Done some research - see first 2 attachments. Now convinced fenbw = 1/2T.
The third attachment is a variation on the same theme and I get the same result i.e. 1/2T ? Which I think should be 1/T, but I cannot see were I am going wrong in the attached derivation ? This time the frequency response of the practical filter is the same at that when you multiply the rectangular pulse of height h and width T, by a cosine of amplitude a, and frequency fc.