D800 Autofocus Target Quality

1. Introduction

At the dpreview forums there is a lot of handwaving about (the lack of) autofocus target quality. Without any evidence at hand, all kind of claims based on generic guidelines are generalised to whatever target is shown. 

There is an easy way to check the quality of an AF target: measuring. I’ve used Reikan’s Focal to check  autofocus consistency of various targets. Although there are certainly some deviations, the outcome is that most targets are sufficiently good for performing AF tests. 

 

2. Test Setup

The analysis of the targets is done with the focus consistency check of Reikan’s Focal software.The outcome of the analysis is a CoF value, determined by comparing (as a percentage) the average quality of all the results against the quality of the best individual shot. Reikan considers CoF values of 95% and above as adequate autofocus (more info can be found here). When I look with my own eyes to the results, CoF larger than 95% hardly gives visible unsharpness, between 92 and 95% gives acceptable but visible less sharp results, and smaller than 90% gives blurred results. 

The test set-up consists of a Nikon D800 with a 24-70/f.28 lens, used at 50mm@f2.8 at approx. 1.5m distance. The target is mounted flat and perpendicular to the camera. Light conditions are constant (cloudy day). I run a minimum of 10 samples per test, and I repeated the test 4x. I also changed distance, small shift of the target, or change the colour temperature of the light (cloudy, sunny, light bulb) to see if there is an impact to the quality of results. Small perturbations can be seen, but the order of magnitude of the outcome is always the same. Several targets have been used for the test. 

The first group of targets are dedicated targets for AF tests: 

  1. Reikan Focal. This is the target provided with Reikan’s Focal software package for doing multiple tests.  
  2. Leonard’s target. A target presented by Leonard Shepherd, claimed to be a good target for AF tests. 
  3. ISO/PIMA target. Used to test resolution of printers and scanners
  4. Thom Hogan. This consists of a collection of various lines, with different thickness (similar one here).  
  5. Single Line. A simple line drawn on a white piece of paper. Two exercises have been done, with the line positioned in the middle of the AF point, and one at 75% of the AF point, both showing exactly the same results. 
  6. Spyder Lenscal, a dedicated target for AF fine tuning. 

Target1

The second group of targets are targets from daily life:

  1. Book shelf. Used to analyse left-field AF problems in the past. 
  2. Book spine. Showing contrasty white letters on a black background. 
  3. Beethoven. A picture of a composer on a flat surface. 
  4. Bobby. A teddy bear from London. 

Target2

 

3. Test Results

The outcome of the test is summarised in the following table, showing the range of CoF values over multiple tests: 

Table

As you can see, most targets show sufficiently good results to perform AF consistency tests. Some conclusions that can be drawn:

  • Reikan’s target, Leonard’s target and the bookshelf show equally good results. (1 + 2 + A). 
  • Although the other targets score a bit lower, the visual impact is hardly noticeable. In practice for visual inspection those targets perform equally well for AF tests. 
  • A single line works as good as multiple lines. (4 + 5). 
  • Real-life targets which are coloured and contain lots of small fine-grain details give less good results than dedicated black & white targets. Nevertheless, the differences are not huge, and still useable for AF qualification. In these cases, taking multiple pictures of those target is recommended. 

To get an idea of the best-case and worst-case results of Bobby, an impression is shown below. The differences are visible (less fine-grain detail at the nose), but please keep in mind we are pixel peeping at 36 Mpixels here. 

Bobby

 

4. Additional tests

Three additional tests have been done that are worth mentioning. 

 

1. Left – Middle – Right AF point Focus consistency

The bookshelf target was tested for AF consistency. The left AF point showed 0.5% better results, the right AF point 0.7% worse results over 3 tests each. This shows the results are sufficiently accurate. 

Bshelf

 

2. Sensitivity to fine-grain detail

With the 24-70 lens set at 24mm/f2.8, the Reikan target was tried at different distances. The differences are about 3% to 5%, where fine-grain detail gives less good results. 

Dis

 

3. Firmware 1.0.1 versus firmware 1.0.2

You can also notice some difference between firmware 1.0.1 and 1.0.2, and some improvements can be noticed in focus accuracy and quality of focus. The improvements are shown in the table below, where the Reikan, Hogan and especially the book spine targets show an improvement in AF consistency. 

Fw

See the before and after results for the bookspine. 

– Before (91.9%) worst (1690) versus best (2255) focus result: 

Before

– After (97.6%) worst (2002) versus best (2256) focus result:

After

 

I’ve also performed some tests in moderate light, and it is obvious that AF quality becomes less (around 90% for targets that otherwise look fine). Hence, good light conditions are required to perform AF tests. 

 

4. Conclusion

The outcome of the test is as expected. All targets provide sufficient accuracy to perform repeatable autofocus focus test. 

 

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: