Hello Michiel,
Thanks for the encouraging remarks. I would still be interested in trying to do some VLBI with you after you try your smaller scale experiment and would interested to know what frequency band you would like to try.
Synchronization is not that hard as long as there are common strong satellite sources for both receivers to use as a reference. I tried a 30 km baseline at 50 MHz for instance, but without good references, I was unable to align signals, but at 140 MHz, many LEO satellites exist that I can use. From Europe, I think it will be challenging to find a common reference signal. Maybe a geosynchronous L or C band visible from both locations would work, but at these frequencies, astronomical sources are very weak and small dishes can only see the strongest sources at these frequencies. Another challenge is that we would need to observe a very compact source that won't get resolved out. For these and other reasons I would pick the 130-150 MHz band as a practical choice for amateur VLBI.
My thoughts are that for transatlantic VLBI, a good reference would be to use signals from the GRAVES radar reflected off the moon at 143.050 MHz. The narrowband, pulsing illumination of the moon by GRAVES should make synchronization relatively easy. I will check some of my latest interferometer data to see if I can detect the GRAVES narrow band signal with my 3 element beam that would be wide enough to simultaneously capture the moon and other astronomical sources.
For synchronization, I use a two step approach where I first do a rough alignment with a Python program that looks at power spectrograms from the two stations' IQ data streams. Stream delays of up to 7 seconds exist from startup between my two LimeSDRs on two separate computers. After the rough alignment is done, I do a fine alignment looking at a phase spectrogram of the cross correlated visibilities. So far, using Orbcomm and other LEO satellites, I note the tilting of the phase across the FX correlator passband and add or remove samples from one of the streams until the phase shift is constant across the passband. This web link provides the math theory behind the approach:
Once some high quality data streams have been captured from receivers in two locations, I think the big challenge will be to figure out how to come up with a sky model to 'fringe stop' the data. Even at 3km/140 MHz (1500 wavelengths) fringes come very fast and the sinc(x) function of a wideband interferometer requires that the FX correlator have fine frequency resolution (~100 Hz or 10,000 frequency bins with 0.05 time resolution). In other words, some computational heft will be required to crunch a large FX correlator visibility dataset.
I think a key pre-requisite for VLBI is to verify that fringes can be detected from a given antenna pair in a 'regular interferometer' configuration first with good S/N. Of course, a good GPSDO is also needed that has an OCXO and not a TCXO reference. For instance, with two Trimble Thunderbolt GPSDOs on an oscilloscope, I can see that they can hold 10 MHz signals in synch over ~100 second time frames. Short term stability is more important than long term stability. I think that will be good enough for VLBI at 130 to 150 MHz at any baseline length with adjustments/corrections made in the digital domain in post-processing.
David