Pointing data with RxA from UT 20021116 are shown below :

They show a very strong trend between the azimuth residual and elevation (central panel in upper row). Similar trends affecting both RxA and RxB data were reported recently :

Date: Wed, 6 Nov 2002
From: Iain Coulson 
To: Per Friberg, Dean Shutt, Tomas Chylek 
Subject: recent pointing oddity

RxB has seen good use while SCUBA is unavailable.
60 data collected over 3 days 01-03 Nov show no systematics -
i.e. no apparent need to change the pointing model. 
RMS residuals in (daz,del) are not worse than 2". 
Then, on the 4th from 18 RxB data, and again with 25 RxA data on the 5th, 
a strong systematic of the azimuth residual with elevation is seen, 
implying, in each case, and if real, a change in NPAE (the 
non-perpendicularity of the major axes) . . .

This systematic is removed (and an azimuth rms of 1.3" achieved) by adjustments to 2 of the TPOINT parameters : d_NPAE=-7", d_IA=-23". The impact of changing IA only is greater than that of changing NPAE only, so maybe the root cause lies with IA (encoder zero-point !?) but these terms (and CA) are highly correlated and therefore hard to separate.

No such systematic has been seen in recent SCUBA data, and an I am not yet able to offer an explanation for this difference.

However, with SCUBA remaining absent for a week more it seems better to accommodate this empirical difference for now. I have therefore installed a pointing model tailored to the needs of RxA/B(W?). There may be small collimation errors unaccounted for, but this large trend should disappear.

The old and new TPOINT models are shown below :

               old           new           change
  IA       -100.3304     -123.0000         -22.7
  IE        +10.0116      +10.0116 
  NPAE      +59.7063      +52.2000          -7.5
  CA       -161.4423     -170.7000          -9.3
  AN        +10.5059      +10.5059 
  AW         -2.9483       -2.9483 
  TF        -14.7109      -14.7109 

20021119
Data the following night showed that the systematic was removed by this change :

However, residual collimation errors were quite large -- of order 30" in azimuth -- and the data had worryingly large raw residuals of (5.2",2.5") in (daz,del). But, the raw scatters are exaggerated by the first two data -- on Uranus, and the early evening data seem generally suspect. For the remaining N=27 data residuals in (daz,del) are (3.1",1.7") - still poor (in azimuth) cf the expected performance, but without obvious curable systematics : running the data through FIT7 improves the residuals only to (2.6",1.7"). The data are riddled with other disturbing pointing changes as reported by the observing crew :

Jim Hoge wrote :

     We saw some significant azimuth changes during the night.  
     On one occasion the change was large enough to throw a grid map off 
     by one offset position (cell was 10 x 10, change was 9.5+). 

The data in question may be either of these pairs of consecutive data :

    az     el      daz    del  hh mm ss.s source           FE
  ------  -----   ----   ----  ---------- -----------      ---
  419.07  38.17   31.4   -0.2  21 25 50.5 CRL618           RXB
  403.75  55.70   23.6   -1.6  21 43 38.2 3C84             RXB

  222.39  64.44   35.8    1.6  03 23 13.7 N2071IR          RXB
  247.62  61.72   26.9    3.3  04 17 14.0 GL865            RXB 

These are v.difficult to explain in terms of model errors, which tend to show up as errors over larger scales in (az,el) space. Tau looked stable throughout the night ! : were other aspects of the weather also stable ?!

[I was on the slopes of Mauna Kea last night watching (for) the Leonid meteor shower and I saw a lot of cloud action in the period 01:00-03:00, although admittedly this may all be well below the summit.]

I am loathe to adjust the model further at this point since the principal rationale for the new model has apparently been justified, despite these other problems. I adjusted the RxB collimations at 12:30 HST 19 Nov 2002; offsets of (0,0) are expected henceforth with RxB.

UT20021120
The next night's RxB data are shown below :

The result is most satisfactory. The largest remaining trends of the residuals are with time (temperature ?) probably as a result of the early start. Twelve (12) of the (33) data included here are of pks0537 as it transitted at elevation of 26 degrees :

The data show no evidence of a step in elevation at transit, implying that our current transit step size of 2" is a good estimate of any such problem. Rms's of the pks0537 data are (1.5",1.1") in (daz,del).

A limited number (12) of pointings with RxW_C were also made :

Apart from the revelation that I forgot to update the RxW collimations yesterday (I have corrected that today), they show good azimuth performance (rms = our canonical 1.5"), but they also exhibit a very strong (-15") decrease with time (temperature?) through the night. However, the temperature of the antenna varied hardly at all through second shift, so the cause does not lie there.

Seven (7) logged data with RxW_D (on Mars, in a 2 hour period, interspersed with the RxW_C data) showed excellent stability -- rms's in (daz,del) of (0.8", 1.5"). Differential pointing offsets were measured twice (although note that it takes >30mins to swap FEs) :

  110.97  27.79   30.8   -6.2  06 04 10.3 MARS             RXW_C
  116.73  36.33   30.3   -9.2  06 43 50.8 MARS             RXW_D
                 ------ ------
         C - D     0.5    3.0

  131.35  49.95   30.7  -13.6  07 53 24.4 MARS             RXW_D
  142.83  55.84   29.3  -11.2  08 30 03.2 MARS             RXW_C
                 ------ ------
         C - D    -1.4    2.4

giving a mean of  -0.5    2.7
                 ------ ------

Conclusion
The model based on the RxA data of 16 November works well for RxB (and hopefully for RxA), and has no significant residual systematics. Collimation adjustments for RxB and RxW should maintain recently determined relative offsets between RxB and RxW.

This model may be well suited for all the cabin instruments, but it is structurally different from the 'SCUBA' model we were using previously. There cannot be real differences in the parameterizations of the imperfections of the telescope -- the value of NPAE, for instance, must be independent of the receiver being used -- so the model differences, if they persist, must be due to errors in the TMU position for SCUBA, or in the SCUBA alignment . . .
I would prefer, when SCUBA returns to service, to use this new (cabin) model and analyse SCUBA data taken with it for such additional imperfections. Certainly, there is no easy way to toggle between two models, nor should there be a need to do so.