[evlatests] Fast band switching test

[Bryan Butler]

actually i rather suspect it should be:

       Duration = DwellDesired
                + Max(CorrelatorSetupTime,
                      AntennaSlew+SettleTime,
                      SubreflectorRotationTime,
                      SubreflectorFocusTime)
                + Other

because correlator setup, antenna move, and subreflector motion all 
start at (roughly) the same time (i *think* this is true, but i'm not 
positive about subreflector rotation and focus happening at the same 
time - and in fact looking at the executor code it looks like there is a 
0.25 sec delay between the two, but that's pretty small).  the Other 
shouldn't necessarily be there at all, but the fact that it takes 5 
seconds to go from reference pointing at X-band to regular X-band 
observing, which should cause none of the setups or motions, indicates 
that *something* in the system is taking that long to get organized 
(it's not executor per se, since it looks ahead by one scan).

the subreflector focus time can be substantial - at least to and from 
L-band.  a quick perusal of the parameters database shows that for the 
other (non-L) bands, the motion is small - of order a couple of cm or so 
maximum.  for L-band, however, it's typically of order 5 cm, and there 
are cases as large as 7 cm.  the motion rate is 0.5 cm/sec, so to and 
from L-band will be 10-15 seconds for focus, and the other bands will be 
of order 5 seconds or less.  one note - i suspect the time to move to 
S-band might be substantial as well, but the parameters database seems 
to have no entries for S-band focus at all, so i couldn't check.  maybe 
i was driving parminator incorrectly.

the subreflector rotation time can also be substantial.  for rick's 
sequence, the subreflector rotation angles and times (with a rotation 
motion rate of about 12 deg/sec) will be about:
  Ku->L: 136    11
  L->S:  172    14
  S->C:   58     5
  X->Q:  177    15
  Q->Ka:  20     2
  Ka->K:  40     3
  K->Ku:  22     2

so, another way of showing rick's table would be:

      Band        Delta    expected    Delta-expected
-----------------------------------------------------
        L          16        11             5
        S          18        14             4
        C          17        10             7
        X           5         0             5
        Q          20        15             5
        Ka         10         5             5
        K          12         5             7
        Ku         10         5             5

where Delta is rick's Duration-Good data (the "setup" time, for all 
intents and purposes), and expected is 
Max(SubreflectorRotationTime,SubreflectorFocusTime), and i've assumed 
moves to and from S-band take similar time as L-band.

i take this to mean that some combination of the correlator setup time 
and the Other is about 5 seconds, so we'd be safe (to +- seconds) with:

       Duration = DwellDesired
                + Max(AntennaSlew+SettleTime,
                      SubreflectorRotationTime,
                      SubreflectorFocusTime)
                + 5

in any case i'm glad to hear it's shorter than the VLA (which was never 
shorter than the built in 20 seconds).

	-bryan


Rick Perley wrote, On 12/7/10 14:40 PM:
>      I ran a short test yesterday -- part of a series of tests in
> preparation for a 'flux densities' run -- to assess if we have
> established good robustness on band changes.
>
>      The test consisted of observing a single source at every band with
> the following sequence:
>
>      L, S, C, RefPnt at X, X, Q, Ka, K, Ku
>
>      This was repeated three times.  The observations were in 'extended
> wideband' mode, including the referenced pointing.
>      Below is a table of the duration requested, and the actual duration
> of good, steady fringes.
>
>      Band        Duration        Good data
> --------------------------------------------------
>        L               50 sec               34 sec
>        S               50                     32
>        C               45                     28
>        XRefPnt   180                 ??  It seemed to work
>        X               40                     35  (no reconfiguration needed)
>        Q               50                     30
>        Ka             40                     30
>        K               40                     28
>        Ku             45                     35
> -------------------------------------------------
>
>      The '18-second' dead zone at the end of observations, when the
> following observation switches sideband, is gone.  Also gone is the
> differing length of blank data at the beginning of a scan for different
> subbands (spectral windows) when changing correlator setup.
>
>      I deduce from this that the actual time it takes to setup for a new
> observation is no more than 10 seconds.  The additional time shown in
> the table above (Duration - GoodData) for some band changes is likely
> due to subreflector rotation.   This is considerably better than the VLA.
>
>      For multi-source observations, the relevant duration to request will
> be something like:
>
>      Duration =  DwellDesired + 10 + Max(SlewTime, SubrRotate)
>
>
> _______________________________________________
> evlatests mailing list
> evlatests at listmgr.cv.nrao.edu
> http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests