Original filter rolloffs
• The original bandpass filter upstairs did not roll off as expected.
• The 300MHz hipass filter did not start rolling off until around 250 MHz. It was only down 16 db and the 205 MHz birdie.
• replacing the original bandpass filter with the 327 30MHz unused filters from the dome solved the problem.
• There still look to be narrow band combs across the band. These are similar to what we see in the dome 327, only stronger..
  

Mods made to the receiver chain:

•  190318: for this dataset, the receiver chain was:
• ant-> amp -> hipass300 -> cable -> bpfilter-> amp->5db pad -> amp -> fiber xmiter
• When looking at the bandpass downstairs on the spectrum analzyer, the bandpass was only about 12 db above the noise floor.
  
• 190322: felix added another 20 db amp upstairs:
• ant-> amp -> hipass300 -> cable -> bpfilter-> amp->3db pad -> amp ->3dbpad->amp-> fiber xmiter
• With the extra amp, the band pass was closer to 18db above the noise floor. This should improve Tsys.
  

Setting up the pointing model

• 20180318
  
• Morning:
  
• The source B2329+296 was tracked ( 5Jy at 327) .. strongest source available during the maintenance time i got
• the signal was square log detected, and then put on on the oscilloscope.
  
• The telescope was commanded to slowly move across the source in az and then za.
• marking the location for the peak of the deflection and comparing this to the transit time gave  the pointing error.
• We ended up with an offset of -2.074 degree offset from the center of the carriage house.
• Afternoon:
• Did crosses on B0240-002 (13.7 Jy at 325MHz)
• Prior to the crosses, the oscilloscope was setup (as above) and some scans were done across the source.
• The oscilloscope offset gave a model correction of -6830 asecs (-6876 asecs relative to ch). This was an offset of 1.91 deg from the center of the ch.
• We used these offsets for the strips
• After processing the strips, the za offset ended up giving a model offset of -6919 asecs (1.93 deg offset from the ch.). this became the new model correction.
• The northern and southern source za offsets differed by about .14 deg.
  
• If people are interested in northern sources, we should probably do some crosses on northern sources.

Doing cross strips on continuum sources:

Links:

Setup
processing the data
    removing rfi
    fitting a gaussian to the strips
The fit results
Summary
        

Plots:
18mar19 za=19 deg
the average,median, and peak hold spectra (.ps) (.pdf)
the strip data and fits for each pattern (.ps) (.pdf)
the fit results (.ps) (.pdf)
22mar19 za=14 deg
the average,median, and peak hold spectra (.ps) (.pdf)
the strip data and fits for each pattern (.ps) (.pdf)
the fit results (.ps) (.pdf)

    Cross strips were done on continuum sources to get the beam width, sidelobe height, pointing errors, and SEFD.
Data was taken on:

• 190318:
  
• 7 strips on B0240-002 (13.7 Jy at 325 MHz).
  
• The strip length was 1.45 degrees
• the za range was 18.4 to 19.75 degrees
  
• 190322:
• 10 strips were done on B0340+048 (9.9Jy at 325Mhz)
• The strip length was 2.175  degrees
• the za range was 13.35 to 14.65  degrees
• the pointing model had the offsets computed from the 190318 pointing errors included.
• Felix had installed another 20db amp upstairs to improve Tsys,
  

Setup:

• cima has no setup for the ch 327 receiver, so we have to fake it out..
• Generate a 327 dome config  in cima then save it to disc.
• From cima setup the  327 receiver and 1 mock spectrometer.
  
• Use the 750  IF,  325 MHz rf frequency, and 43 MHz bandwidth for 1 mock spectrometer.
• Lets call the config file: ch327feed.conf
• A cima  command file is then used to config the dome 327 receiver and then send commands to switch to the ch327. The file is:
  
• #
• #  run this once after starting cima. it will setup the 327 receiver and the mocks..
• #  and adjust the power levels.
• #
• LOAD ch327feed.conf
• EXEC vw_send "if2 if2_sigsrc ch"
• EXEC vw_send "if2 if2_inpfrq 300"
• #   setup pointing model comment out for testing..
• EXEC vw_send "trkinit chlb"
• EXEC vw_send "pnt mode stow 2" 
  
• ADJUSTPOWER
• Notes:
  
• the if2 commands switch from the 750 IF to the 300 MHz IF downstairs (since we're supplying 325 MHz rf).
• trkinit chlb     .. switches to the pointing model for the 327 feed on the ch. I'm using the same file as the old seti lband feed on the ch (since it's already in the software.).
• pnt mode stow 2 ... sends the dome to its stow position (the ch model was made with the dome at stow).
  
• The setup parameters were:
• iflo: cfr 325 MHz, nocal
• mocks:43 MHz bw, 8192 chan, 1 second dumps
• The cross pattern were:
• 60 seconds per strip: 60 1 second spectra
• 6,9  beams long: 1.45,2.175 deg (18,22mar19). The 18mar19 lengths were a bit short to get a good baseline on the gauss fits.
  
• All angles were great circle.
  

Processing the data:

    Removing rfi:

• 20190318: plots showing the average,median, and peak hold spectra (.ps) (.pdf)
• 20190322: plots showing the average,median, and peak hold spectra (.ps) (.pdf)

• All of the patterns were combined into a single spectra.
• Black is polA, red is polB.
  
• Top: this is the average of 60*2*npatterns  spectra
• vertical scale is .1 to 10 Tsys (log plot)
  
• middle: the median spectra
• bottom: the peak hold spectra (max value in each channel for all spectra)
• vertical scale is .1 to 1000 Tsys (log plot)
  
• The rfi  308-213 MHz are the 3rd harmonic of the fm stations: 102.9,104.1
• The narrow comb  like spikes are probably coming from us (az,za encoders?)

To remove the rfi:

• for each cross pattern compute the (rms by channel)/(average channel value) using the 120 spectra (the rfi was a lot stronger than the source deflection).
• a 4th order harmonic and 1st order polynomial was fit to the rms
• values outside of 3 sigma were flagged as bad (as well as the edges of the band).
  
• The total power over the bandpass was computed for the good channels from the fit. This was used for the fitting.
  
• about 70% of the band was used for the total power

Fitting a gaussian to the strip total power.

A gaussian was fit to each strip. The sidelobe regions of the strips were excluded:

• The fit was:
• y=c0 + c1*x + c2*exp(-((x - c3)^2/c4)
• c0 : Tsys
• c1: slope tsys (mainly when moving in za)
• c2: source deflection
• c3: pointing offset
• c4: beam width (sigma ->fwhm).
• 20190318: strip data and fits for each pattern (.ps) (.pdf)
• There are 7 pages. 1 for each pattern. The strip length is 1.45 degrees.
• 20190322: strip data and fits for each pattern (.ps) (.pdf)
• Top: az strip. black - raw data, green fit. linear scale
  
• middle: za strip: red - raw data, green fit. linear scale
• the c1 term (linear in za motion) is non zero as we move up in za (spillover increases).
  
• bottom: raw data az strips:black, za strips:red. db scale
• the data is scaled to the peak value
• For 18mar19 the  za bm width is greater than the az bmwidth (za=19degrees)
• For 22mar19 the za,az bm widths are the same (za=14 degrees).There is much less za beam spillover..
• Both az sidelobes are about the same size. za=19 about 8db, za=14 about 10db down.
  
• the za sidelobes before transit are not present.
• 18mar19 (za=19) the za sidelobe after transit is larger than the az sidelobe. for
  
• 22mar19 (za=14) the za,az sidelobe after transit are about the same size.

The fit results:

    The  fit results include pointing error, beamwidth, and SEFD.

• 20190318: Fit results (za=19deg) (.ps) (.pdf)
• 20190322: Fit results (za=14deg) (.ps) (.pdf)

•  Page 1 pointing error, beam width
• top : pointing offsets
  
• black: az offset, red za offset.
• A negative value --> the source came too soon in the strip.
• You need to add the negative value to the model correction.
• 
  

• model offsets  before and after fits:

  		az offsets (asecs)	za offsets (asecs)
  18mar19 avg za=19	model before fits	67.53	-6830
  	fit offsets	-46	-89
  22mar19 avg za=14	model before fits	21.4	-6919
  	fit offsets	-10.5	-59.6
  ch Model		67.53	-46.37
  327ch za offset from ch (ch - 327ch)	18mar19: za=19	56.1	6872
  	22mar19; za=14	46.1	-6978

•  
• The 327 ch feed points
• at za=19 deg: 1.91 degrees uphill
• at za=14 deg: 1.94 degrees uphill
• I left the ch327 model with the offsets computed from 18mar19 (since dale will be looking at high za).
  
• Bottom: beam width fits
• black: az strip, red: za strip, green: average gr327 beamwidth.

• za deg	ch327 bmwidth Amin		gr327 Amin	ch327/gr327 avg bmwidths
  	az	za	avg za,az
  19	17.5	20.5	14.2	1.33
  14	17.8	18.1	14	1.28

  
  
• The average beamwidth for the ch327 feed is about 1.3 times wider than the gr327 receiver.
  
• the gain goes as the square of the (beamwidths)
  
• (1.30)^2 = 1.77= 2.65db
• This assumes that the beam efficiencies are the same.. (which they probably aren't since the ch327 sidelobes are higher).
  
• 
  
• Page 2 SrcDeflection, SEFD
• Top: Src deflection/Tsys.
  
• The fit to amplitude and Tsys were used to compute srcDeflection/Tsys for each pattern
• Bottom: SEFD (Tsys/Gain)

• za	sefd  (Jy)		sefdRatio: ch327/gr327
  	ch327	gr327
  19	54.6	16	3.41
  14	40.3	12	3.35

  
  
• the sefd for the ch327 is about 3.4 times worse than the gr327. You'll have to integrate about 11.5 times longer for the same signal to noise.
  
• 
  

Summary:

• The measured beam widths  are 15  to 20 % smaller than the simulations.
  

Comparing the gains using the hpbw and the beam efficiency

• Directivity, maxGain= (4 * pi *EffectiveArea)/(lambda^2)
• hpbw=1.02 * lambda/Diameter
• the hpbw gives an area =  pi * [(1.02*lambda)/(2*hpbw)]^2
• The hpbw is only measuring power in the main beam.
  
• The directive gain uses the entire beam solid angle (integrated over 4pi).
  
• You need the  beam Efficiency to convert from main beam to beam solid angle: bmEff= mainBeam/beamsolid angle.
  
• So Dir = ((pi * 1.02 * lambda)/hpbw]^2 * beamEfficiency
• For post maria the greg dome 327 beam  efficiency is about .55
• when computing the max gain for the ch327 feed, i didn't know its beam efficiency, so i used the same value as the dome.
• I know that the beam efficiency of the ch327 feed should be less than the gr327 since the sidelobes are higher.
• So the ch327 maxGains are an overestimate.
  

 home_~phil