This is a long and winding trail, as is usual when something materialises out of happenstance - and so I am not sure I have the details correct.
I am interested in searching for transient radio bursts and - more recently with the prospect of having > 1300 hours of observation filterbank data sitting in front of me - I was interested to know whether the work to get Heimdall (searches for transients) running here at HawkRAO was worth it.
To help decide, the Astropeiler Stockert team very kindly downloaded a 3.5 GB filterbank file (circa 2017) of mine and ran it through Heimdall. They reported that the file was processed OK and sent candidate files. One was the whole Heimdall enchilada (>3000 candidates !!!) and a second one which was a curated list (~ 30 candidates) after eliminating obvious RFI of various kinds.
Some HawkRAO code was written to produce a dynamic spectrum for each candidate in the lists. Nothing remarkable turned up in the curated list - but I thought "Hey - why not run it on the >3000 candidate list ?" After doing this I viewed the spectra (all 3230 of them) by eye and - to my surprise - found about 1/2 dozen dispersed pulses - e.g...
The dispersion was estimated a DM=47.46 using HawkRAO code - but when the Astropeiler team was informed of the discovery of the dispersed pulse, they firstly confirmed the dispersion on one of the pulses (at ~1730 seconds) using psrplot...
...and then ran prepsubband/'single_pulse_search.py' on the filterbank file ...
...which clearly shows 3 dispersion 'christmas trees' at about DM=52.
They also supplied a 'exploredat' view of the pulse at 5533 seconds...
...which looks remarkably like the results from Crab giant pulses. In fact, one of my professional mentors suggested it was actually the Crab pulsar in a sidelobe - but the DMs don't match (Crab DM=56.8).
At this point the Astropeiler team rescued me from PRESTO 'installation hell' and so I could do my own PRESTO analyses (the installation procedure the I wrote 2 years ago no longer works - 3 different sets of errors on three machines...).
I looked at a file recorded 310 days later - but this time there was a major peak at DM=42. However, when I restricted the time and SNR ranges I found that underneath the other signals there was a clear indication of a pulse at DM=52.
Terrestrial RFI ?
The most probable source of these puzzling pulses is a terrestrial RFI - but I am mystified as to what terrestrial source would produce single ~0.5 ms pulses at a rate of 1 to 2 per hour with a dispersion DM=52. Chirp radar would surely produce a train of pulses - not single ones separated by tens of minutes.
Cautionary Tale ?
At the very least this is a strong lesson as to the ability of RFI to convincingly mimic cosmic signals and the need to verify any claims by a wide range of tests - and especially via professional applications. As a writer of 'home-made' code myself I know the real danger of manufacturing signals out of random noise using our own code.
At this point in time I had no idea where this all would lead to.
Are these pulses terrestrial RFI ?
Are they of cosmic origin ?
Would some turn out to be Crab Giant Pulses (GPs) ?
A clearer picture should emerge as the rest of the 1800-odd hours of observations is processed...
This requires coding of bespoke software to search, curate and analyse the archived filterbank data.