The following relationships between the elevation pointing residuals, del, and the leg temperatures Tf(ront), Tb(ack) are assumed possible within the telescope control software:

                del  =  c1 * (Tf - Tb)   + const
                del  =  c2 * 0.5*(Tf+Tb) + const

Since June 2000 the TEL (now THI) code has used a value for c1 of +6.0"/deg, although the current value of c2 is +0.0"/deg rather than the +0.3"/deg when last updated on the VAX side (in TEL.IFL). The validity of these coefficients c1 (TEMP_SLOPE in TEL.ifl language) and c2 (TEMP_MEAN_SLOPE) are checked occasionally by plotting the residuals against these temperatures. The results for RxA pointings in the month of May were as follows:

The suggested changes to the coefficients may be tabulated thus :

  Month     Rx    No.obs      c1             c2
 --------  ----   ------  -----------   -------------
  200405    A      302    -5.4 +- 0.9   -0.33 +- 0.06
            B      219    -8.2    0.7   -0.54    0.07

The ranges of temperature covered, viz -0.5 < (Tf-Tb) < 0.3 and -3 < 0.5*(Tf+Tb) < 10, are quite typical. The trends seen in the May data appear extremely (>5-sig) significant, and indeed, the size and sign of the suggested corrections even suggest that the parameters TEMP_SLOPE and TEMP_MEAN_SLOPE are not being applied at all.

How long might this have been happening ?

Data for each month of 2004 are presented below, with 3sigma significance marked in red :

  Month     Rx    No.obs      c1             c2
 --------  ----   ------  -----------   -------------
  200401    A      221    -2.7 +- 0.5   -0.04 +- 0.07
            B       22     2.9    2.5    0.22    0.88
            S      350    -0.4    0.3   -0.21    0.05
  200402    A      111     1.9    1.0   -0.15    0.17
            B       28    -0.7    0.7    0.22    0.30
            S       82    -0.5    1.0   -0.41    0.10
  200403    A      114     0.2    0.8   -0.11    0.12
            B       31     2.1    1.1    0.23    0.12
            S      281    -1.2    0.4   -0.22    0.05

  200405    A      302    -5.4 +- 0.9   -0.33 +- 0.06
            B      219    -8.2    0.7   -0.54    0.07
  200406    A      209    -9.4 +- 1.0   -0.73 +- 0.08
 (to date)  B      346    -4.8    0.7   -0.87    0.07

• NB : The ENVIRO task was released on 07 April, so data from April are not included.

Significance in both c1 & c2 tends to occur together only latterly, so the earlier, solitary 3- or 4- sigma results might be spurious. That conclusion may apply also to the S-CUBA results in March. On the other hand the significance of the results in the last two months has risen to between 5 & 10 sigma.

Either

• the corrections are not being applied:
  • the coefficients are being read and used as zero, or
  • all the relevant temperatures are being read and used as zero, or
• the thermal behaviour of the telescope has changed - by a factor of x2 in the last couple of months - and we need to change the coefficient(s) to c1 = 12.0 (and/or c2 = 0.3).

Matt has sniffed around the software to check on the first two options and is convinced that the coefficients c1 & c2 and the leg temperatures are being read and implemented as intended. The third possibility seems unphysical and checking the efficacy of new values of c1 (and/or c2) would require further data acquisition.

20040625
Matt identifies ENVIRO:OPTO:BACK (actually in lower case) as the instantaneous temperature of the 'back legs' of the antenna (and OPTO:FRONT for the front legs), and NTM confirms that these are A-frame temperatures. The use of OPTO:BACK and OPTO:FRONT instead of the 15-minute post-event averages (BLEGAVE and FLEGAVE, which I had used previously), changes the fitting to the coefficients only slightly. The new fits for the June 2004 data are compared with the 'old' ones:

  200406    A      209    -9.4 +- 1.0   -0.73 +- 0.08
 (to date)  B      346    -4.8    0.7   -0.87    0.07
NOW
  200406    A      209    -8.7 +- 1.0   -0.79 +- 0.08
 (to date)  B      346    -4.7    0.7   -0.89    0.07

So the significance of the trend is barely disturbed by this important, but minor, correction.

20040702
I used a hairdryer to raise the temperature of each A-frame sensor in order to confirm that the ENVIRO task was calculating front and back mean temperatures correctly. The results showed that the IDs had been rather scambled :

     Number ?            0    1    2    3     4    5    6    7
       or                1    2    3    4     5    6    7    8
  Current description   llr  lrr  ulr  urr   ulf  urf  llf  lrf
  Actual location       LRR  ULR  URR  ULF   URF  LLF  LRF  LLR

See also this plot. There is a slippage of one place in notation: the actual locations may correspond to codes 1-8 while a 0-7 notation is required ?!

It should be possible to calculate if this confusion of probe identities has indeed caused the observed trend. This plot shows the error induced in the calculation of (Tf-Tb) by the identity error. The incorrect temperature difference is plotted as a function of (the actual ?) (Tf-Tb). When the front legs are at their warmest w.r.t. the back legs (upper-right) the erroneously calculated (Tf-Tb) is not so warm : almost no correction is made The impact would be to underestimate the elevation pointing correction, almost to the point of not applying it all. The apparent need for almost the full 6"/deg of correction is thus explained.

20040702 - continued
The identities were corrected at about 17:00 HST on 02 July, as shown in the plot below - which also includes the tests done at about 11:00 HST.

However, the calculations of the mean temperatures for the front and back legs were not similarly adjusted until approximately 16:00 HST 06 July. The elevation residuals that night gave these results :

    UT      Rx    No.obs      c1             c2
 --------  ----   ------  -----------   -------------
 20040707   A       12    -0.5 +- 0.8   -0.22 +- 0.11

with an rms value of 1.6". Some sense of normality has returned.

20040723 - update to TEMP_MEAN_SLOPE
On 06 July the antenna leg temperature probe IDs were rectified within the ENVIROMON task. The pointing data since then are analyzed for relations of the form

             del  =  const_a * (Tf - Tb)
             del  =  const_b * (Tf + Tb) / 2

   FE    No.data   const_a  +-       const_b  +-
 -----   -------   ------------     -------------
  RxA       111    -1.30  0.41       -0.41  0.08
    B       205    -0.74  0.23       -0.33  0.05

20040729
Following TMU problems, and some bad weather it would seem that a sign change is needed (sigh). Implemented today.

20040805
Data from RxB last night in the T_leg range 0-9degC show good implementation of this new value.

20040806
Matt Rippa and I completed the (re-)calibration of the antenna leg temperature probes, fitted straight lines to plots of (ENVIRO-Optomux-) voltage output -vs- temperature (as measured by our newly purchased thermometer), and installed the new linear coefficients into the ENVIRO task. The impact upon the temperatures output by ENVIRO was immediate :

The plot shows the calibration experiments ending before 2pm and the reboot at about 5pm. The subsequent temperatures are considerably tighter, confirming the concerns I had earlier in the year. These measures compare well with those used by the Surface Project. The impact upon the means-of-4 measures called 'T_front' and 'T_back' is shown below:

The front legs now show relative heating during the period HST 06:00-18:00, resulting, one imagines, from insolation (assuming the carousel main doors are open). The range in (T_f - T_b) is approximately -1 to +1 degC. The algorithms correcting elevation pointing for this value - and for the overall mean temperature - will need adjusting once sufficient data are collected in this new regime.

20040809
The weekend's data (N=26 with RxB) suggest elevation pointing is not suffering systematic error as a result of the current algorithms. These apply corrections of 6" for each degree of (T_front-T_back) and 0.3" for each degree of the mean antenna temperature, although there is weak evidence that setting the latter value (TEMP_MEAN_SLOPE) to its theoretical value of zero would be equally good.

On the other hand : focus data show the need for significant corrections. More data needed.