Pointing model update dec06
21dec06
Sections
Intro
Measurements
Conclusions
Plots
The pointing
errors for 2004,2005,2006 (.ps) (.pdf):
The 1az
and sin(za) fits to the pointing errors (.ps) (.pdf):
Pointing model 15 was installed on 14may04. Since then
we have had some azimuth and tiedown encoder failures. These were on: (note
that this list may not be complete)

17mar05: td12 encoder replaced

17mar05: td12 jack replaced

06mar06: td12 encoder replaced

12aug06: td8 encoder failed.

14oct06: azimuth encoder 1 failed.

14nov06: azimuth encoder 1 failed.
A tiedown encoder failure requires
us to reset the encoder position of the jack. This is done with a tape
measure and it is probably not accurate to more than .1 inches. If the
jack position changes (since the model was made) because of the resetting
of the encoder, you will see a 1 azimuth term in the pointing errors (with
the maximum or minimum aligned with the tiedown azimuth). A .1 " error
in the tiedown height would give a 5 arcsecond 1 azimuth error (.1/(192'*12)/1.76
tdInchPerPlatformInch).
An azimuth encoder failure requires a resetting of
the azimuth encoder position using a mechanical pointer/index mounted next
to the encoder. The resetting can probably not be done to better
than .05 inches (the radius is about 60 feet). This gives an error
of 5 asecs at 20 degs za (or 2.5 asecs at 10 degrees za). The error on
the sky changes as the sine of the zenith angle.
A third source of pointing error is the UT1UTC approximation
that is used in pointing. This error was about 1 asec jun06 to nov06 and
then climbed up to about 2.7 Asec by dec06 (see
ut1utc updates).
The more recent pointing errors showed a 1 azimuth
term in the za pointing errors. To investigate this the following was done:

Plot the pointing errors by year since the last model was installed (may04)
to see when the errors started.

Using the pointing data since the last azimuth encoder failure (14nov06):

Remove the 1az term caused by the UT1UTC errors (a new value was fit for
dec06 through jun07).

Fit for a 1 az term in azimuth and za (to fix any tiedown position errors).

Fit for a sin(za) term in the azimuth errors. This will correct any azimuth
encoder errors caused by resetting the azimuth encoder.

Include constant terms in the fits.
The data used came for the x102 calibration runs. Receivers
with frequency greater than 1 GHz were used. Any data with
low tiedown tensions (less than 5 kips) as well as data taken 9am to 6pm
where not included.
The Measurements top
The first set of plots show
the pointing errors for 2004,2005,2006 (.ps) (.pdf):

Page1 2004 Errors:. This data was after 15may04 when model 15 was
installed. The only obvious structure is in the az errors vs az which weren't
fit very well by the model (we need an azimuth encoder table).

Page 2 2005 Errors: The za errors show a 1 azimuth dependence with
a maximum around az=40 degrees. The az errors at az=30 have increased.
This may be the structure from 2004 with the inclusion of the 1 az term
for the azimuth errors. The azimuth errors 1 azimuth term should have the
same amplitude az the za term and its peak should occur 90 degrees before
the za peak.

Page 3 2006 Errors: The 1 az term for the za errors is more pronounced.
The data after 14nov06 was used to fit a 1 az
term and a sin(za) term in azimuth. The error in the UT1UTC conversion
was removed before the fits (a .167 second time second or 2.5 arc seconds).
The plots show
the 1az and sin(za) fits (.ps) (.pdf):

Page 1 The az,za Errors: This is the data from 14nov06 thru 10dec06.
The UT1UTC error has not yet been removed. The rms errors for za and az
are: 7.07 and 6.83 asecs with offsets of .62 and 3.91 asecs. The
za error 1az term is similar to the rest of 2006.

Page 2 az,za arrow plot of errors. This shows the errors plotted
vs az and za. The length of the arrow is proportional to the error (10
asecs is one division). The az,za coverage is not complete.

Page 3 The fits: The UT1  UTC errors have been removed from the
data points before fitting (since the UT1UTC data was updated 11dec06).
The black * are the data, the red * are the fits.

Top za errors vs az.

The fit: .7 + 5.8cos(az) + 3.8sin(az) .

The 1 az term has an amplitude of 6.9 asecs and a phase of 33.4 deg (peak)

Middle Az errors (1azfit): The constant and sin(za) terms
were removed from the data points and then plotted along with the 1 az
fit in azimuth.

The fit: 2.0 + 1.6cos(az)  3.9sin(az)

The 1 az term has an amplitude of 4.2 asecs and a phase of 67.1. This
is 100.5 degrees rather than 90 degrees from the za fit.

Bottom, sin(za) fit to az errors. The constant and 1 az terms were
removed from the data points and then plotted. The sin(za) azimuth term
was then over plotted in red. The amplitude is 35.2*sin(za)

Page 4 fit residuals: The fit residuals are plotted versus azimuth
and za. The 1 az term in the za errors is now gone. The az errors vs az
still show structure around az=30.
The fits were done separately for azimuth and za errors.
A tilt of the platform should give the same amplitude and phase for az
and za errors (with a 90 degree phase shift). I tried fitting the az,za
errors simultaneously. The results did not differ a whole lot from the
separate fits.
Conclusions top

The UT1UTC error was updated 11dec06. On 10dec06 before the update the
error was about .167 seconds of time. This changed the measured data by:

azErr=azErr  2.5*sin(az+90)

zaErr=zaErr  2.5*sin(az)

The pointing models for all receivers were updated. The following
terms were subtracted from from the model15 values:
errors 
C0 
cos(az) 
sin(az) 
sin(za) 
azErrors 
2.0 
1.6 
3.9 
35.3 
zaErrors 
.7 
5.8 
6.9 
0 

The residuals for the 14nov06 to 10dec06 pointing errors change by :

rms az (Asecs) 
rms za(Asecs) 
Before 
6.83 
7.07 
after 
5.16 
3.28 

The za error rms improved by 4 asecs primarily because of the 1az fit.
The az errors only decreased by about 1.7 asecs

The 1 az term in the za errors was there for most of 2006. The change must
have come from one of the tiedown failures in 2005. Another possibility
could be extra weight that was placed on the platform (in an asymmetric
manner).
processing: x101/061207/chkpnt.pro
page up
home ~phil