Flare

By Rishi Sanyal

If you've been paying attention to the Internet over the past few weeks you'll likely have heard a lot of talk about 'flare' in connection with Nikon's D750 DSLR. Not all D750s seem to display the issue, and on those that do, it's only noticeable in a very narrow band of shooting scenarios. We shot literally thousands of frames on two D750 bodies during the course of our own testing and didn't have a problem with it, but some people clearly do. And Nikon (probably still a little gun-shy from the D600 oily sensor PR disaster) acknowledged the issue in late December and pledged to repair affected cameras. In mid-January the company asked retailers to discontinue sales of the D750 until new, problem-free models could be stocked, and now Nikon has issued specific instructions on how to check, and get service, for an affected body.

So what is 'flare'? The problem shows up as an unusual pattern - a little like optical lens flare - that can occur when bright light sources are just outside of the frame. If you see it at all, you'll see it at the top of the image, with a shadow in the flare manifesting itself as a darker band across the top of the frame.

This is how the flare issue manifests itself - flare along with bands resulting from regions that are masked from flare. The above photo was actually taken on a Canon 1D X. Read our analysis below.

The flare appears to be caused by internal reflections from the autofocus module that make their way past the light baffle (above the AF module) and onto the image sensor. And while this only occurs in heavily backlit situations where you're already likely to experience flare, the oddity is that the flare is abruptly cut off near the top of the image, leading to what looks like a band across the upper portion of your image. Dave Etchells at Imaging Resource did an excellent piece explaining exactly why this issue occurs, and it appears that the severity of the problem - and the reason it varies from body to body - is linked to the placement of the AF module below the mirror. Some bodies have the AF module recessed lower into the body, and it's hypothesized that bodies with lower AF modules have less of an issue. We ran through various tests before we published our D750 review, but in light of Nikon's recent decision to address the problem directly via a service advisory, we decided to dig a little deeper, and collate some of our experience from late last year - and some new tests - in an attempt to really get to the bottom of things.

To do this we picked a number of cameras we had in the studio to see how they performed in this regard. We used the pro-grade EOS 1D X, and tested a pair of D750s along with the D810 and D4S. Although we had not seen any flare/banding problems in the thousands of shots taken by either D750 over the course of our testing, we wanted to see (1) if we could recreate it under a controlled scenario, (2) if the severity of the problem differed between the two copies of the D750 we had, and (3) whether it was a problem on other DSLRs as well.

We shot the test scene below at two apertures (F2.8 and F5.6); flare is generally an issue at wider apertures, and so we also wanted to see what happens as we stopped down. For the Nikons we used the Nikkor 24-70 F2.8 lens, while the Canon EOS-1D X was equipped with Canon's EF 24mm F2.8 IS USM. Our light source was a rather bright LED flashlight.

The results show that the flare issue on D750s do indeed vary from camera-to-camera. In addition, we found that several other high-end DSLRs have similar problems if you try hard enough to find them. In fact, the Canon 1D X appears at least just as egregious as the D750 in this regard. Have a look below.

F2.8
F5.6
Canon EOS 1D X: Wide open we see three areas of strong banding. The banding begins to disappear as the lens is stopped down to F5.6, but it's still quite clearly there.
Nikon D4S: There's no noticeable banding at either aperture.
Nikon D750 #1: At F2.8 there are two distinct areas of banding, with fairly harsh transitions. The transition itself gets smoother, and becomes less of a problem, as the lens is stopped down. Note we really had to point the light source at such an angle as to generate a lot of flare before banding was obvious.
Nikon D750 #2: Our second D750 appears to have less of an issue than the first (the one above), with a less harsh transition in the band caused by the flare. By F5.6, there's slight banding, but nothing too offensive.

Nikon D810: The D810 shows very slight banding and flare at both apertures, but we find it so slight as to not be much of a practical concern at all.

Not shown: Canon 5D Mark III did not appear to have this issue.

You'll first note that the problem can vary with aperture. Recall that the qualities of flare change as you stop down. We also found that the nature and severity of the banding also changes with focal length, as well as the size of, and distance to, the light source. So take the samples above only as representative, and not a comprehensive look at exactly how the issue might present itself in your particular body.

The D4s appeared unaffected by this issue, and the D810 was only very slightly affected. One of our D750 bodies was worse than the other, with a harsher band marking the transition between the area with flare and the area masked from flare. The Canon 1D X seemingly fared just as badly as the worse D750, with three distinct bands when shot wide open. You can even get some flare and banding from side-lighting with the 1D X, if you try hard enough, visible on the left side of this photo.

We do need to stress, though, that these issues only occurred over a narrow range of incident angles when the light source was just outside of the frame. It's also dependent upon a number of other factors, and even after we saw the problem, it wasn't always exceedingly easy to reproduce in an offensive manner. A very, very big, bright light source in our studio did make recreating the problem a bit easier, but even then we had to set up the camera at just the right angle to reproduce it. This angle guaranteed you'd have flare anyway, regardless of whether your camera has the issue or not. Given this, we'd venture to say that this issue won't really be an 'issue' in all but the rarest of circumstances. It should come as little surprise, then, that the problem has not been found over the course of nearly 3 years of the Canon 1D X being available. The cynically-inclined might conclude that professionals have bigger things to worry about...

What causes this issue?

A video by Youtube user Tomasz Piotrowski indicated that the placement of the AF module had something to do with the severity of the flare problem, so we investigated our two D750 bodies. Indeed, as you can see in the rollover below, the 'worse' D750 body (#1) had its AF module positioned slightly higher than in the 'better' D750 body (#2). It's likely that internal reflections caused by light entering the AF module and exiting at angles so as to hit the image sensor is the main culprit for the flare, with cameras containing AF modules higher up (or closer to the image sensor) exhibiting this issue more severely. Perhaps the reflections from light entering modules closer to the image sensor make their way into the image more easily. Or, perhaps this sort of flare is more prevalent than we might think, and we just dismiss it as lens flare. It's hard to say.

D750 #1
(note how the AF module sits higher up compared to our 2nd body)

D750 #2
(note how the AF module sits lower in the body compared to the 1st body)

What makes the flare particularly egregious in the D750 (and 1D X) examples is the masked area that leads to banding. We imagine that, depending on the placement of the AF module and sensor, and other body elements, the area masked from the flare (or stray, reflected light rays) varies, which is why we see different flare patterns across different bodies.

The D750 in particular might be more sensitive to this issue than other cameras given its small body - there's a lot packed into the tight space, and a small change in placement of the AF module may be just enough to make the (almost unavoidable) internal reflections particularly deleterious. It's interesting to note that there's a light baffle in there (which you can see in the image below above the AF optics); the effectiveness of that baffle might determine how resistant your camera body is from flare. In most cameras we've checked, the baffle never fully covers the AF module when the camera is taking a shot (and the mirror is up), but perhaps it needs to cover just enough of the AF module to avoid this issue. We'd imagine that the effectiveness of this baffle is also sensitive to fine changes in the placement of the AF module, so really it's a perfect storm of factors that need to come together to recreate this flare phenomenon. It's no surprise, then, that the problem itself is so rare. And, again, not unique to the D750.

We'd expect the fix that Nikon is implementing to reposition the AF module to mitigate these flare issues.

Implications for autofocus accuracy

Given that the problem arises from the placement of the AF module, we were curious to check for any effects on focus accuracy. Since we were lucky enough to have two D750 bodies with differences in severity of the flare issue, we tested both bodies for front/back-focusing issues relative to one another, with two different prime lenses: both a Sigma 35mm f/1.4 Art lens, and a Nikkor 35mm f/1.8G (to rule out any possibility that it's just the lens that's at fault).

We used a LensAlign target (developed by Michael Tapes Design) for our test, aligning the camera to the target at a distance of 25x the focal length of the lens. Proper focusing on a target just to the left of the scale shown below yields a sharpest plane of focus around '0' on the scale below. Any front- or back-focus is then easily seen as a sharpest area of focus in front of, or behind, '0' on the scale.

What's interesting is that body #1 (the 'worse' D750) front focuses consistently relative to body #2 (the 'better' D750), which you'll see if you roll back and forth over A (D750 #1) and B (D750 #2) below. The red arrows indicate the point of optimal focus. You'll note below that both lenses front-focus with both bodies; however, the body (#1) with the AF module placed slightly higher front-focuses more, in both cases. You can also see the optimal planes of focus if you hover over C and D, which show the results of the more accurate contrast-detection autofocus (CDAF). Furthermore, we verified that the first D750 consistently front-focused relative to the second one with two different lenses, so it's unlikely that it's the particular lens solely at fault here.
Sigma 35mm f/1.4 Art
Nikkor 35mm f/1.8G
(A) D750 #1 (PDAF)
(B) D750 #2 (PDAF)
(C) D750 #1 (CDAF)
(D) D750 #2 (CDAF)

Roll over from A to B to see how D750 #1 front-focuses more than #2, with both lenses, when using AF through the viewfinder (PDAF, or phase-detection autofocus). Red arrows indicate focus.

With the Sigma 35mm lens (left column), D750 #1 appears to be sharpest at '12', which is well in front of '8', where D750 #2 is sharpest (roll over from A to B, focusing on the left column). Optimal focus is around 0 or 1 for both bodies (see CDAF examples C and D).

With the Nikkor 35mm lens (right column), D750 #1 appears to be sharpest at '5', which is well in front of '1', where D750 #2 is sharpest (again, roll over from A to B, focusing on the right column). Optimal focus is, again, around 0 or 1 for both bodies (see CDAF examples C and D).

Hover back and forth between C --> A, or D --> B to see the difference between CDAF and PDAF for the same body+lens combination to get an idea of how inaccurate phase-detection AF (through the viewfinder) is with either body. In this case, you could get away with not fine-tuning body #2 and this particular copy of the Nikkor 35mm f/1.8G lens. Both lenses need to be calibrated with the 'worse' D750 #1, though.

Curiously, front focus is exactly what one might expect from an AF module placed higher up in the body, closer to the mirror assembly. This slightly misplaced AF module would experience a shorter effective optical path than the path to the image sensor, and the intended subject the AF module focuses on would then fall into focus in front of the image sensor - meaning an object slightly in front of it would then focus on the image sensor, which is further behind.

Now, we're not suggesting that the physical shift in the AF module is the entire cause of the front-focus, nor that Nikon doesn't already correct for shifts like these in the AF calibration process every camera undergoes (even for each of the individual 51 AF points, as far as we know). What we're suggesting is that there may be some residual effects of the displacement of the AF module on autofocus that aren't fully corrected for in the factory calibration procedure.

Whatever the cause, AF fine tune fully takes care of this relative front-focus, even in the body with the visibly displaced AF module. And the AF fine tune values I had to use were nothing uncommmon: +12 and +4 for the 'worse' and 'better' bodies, respectively, with the Sigma lens. I've routinely had to use AF microadjustment values close to or in the teens for many cameras, including 5D Mark IIIs, D810s, 1D Xs, etc. We would normally expect factory calibration to account for slight differences in placement and optics of AF modules; however, even with factory calibration it's difficult for any camera maker to guarantee that any particular body will work perfectly with any lens when it comes to accurate phase-detect AF - especially with fast primes.

Key takeaways:

  • Our 'worse' D750 consistently front-focused relative to the 'better' one
  • The magnitude of this front-focus was nothing out of the ordinary, presumably due to factory calibration that is designed to deal precisely with such variations.
  • AF Fine Tune was able to account for any residual focus inaccuracies with a number of lenses we calibrated on both D750s.
  • Many cameras need just as much Fine Tune with some lenses as what we observed with our worse copy, so the AF implications here are more interesting, and less something to be concerned over.
  • We can't help but wonder how many cameras have small variances in AF modules and optics that lead to consistent inaccuracies in AF due to not being fully calibrated in the factory.
  • If you shoot fast primes, calibrate at home.

Should you be worried?

We shot thousands of photos on two Nikon D750 bodies and never experienced the flare issue. We took shots directly into the sun, or with the sun just outside the frame, for dynamic range tests and for our D750 video at the Museum of Flight, and still saw no issue. When we directly went looking for it, and tried hard to reproduce the flare, though, we were able to reproduce it. And, yes, both of our bodies are listed as potentially affected by Nikon.

Does that mean it won't be an issue for you? Not at all - it absolutely might, especially in situations where you're purposefully heavily backlit aiming for a shot with flare. In such situations, if you're unlucky and have an affected unit and happen to hit the right angle with a bright light source just outside the frame, you may experience this issue. Even then, you'll be able to avoid it by slightly shifting the angle of your camera, but if that changes your shot, then you may find it irksome.

After all, a ruined shot is a ruined shot, and we photographers often prefer to have our own selves, rather than our cameras, to blame when it comes to reject photographs. So it's a noble goal for camera manufacturers to avoid such issues altogether. The bigger take-home message here, though, is that these are complex devices, and you're going to find problems if you dig for them. Flare might be one of them. Out-of-focus photos may be another. We had to fine tune a number of the primes we shot with the D750, but then that's par for the course when it comes to shooting fast primes on full-frame cameras using traditional phase detection AF systems. We can tell you that none of these issues would prevent us from picking up a D750, but ultimately you'll have to decide how important, or unimportant, these considerations are to you.

Thankfully, all the issues addressed here can be, well, addressed. Nikon is fixing the issue with flare, so concerned owners and purchasers need not worry. As for focus inaccuracies, fine tune (or microadjustment) can be used to eliminate consistent front- or back-focus issues that are not fully tuned away in factory calibrations. Since the need for calibration arises at some point, with some lens, on almost every DSLR we've tested (the manufacturer can't optimize your body for every lens out there), the focus inaccuracy we show above with the D750 is nothing particularly egregious. We will, however, be curious to see if our D750 bodies show a change in AF calibration when they're returned to us from Nikon service. But either way, given the technical marvels cameras are these days, we'll only have ourselves to blame for bad photographs.


*Any deviation from '0' for the CDAF routines simply indicates slight misalignment between the camera and the target. We didn't have any major misalignments over ~1 unit; hence, for all intents and purposes, '0' represents ideal focus on the LensAlign scale.