Monday, November 2, 2020

Focusing a Tilt-Shift Lens

 WARNING: This post assumes a working knowledge of a Tilt/Shift lens, plus this post is aimed at DSLR Tilt-Shift users, ie not Technical Camera users

It seems everyone who gets a tilt-shift (TS) lens goes through the same baptism of fire: that is, how the hell do I use this damn thing!

In this and subsequent posts I will talk about how I use a TS lens, as I was not content with some of the broad advise that others have posted, eg "Focus for the Forground, then Bend (tilt) for the Background".

These posts are aimed at the TS owner who understands the basics, ie tilt angles and hinge heights, but wishes to get more out of their expensive lens.

To help understand TS lenses I wrote a TS simulator, that you can access on the right (TiltSim). Note it was written in DesMos, and thus requires you to give a name when you use it, but you don't need to give your real name and you certainly don't need to give any personal data. 

TiltSim is, hopefully, intuitive and easy to use, however, here are a few pointers:

  • Tilting and shifting is modelled;
  • Additional settings, ie focal length and CoC etc, are accessible in the 'Set Up' folder;
  • Camera tilt is modelled via tilting the ground plane;
  • A tilt of zero allows you to use TiltSim as a normal depth of field simulator;
  • The Object and Extend are there to help model objects in the field of view (which I'll cover in subsequent posts);
  • Focus can be achieved via the slider or moving the optical axis point of focus.

Let's first take a look at the TiltSim screen:

Here we see a non-tilted and non-shifted simulation of a 24mm lens, at f/8, focused short of the hyperfocal (the red vertical line). We see (green lines) the near and far depths of field, as well as the point of focus. The ground plane is parallel with the camera's optical axis and positioned 750mm below that axis. As I wish to achieve a high quality image, I've set the circle of confusion to 12 microns, ie about 2 sensor pixcels on my 5D3.

In order to demonstrate the non-tilted functionality, let's simply focus at the hyperfocal and see what we get:

Here we see, focused at the hyperfocal, H/1, the infinity blur is 12 microns, as expected. We also see that the near DoF is positioned at H/2. Note the red dots are positioned at H/5, H/4, H/3, H/2, H, 2H and 3H. 

Let's now assume we are simulating being in a cathedral and wishing to capture the floor from near to far (infinity). 

As we can, see, we can't really achieve that in a single image, as the closest part of the floor we can see is short of the near DoF. Without a TS lens we would take and additional focus brackets at H/3, ie to cover the focus shortfall. As shown here:

But, of course, we have a TS lens and thus we can achieve what we want in a single image, ie a perfectly focused image of the floor of the cathedral.

But how?

The way I do it is to use two simple 'rules':

  1. Set the point of focus (using zoomed-in LV) where the plane of interest, ie the surface I wish to see in perfect focus, passes through the centre of the optical axis. 
  2. Then use zoomed-in Live View to tilt the lens to ensure the rest of the plane of interest comes into focus.

Note: that the optical centre is only in the centre of the LV screen, if the lens is not shifted. If shifted, the optical centre of the LV screen could be, say, at the edge of the frame. I'll cover this in more detail in a subsequent post.

Sounds simple, but in the above case the floor is parallel to the lens: so where do we set focus? 

Simply where the plane of interest intersects the optical axis; which, in this case means setting focus to infinity, ie where the planes 'converge':

Here we see we are focused at infinity and the near DoF is, as expected, at H/2. Let's now tilt down.

But by how much I hear you say?

Let's guess 0.5 degrees:

Clearly this is not enough, but we see one of the key things to notice. If you are focused at infinity then the plane of sharp focus will be parallel with the optical axis, ie the floor in this case.

Rather than worry about the tilt angle and J heights, all we need to do is use LV, and zoom in to the closest part of the floor we can see in the Field of View, and tilt until the floor achieves the best focus:

Here we see that 1.8 degrees is needed to achieve what we are seeking, ie the floor in perfect focus from near to far. That is the hinge, also know as the J height, around which the tilted plane of focus rotates, is now coincident with the floor; and, because we are focused at infinity, the plane of sharp focus is parallel with the optical centre of the lens.

We also see another key piece of the TS focusing jigsaw, namely that, at the hyperfocal, red vertical line, the DoF, either side of the plane of sharp focus, is always positioned at J, ie the hinge height.

But can you do better than the above? As half our depth of field is 'wasted' below ground level.

Assuming you have been following my posts, you know the answer. Namely, we simply refocus at the hyperfocal:

Here we see that we are now maximising the available DoF, which is always 2*J at the hyperfocal, ie near to far or upper to lower.

BTW the green dots on the hyperfocal simply show the distance at J away from the tilted DoFs. Or, put another way, at a blur of twice the CoC.

Let's now assume we also wanted to capture the rest of the cathedral in the FoV. In other words we need to carry of tilted focus stacking.

Once again, assuming you have been following my blog, you know where to focus, ie at H/3 and at H/5.

That is the DoF 'rules' we use with non-tilted lenses, also work with a tilted lens.

In these nest two screen shots we see the tilted focus brackets to achieve what we are after:

An alternative strategy, as we started at infinity, ie H/0, would be to focus at H/2 and H/4:

Finally, for the Magic Lantern Canon uses, assuming you are using my DOFIS script, all you need to do is use DOFIS to tell you where to focus, eg:

  • Take you first image at tilted infinity focus, to get the floor as sharp as possible;
  • Then use the DOFIS feedback to tell you when to take your next image, ie as you focus away from infinity.
In future posts I will use TiltSim to show more complicated uses of TS lenses, however, the key take ways from this post, ie assuming the camera is not tilted, are:
  • The tilted plane of sharp focus always passes through the hinge point and the point of focus on the lens optical axis;
  • If focused at infinity, the plane of sharp focus will always be parallel to the optical axis of the lens and the near (upper) DoF will be positioned at J at the hyperfocal;
  • If focused at the hyperfocal, the far (ie lower) DoF will be parallel to the optical axis;
  • At the hyperfocal distance the near (upper) and far (lower) DoFs are positioned at J either side of the plane of sharp focus, ie the total DoF is 2*J when focused at the hyperfocal;
  • At the hyperfocal, at twice the J distance either side of the plane of sharp focus, the defocus blur is twice the CoC.
In future posts I will look at more complex use cases for TS lenses and how TiltSim (and DOFIS) can help you achieve optimum focus.

As usual I welcome any feedback on this post.

No comments:

Post a Comment