Saturday, 2 November 2013

The Nikon D800 has serious AF accuracy problems if the AF assist light of the SB-900 is enabled (part 2)

As mentioned first in a previous article, the D800 has a serious auto focus accuracy problem if the camera is using an external flash and the flash AF assist light is used.

I contacted Nikon after I wrote the first article and just a few minutes later my case was closed with a short message saying the following:

"I am sorry to learn about your issues with the D800 focusing when the AF assist light is used.
I would like to confirm that this phenomenon is known to us. It is caused by the following:

1. The wavelength of red light is longer than approximate wavelength of white light.
2. Lenses (including the one in the AF sensor) are calibrated to provide best sharpness for the whole spectrum, however are not ideally corrected for all colours.

This leads to a wider degree of tolerance when focusing using red light. This will be common to any flash that uses red light as a focusing aid, but assuming the camera is not already back focusing this is within specification. Shooting wide open (especially with fast lenses) on high resolution cameras will highlight this issue, which is why we would recommend stopping down between 2 to 4 stops from maximum aperture, especially on a D800.

We apologize for the inconvenience this may cause you."

Of course, I was not happy about the answer or the fact that Nikon decided to close the case without doing anything more, so I explained that their answer is not satisfying and is totally incorrect since the AF problem is there, even when the AF assist light is covered, which rules out the wavelength theory, and that stopping down the lens is not helping since the camera has a problem even with the 24-120/4 lens, so stopping down the 50/1.4 to f/2 would make no difference at all. Besides, without the AF assist light, I have no problems getting accurate focus with the 50/1.4G at f/1.4. After this message my case was reopened and Nikon Support asked me to take a few test shots according to their own recommendation on how to test auto focus.

Three unedited raw images was requested, each taken on tripod with the 50/1.4G at f/1.4, one without flash, just using available light, one with the flash and AF assist light and a third with the AF assist light covered. I decided to provide them with two "bonus" images as well, one with the flash but with AF assist light turned off in the flash, and one with the camera in P mode. I choose P mode because in this mode the camera should set whatever it chooses regarding ISO, shutter speed and aperture, knowing that it will stop down the aperture to f/5.6. I decided also to take six images with each settings, and would force to start focus in three from infinity to target and the other three from nearest focus to target. This would rule out the possibility that focus direction would be the cause of the problem.  I considered that if all six images in each set demonstrates the same behaviour then the case is clear.

I was no longer surprised by the results, I knew that the focus problem was not caused by my own hand shake or camera movement, and using tripod did not improve the results. To eliminate any possibility of focus error caused by movement, I even used a wire trigger so that I did not need to touch the camera at all after I offset the focus between each shot.

However, there was one thing which surprised me a lot. That was the shots in P mode, with the aperture set to f/5.6. I expected that this would hide any focus problems, so seeing that severe back focus was hard to believe, but it was there, showing that stopping down is not helping at all, the focus is just moved even more back, so the end result is even worse than at f/1.4 since the target is even more out of focus. This is actually indicating a pure firmware error, so there is hope...

After I submitted my test images, Nikon changed the status of my case to "Researching" and it is currently in that state.

Hopefully this will soon result in a firmware update and a fix of at least this problem.

The submitted test images and the results

This is the requested test target scenario. The focus was on "D800" of the D800 / D800E user manual. The ruler is there to clearly indicate the direction of focus, not to measure the back focus.

Image 1: D800 without flash, on a tripod, 50mm f/1.4 lens with wide open aperture, focused in viewfinder mode (not Live View) and with the central AF field on the "D800" of the D800 manual. The white balance is wrong because the ambient light was too low and the camera did not manage it well but once that is corrected you will see that the image is OK if you disregard the high ISO noise.

Image 2: Taken as above, with help of the SB-900 AF assist lamp and with flash.

Image 3: Taken as above, with the SB-900 AF assist lamp covered and with flash. It seems that by covering the light the focus is moved very far back. I don't know how that can be, but this is the case always when the AF assist light is covered.

Image 4: Taken as above, with the SB-900 AF assist lamp off and with flash. No question about that the auto focus is accurate if the AF assist light is not on.

Image 5: Taken as above, with the SB-900 AF assist lamp on and with flash. This image is clearly the worse, even though the aperture was f/5.6. This is actually indicating a pure firmware error since it is obvious that when the lens is stopped down the focus is moved even more to the back, so much that the target is seriously blurred. As you can see, best focus is very far back, on the ruler it is somewhere between the digits 19 and 20 on the scale nearest the books. It is simply horrible.

What is interesting also that reading the EXIF data using Kuso EXIF viewer shows a focus distance of 0.84m for all five images, which is obviously not the case since there is several cm back focus between the accurate and the back focused images.

This problem is now confirmed by several other people, and not only when the SB-900 is involved but also with other flashes, even third party. Hopefully Nikon will take this problem seriously and fix it very soon. There is no doubt in my mind that this has nothing to do with the light spectra, just on the fact that the external flash AF assist light is turned on by the camera.