[evla-sw-discuss] Getting pulsar parameters to WIDAR

[Walter Brisken]

I've added our in-house pulsar timing experts to this message against 
their will.  Unless they opt out, it might be good to keep them in the 
loop as they have the tools to quickly evaluate the adequacy of 
requirements we set forth.  Some quick responses below...

-W

On Thu, 17 Jul 2008, Barry Clark wrote:

> I'm afraid I don't understand things very well, but somebody should
> be talking about this, so here goes.
>
> I believe the standard for predicting pulsar timing is the program
> tempo, maintained at CSIRO and/or Princeton.  (There is some attraction
> to pulling out the guts and integrating them separately as we have
> done for CALC, but with obvious disadvantages as well.)

Princeton no longer has a pulsar group.  ATNF/CSIRO is the only maintainer 
now.  There are now two programs that are used for pulsar timing and 
predicition: tempo1 and tempo2.  The later is likely to completely 
overtake tempo in the next few years, if not sooner, but there will be 
holdouts.  tempo2 uses a par file that is mostly compatible with tempo1 
and will do the right thing if it is told which type of par file is 
submitted.  It is not always possible to determine which variety a par is 
by looking at it, and some fields use different units now, so I think we 
will either want to force users to submit tempo2 files or to force them to 
say which format the par file is.  Paul thinks the second option is 
better.

> I believe the thing an observer should supply is a tempo parameters
> set, which would need to be passed through the OPT (without it having
> to look at it, really).

For full-period pulsar binning this would be fine.  For gating or binning 
over a narrow phase window, this is very dangerous.  A par file for a 
typical pulsar will only be able to predict phase good enough for gating 
for about 1 month or so, and some experiments may demand better precision. 
There will need to be a mechanism for the user to submit par files on the 
day of observation, possibly with pulse phase adjustments determined by 
observations made the the VLA that same day.

> I think further processing needs to be done during or near observation.
> To the supplied parameters, we should add TAI-UTC, UT1-UTC and possibly
> the other earth-orientation parameters.  True, we could force observers
> to make corrections for these after observation, but to do so seems cruel.

Depending on the observing mode, these corrections may not be possible.

> This processing could be done at OPT_model => script conversion time
> or during observation.  Tempo's output is a polynomial.  Walter, can
> we make a polynomial that will last for an entire scheduling block of
> a few hours?  It looks close for some of the nastier doubles.  If not,
> it is perhaps better to run tempo at observe time, one observation at
> a time - seems less complicated than having the script writer supply
> several sets of polynomials.

For timing purposes, most pulsar folks use 15 polynomial terms over 30 
minutes for the most extreme pulsars.  I saw down with Paul Demorest and 
we played a bit.  It is clear that 12 terms over 60 minutes produces 
polynomials which are marginally insufficient for the most extreme cases 
(e.g. J0737-3039).  In any case, a single polynomial over even a few hours 
is likely to be insufficient for some interesting known cases.  How about 
just stating that each scan will have one polynomial and one could 
schedule multiple back-to-back scans with different polynomials if the 
duration is longer than a valid polynomial can be made?

> The polynomials output from tempo are in terms of phase and derivatives,
> not period and derivatives, as shown in the draft VCI.

This continues to be true for tempo2

> For transferring the model across the VCI, we can either send along
> lots of coefficients (up to a dozen, perhaps) at (sub)scan setup time,
> or have the AntennaPhysicals send along three coefficients to the
> station boards, packaged with the delay models, every 10 seconds or so,
> which looks adequate.

Would there be a limit at 12 terms?  15 terms is fairly common in the 
timing community.

Paul informs me that in folding high precision timing data, they typically 
take a 2 term polynomial (delay and rate) over 1 second intervals.  Would 
it make sense to do this as well?

-Walter (with advice from Paul)

> _______________________________________________
> evla-sw-discuss mailing list
> evla-sw-discuss at listmgr.cv.nrao.edu
> http://listmgr.cv.nrao.edu/mailman/listinfo/evla-sw-discuss
>