How many samples are needed to transfer the model between rcvrs?

feb,02

     The model is made with the sband narrow reciever. After the model is completed, a source not used to make the model is tracked using sband narrow. This tells us how well the model is working and is then used to transfer the model to the other receivers (when there were only two other receivers I actually used 2 sources, 1 north and 1 south of our latitude).
    The model for each receiver differs from the sband narrow model by an offset in azimuth (great circle)  and zenith angle .These offsets are computed by tracking the above source (or sources) rise to set and computing the difference in pointing errors between sband narrow and the new receiver. There are 9 installed receivers 7 of these are at lband or above and need relatively good pointing. Tracking the verification source would then take 7 days (1 for sbn and 6 for the other receivers).
    The plots shows the difference in pointing errors for CTA21 using 4 receivers (.ps) (.pdf) (lbn,lbw,sbn, and cband) after model 12 was completed. This data was taken during the day. It was not one of the more succesful attempts at verifying a model. It was the first time that active tiedowns were used to keep the average platform heigth fixd.The figures have:
  • The black line is the difference in the azimuth errors
  • The red line is the difference in the zenith angle errors
  • The dashed line is the median value of the difference for the entire strip.

    •  

     

    The table below has the rms pointing error for each receiver on CTA21 and the  rms of the difference in pointing error between the receiver and sband narrow:
     
    receiver
    Rms azErr
    Rms zaErr
    RmstotalErr
    		 rms azDif rms zaDif
    sbn
    7.5
    3.7
    8.4
    		 - -
    lbn
    6.5
    2.8
    7.1
    		 4.4 4.0
    lbwide
    10.1
    3.0
    10.5
    		 8.5 4.8
    cband
    7.6
    4.5
    8.8
    		 7.5 4.4

    The model rms was 6.7 asecs so this source was not fit very well by the model. To speed up the verification you could track only a portion of the source on each receiver, and get all the recievers done in one night. Looking at the curves shows the perils in doing this. A single point can be up to 15 asecs from the median value. There are also trends that last for a long time: lbn-lbw azerr difference change at sample 40 when the source transits. This could happen if the platform tilt changed north/south between the two days. I'm starting to wonder whether the large, fast jumps in azimuth and then slow recovery is the wheels driving off the rolling surface and then jumping back when the rollers finally hit the limit. Remember that 1/8" is 5 arcseconds at 435 feet...

        If the offsets between receivers remained constant then we could use that to compute the new offset for each model. Unfortunatley, most models have had the horns repositioned (via a survey) before building the model. The offsets between model 11 and model12 (.ps) (.pdf) are shown in the figure. Sband narrow offsets changed by -6 and 2 asecs. Lbw had the largest variation changing by -8,+5.5. This differs from lbn by about 3 asecs.  For a more complete history of the turret positions and the pointing errors see: turret positions and pointing errors.
     

    processing: x101/model/testverify.pro
     home_~phil