Epoch times in *.inf files

[aaronpearlman]

Hi @scottransom,

I was wondering the reason why a dispersive delay time is subtracted from the start time of the observation when running prepdata with a non-zero DM. This only happens when the barycentering flag is enabled.

This seems to suggest that you would have different barycentric start times (given in the *.inf files) for different DMs. If this were true, then you should see a slope in the DM vs time plots for say a single pulse detected at multiple DMs (assuming dispersive effects are negligible).

Since you don't see this behavior, should the start time for observations de-dispersed at non-zero DMs really be the barycentric time output by TEMPO with DM = 0.0?

Thanks,
Aaron

[scottransom]

Hi Aaron,

In general, if you barycenter data, you do it because you want the data to be at some basic inertial reference frame so that you can compare other data to it. By correcting for the dispersive time as well, you make that even stronger as all frequencies can then be referenced to the barycenter.

As for your 2nd paragraph, yes, you are right, you should see a diagonal line in the DM vs time plane. And you do! As long as there is enough dispersion (which is often not the case), and you zoom way into the single-pulse plots, there is definitely a diagonal line of pulse arrival times.

For topocentric data, we usually want to keep the data as close to "raw" as is possible. That allows us to use (for example) TEMPO on TOAs made from such data.

Hope this helps,

Scott

[aaronpearlman]

Hi Scott,

Thank you for the quick reply - very much appreciated.

Correct me if I am wrong, but I still believe that PRESTO's single pulse search and DM vs. Time plots ignores the different start times in the *.inf files if you de-disperse with different non-zero DMs. The diagonal line of pulse arrival times you are describing would only be due to dispersive effects (see my example below).

I am interested in this because this leads to two important questions:

1. How would you go about assigning a ToA to a single pulse if you don't have an accurate measurement of the DM of the object (or if the DM of the object changes)? If the barycentric time in the *.inf files is indeed the start time for the de-dispersed time series, you will get different ToAs depending on the selected DM.

2. If the ToA of a single pulse is tied to the DM, then why does this effect seem to be ignored by PRESTO's single pulse search?

To illustrate what I'm finding, here's a bright single pulse from Vela taken at S-band. The time sampling of the data is 256 us and the nominal DM is ~68 pc cm^-3, which would produce a dispersion delay of ~5 ms across the bandwidth shown, which you should see across ~19 time samples. This is indeed the case for the bright single pulse shown below:

bright_vela_sp.pdf

Now, I de-dispersed the data with DMs from 0.0 to 300.0 (including the nominal DM) in steps of 10.0. I performed a single pulse search and filtered the lists so that they only include this candidate. PRESTO's DM vs. Time plot would look something like this:

dmtime_vela.pdf

You can clearly see the diagonal drift in the ToAs at the various DMs, but this drift is due to the smearing of the pulse at the various DMs. The difference in the reference times is ignored here. This seems to be because the "Time" axis is the sample time relative to the start of each de-dispersed *.dat file.

If I plot the times in the *.inf files relative to the reference time from de-dispersing with the highest DM, you see that this different drift (due to different start times) is an order of magnitude larger. The relative size depends on the choice of DM and the frequency at the top of the band.

tstart_vela.pdf

These times are determined by prepdata by subtracting the dispersive delay (for each DM) with respect to the top of the band relative to infinite frequency from the barycentric start time given by TEMPO with DM = 0.0.

I would have thought that the DM vs. Time plots would have incorporated the effect of different reference/start times for each DM. I'm also curious what you recommend for handling the issue in (1).

Also, I still don't understand why this DM delay is subtracted. I would have thought that PRESTO skips/shifts the data in each channel by a small amount to ensure that a pulse is accounted for regardless of the DM. However, the length of the time-series is always the same regardless of the DM used for de-dispersion.

Thanks,
Aaron