K = "efficiency factor" The factor, K, is slightly greater than unity for reasonable "samples per row" (DFBn_S_nsamp) factors like 2048 or 4096, and allows for the auxiliary tables and header info included in the CFITS files. To a first approximation it can be set to unity, with errors of a few percent. (However K can be very large if you accidently set samples/row to 1 or some other low value, leading to files bloated with header info). If you want more accuracy than this then a little more work is required.