Processing notes

These notes are to keep track of processing changes with the aim of obtaining the best possible rms timing residuals for our sample of pulsars.

The following table contains for each pulsar:
1. name
2. rms timing residual given on PPTA poster (in brackets - the rms in Hotan et al 2006)
3. rms timing residual after simple processing (running PPTAscript on most recent data): c20 = cpsr2 20cm, d20 = dfb 20cm, w20 = wbc 20cm etc.
4. Used John S's most up-to-date zap file and fixed the DFB jump
5. Improved the standard profiles

(More information: PPTA Summary Table)

Test 1:

(corresponds to column 4 in the above table).

• Fix DFB jump in most recent data

Up to now, the differential delays in the dfb between configs with different numbers of channels have not been compensated for. On 7 Aug (MJD 53954, also indicated by PROC_CMD in the HISTORY table being set to "PSRDFB:2006-08-07_14:55", or later) these differential delays were taken out in the on-line program. N.B., this will affect all TOAs created from PSRFITS files generated after this date. Integration start times (STT_OFFS in the main header) have been moved to earlier times by the amounts shown in the following table:

32-channel configs: 1.2969e-6 s
128-channel configs: 3.5547e-6 s
512-channel configs: 12.2813 e-6 s
2048-chann configs: 46.8829e-6 s
This has been fixed in the script by adding these time-offsets into DFB data prior to 7 Aug in the PPTAscript. The raw data have not been changed.The software to do this is called fixDFB.

For J1909-3744 DFB data, the DFB rms at 20cm was 4us and is now 0.7us. At 10cm it was 4.5us and is now 1.0.

We also fixed the timing models. For 0613-0200 we got a large improvement by fitting for PMRA/PMDEC.

1024-0719 seems that the fitting algorithm for weak pulses locks onto the wrong pulse profile component (i.e. the points are well away from the zero residual and the errors are relatively small). For now we have deleted these observations. Fitting for PMRA/PMDEC significantly reduces the timing residuals. The DFB data seems to be giving an rms ~0.9us.

1045-4509: with the DFB we are getting (after fixing up the offset) about 1.7us which agrees with the poster value. However CPSR2 data doesn't look as good.

1600-3053: get a huge improvement by fitting for PMRA/PMDEC. The best fit (with DFB/not corrected yet) is 0.968us - the DFB data only goes to May 2006 - where has the rest of the data gone?

0711-6830: the DFB data alone gives much better fitting than CPSR2 of ~1us.

1603-7202: For this pulsar the DFB data alone (after deleting the points with a jump) gives an rms of ~1.7us

1022+1001: DFB data alone (after deleting points with a jump) gives an rms of ~1.4us

1824-2452: This is a bit strange. We need to add in an F2 and F3 (but don't fit) from the literature (from Cognard and Lestrade 1997) and the rms significantly reduces to about 0.8us! The earliest DFB need to be deleted - reason unknown

1643-1224: Get a huge change when we use the DD model instead of the ELL1 model

Test 2:

• Only include DFB data after the levels have been fixed
• Throw out bad WBC data
• Run full zapping using John's scripts
• rfi_clean
• Try polarization TOA fitting
• Use analytic templates
• Correct for DM
• Average data-points together to get average over 2 week intervals

J1939+2134

The timing residuals for this pulsar shows a lot of timing noise. The early DFB data were corrupted and need to be deleted. The WBC profiles have a dip before and after the main pulse. Fitting for DM corrections, XP can get timing residuals at 20cm of 200ns over a couple of years.