Subject: PMTRAS Diagnostics

As some of you have discovered, the PMTRAS analysis software now
generates a warning message when it has reason to believe that the roll
solution is suspect. In such a case, you can make diagnostic plots to
help resolve the problem.

The attached write-up (which will also be posted on the RHESSIDATACENTER
web site) offers guidance for interpreting these plots.

gordon

 

11-July-2002
 ghurford

Interpreting PMTRAS Diagnostic Plots

The PMTRAS software issues a "WARNING PMTRAS solution is suspect"
message if any one of several internal checks is not satisfactory.
Generating a complete set of diagnostic plots by using the
"pmtras_diagnostic = -1" keyword may help identify the reason for the
problem.  In response to pmtras_diagnostic = -1, you will typically get
some extra screen output and 11 plots, of which 5 may be relevant in
this context.  The purpose of this writeup is to suggest how these
diagnostics might be interpreted to help resolve the problem.

The 4th plot, "Blip Total vs. Time"  should be examined for any gaps
(>=1 minute).  Such gaps are probably due to missing packets, in which
case obs_time_interval and/or ras_time_extension should be modified to
avoid inclusion of such a gap.  (The solution is based on PMTRAS packets
starting within +-ras_time_interval of the obs_time_interval.  Default
value of ras_time_interval is +-1400s.)

The 6th plot, "Preliminary Estimates of Rotation Period" has data points
which are normally spaced at 3 minute intervals.  This should be checked
for any missing data points.  Gaps here indicate a virtually blank star
field for that part of the orbit.  Again, obs_time_interval and/or
ras_time_extension should be modified to avoid inclusion of such a time
gap. 

Over an orbit, the range of the Rotation Periods should typically be ~2
milliseconds.  Except for characteristic hiccups at sunrise and sunset,
the successive periods should represent a reasonably smooth curve (with
'noise' at the 0.1 millisecond level).  If this is not the case, using
the blip_min=x keyword (where x is higher or lower than the Blip
Intensity cutoff indicated in the output) might help.  Alternatively,
obs_time_interval or ras_time_extension might be modified to avoid
inclusion of the offending point.

The 11th plot, "File-Averaged Phase Residuals" shows the residual phase
of all the star-associated blips, after removing the best fit uniform
rotation period.  Since all stars should lie on the same curve, this
plot must be single valued for a reliable solution.  Double-values
anywhere suggest that the entire solution may be suspect (due to star
misidentification).  Red +'s represent discarded outliers and can be
disregarded.  Gaps are ok here as long as they do not include times that
overlap with the actual imaging time_range and as long as linear
interpolation between adjacent data points seems appropriate.  If there
are gaps or double-valued regions,  then again, obs_time_interval and/or
ras_time_extension should  be modified to try to alleviate the problem.
(The time range has some affect on which stars are identified and then
used in the final solution.)

The 9th plot, "Intensity/Phase Grouping" is very  instructive for
getting insight into what is actually happening, although there is no
one-to-one correlation between plot characteristics and "fixes".   In
this plot, individual star blips are shown as white dots, plotted as
intensity vs. relative rotational phase (or position angle).  A given
star will always have the same phase and a (~x2) range of intensities
and so will show up as a narrow vertical bar.   (PMTRAS is not a good
photometer!)   The orange boxes surround groups of blips that are then
assumed to come from the same star.  The yellow diamonds represent the
expected stars that might be observed, shown shifted in phase to best
match the observed groups of blips.  When a star has been associated
with a blip group, a red X is plotted as well.  The predicted and
observed star intensities should agree to within ~x3.  The robustness of
the star association (viz. correctness of the position angle of your
image) can be judged by the number of associated stars and the
(undesirable) potential for alternate but equally valid associations. In
this regard, more associated stars is good, Closely spaced stars or
closely spaced blip groups are not good.

The 7th plot (Preliminary Timing Residuals) shows individual blip phases
as a function of a timing residuals, assuming the preliminary rotation
periods.  (The timing residuals are closely linked to the relative
rotational phase in the Intensity/Phase Grouping.)  All blips for an
individual real star should show up as a horizontal line.  (Diagonal
lines may be city lights.)   When interpreted in conjunction with the
other plots, it can provide insight into why gaps may have occurred at
different phases in the orbit.

Questions concerning interpretation of pmtras data should be sent to
ghurford@ssl.berkeley.edu