First fringe plots showed delays corresponding to several positions in the field, but nothing exactly at the phase centre.

These plots show the delays for the baseline TS001-UK608:

All baselines to TS001 can be found in the PDF version The strongest signal clearly has a non-vanishing delay.

Because the pulsar should be bright enough to see it clearly, we investigated this further. Adam Deller found that the phase centre used in the observations was off by ~7 arcmin, which on the longest baselines corresponds to about 8 microsec.

We phase-shifted (with NDPPP) to the correct position and made new plots:

All baselines to TS001 can be found in the PDF version Now the delay of the strongest signal is consistent with zero as it should.

Adam also found that UVFIX in AIPS should not be used to shift the phase centre. It re-calculated the UVW coordinates using (probably) different conventions and seems to mess up the result in general.

This is a data set without international baselines that was taken to calibrate the M51 observations that David Mulcahy is working on.

Andreas Horneffer converted the demixed and flagged data to circular (with both converters) and calibrated with BBS using a point source as sky model. Only the diagonal elements were calibrated. BBS did not crash or complain about the unexpected polarisation basis!

For the imaging, an upper baseline length limit of 6km was applied.

This is the version converted with Tobia's converter without applying corrections for possible X-Y offsets:

There are significant artefacts around the target, possibly caused by X-Y offsets that corrupt the conversion to circular.

Andreas determined the X-Y offsets for baselines including at least one remote station using Olaf's script (results here) and converted again to circular with the correction applied:

It is believed that the differences here are not due to the use of different converters but really result from the correction for X-Y offsets.

This source is nominally about 10Jy (although in this case the noise is likely dominated by source noise). A series of experiments were undertaken to try to find out how faint a source can be successfully fringe-fitted; this was done by adding different amounts of noise to the observation and, for each iteration, phasing the superterp, converting to circular polarization, and reading into uv-fits for fringe-fitting.

1) Tripling the noise (source flux → 3Jy) produced a dataset which could be fringe-fitted without problems using a solution interval of 1 minute.

2) Adding an extra factor of 12 in noise (source flux → 1Jy) produced a dataset which could just be fringe fitted (one station failed; the fringe fit fell apart if the uvrange was restricted to 80-1000km)

3) Adding a factor of 50 in noise (source flux → 200-300mJy) produced a dataset which could not be fringe fitted.

After fringe fitting an attempt was made to map the data using the Dutch baselines only. However, this was not very successful probably due to the restricted u-v coverage which is more or less entirely in the N-S direction (cf. the map made by Olaf Wucknitz with a fuller dataset).