Back Focus, Front Focus or Spot-on? Do the Domino Test

Now that even the most modest Nikons feature 24 Mp. sensors, optical quality and lens performance (let alone photog skills) are increasingly coming under pressure.
It is no coincidence that many of the latest Nikon mid-range and top-of-the line cameras include a feature called “AF Fine Tuning” which allows users to adjust the auto focus of their sample of a particular lens to their particular camera.

On Flickr some users are commenting on dialing in as much as +/- 20 AFFT correction on their lenses, which, to me, seems excessive.
Although it is reasonable to assume that lenses might be off a couple or even a bunch of millimeters between one sample and another, claiming that they are 5 or more centimeters wrong seems like a lot.
The large majority of users also say that their lenses back-focus, which is curious because, according to my tests, the rule of thirds applies here: one third of my lenses back focuses, one third front focuses – mostly my old AF-D and MF (AI, AI-S) lenses – while another third focuses spot-on, most notably my third party samples.
Note that MF lenses cannot be corrected, and that correct focus was acquired using the focus confirmation dot in the view-finder, which could be the reason they are a tad off.

On their European website Nikon offer a procedure for AF fine tuning, which could be called the “Ruler Test”
It consists in mounting a ruler in a 45º angle next to a stationary object that serves as a central focusing point – a book – and focusing on this center point.
However, in my experience, this test is extremely difficult to interpret, because of the gradual transitions between the exact focusing point and the out of focus that can – supposedly – be observed on the ruler.
Thus, I have devised a different test in the form of a “trapped” model, which consists in more pronounced focusing steps of 6 mm each, which is the thickness of the domino stones this test is called after.

Set-up

15 Dominoes were set up on a flat surface, aiming to keep them as parallel as possible. The camera was set up on a tripod at the height of the table, keeping it horizontal (without tilt or yawn) and trying to maintain it as parallel as possible with the subject.
A 130 W cold (0.6 kW incandescent) continuous artificial light source was set up at seven o’clock, while a white bounce was positioned at three o’clock, in order to minimize shadows.
The D7100 was set to 100 ISO, JPG Fine (optimal quality), AF-S on the center focus point and at a one second exposure delay to rule out any manipulation related camera vibrations.
All lenses were shot at their maximum aperture, to allow for the most pronounced depth of field fall-off.
Focusing distances were between approximately 60 cm. (28mm f/2 Nikkor) and two meters (100-300mm f/4 Sigma), since close focusing is most critical in terms of depth of field.
All zoom lenses were tested at their maximum focal length, because it turned out that at the short end most were less critical in AF performance.

According to my results, every 6-mm thick domino corresponds to approximately one +/- AF correction step, which allowed me to calculate AF correction according to the number of dominoes AF was off.
With the focus point on the center domino, any deviation towards a domino on the left represents approximately one plus point in focus correction (front focus), while every domino to the right one minus point (back focus).
Ideally, focus fall-off for any lens would proceed linearly to the left and right, meaning that domino 1 and 15 ought to be similarly out of focus; any deviation means that focus is not on the center domino - always in the supposition that the camera is parallel to the subject, while also that the lens does not suffer from relevant optical defects.
Although this test is hardly conclusive, it allowed me to evaluate and correct 8 lenses in a two-hour time span, while also to roughly assess their optical performance and possible defects.
For those who wish to be more thorough, my friend Carsten suggests to test at different f-stops up to two down from maximum aperture to rule out focus shift, which, as he points out, can be substantial on certain lenses. You might also want to repeat the test at infinity, especially for wide angles.
Finally: keep in mind that my results are particular and that yours may be very different depending on the lenses and camera(s) involved.

Tested lenses & Auto Focus Fine Tuning

AF 12-24 f/4G Tokina. AFFT: 0
AF-S 17-35 f/2.8D Nikkor. AFFT: -2 (Back Focus)
AF-S 18-55 f/3.5-4.5G Nikkor. AFFT: 0
AI 28 f/2 Nikkor. AFFT: NA (Front Focus)
AI-S 35-105 f/3.5-4.5 Macro Nikkor. AFFT: NA (Front Focus)
AF 35-105 f/3.5-4.5D Macro Nikkor. AFFT: +5 (Front Focus)
AF-S 55-200 f/4-5.6G VR Nikkor. AFFT: -3 (Back Focus)
AF-S 100-300 f/4D Sigma. AFFT: 0

Beyond auto focus fine-tuning: optical quality

Unsurprisingly, the AI 28 f/2 Nikkor is the sharpest and most defect-free lens in my current arsenal. Surprisingly, my sample of the 55-200 f/4-5.6G VR (top image) is a close runner-up.
Meanwhile, the 18-55 f/3.5-4.5G is – by far – the weakest lens in this line-up, which is why I will relegate it to the D40, where it performs “good enough”.
On the D7100, on the other hand, its optical flaws become painfully obvious. Although it required no AFFT correction, its sharpness and contrast are a little disappointing, even though it does not suffer from any other major optical defects, such as chromatic aberration, for example.
My sample of the AI-S 35-105 f/3.5-4.5 clearly outscores its AF-D descendant in sharpness, although both are free from obvious optical defects, even wide open.
Meanwhile, on the D7100, at least, the 17-35 f/2.8D does simply not live up to its $1.900 price-tag; it is not quite as sharp as it ought to be, while it also suffers from chromatic aberration throughout and spherical aberration towards the extremes of the image.
This is sad news – even if not quite unexpected – given that DxO Mark has been rating it fairly poorly on the majority of Nikon DSLRs it has been tested on.
Fortunately, the lens apparently does better on recent FX cameras, like the D600 and D800, even if it’s still mostly outscored by the more modest 18-35 f/3.5-4.5G, which costs less than half (SRP $750), while it is more or less on par with the – now discontinued – D version of that same lens.

Both my 12-24 f/4G Tokina and the 100-300 f/4D Sigma are virtually defect-free – apart of some very mild CA – focus spot-on, are sharp, and perform in every other way as expected, which cannot be said of their Nikkor brethren, which are kind of a mixed bag.
The AI 28 f/2 is outstanding, the $300 55-200 VR a very pleasant surprise, the $120 18-55 clearly gives away its price tag, while the $1.900 17-35 is disappointing, to say the least.
Why a $300 “amateur” mid-range zoom outscores a supposed pro-lens is a riddle to me; I can only assume that it may have to do with my rather unscientific testing.
All in all, my 40 year-plus manual focus AI and AI-S lenses perform as good as or better than their descendants, which is something to ponder on, given that most of them can be had at a fraction of their original retail price at Amazon or E-Bay.

Conclusion

The domino test does not only allow you to fine tune AF on your auto-focus D and G lenses, but also to roughly evaluate their optical quality.
Even though the rule that every buck spent comes back in optical performance remains true in general, you may find that some legendary lenses can fall off of their pedestal, while some heavily underrated ones step it up a couple of notches.
Be ready to be surprised.

You might also want to read:

The Quick & Dirty Focus Test