NRAO Home  >  Green Bank  |  Wiki Topic:    GB > PTCS > TemperatureSensorResults

Changes | Index | Search | Statistics | Go

First Results from the GBT Antenna Temperature Sensors

At the PTCS Conceptual Design Review held in April 2003, we identified additional antenna instrumentation as a key aspect of our new, system oriented approach to achieving effective 3mm operation. In particular, we identified as the highest priority item a system of ~20 precision temperature sensors to be deployed on the antenna feedarm, primary backup structure (BUS) and alidade. The workpackage to design, construct and install these sensors was completed, on schedule and within budget, in August 2003. We are now starting to obtain our first results from these temperature sensors, and the initial indications are extremely promising.

In late August, we devised an experiment to isolate the effects of thermal gradients on antenna pointing and radial focus. We identified a NVSS pointing source, 0117+8928, which lies within one degree of the North Celestial Pole, and so remains at an essentially fixed azimuth and elevation of (0,38). By performing repeated pointing and focus measurements on this source under stable, low-wind conditions, we can exclude essentially all sources of pointing/focus error (for example any gravitional terms) apart from thermal gradients in the antenna.

An extremely successful experiment of this sort was performed for approximately 22 hours, from 1pm EDT on Friday 5th September to 11am EDT on Saturday 6th. An example of a subset of the temperature sensor data is shown in Figure 1. This run was followed by a second, performed under rather less stable conditions, on September 11th. The results for azimuth, elevation and focus are shown in Figure 2, Figure 3 and Figure 4. The X axis is sequential scan number (we obtain two azimuth and elevation cross-scan results for each focus observation), the Y axis is residual from the standard pointing/focus tracking models in arcsec or mm respectively. The discontinuities in the approximate center of the plot, are as a result of merging the two successive datasets; the increased scatter in the second dataset reflects the poorer conditions under which they were obtained.

The solid line in each case represents the results of performing a linear model fit of the data from all of the temperature sensors (including a mean bulk temperature) - this calculation is the discovery of the best linear combination of temperatures that yields the offsets, and so the solution is just the generalized eigenvalue problem. The results are outstanding. The second panel in each figure shows the residuals from the prediction; the RMS residuals are 3 arcsec, 3.5 arcsec and 3mm respectively (and significantly better if only the first dataset is included). In one sense, this should not be surprising; the assumed physics is that temperature perturbation influences on pointing and focus are linearly independent from node to node within the structure, something that we have first-principles reasons to believe. The goodness of fit is gratifying however.

Of course, the fit to the data for which the fit was made is expected to be good. We have attempted to use these results to predict the observed pointing and focus offsets for a third dataset. The prediction in this case is the dot product of the results of the above linear regression analysis with the median value over each scan of each of the 20 temperature sensors. The results, again for azimuth, elevation and focus are show in Figure 5, Figure 6 and Figure 7 The results for focus are really excellent. The azimuth and elevation results are not quite so good. The elevation results could be significantly improved by allowing a single "local pointing correction" at the start of the run, but the system should really predict this. What does this mean? Roughly, we have to get more data.

The specific experiments and analysis described here were performed by Kim Constantikes, Dana Balser, Richard Prestage and Jim Condon. The development of the instrumentation and techniques needed to execute the work was a collaborative effort of all members of the PTCS Project Team.

UPDATE: In order to better understand the phenomenology of temperature-induced structural distortions, we've generated an AVI movie (both a 15 fps frame rate fast version and a 1 fps frame rate slow version) of node temperatures and the differences between some adjacent nodes (a thermal network). The attached AVI file from the 5 September experiment has the sensors in accurate positions within the GBT base coordinate system and the GBT in the correct pose (0 az, 36 el). The nodes are pseudo-colored using the colormap attached to the bottom of the graph (ignore the axis associated with the color bar), and the edges that connect nodes are colored in the same fashion, except the scale factor has been adjusted to better illustrate temperature differences. The graph title in each frame tells you what the temperature difference is between the coldest node (blue) and the hottest node (red)- in this animation 23C, which is a very large diurnal variation and indicative of the direct solar heating that occured during the daylight observations under a clear sky. The colormap scale for edge coloring changes from frame to frame, and is typically about 6C. The maximum temperature difference s are on the connection from elevation bearing to horizontal feedarm sensors, and exceed the design temperature gradient (5C across the extreme dimension of the GBT, see Loral Tech Memo 52) but a large amount.

It is easy to visualize the sun rising,transiting, and setting while looking at the back of the feed arm and noting the temperature differences from front to back and side to side. Also note that the elevation bearing weldments change temperature very little and slowly (compared to the feed arm), so that there not much time where the horizontal feed arm, BUS, and alidade legs don't have a large temperature differential.

The second AVI file (fast version slow version) attached is temperature data from a recent three day period which had substantial overcast (Isabel) periods during the day. The total temperature variation of the GBT in this period is less than half of the preceding example, and temperature differences in the structure are also much less. If you watch closely the differentials in the feed arm as compared to time-of-day (EDT) you can see periods where the sky clears and gradients are introduced by solar illumination. Note that the "blacked-out" period in the animation: Some sensors were briefly shut down.

-- KimConstantikes, DanaBalser, RichardPrestage - 22 Sep 2003

Attachment:	Action:	Size:	Date:	Who:	Comment:
GBTTempsMovie.avi	action	4196352	21 Sep 2003 - 22:23	KimConstantikes
GBTTempsMovie2.avi	action	4071424	21 Sep 2003 - 22:29	KimConstantikes
GBTTempsMovie2Slow.avi	action	1827840	22 Sep 2003 - 16:46	KimConstantikes
GBTTempsMovieSlow.avi	action	1827840	22 Sep 2003 - 17:11	KimConstantikes
soln_05p11_az.pdf	action	19728	22 Sep 2003 - 20:58	RichardPrestage
soln_05p11_el.pdf	action	20012	22 Sep 2003 - 20:58	RichardPrestage
soln_05p11_focus.pdf	action	11248	22 Sep 2003 - 21:00	RichardPrestage
st.pdf	action	40123	22 Sep 2003 - 21:00	RichardPrestage
predict_030824_az.pdf	action	4925	22 Sep 2003 - 21:01	RichardPrestage
predict_030824_el.pdf	action	4781	22 Sep 2003 - 21:01	RichardPrestage
predict_030824_focus.pdf	action	3364	22 Sep 2003 - 21:02	RichardPrestage
tempMovie030924.avi	action	6320128	24 Sep 2003 - 22:32	KimConstantikes
tempMovieSlow030924.avi	action	2605568	24 Sep 2003 - 22:32	KimConstantikes
loralTM52.pdf	action	4390127	24 Apr 2008 - 16:39	ToddHunter