Azimuth encoder jump 14oct03
16oct03
Background:
There are two encoders mounted 180 degrees apart on the azimuth
arm. These absolute encoders have a gear on the end of their shaft
that is mated to the the azimuth encoder rack gear via a spring.
The encoder on the dome side is used for pointing. The encoder on the carriage
house side is used to measure any bending in the azimuth arm. This is done
by measuring the relative change in the two encoders as the arm swings.
The failure:
On 15oct03 barbara catinella noticed that the azimuth
error of the calibrator source she had done the previous two nights
had jumped by -16 arcseconds. Luis murray also noted on the 15th that the
azimuth encoder difference was larger than normal.
The vertex motor data is logged once a second to
a file. This allowed us to go back and look at what was happening during
the failure. I checked the azimuth encoder difference for 14oct03 and there
was a jump of one of the encoders. I then looked at the first of the month
for the previous 8 months and found no other jumps. The figures shows the
azimuth
encoder values before and during the jump (.ps) (.pdf)
(watch out.. it's 4mb!!)
-
Fig 1 is the azimuth encoder difference (domeEnc - chEnc) versus
second of day for 14oct03. The units are encoder counts (1 encoder count=.000144
degrees). The Top figure is the entire day, the bottom figure is a blowup
during the jump with the red dots flagging where the jump occurred.
It looks like there may have been two jumps. The slow changes of the encoder
difference come from the flexing of the azimuth arm as it moves in a particular
direction. The jump differs because it occurs within two readings
of the encoder (once a second).
-
Fig 2 plots the rate of change of encoder difference (per second).
The top plot is the entire day, the lower plot is during the jump. The
green dotted lines would be jumps of one and two teeth of the encoder (64
encoder counts per tooth). The red dotted line is the second where the
the encoder difference jumped by two encoder teeth.
-
Fig 3 has the azimuth (top) and gregorian dome za (bottom) positions
during the jump. The jump occurred at az 42.186, za=12.0229, ch=stow. The
azimuth was moving at slew rate and was just starting to slow down. The
dome was moving downhill at slew rate.
-
Fig 4 Top The velocity is measured using the azEncGreg and azEncCh
by differentiating the encoder positions. These are the black and red lines.
The velocity is also measured independently by the motor amplifiers using
the motor resolver (green line). The amplifier velocity and the azEncCh
agree so it must be the azEncGr that is jumping.
-
Fig 4 bottom: A linear fit was made to the encoder positions (azEncGr
black, AzEncCh red) during the constant velocity motion and then subtracted.
You can see that the azEncGr has jumped.
-
Fig 5. I corrected the 14oct03 data with jumps of 2, 3, and 4 teeth
and then recomputed the azencoder difference. I sorted the difference by
azimuth and smoothed it by 7. This data was plotted vs azimuth for jumps
of 2,3, and 4 teeth (top,middle, bottom plots black lines). I over plotted
the azimuth encoder difference for 13oct03. The middle plot with a jump
of 3 teeth agrees with the previous days data.
A jump of 3 teeth=192 encoder counts=99.66 arcseconds
in azimuth (this is a little circle error). Barbara's calibrators were
taken at 9.5 degrees za. The jump should have changed the az error by 99.66*sin(9.5)=16.4
arcseconds with the sign being negative...so things agree.
On 16oct03 at 14:00 we moved the azimuth encoder
3 teeth moving the azimuth encoder difference from 268 to 64 encoder counts.
We need to get an index on the vertex encoders (especially the ch and
dome which don't have dual encoders) so we can respond faster to these
encoder jumps. I'll also generate a daily monitor page that shows
the azimuth encoder difference for the previous day and compare it to the
expected values.
processing: x101/031015/agc.pro
home_~phil