(Please click on the image to see the movie for the 6 hour observation. This used all 248 subbands.)
Results of the delay/rate fits are shown here:
(Please click for a postscript version. All 248 subbands were used.)
We see a non-dispersive delay of -0.3 to -0.4 microsec. This seems to be a combination of a clock offset (less than in Effelsberg, where we typically had about 1 microsec) and maybe a small error in the station position. Whether the small slope of the curve is due to clock drift or the station position still has to be decided. The ionospheric (dispersive) delay varies between 0 and about 0.25 microsec, not too bad! Phase rates are also well-behaved, they are generally below 10 mHz. (Dispersive numbers calculated for the middle of the band, ca. 55 MHz.) The increased noise around 22:40h is a result of the weak signal at that time. Because XY/YX are consistently higher than XX/YY, swapped polarisations seem to be the more plausible explanation. Faraday rotation would change with time (similarly to the dispersive delays, multiplied with the magnetic field). But, of course, it may just be a coincidence that Faraday rotation turns X into Y and vice-versa. In order to understand this better, I split the data into 4 frequency ranges, A-D, each 12 MHz wide and repeated the fits for them. Here I only show the flux of the peaks in the delay/rate spectra:
(Click for the postscript version with higher resolution, see page 1.)
We see that the ratio of XX or YY to XY or YX systematically decreases with increasing frequency (D is very weak). If we extrapolate to higher frequencies, we expect almost no XX/YY but high YX/YX. In this limit Faraday rotation would decrease, so that we would not expect real cross-polarisations. My conclusion: The polarisations really seem to be swapped somewhere. Differential Faraday rotation is clearly detected at the lower frequencies, but it cannot account for the high XY/YX at higher frequencies.
To get something like total intensity, I also added the power of all polarisations:
(Click for the postscript version with higher resolution, see page 2.)
Here I do not correct for different gains or projection effects of the dipoles. Nevertheless, the curves beatifully show the beating due to the double structure of the source. The fringe-spacing decreases for higher frequencies as expected. I don't know why the signal is so weak at the highest frequencies in D. This is probably a combination of lower sensitivity and source structure.
Check for possible updates at the LOFAR USG forum.