[evla-sw-discuss] Flags

[Barry Clark]

Chunai asked me a couple of questions about flags Friday that raised
design considerations, so I thought I'd write down everything I could
think of about flags, so we could have an informed discussion.
Specifically, I want to list the EVLA antenna flags I know about,
and what one needs to know to use them.

Flags are generated in a distributed fashion, and passed to a flag
collator, which may be considered part of the DCAF.  I'll use DCAF
as a generic term, for the Modcomp version and the VLA correlator version
as well as for the WIDAR version.

I.  Flags that flag an antenna bad.

1.  Antenna not at commanded position flag.  This is pretty straightforward, 
and can come directly from the MIB.  Except for one case - ACU MIB dead or 
otherwise not responding.  I have thought about whether we should set the
limits for this as a function of band, and at the moment am feeling negative
about doing so.  Some 20cm observations - deep, high dynamic range, mosaiced
observations, such as those that Eva Schinnerer et al. are starting next 
month - are as touchy about pointing as the higher frequencies.  I'd 
rather the on-line system not know that much about observations.  True, 
90cm and 4m observations don't worry much about pointing, ever, but we 
don't do enough of these bands to worry about loosening specs for them.

2.  Subreflector out of position.  Rotation has much the same properties
as antenna position.  Focus, I think I'd be inclined to just flag on focus
in motion, and not worry too much about location.  An alert if it stops 
motion without getting close to the commanded position would be in order.

3.  L305 oscillator out of lock.  (I wouldn't flag on MIB not responding,
though - MIB isn't needed for correct operation.)

II.  Baseband dependent flags

There are four or eight basebands generated and returned to the CB,
depending on the settings of the switches in the T304.  Which of these
are actually processed by the correlator would, in a reasonably structured
world, not be known by the Antenna.  So, most reasonably, the Antenna
would cause flags to be issued for all basebands, and forwarded to the
DCAF, which surely knows what basebands it is writing out, and can 
associate flags appropriately.  Operators looking at flags, though,
might get confused if we are flagging something that isn't being used.
This is solved for the VLA by passing the flags through the TELCAL
equivalent on the way to the operator displays.  It isn't solved for
the VLBA, but then the oscillators there seem to be very reliable, so
the question doesn't arise much.

1.  L301-1 out of lock.  This may affect all, none, or some of the basebands,
depending on the LoIfSetup.  The EvlaAntenna, of course, knows this 
information.  But how does it get to the user?  I see four possible ways:
a) the EvlaAntenna receives the alert from the device and analyses it
and generates a new alert (flag) saying what basebands are flagged bad;
b) the Subarray broadcasts the connectivity to flag users telling how to
interpret the alert; c) flag users, when they receive the flag, ask the
antenna for its baseband setup; or d) the EvlaAntenna tells the L301 what 
basebands it is helping generate, and the L301 includes that information 
in its alert.  The last has a minor advantage in that if the L301 knows it 
is not being used for any baseband, it can tell a technician working on it
that he is free to muck with it without disturbing an observation.

2.  L301-2 out of lock.  Essentially the same as L302-1.

3.  L302-n out of lock.  Relatively straightforward compared to L301,
but the same sort of thing.  L302-1 generates basebands A1,C1 in three 
bit mode or A,C in eight bit mode.  Similarly L302-2 generates B1,D1
or B,D.  L302-3 and L302-4 are used only in three bit mode, and generate
A2,C2 and B2,D2 respectively.

4.  No IF power.  Straightforward.  Which baseband(s) is affected can be 
got from, for instance, the MIB name issuing the alert.

5.  Deformatter out of lock.  There are three bits per deformatter.  In
three bit mode, each bit corresponds to a baseband.  In eight bit mode,
if any bit is set, it is best to flag the single baseband being handled
by the deformatter.  (Not clear to me whether the deformatter knows it
is in three or eight bit mode - the current deformatters only do the
eight bit.)   I guess the information about which IF is affected is
encoded in the deformatter CMIB name.

6.  No baseband power (from T5 Device).  Should never happen by itself
except when a cable is unplugged, but still worth checking.

7.  Varying switched power in T304.  I regard this as a transition flag.
The two usual causes go away eventually.  They are interference (WIDAR
is designed to live with this) and integrator recovery when you get on
source (T304 will have a boxcar, rather than an exponential, integrator
(or at least approximately a boxcar)).

8.  Absence of switched power in T5.  Again, obviously a transition issue.

9.  Varying power in T5.  Ditto.

Although the presence of switched power is a powerful diagnosic, I rather 
think we don't want to have the T304 flag on it.  The switched power there
can do strange things with interference present, and the equivalent, but 
much richer information is in the WIDAR data.