Azimuth encoder difference
There are two azimuth encoders. Encoder 1 is on the
gregorian side and is used for pointing. Encoder 2 is on the carriage
side and is used to measure azimuth pending (as well as a check on the
gregorian side encoder).
On 09mar06 an azimuth spin was done from az=-90
to az=+270 and then back at slew rate. The encoder difference
was sampled once a second.
The first plots show the
encoder difference versus azimuth (.ps) (.pdf):
There is a positive going bump in the encoder
at az=160 degrees and a corresponding negative going bump 180 degrees
On 21jan00 the
runout of the azimuth encoder rack gear was measured. It showed
the azimuth encoder rack gear was bowed inward (by about 1 inch)
an azimuth of 160 degrees (unfortunately the computed azimuth error was
off by about a factor of 10!).
Top plot: The encoder difference (in encoder units) versus azimuth
angle. Black is the clockwise (increasing azimuth) spin while red is
CCW (counter clockwise spin). The average difference between the CW and
CCW spins in 148.5 counts. This is caused by the flexing of the azimuth
arm when driven in a particular direction. The Enc1 side is driven by
encoder1. The encoder 2 side is not servoed.
Bottom plot: the same plot with the vertical scale in degrees rather
encoder counts (there are about 6935 encoder counts in a degree).
The next plots show the repeatability
of the azimuth encoder rack gear error (.ps) (.pdf):
The data and the fit is stored in the idl save file encDifFit.sav .
Top : The black line is the CW encoder difference. The Green line is
difference between the CW and CCW spins. The red line is the CCW
difference with the average CCW -CW difference removed. The two lines
2nd plot: The differences for the CW and CCW spins were averaged and
the mean was removed. The average encoder difference should be 0 since
we made a complete revolution. There looks like there is a 180 degree
(which you'd expect). The red plot is (-average of spins) offset by 180
degrees. The 180 degree symmetry is repeatable.
Bottom: A third order baseline was removed from the average of the two
spins (black plot) and then a gaussian was fit to the bump around
Assuming that most of the bump is coming from az=160 (and not
this is supported by the runout measurement) then the fit at 160
can be used to correct the pointing model around this area. The current
pointing model does not fit a bump like this very well.