Telescope Jumps: One Cause is Found

It looks like we now know the cause of at least one of the problems plaguing the poor 3.5-m telescope! Many thanks should go to Russet for her pointing out the utility of the 3.5-m guider data and to Russ O. for his help in deciphering the TCC logs. This has indeed been a valuable source of information on some of the telescope jumping problems.

As implied by my earlier jump report, we have more than one problem with the jumps that the 3.5-m is currently experiencing. You can classify the problems as those that effect telescope imagery and those that impact the telescope collimation. The largest mirror motions are clearly associated with the telescope collimation, however, the impact on the users of the smaller motions is more noticeable because they are occurring regularly and have an immediate and noticeable effect on images taken. The problem that we believe we now understand are the jumps that are affecting telescope imagery. I will discuss those problems below along with the fixes that we should probably pursue.

The jumps that are affecting telescope imagery are being caused by the collimation process, but only under certain circumstances. The collimation process currently has two components. One tries to take out the sag of the secondary cage as the telescope is pointed toward the zenith. The second component tries to take out changes in the secondary piston caused by temperature changes of the truss structure. Soon we hope there will be a third process which compensates for translation of the secondary cage as the telescope is pointed toward the horizon. Temporary efforts to include this third component have been made, but this is not currently installed in the software and can be ignored in the context of this discussion. The jumps occur when the temperature component drives the secondary in a direction opposite to where it is being driven by the sag component. I will illustrate this with four plots.

Figure 1 below shows the x and y centroid positions of a star on the guider during a 2 hour period on the night of March 3. At this time the guider was set to take 30 second exposures. The circles show the x-centroid positions in binned pixels and the squares show the y-centroid positions. Marked a through k are several telescope jumps that are clearly observed in the y-centroid positions. The corresponding x-centroid positions are also marked in the same way. Between 3.2 and 3.4 hours UT, the telescope operator was executing a series of instrument offsets which created the slopes seen in both the x and y centroid data. The jumps seen in this figure range from 0.7 to 1.8 arcseconds. This range is quite consistent with all of the documented image jumps that have been studied so far. It is immediately clear that the jumps are clustered in a systematic way.

Figure 2 shows an expanded view of the first four jumps shown in Figure 1 and Figure 3 shows a similar view of the second four jumps. Again, the x-centroids are shown by circles at the bottom of the plot and the y-centroids are shown in squares near the top. In triangles we include a plot of the secondary piston position as commanded by the collimation routine. The constant given in the legend has been subtracted from the actual piston values to enable them to be displayed in the same graph as the centroid data. The units of the piston changes taking place are in microns rather than pixels as shown in the axes labels. At this time the TCC was set to update the telescope collimation every 60 seconds. The general slope of the piston data is set by the sag component of the collimation process. Over the period of time displayed in Figure 2 the secondary piston was changed by a total of about 20 microns to account for the sag of the secondary vanes. The upward and downward discontinuities are caused the temperature component of the collimation. About every 5 minutes the weather data is updated in the TCC database. When this happens, small changes in the secondary truss temperature are translated by the TCC to small changes of the secondary piston position. The discontinuities in Figures 2 and 3 owing to temperature changes range from 0 to 5 microns of piston motion. Two things are obvious from these figures. First, the jumps are perfectly correlated with the changes in piston position caused by the weather data. Second, the jumps only occur when the temperature data causes motions of the piston which go in a direction opposite to the direction in which the sag component is driving the secondary.

Figure 4 shows the variation of the air and secondary truss temperatures during this night. The air temperatures are shown with square symbols and the truss temperatures are shown in circles. As expected, the air temperature changes lead the truss temperatures in time and also show greater variability. The data from Figures 1 through 4 were acquired between 3.3 and 4.5 hours UT during a time where the temperatures were actually quite constant! The bar in the lower left of the figure marks this interval. This illustrates that the jumps seen in Figures 1 through 4 are not being driven by unusual conditions. The temperature differentials that are causing the discontinuities seen in these figures were all less than 0.07љ C!

The small size of the temperature differentials which are causing the piston discontinuities shown in Figures 2 and 3 indicate how sensitive the system is to changes in temperature. The smooth trends in the temperature data shown in Figure 4 argue that these are real temperature changes in the truss structure and are not caused by errors in the sensors.

The fact that the direction of the piston motion must be reversed for the jumps to occur does indicate that some sort of backlash in the actuator mechanisms is responsible for the telescope jumps. This mechanism cleanly explains why the jumps are not in any natural actuator direction as was reported earlier because the collimation process simultaneously moves all three actuators. The jumps are the result of unequal amounts of backlash in all three of the mechanisms. This allows the direction of the jumps to be consistent in orientation and yet not aligned with any natural actuator axis. It also explains the frequency of occurrence and the imperfect correlation with the collimation motions because it is the combination of both a temperature change and a change in the correct direction that are required for the jumps to occur.

One very odd feature of this mechanism is the fact that it appears that earlier tests seemed to indicate that reversed motions of the secondary piston had little or no effect on the mirror motions. One has to wonder how this can occur. So far, the only difference between the pistoning tests done earlier and those caused by the temperature changes shown here is the magnitude of the piston motions. The smallest motions made in the piston tests earlier were 15 microns, which correspond to 1.4 full turns of the stepper motors. The largest motions seen here are only 5 microns, which correspond to about 0.5 turns of the steppers. The size of the jumps are not clearly correlated to the size of the temperature discontinuties. In Figure 2 a very small downward discontinuity at 3.48 hours UT caused a significant mirror motion of about 1 arcseond while in Figure 3 at 4.26 hours UT a similarly small discontinuity caused a jump so small that it was not initially recognized as significant (and therefore when unlabeled) until the correlation was discovered. Perhaps there are some unwanted torques which are somehow relieved in the mechanisms when the stepper motors move more than a full turn or the backlashes average out over multiple turns of the steppers. Clearly this is just speculation at this time.

There are several possible easy short term solutions to this problem. The quickest thing to do is simply to increase the time interval of weather updates. Increasing the weather update interval to 10 minutes will significantly decrease the occurrence of the telescope jumps but will not eliminate them entirely. The attraction of this approach is that it requires a minimum amount of effort on the part of the observing specialists. The downside is that some defocus can occur owing to secondary vane sag over this interval. Figures 2 and 3 show typical piston changes in this regard. These figures show that a defocus of about 5 microns can be expected by this action. This is probably tolerable. A somewhat harder approach would be to have the observing specialist turn collimation off during exposures and manually reinitialize it between them. This is obviously a headache for the observing specialists but it will probably be difficult to have the TCC do this automatically. Russ's feedback here is required.

Part of the good news here is that hardware solutions to this problem are already under way and therefore the software fixes will only be needed temporarily. The addition of the Heidenhain encoders should allow such motions to be both monitored and corrected. Since these encoders are due on the telescope within a few weeks, it is probably not necessary to implement other efforts to fix these jumps.