June 30, 1998

Note to reader: This was originally a letter to George and Imai, reporting the results of the analysis of an encoder test they ran.

The test, run in SuperTech, used primitive testing routines to move the telescope and read back the declination absolute position encoder (APE) and the difference in displacement counts between the APE and the relative position register (RPR) after each slew. The RPR accumulates displacement counts from the incremental encoder, part of the telescope drive loop.

The slew pattern was as follows:

4 round trip slews at 1 hour HA between -15 degrees and 45 degrees declination.

4 round trip slews at 0 hour HA between the same declinations.

4 round trip slews at -1 hour HA between the same declinations.

The declination incremental encoder changes its ratio of displacement counts per arcsecond with changing hour angle. The results of the tests can only be made consistent by assuming different dec encoder ratios at each of the three hour angles tested. The telescope pointing and drive systems all expect a constant encoder ratio on each axis, independent of the other axis. The mechanical assembly of the declination incremental encoder needs to be examined for the cause of this problem.

Also, there is a minor degree of encoder roller slippage to the north, of about one displacement count per degree (.05 arcsec).

George & Imai:

Here is the raw graph of the back and forth dec scans done at hour angles 1, 0, and -1 hours.

The line scans done at one hour west HA are data points 1-11; at 0h data points 12-19; and at 1h east are 20-30. There are three obvious things here. If the APE and incremental encoder were perfect, the graph would be a horizontal line. Instead, we see -

• A general upward slope, indicating that the APE reading is increasing as we do back and forth scans, so we are drifting north. The encoder roller is slipping more on the north legs, causing us to overshoot in that direction on the absolute encoder reference.
• The back and forth line scans show triangular peaks. The encoder ratio in SuperTech is off, as expected. A count difference accumulates as we slew out, then the difference is taken up as we slew back. I believe the SuperTech dec ratio is 19.9466 counts/arcs.
• This is the most important: The true encoder ratio is different for each hour angle! There is a distinct difference in amplitude of the triangular peaks for each of the hour angle runs. And the change in encoder ratio is not symmetrical about the meridian in hour angle, but increases as we go west to east!

In fact, the first run at one hour HA shows a changing ratio for the first three round trips. The runs at the other two HA positions are pretty much constant throughout each run. Why the encoder ratio is changing like this is the question, if the changing encoder ratio is the actual cause of the difference in the plots.

Here is a plot of the same data, only with a slippage of 1.05 counts per degree dec displacement subtracted out:

Notice that the plots are now pretty much horizontal and the three different runs show up clearer, with obviously different encoder ratios. (The vertical scale is different from the previous plot.)

Here is a graph of the three runs on separate series (plot lines), with slippage taken into account, making the graph fairly level, but everything with the same encoder ratio:

Here is the above graph but with optimized encoder ratios for each series, to the same scale as above:

By changing the dec encoder ratio for each HA position, we now have optimized the runs. Except for the starting point of the first run and the ending point of the last run, the points all lie within about 100 counts of each other, or 5 arcsec.

So, why is there a different dec encoder ratio for each HA position? What is going on with the encoder roller or surface to cause an apparent change of ratio as HA changes? Or is the dec APE sensitive to HA? I doubt that, because a pointing test run during engineering time a few weeks ago showed a direct relation between dec slew error and HA position. The on-line software does not use the APE. At the time I thought it was because of some software glitch. Now, with this data, it looks like there must be some mechanical reason.

So what do we do now? One possibility is to repeat this test, only do just 2 dec round trips each at HA positions -2h, -1h, 0h, 1h, and 2h to see if the relationship between dec incremental encoder ratio and HA position is real and if so, what its rate of change is.

Otherwise, we can just accept that there is this dependency of dec encoder ratio on HA and try to look for it in the mechanism. Maybe you can show this to Paul and see if he has any ideas. Something is changing in the encoder mechanism as it tilts with HA. Is the roller pressing with more force when tilting in one direction rather than the other? Is it sliding axially so that it is running on a different part of the surface if tilted one way or the other? Is the drive gear slipping axially when the yoke tilts? Is there anything loose in there? Is there something about the roller alignment that will make it sensitive to HA position? I think a thorough inspection of the dec incremental encoder and drive gear is in order here. Before we adjust anything, let's see if we can find something wrong. Perhaps you could mount a TV camera and light in the yoke and watch the incremental encoder on a monitor as the yoke tilts through a range of motion. Any shift of a mechanical component may be too fine to see on a TV screen; you may have to mount a dial gauge and watch that with the TV. This is starting to get complicated, I know.

Finally, the previous test using a square pattern indicated that there might be spurious dec counts being generated. This was not seen in this test. My interpretation that what we were seeing was spurious counts may have been wrong. It may be an artifact of this strange dependency of dec ratio with HA when doing a square pattern.

Jim