[evla-sw-discuss] pulsar data

[Michael Rupen]

Hi Brent,
   nice to have you back!

> 1. There is indeed one pulsar timer per baseband pair, meaning 4 timers in 
> total.  Each of these can be set for a different pulsar ephemeris.  The 
> restriction is that all of the correlations happening within one sub-band 
> correlator *must* use the same ephemeris.  With 16 sub-band correlators, this 
> leaves plenty of flexibility to track 4 pulsars simultaneously.

So the four timers implies a maximum of 4 pulsars, each of which would have
to be observed with a different BB pair.  The additional restriction on
sub-bands implies that in this mode one would NOT obtain a full BB pair's
worth of bandwidth on each pulsar, but would have to split the 16 sub-bands
amongst them, e.g., 4 sub-bands on each of the four pulsars, or a 8-4-2-2 
split, or some such.  Is that right?

I fear I confused the issue with all that talk of phasing the array.  Since
you get all the correlations in all the bins, there's no need to phase the 
array.  I was thinking of too many things at once :}

Does one get an "untimed" bin at the same time as all these other bins --
i.e., the regular correlator output?  or can one set things up that way?
I'm thinking of doing a time-binned pulsar observation, while simultaneously 
making as deep an image as possible of the unpulsed sky in that direction.
I know one could just sum up all the "boring" time bins, but that costs
a lot in terms of output data rate.

> 2. Pulsar gating and pulsar phase binning are indeed two different modes, and 
> operate as Michael describes.  Given the pulsar phasing capability, I can't 
> imagine why one would want to use gating...but it is there with the ability 
> to generate different gates on different sub-bands to track dispersion.

I imagine this will be used mostly to avoid humongous data volumes.  Gating
gives one output data stream, time bins gives many, and if you want lots of 
channels the data rate restriction might kill you, esp. in the larger 
configurations and/or lower frequency bands (where you might want high
time resolution to avoid time-averaging losses and/or RFI).

> 3. Officially, the correlator does not support "bunching" of phase bins at 
> some particular phase of the pulsar.  However, the signaling to generate 
> phase bins is controlled by software, and so this is entirely possible...but 
> you should decide if this is indeed an important mode reasonably soon so the 
> real-time s/w people can program in the capability.

That's one for Walter!

I suspect that, even if we don't do "bin bunching" at the correlator, we
will want to be able to select/average the appropriate time bins at the 
output, to lower the data rate.

> 4. With my limited experience, having to get the pulse phase correct to track 
> the pulsar seems like an nightmare to me.  I think a better long term 
> solution, if bunching of phase bins is required and/or pulsar gating is 
> required, is a feedback mechanism whereby for the first few minutes (or 
> however long it takes to get a detection), phase binning is used with even 
> distribution of bins across the period to find the epoch, and then that epoch 
> is immediately subsequently applied for phase bin bunching or gating.

Yes, I was trying to suggest something like that at the end of my last 
message.  Get a few minutes to set things up, then zoom in on the interesting 
bit.  If we do this at the correlator output that might not be so bad; part
of the online system?

>
> 5. The 65536 phase bins mode is possible, however due to CBE and output 
> bandwidth limitations the time resolution would be on the order of 100 msec 
> (for all spectral channels...decreasing with fewer).


Cheers,

            Michael