View Single Post

Fast Moderate Wides, Focus Shift and Pixel Peeping
Old 04-12-2015   #1
Registered User
ferider's Avatar
ferider is offline
Join Date: May 2005
Posts: 11,249
Fast Moderate Wides, Focus Shift and Pixel Peeping

I love my 35/1.4 Nokton and 28/2 Ultron lenses, even on the almost new to me M240.

Now, plenty of reviews have questioned their usability on digital Leicas due to "focus shift", after testing the lenses against a tape measure or flat focus target.

Annoying for me, as in 9 out of 10 photos, these tests are not representative of what I do: focus and recompose.

Let me explain: typically, when shooting a portrait, I use the RF patch to focus, and then recompose. Get the eyes out of the center of the picture, or similar.

A picture says more than 1000 words, here is what then happens geometrically:

- I fix the focal plane at the object distance "o"
- I rotate the camera to put the focus object out of center. For example using the 1/3rd rule. Then I shoot.

By doing this, I also rotate the focal plane of the lens (blue vs. red in the picture), and introduce a focus error, "d". This error is roughly proportional to the square of the angle of rotation. For normal and longer lenses, that angle is very small. For very wide lenses, there is no problem, since the DOF covers any error introduced. Well, unless you use a 21 Summilux, maybe, which I don't .

Typically, the largest focus error is introduced for fast (shallow DOF) moderate wides, 35 and 28mm for example. The above picture shows the calculation for both focal lengths. Now consider that a 28/2 at min focus has a DOF of 7.22cm. A 35/1.4 at the same distance has a DOF of 3.23cm (see In both cases, the focus error introduced by a simple rule of 3rd composition can be roughly half the DOF for 1/3rd composition, and worse when you rotate the camera more.

How does that look in practice, say, for the 28/2 Ultron ? The following are 100% crops of test photos when my copy of this lens is focused in the center, at f2.0, 2.8 and 4.0:

Clearly the lens shifts, and for this test, it is optimal at f2.8.

Here the same test, but - after center focus - tilting the camera so that the object touches the corners of the Leica's 90mm frame-lines (roughly rule of 3rds):

You see, that by recomposing, the lens is now optimal for f2.0. Which suits me fine.

Note that the 28 Ultron seems to have a nice flat focal plane, so the calculated numbers fit. Lenses with more field curvature will behave differently. I have lenses where the field curves back, and some lenses where the field curves forward, either making the above effect smaller or larger.

The morale ? Numbers are just that, they might not say anything about how a lens performs in the way you use it ... look at real usage examples, or try before you buy.


  Reply With Quote