Nikon D7000 Focus Issues

2011-02-28

Updated: 2012-04-07

The Nikon D7000 [a] is, according to all reviews I've read, one amazing piece of hardware. However, as I describe in the article Camera Technique, getting all that amazing technology to work for you is not trivial. The D7000, with its high-resolution sensor, is more unforgiving than most cameras. At $1200, it is really built for lenses in the corresponding price class. It is therefore quite depressing to see people attaching a $200 Nikon 35mm/1.8 DX [b] and then wondering where the $1200 technical image quality went. A hint: It has to do with chains and weakest links. It is equally painful to read other complaints about the autofocus system on the D7000. For all I know, it isn't perfect. Maybe it is a bit wonky - all I know is that having a D7000 with a broken autofocus system is way more common among beginners with DSLRs than it is among experienced users. There is something about beginners that seems to attract these "broken" cameras. After reading this thread [c], I felt I couldn't be quiet anymore. DPReview is a gearhead forum, but strangely enough their collective grasp on gear and engineering methodology in particular is weak. When I think of a "gearhead", I visualize some hard-core tech people who may not be able to produce great art, but who know these machines inside and out.

If you really want to test your lens, as opposed to, you know, do like the rest of us and go out and take a bunch of actual pictures and see if they look all right, you should do it properly. LensRentals has a thorough article titled, appropriately enough, How to Test a Lens [d], that describes various ways to see if your lens is broken. Among other things, it shows what level of front- and backfocus you can reasonably expect. Both camera bodies and lenses are made to specific tolerances, and sometimes both are "off" in such a way as to "bring out the worst" in each other. If you find that your combination of lens and body is particularly bad - just send it to your Nikon service center for evaluation [e] and they'll help you. Just don't whine about it on forums. That ain't helping anyone.

Needless to say, nobody on DPReview bothers with the kind of testing done by LensRentals. So I fired this one off, in the hope that we'd at least talk about focus issues in a sensible manner. In the dystopian belief that it will soon be forgotten, I replicate it here. Maybe someone will end up here when googling for "d7000 focus issues" (slightly edited emphasis for readability):

Hello,

I have seen a number of posts regarding the D7K autofocus system. What I would like to bring up are the underlying assumptions people tend to make, because I'm not quite sure they hold.

I'm not saying that those with AF issues are wrong - you would expect a certain percentage of cameras to be duds, that's just the way manufacturing processes work. But I would like some reasoning about the whole system - lens, camera, AF, error tolerances and trade-offs.

Assumption 1: Autofocus, when properly working, will always give you the optimal focus when it finds focus - irrespective of focal length, lens quality or light.

Assumption 2: Phase-detect autofocus issues can be detected by comparing with contrast-detect autofocus.

I find no real reason to hold these two assumptions. The reasoning is this: Contrast AF (CAF) works on the actual sensor data. When the CAF system has determined the focus distance, it is the best distance (well, given the accuracy of the focus stepper motor) because the image contrast has been measured directly using sensor data. Phase AF (PAF) has to predict what the image will look like, as it doesn't use sensor data. Therefore, it is dependent on having very accurate information about the lens and camera in order to find the distance to the subject.

I haven't been able to find anyone authoritatively stating that PAF can be expected to have the same quality as CAF, and it seems to me that PAF will always on average give worse results than CAF, just based on it being exposed to various imperfections in the lens in ways that CAF isn't.

In particular:

Lenses have chromatic aberration. This means that the PAF system will yield different results depending on lighting, as the peaks will end up on different places on the PAF sensor. A good PAF sensor should compensate for these shifts in ambient light, but I think it is silly to think it can be compensated for completely. CAF has no such issues as it measures the contrast directly from the sensor.

Not all lenses are completely symmetric. Manufacturing imperfections sometimes result in lenses where the left and right sides of the lens is off by a very little or so. Since PAF works by comparing the image received from the left and right sides of the lens, this will result in it being wrong. CAF isn't affected by this.

Lenses have distortion. This, again, introduces errors in the distance measurement of PAF.

Camera bodies have imperfections. Like everything else, they're manufactured to set tolerances. Sometimes, the error in camera manufacturing and lens manufacturing cancel out and you get perfection. Other times, they add up, and you get much worse results.

There's also the stepper motor in the focusing system. Being a stepper motor, it is only so precise. If perfect focus exists between two steps - well, you can't have it.

Based on this, we would expect a PAF system to:

...not work exactly the same in different light (chromatic aberrations can't be completely corrected)

...not work the same in all focal lengths (distortion can't be completely corrected for)

...not work the same with all lenses (lens imperfections can't be completely corrected for)

...not work the same with all cameras (camera imperfections can't be completely corrected for)

Which, I think, is exactly what we see. What we can expect from a PAF system is that it can be calibrated differently. For example, optimized for a certain light, lens and focal length - but then at the cost of not working that well in other conditions.

So, my first question is this: Can anyone authoritatively state that PAF systems are even capable of giving the same quality of focus as CAF systems? Because if not, then we're chasing a level of perfection that simply isn't possible with today's technology - at least not at the price we're willing to pay.

All of you who have dialed in AF fine tune indoors, only to find that the lens front-focuses in daylight: Do you have a lens with zero chromatic aberration? The same goes for all of you who sent the camera to Nikon for adjustment and now have perfect focus indoors, but find focus errors outdoors.

All of you who have dialed in AF fine tune at one focal length, only to find that the lens front-focuses on other focal lengths: Do you have a lens with zero distortion?

Assumption 3: Perfect focus is possible, with all lenses, all cameras, all conditions, all the time.

This is the final assumption that seems to be present. Given what we know about optics, is it even realistic to expect perfect focus? Are our expectations set correctly?

In many D7K focus threads, someone shows a photo and points to some defect in it: "I focused here, but the sharpest point is here".

Well, is your lens really so perfect that your plane of focus really is a mathematical plane? How much did you pay for that lens? Or could it be that imperfections in the lens and imperfections in the rest of the system (camera, AF, etc.) just so happen to cancel out at the sharpest point, while they don't at the focus point?

In particular, is the point where you put the focus out of focus to such an extent that it really is an AF problem, or is the AF working within the tolerances one can expect from an AF system that must produce a result within reasonable time in a very wide spectrum of conditions?

My second question is this: Using authoritative sources and not just saying "what you think it oughta be" - what level of error can we expect from a modern PAF system using consumer grade lenses and bodies? What level of error can we expect with pro lenses and bodies?

— Nikon D7000 Focus Issues on dpreview.com [f]