I am being stupid. Either that, or less likely, the documentation in PDL::IO::FITS is wrong.

I have a FITS image in which each pixel is a 32-bit float. That is, BITPIX == -32. I wish to convert it to 32-bit fixed-point format, where BITPIX == 16.

This is my script.

#! /usr/bin/perl

use strict;
use warnings;
use PDL;
use PDL::IO::FITS;
 
my $x = rfits('foo.fits');
$x->wfits('bar.fits', 16);

And here is what happens:

pcl@nut:~/Astro/Code$ imhead foo.fits | grep BITPIX
BITPIX  =                  -32 /
pcl@nut:~/Astro/Code$ perl to_int.pl
pcl@nut:~/Astro/Code$ imhead bar.fits | grep BITPIX
BITPIX  =                  -32

The output file is not converted to 16-bit unsigned format, despite what the documentation appears to say.

What am I not understanding and how do I get data conversion?