Friday, September 1, 2023

Digitally emulating a 30mm Hasselblad XPan

I met up with a good photography friend this last week and was suitably envious when he brought out his 'new' XPan (Mk1) system, in near mint condition, that he had just  purchased with a full set of lenses (this picture is not of his kit).

As I can't justify the cost, some 6-8,000 pounds, purchasing such a camera system, and don't shoot film; I thought I would see how I could digitally emulate his XPan, fitted with the widest XPan lens, a 30mm.

As it would be near impossible to fully emulate the XPan, I decided to focus on two attributes:

  • Getting close to the image size of 65mm x 24mm
  • Getting close to the Angular Field of View at infinity, which is around 94 degrees, horizontally, using a 30mm XPan lens

As I also wanted to create something different, I decided to use the EOS M mount as my digital back, as I have full spectrum, visible and IR M mount cameras: several Ms, an M10 and an M100. All my M mount cameras are converted by Alan Burch.

My starting point is an EOS EF to M RhinoCam Vertex Rotating Stitching Adapter, which allows me to create a 7780 pixel square, stitched, 'quad'. Note the adapter is not electronic.

To provide me the greatest flexibility in focal length optimisation, I decided to use my EOS 24-105mm, F/4L, locked by using the lens twist technique, to the required aperture. In the test I decided to use f/8, which on my IR converted M100, was a little too stopped down, ie diffraction.

Mounting the above on an available nodal rail, a pano, clickable, rotating base, and an Acratech reversed ballhead for levelling, resulted in the following set up.

The position of the infinity, no-parallax, entrance pupil at 30mm, in blue below, was estimated from the PhotonToPhotos optical hub.

Using a field/angle of view calculator from scantips gave me a hint that I should rotate the 30mm lens twice by 30 degrees, which gave me sufficient overlap of the quads, and got me close, but wider, to the XPan 30mm FoV of 94 degrees. That is, ignoring pupil magnification, an infinity AFoV in degrees of around 2*30 + (360/pi)ATAN(quad-sensor-width/(2*focal-length))

After stitching the 12 images, of three quads, in Lightroom, I was then able to extract/crop out a 617 image with about the right field of view of a 30mm XPan.

The proof of principle (IR) test capture, with my converted M100, was, as usual, in my garden. In this test I should have spent more time refining hyperfocal based, infinity focus, but as I was only establishing the workflow, that wasn’t too important.

One of the uncorrected IR quads (7820 pixels square from four rotations of the adapter, that is simulating about a 29mm square sensor) looked like this:

After stitching in Lightroom, using one of my IR profiles, carrying out some toning and a hue-based 'channel swap' etc, the resultant 617 image, 15708 x 5544 pixels in size, looked like this:

If we assume the Canon M100 pixel pitch is 3.7 microns, the resulting 'XPan emulated digital image' represents a sensor some 58 x 20 mm in size, a little bit short of the XPan, but still pretty impressive.

Obviously the above is not really a digital equivalent of an XPan. The need to stitch 12 images is the real problem, but, IMHO, it's not a bad emulation.

Bottom line: if you don't have 6-8 thousand pounds burning a hole in your pockets, with a few pieces of cheap equipment you can get close to digitally emulating a 30mm focal length XPan, using a cheap Canon EOS M mount camera.

As usual I welcome any feedback or comment on this post, or any of my posts.

No comments:

Post a Comment