# Regenerating the imaginary data of a 2D NMR spectrum with a hilbert transform

2D spectrum take up a large amount of hard disk space. At my last lab it was standard practice to delete the imaginary portions of the spectrum with xfb n (We used TopSpin) and if needed later, to regenerate it from the Real part of the data with a Hilbert transform (Related, the TopSpin Commands and References, ~pg P-233 in version 2.1.2)

My first question: If the raw data is still stored in the ser file (TopSpin Commands and References v2.1.2, pg P-8) why can't you just reprocess the raw data? Is all the data not there? (I suspect this somewhat, as those files are much, much smaller then the processed files, and TopSpin processes 2D spectra as it acquires them, unlike 1D spectra)

My second (highly related) question: Is there a downside to regenerating the imaginary data from the real data? Is it a perfect replication of the original data, or is it an approximation? Why do we ever bother to store this data if we can recreate it at any time?

-

If you have the raw data (the FIDs in the ser file) still available, there is no reason to use a Hilbert transformation, you can just process with xfb instead of xfb n again. xfb n also doesn't really discard the imaginary part, it processes the spectrum from the raw data again, but doesn't save the imaginary processed data.
The ser file contains all the recorded FIDs, that data is sufficient (together with the parameter files) to recreate the full spectrum. That your Topspin processes 2D spectra automatically is just some specific configuration, it doesn't do that by default.