LOFAR: first international fringes

Please note the press release of the MPIfR and the one of ASTRON. Our findings are also featured by the EVN newsletter.
Some more details are shown in a presentation.

Fringes between the first international LOFAR station in Effelsberg and stations near Exloo in the Netherlands have now been detected after several unsuccessful attempts. The long-awaited signal was finally found in observations of the source 3C196 taken with the low-band system (LBA) on 20 August 2009 for the "First Imaging Busy Week" (George Heald et al).

Neal Jackson and James Anderson were the first to claim seeing hints of a weak signal in this data but were not sure about its true nature. This motivated me (Olaf Wucknitz) to have a closer look.

First delay spectra (28 August 2009)

There were indeed consistent (but noisy) phases visible in parts of the data (frequency averaged over subbands), but these alone could also be caused by other effects. One week after the observations I produced the first delay spectra of NL-NL and DE-NL baselines (one subband around 45MHz, about one hour of data 11:15 to 12:15):

NL-NL baseline DE-NL baseline

Higher resolution delay spectra (30 August 2009)

It is rather obvious that there is a signal at zero delay on both baselines, but it is much stronger on the shorter Dutch baseline. However, it is suspicious that the maximum signal is located at a delay of almost exactly zero, since we would expect some ionospheric delay that is not corrected during the correlation. This can be investigated further with the following plots that have been prepared with higher resolution in delay-space (again for NL-NL and DE-NL baseline, respectively):

NL-NL baseline DE-NL baseline

Please click on the images to see a postscript version that also shows some other subbands of the same observations. These plots (as the ones above) were produced by taking one subband (195kHz total, 256 channels of 763Hz each), coherently averaging 1 to 32 of the 5sec integrations, Fourier-transforming to translate into delay space and subsequent incoherent averaging of the power as function of delay. Again the international LOFAR signal is considerably weaker than the Dutch one. More importantly we see that the delay is not exactly zero but around -1 micro-sec, about the value expected from the ionosphere and clock offsets. We also see that the signal gets weaker for longer coherent averaging intervals (the lower curves) because the non-zero phase rates cause decorrelation. Still it is good to see that the signal is still clearly visible even after averaging over 160sec. The ionosphere seems to behave well even during daytime.

Delay/rate spectra (31 August 2009)

We can take the rates into account by producing two-dimensional delay/rate spectra. In these plots a peak should appear at the corresponding delay and rate, while its strength would automatically be corrected for the two effects. The movies below show these spectra as a function of time. Each time-interval consists of 64 of the original 5sec integrations. A different subband around 69MHz was now used.

NL-NL baseline
A postscript plot with flux, delay and rate as function of time for this NL-NL baseline can be found here.

DE-NL baseline
A postscript plot with flux, delay and rate as function of time for this DE-NL baseline can be found here.


It is rather obvious that we do indeed see an astronomical signal. The variation of flux with time is also consistent with the structure of the source 3C196. Once we take into account delays and rates, we even find that the signal on the international baseline is not that much weaker than the pure NL-NL signal as we originally thought.

Later results: multi-band fits

Later I produced movies and plots in which I either averaged incoherently over all bands or fitted for consistent multi-band delays to all subbands simultaneously. The latter also provides much higher resolution in delay and even allows to distinguish between dispersive (ionosphere) and non-dispersive (clock errors, position errors of target or station) delays and rates. The parameter space gets quite large when fitting for everything, so the computations take "slightly" longer.
Here is a movie of coherent multi-band delay/rate for a DE-NL baseline: DE-NL baseline multi-band
Note that the scale in delay (but also rate) is now very different from the previous movies. We also see that the response is slightly dirty in the delay direction. This is partly due to the fact that I only fitted for a non-dispersive delay in this case.
By varying all four parameters (dispersive/non-dispersive delay/rate), I produced the following plot:

DE-NL baseline multi-band
Please click on the image to see the postscript version.

What do we learn? The flux curve does now actually look like a part of a sinusoidal beating pattern. The delay has two contributions, a constant non-dispersive part due to a clock offset and a time-variable dispersive one due to the ionosphere. Both are at the expected levels. The rate, on the other hand, is clearly dominated by the dispersive ionospheric part. The clock offset is pretty much constant.

To my homepage
To the AIfA
To the University of Bonn

This document last modified Wed Feb 10 15:42:45 UTC 2010