[Psrfits] Proposed (small) changes to PSRFITS definition

[Matthew Kerr]

Hi All,

Related to the "NSTOT" idea --

In fold mode data, it's not currently possible (I don't think) to
consistently determine how much data is in a file in the sense of being
able to accurately compute the radiometer noise level.

For DFB data, one can recover it from DAT_WTS as they all are set to 1
initially and no online RFI zapping is performed.  However, I think this
behavior isn't guaranteed.

For dspsr instruments, the weights don't seem to be normalized in this
way.  (Presumably they account for RFI zapping etc. in a relative sense,
though.)

I'd propose something like "NSAMP" for each subint -- a dimensionless
number giving the product of the total (actual) bandwidth and observing
time in the subint, i.e. the number of field samples.  Then, provided the
DAT_WTS are at least proportional to the actual channel bandwidth, this
number can be updated through scrunching and zapping operations, and one
can always accurately determine the radiometer noise floor.

If DAT_WTS couldn't be used in this way (e.g. if one had S/N weighted?),
another option would be to introduce a DAT_NS array shaped like DAT_WTS but
giving nsamp for each channel & subint.  A cost of 256 KB for a typical
64-subint, 1024 channel observation.

Accounting for npol could either be left as an exercise for the user or
incorporated in the above.

Cheers,
Matthew


On 3 May 2014 02:39, Willem van Straten <vanstraten.willem at gmail.com> wrote:

> Hi Dick,
>
> LEVSEPN and NSTOT sound like very sensible additions to me.  OBSDESCR also
> sounds like a useful way to log the unexpected and anomalous.  Also,
> patterns might arise in the use of this free-format text that lead to the
> identification of new, useful parameters that are added to the definition
> (and can be parsed by automated tools).
>
> Cheers,
> Willem
>
>
>
> On 2 May 2014 12:32, <Dick.Manchester at csiro.au> wrote:
>
>>  Hi all
>>
>>  Partly as a result of some discussion about encoding and decoding of
>> 2-bit data, George Hobbs has requested that the PSRFITS definition include
>> a specification of the digitisation. I'm proposing a new version of the
>> definition (v5.4, attached) which includes a parameter "LEVSEPN" in the
>> Digitiser Counts table. This is defined as the separation of the
>> digitisation thresholds in units of the input rms value. Also at George's
>> request, I'm proposing a new binary OBSDESCR table, which is simply
>> free-format text where any information related to the observation or the
>> processing can be stored, and a new parameter NSTOT in the SUBINT table
>> main header, which is the actual number of valid samples in a search-mode
>> file (the last subint is often incomplete).
>>
>>  Please send any comments (using "reply all") on these changes or other
>> suggested changes to the definition by Tuesday May 6.
>>
>>  Regards
>>    Dick
>>
>>
>> _______________________________________________
>> Psrfits mailing list
>> Psrfits at lists.pulsarastronomy.net
>>
>> http://lists.pulsarastronomy.net/mailman/listinfo/psrfits_lists.pulsarastronomy.net
>>
>>
>
> _______________________________________________
> Psrfits mailing list
> Psrfits at lists.pulsarastronomy.net
>
> http://lists.pulsarastronomy.net/mailman/listinfo/psrfits_lists.pulsarastronomy.net
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.pulsarastronomy.net/pipermail/psrfits_lists.pulsarastronomy.net/attachments/20140505/f8965684/attachment-0001.html>