- Putting in an alternative FITS output format to RPFITS! Not because RPFITS is so bad, but because ATLOD for AIPS is so bad. FITS-IDI functionality is planned sometime - but if anyone wants to contribute...
- Turning on polyphase filterbank functionality - at the moment this is an option in the input file but not implemented in the code. I had polyphase filterbank functionality working an earlier, pre-MPI version of the code, so it shouldn't take long to put in and test. Let me know is this is a critical functionality for your application.
- Documentation for the DiFXGUI source code so its more easily extensible.
- Developing an intermediate step to enable reading directly from Mk5 machines - this would require something running on the Mk5 I think. We mighht be able to use the eVLBI reading from a socket approach, if the Mk5 could squirt data out directly into a socket, and this might be an easier way.
- Optimising the fringe rotation when linear interpolation is used. Presently this still uses a sin/cos to get the complex fringe function, but this could be more efficiently implemented with a complex multiplication, especially for large numbers of channels.
- Optimising the interpolation of delay points in Datastream - this has the potential to become a bottleneck at high datarates.
- Allowing continuum experiments with small numbers of channels to be more efficiently implemented using and optimal length FFT (eg 512) and averaging channels before writing to a FITS file.
- Allowing essentially the above point for multiple sources per integration - letting you reduce the output data size if you have several known sources in your beam and want to image them separately.
- Creating a version of the gencalc_delays model generating program which uses the VTD library (http://vlbi.gsfc.nasa.gov/vtd/) as an alternative to CALC, and comparing accuracy.