Robert has already stated that you can export each sheet as a csv (or some other delimiter if your data contains commas). Each CSV file can then be ftped or NDMed, or whatever, to the mainframe. You can then use your sort product to parse the data into fixed length fields. Examples abound - I know because I did this a couple of weeks ago using the postings on this, and other, forums.
The problem I have is that people can explain things quickly but I can only comprehend slowly.