This is a continuation of a test of the following lenses on the Sony a7RII:
- Zeiss 85mm f/1.8 Batis.
- Zeiss 85mm f/1.4 Otus.
- Leica 90mm f/2 Apo Summicron-M ASPH.
- AF-S Nikkor 85mm f/1.4 G.
- Sony 90mm f/2.8 FE Macro.
The test starts here.
In my lens testing, I have been focusing wide open no matter the f-stop in use for making the exposure. SLR lenses are designed to be used that way, and the best ones don’t shift the focus point as you stop down. But mirrorless lenses aren’t necessarily designed to minimize focus shift, and there have been a couple of places in this test where it looked like the Batis and the Sony macro had a material amount of such shift. It just isn’t credible that the Batis is sharpest at f/2, for example.
It occurred to me that the motorized rail setup that I’m starting to use for the LoCA testing would be an ideal tool for three things:
- measuring the amount of focus shift in a lens
- providing, through automated trial and error, repeatable MTF results with top-notch lenses, a feat that had eluded me
- finding the best MTF for each f/stop
I mounted a Sony a7RII to the Cognisys computer-driven focusing rail. I set the controller up to use 156mm of travel and make 40 exposures 4mm apart. I set the assembly 3.3 meters from the target, mounted a Sony 90mm f/2.8G macro lens, and made a set of images at each whole f-stop save f/22.. I developed the sets in DCRAW using AHD, and had it convert the files to 16-bit sRGB TIFFs. I told Imatest to calculate the on-axis MTF50 for luminance for a horizontal edge, and plotted the results in cycles per picture height vs distance from the point furthest away from the target.
Here’s what I got:
The blue f/2.8 curve peaks at 1000 cy/ph. The red f/4 curve peaks at 1400 cy/ph 40mm closer to the target. If the camera were focused at f/2.8 and left there for an f/4 exposure, it would look like the lens could only do 1200 cy/ph. It gets worse. The gray f/5.6 curve also peaks at 1400 cy/ph, but 64mm closer to the target than the f/2.8 peak. If the camera were focused at f/2.8 and left there for an f/5.6exposure, it would look like the lens could only do 1100 cy/ph. F/8 is not as critical, but the peak is at about the same place as f/5.6, and the error is less than 100 cy/ph. By f/11 it doesn’t matter where you focus within the range of the rail.
As you can see, this calls into question all the MTF numbers I’ve already posted for this lens, and makes me suspicious of the numbers I’ve posted for the other four lenses that I haven’t yet tested in the manner described in this post.
It also appears that I want to increase the amount of travel I’m using with the Cognisys rail up to close to the maximum of about 200mm. I’m also considering changing my wide-open focus point on the rail from about mid-travel to a point further back. However, at this point I only know that these this lens focuses closer as it is stopped down. I don’t know if the other four behave the same way.
It’s too bad that this testing at multiple distances is so time consuming and eats up so much disk space.
David Braddon-Mitchell says
Good work Jim!
This makes a lot of sense of some of the DXO tests of mirrorless cameras. They have the Zony 16-35 zoom peaking wide open, which has never been my impression in the field, and I think they have the edges of the 70-200 getting worse as you stop down.
It also makes sense of the Sony focusing algorithm which seems to open the aperture and then close it while focusing on some lenses. Maybe it’s getting quickly in the zone wide open, then achieving accurate focus stopped down?
Jack Hogan says
Excellent demo of focus shift Jim.
Michael Demeyer says
I presume this is all in the center of the field?
Agree, this is a great way to understand the practical effect of focus shift, at least at close distance. Any idea how you might test it at mid- or long distance?
Yes. I’ve added the words “on-axis” to the post to make that clear. Thanks.
I’m in the market for a 10-meter focusing rail. Seriously, no.
Oo, very nice.
1000 cycles per picture height is a rather extreme threshold for depth of field though… at f/2.8 the Sony 90 would have no depth of field at all.
This refers to the Batis test, which has been updated and moved here:
And, CarVac, I do take your meaning here.
Max Berlin says
Since I started Imatesting with a Sony, Novoflex adapters and a number of manual Leica and Contax lenses, I never knew enough to assume that focusing at wide open would get me a good result at f4 or 5.6 etc. I always had as part of my time consuming testing to refocus for each f-stop and take the best of 3 to get a valid Imatest score.
Nikon is problematic for this because one has to use a loupe and live view to with live effect on to be certain that they are actually focusing on the right plane.
The thing is though that depending on the subject distance and DOF that the OVF focus indicator can be accurate enough up to f4 as it does shift as your testing indicates but still falls within the DOF.
But in reality I don’t worry about it too much anymore. If I want to shoot a live moving subject f2.8 – f4 works well enough and fast enough with Sigma Art and some Nikon lenses.
So many photos (portraits especially) precisely taken with the Otii require a lot of post before I can deliver a full size image.
Few clients can accept such high level of detail that exceeds their 10x ‘beauty mirrors’ they use to pick and pluck prior to the shoot.
It’s easier to catch a nice expression with a lens that doesn’t show every flaw seen and previously ‘unseen’.