Nikon D800 / D800E autofocus Precision
But after testing today with my D800 and D800E both just back from Nikon, my conclusion remains the same as what I initially showed in my review of the Nikon D800: the D800 /D800E has poor autofocus precision, and that’s with the center focusing sensor, which is the best one (one example). In short, the asymmetry issue is a footnote issue compared to the fundamental problem of poor precision in general.
It is important to note that I use the word “precision” in its proper scientific meaning. The Nikon D800 / D800E autofocus system lacks precision at moderate to far distance (it seems to lack accuracy also, but this I am less sure of, as proving that would require several hundred shots and extensive analysis, and still be of no practical value, e.g. an accurate average is of little use for any particular image). Poor precision means you get sharp or soft or mush under identical repeat conditions. Which is what I observed in my testing with my freshly serviced Nikon D800 and D800E.
If the Nikon focusing system had high precision (it does not), then any accuracy issue could be dealt with using AF Fine Focus adjust. Which is why so far I have found AF Fine Focus Adjust to be of marginal value. After all, AF Fine Focus Adjust is designed to compensate for accuracy errors, and thus has no bearing on poor precision. In short, all AF Fine Focus Adjust can do is to make accuracy better on average. It won’t improve precision.
Today’s tests with my factory serviced D800 and D800E + Nikon AF-S 85mm f/1.4G were a huge disappointment, showing me that using autofocus at a distance of ~80 feet is joke, showing hit and miss results ranging from sharp to soft to badly blurred. All on a tripod with everything fixed — just repeat the same thing and roll the dice on what you get. Which is entirely consistent with the numerous past tests with other Nikon bodies and other Nikon AF-S ƒ/1.4 lenses, which I’ve shown and proved over the years in my publications. The Emperor has no underwear.
At closer range, my past experience has seen better results, e.g. focusing on eyes for portraits. So my comments must be taken in context of “medium and far” distance, and also with f/1.4 lenses in the 24mm - 85mm range.
In my testing today under the same conditions, the Canon 5D Mark III + 85mm f/1.2L II achieved both high precision and high accuracy, similar to what I found with the Canon 1D X autofocus. If anything, an ƒ/1.2 lens is more of a challenge than an ƒ/1.4. lens. Were I relying on autofocus, I’d surely be taking the focusing precision and accuracy competence of a camera quite seriously in making a purchasing decision. But I manual focus most of the time, so for me it is less of an issue than say, a wedding or sports shooter (and BTW I have not researched Nikon AF recently with super teles, but they tend to be better due to the high level of defocus one gets with such long lenses, which seems to help the AF system).
I have no doubt I’ll get emails claiming good results, which doesn’t mean much to me: other contexts, other subject matter, improper testing, good testing under specific circumstances, etc. After all, back when Leica told me the S2 had no autofocus issues (after my report of such issues) I took the Leica rep out with his camera and lens and in 10 minutes made a dozen badly-focused images while he watched (all supplied to Leica, complete stonewalling, never heard back). I have so much context now on the 24/35/50/85 Nikon fast primes on D3x/D3s/D800/D800E that the precision issue just cannot be dismissed. I am satisfied that this is a fundamental limitation of Nikon autofocus. Period. Of course, if you’re shooting a D700 or D3s, you have 1/3 the pixels of the D800, and your margin for error is somewhat higher, but the issue remains identical. Except that the D700 always did seem to do better than the D3x in the few times I tested both.
So here’s the bottom line: can I point a Nikon D800 + ƒ/1.4 prime at a subject at a moderate to far distance, hit the AF button and consistently get a sharply-focused image at ƒ/1.4 or ƒ/2? The answer is a resounding “NO”.
Please hold detailed and non-sequitur comments. I am at 12,000' in the mountains and cannot respond in any detail, or at all to some messages.
Alan Y writes:
Your comments on the lack of precision of the D800 AF is spot on with my observations. I often use the 24/1.4G and 85/1.4G wide open and accurate focus is hit or miss.
I too think it was similar with the D700, but now with higher resolution, you can really see and appreciate how far off the focus really is. I thought Nikon had superior AF accuracy and precision to Canon bodies- a main reason I bought Nikon gear- and sad to hear that may not be true. AF is incredibly important--actually essential-- for candid environmental portraits and I hope Nikon works to better this technical flaw.
DIGLLOYD: I first saw these problems with the advent of the Nikon 24mm f/1.4G several years ago. While apparently a little better with older focusing systems (eg D700), it has been a persistent problem on the D3x and later.
Cliff L writes:
It's nice to see there's at least one person out there who's speaking up about the D800/D800E AF problems.
Now, you'd better be prepared for the onslaught of email from Nikon fanboys, who will tell you your tests aren't valid because you didn't use a Nikon factory-approved test target... LOL
DIGLLOYD: Out in the field, I’ll be damned if I can use Nikon autofocus with the ƒ/1.4 primes, rendering autofocus a non-feature, which is one reason I just shoot completely manually— errors are thus my own.
As for the Haters, most of that savagely anti-intellectual dreck fills up discussion boards, which is why I do not frequent them. The day I get precise and accurate Nikon autofocus it will be reported here, as I have no axe to grind— only a desire to have autofocus be a useful feature I can use in the field. In this respect I am thrilled with the Olympus E-M5.
Roland W writes:
“The Emperor has no underwear” LOL! Thank you so much, this honest review fits exactly in my picture. My wife uses Nikon, a D300 (+ old film SLR bodys), and I Canon (5D3, 7D). We both shoot a e.g. lot of wildlife with comparable lenses (e.g. tele range from 70-200/2.8 up to 500 mm superteles). So I know both AF systems quite well. When it comes down to real action such as catching birds in flight the Nikon has much more OOF issues than my Canons. In particular my 5D3’s AF performance is stunning (with a good lens). To be fair, the 7D’s AF performance is worse but still significantly better than the D300’s – with a 6 MP higher resolution and therefore need for more precision. So my wife, a skilled photographer and an passionate Nikonian since her first beloved FM-2 (a real classic Nikon), seriously thinks about switching to Canon and going for a 5D3.
She recently tested a D800 in a local shop and even this short test did not convince her that she’d get a much better AF performance with such an upgrade. Now she thinks about giving Nikon a last chance with a D4 before she switches to Canon – even she does not like those bulky pro bodies. Did you ever test the D4’s AF performance with a really fast lens? Would like to read that one…
I am so relieved reading your blog because if you check Nikon forums everything is sooo shiny, and if someone comes up with a critical issue she/he receives really nasty or useless comments.
DIGLLOYD: Autofocus in the realm of 20+ megapixel DSLRs is now a lagging technology apparently not up to the precision demands of the sensor resolution. Though this issues I’ve seen would be a problem even at 12 megapixels. There are some exceptions— it seems that long lenses can do quite well, presumably because of the blur/sharp thin-ness. Though I'm not so sure this would be true at the same magnification as something like an 85/1.4 (eg a 300mm about 3 times farther away would have roughly the same magnification as the 85mm).
Also, with some more testing, it seems that if I give the D800 a very high contrast subject (like a silver car wheel silver with a black backgound), it does OK. Unfortunately I prefer to photograph more than car wheels and black/white charts! Other stuff is very disappointing in the field.
But I think there is also another explanation (which is not to exclude several other issues): the relatively low contrast of f/1.4 primes in the 24mm - 85mm range has an inherent ambiguity (uncorrected contrast-robbing optical aberrations)— seems to me that a “smart focus” might stop the lens down just a bit, e.g., ƒ/2. For that matter, compensation for focus shift should be programmed in.