[evlatests] Time taken to change bands

[Bryan Butler]

P to L is the largest focus change (P to S is close).  from P to L is 
almost 17 cm (average - worst case is 21 cm!), which means ~30 seconds 
to move the subreflector.

from the parameters database, currently:

               focus
band   median   min   max
----   ------   ----  ----
  P      11.9    11.8  11.9
  L      -5.0    -9.0   3.0
  S      -2.5    -6.0   2.0
  C       0.9    -3.0   5.3
  X       1.5    -2.5   5.8
  Ku      2.2    -2.0   6.2
  K       2.2    -1.9   6.2
  Ka      2.2    -2.0   6.4
  Q       2.2    -1.8   6.4

the wide range explains why you see a large variation between antennas.

yes, 20 seconds is conservative, but we *need* to be conservative when 
doing things in the OPT, because we don't know even what the worst case 
is for a band change, because it depends on the current values in the 
parameters database (when the SB is run, vs. when it is constructed in 
the OPT).  if we tune it to the average, then half the antennas won't 
make it, etc...

and it really affects very few projects.

we've had this discussion before (december 2010, most recently)...

	-bryan


Rick Perley wrote, On 2/14/14 12:01 :
>     In preparation for a calibrator run this weekend, I've carefully
> checked the time it takes to change bands and correlator modes.  In
> general, it's less than you might (and the OPT does) think.  It's also
> very repeatable amongst the 7 sources in the script.
>     The sequence of operations is as follows.  Note that the sequence is
> selected to minimize subreflector rotation time.
>     1) Move to source:  This was done at L-band.  The preceding band was
> 'P'.  Correlator mode change required.
>     2) L-band to X-band referenced pointing.   Correlator mode change.
> Large subreflector focus motion.    3) X-band referenced pointing to
> X-band observation.
>     4) X to C        Correlator mode change.    5) C to S
>     6) S to Ku        Correlator mode change
>     7) Ku to K        Correlator mode change
>     8) K to Q
>     9) Q to Ka    10) Ka to P      Correlator mode change.   Large
> subreflector motion in focus.    11) P to L -- move to the next source.
> Correlator mode change.
>     For all observations, the integration/dump time was 1 second.  All
> observations were taken with the 8-bit samplers.    For all observations
> *except P*, I requested 35 seconds On Source time.    For 'P' band, 40
> seconds of On Source was requested.
>     In all cases, the OPT simply adds 20 seconds to the OnSource request
> time.
>     Below is a table giving the time taken to actually effect the
> change.   In this table, 'Time to Change' is the duration between the
> last good data of the originating band, and the first good data in the
> new band.
>     Important Conclusions:
>
>     1) The OPT's built-in addition of 20 seconds to change bands is very
> conservative, and is presumably based on a worst-case motion change.  It
> would be nice if the OPT were made smarter, taking note of the actual
> pair of bands involved, (and whether the correlator changes mode) since
> the change time between high frequency bands is no more than 10 seconds.
>     2) A more careful look at the L to X band, and Ka to P band setup
> times reveals a large variation between antennas.
>         a) L to Xband:   The time taken is 12 to 24 seconds.  The same
> antennas are always the slowest:  ea 10, 12, 16, 20, and 22.  Of these,
> ea20 is always the last.  Either it has the slowest focus motion, or has
> the longest travel distance (or both).   Perhaps up to 10 seconds is
> taken up the the rotation, the remaining variation must be due to focus
> change.
>
>         b) Ka to Pband:  The time taken is between 21 and 39 seconds. It
> is clear from looking at the variation in visibilities during this
> period that the antenna gain change is a very weak function of the
> subreflector position!  (This is good news for VLITE ...).  The same
> antennas always take the longest time:  ea16, 19, and 22.
>     Very unfortunately, ea21 (with the new FRM system) was *not* in the
> array, since it had been moved earlier in the day.
>     3) It appears the correlator mode change times are not an important
> factor here.  The C to S move -- which does not involve a correlator
> change, took 15 seconds, about the same as the S to Ku move, which does
> require a correlator change.  The later move is slightly shorter in
> subreflector motion, and took only 2 more seconds, presumably due to the
> correlator change.
>
>
> _______________________________________________
> evlatests mailing list
> evlatests at listmgr.cv.nrao.edu
> http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests
>