D500 Cross Sensor Limitations - Light or Physics?

[James Farrell]

Help! I am totally confused. Similar to the D5, the D500 has a chart of AF cross sensor limitations for various Nikon lenses on page 98 of the D500 User Manual. A snapshot of that chart is attached. I am totally confused by this chart. I simply don't understand the limitations and the reasoning that some of Nikon's best and most expensive lenses have such limitations. That said, I do understand that there have always been limitations as to auto focus performance when using teleconverters, especially on older Nikon bodies.

Maybe I could understand this better if I knew whether this was an issue of the amount of light needed for the various rows of cross sensors that f/4 or slower lenses provide for the sensor or whether the design of a specific lens, regardless of its largest aperture, limits the AF ability of some AF cross sensors due to angles or something having to do with physics. Why are some f/4 lenses limited, but others are not? I am treading in dangerous waters; I am not that versed in light vis a vis physics so I am trying to get my arms around this issue and understand the aforementioned limitations in plain old, easy-to-understand language that my simple brain can digest. 

[Frank Fremerey]

Lucky guesses of possible physical limitations:

1)  certain lenses drawing characteristics trick phase detection to focus on other parts than the sharp parts

2) certain lenses field curvature corrections trick phase detection to on other parts than the sharp parts

Motivation: I see the AF-S 2.8/60 Micro in the chart.

Now I am listeing.

[Bjørn Rørslett]

The effective aperture of the Micro in the near range is much smaller than f/4.

[Frank Fremerey]

The effective aperture of the Micro in the near range is much smaller than f/4.

Du you see "near range" mentioned in the list?

PS. AF calibration is mentioned as troublesome with the 600mm lenses. A possible connection?

[Roland Vink]

The AF-S 105 VR micro speed also drops below f/4 in the near range, similar to the AF-S 60/2.8, but it's not listed ...?

[James Farrell]

The AF-S 105 VR micro speed also drops below f/4 in the near range, similar to the AF-S 60/2.8, but it's not listed ...?

Ahhh ... it's an f/2.8 lens to start with, right ...?

Folks: I'm thinking here that the central question of what the chart in my original post is all about is being ignored as some of you have focused (pardon the pun) on two micro/macro lenses. Does anyone have any thoughts on the bigger picture that I am asking about in my original post?

[Roland Vink]

The AF-S 105 VR micro is f/2.8 to start with, but so is the AF-S 60/2.8 micro. No mention of the DX 40/2.8 or 85/3.5 micros either.
Also not sure why the outer cross sensors don't work with non AF-S lenses.

Back to your original post ... obviously the outer cross sensors require f/2.8 or faster. They work with some f/4 lenses also - the 16-80/2.8-4, 24-120/4 and 70-200/4 are not mentioned in the restricted lists so I assume they are usable. Why there are limitations on the other f/4 lenses is hard to say.

It could be that vignetting means there is insufficient light towards the edges for the cross sensors to work. That may be why the DX 12-24, with its smaller image circle, is included on the list. But the others are FX lenses where vignetting is small relative to the DX sensor. Also most of the lenses mentioned are telephoto primes which tend to have less vignetting than zooms (which are not on the restricted list).

The restricted lenses are mostly super-telephotos where the image forming rays from the rear are nearly parallel, and perpendicular to the sensor. Maybe the cross sensors are optimised for shorter lenses where the ray angle is a bit greater?

[Øivind Tøien]

...
It could be that vignetting means there is insufficient light towards the edges for the cross sensors to work. That may be why the DX 12-24, with its smaller image circle, is included on the list. But the others are FX lenses where vignetting is small relative to the DX sensor. Also most of the lenses mentioned are telephoto primes which tend to have less vignetting than zooms (which are not on the restricted list).

The restricted lenses are mostly super-telephotos where the image forming rays from the rear are nearly parallel, and perpendicular to the sensor. Maybe the cross sensors are optimised for shorter lenses where the ray angle is a bit greater?

I would not want to use the sensors at the extreme edge anyway with my 12-24mm, there is too much field curvature hidden within the depth of field.
I think you are right with respect to parallel rays with the long lenses listed, perhaps the baffles get in the way or perhaps the secondary mirror is not big enough? (Remember how the mirror sometimes can cut off the viewfinder image with long lenses; not sure if that is the case on modern bodies).

[bjornthun]

The 10-24/3.5-4.5 is not on the list, but it may be for those focal lengths with f/4-f/4.5.

[James Farrell]

The AF-S 105 VR micro is f/2.8 to start with, but so is the AF-S 60/2.8 micro. No mention of the DX 40/2.8 or 85/3.5 micros either.
Also not sure why the outer cross sensors don't work with non AF-S lenses.

Back to your original post ... obviously the outer cross sensors require f/2.8 or faster. They work with some f/4 lenses also - the 16-80/2.8-4, 24-120/4 and 70-200/4 are not mentioned in the restricted lists so I assume they are usable. Why there are limitations on the other f/4 lenses is hard to say.

It could be that vignetting means there is insufficient light towards the edges for the cross sensors to work. That may be why the DX 12-24, with its smaller image circle, is included on the list. But the others are FX lenses where vignetting is small relative to the DX sensor. Also most of the lenses mentioned are telephoto primes which tend to have less vignetting than zooms (which are not on the restricted list).

The restricted lenses are mostly super-telephotos where the image forming rays from the rear are nearly parallel, and perpendicular to the sensor. Maybe the cross sensors are optimised for shorter lenses where the ray angle is a bit greater?

Your hypothesis makes good sense. For me, the most credible explanation I've read (so far). Hope this discussion continues and more revelations appear. Thank You. 

[Bjørn Rørslett]

There is an obvious conflict with the "parallel light" hypothesis. Why have the shorter 500/4 lenses more restrictions than the longer 600/4 models? Makes no sense.

[James Farrell]

There is an obvious conflict with the "parallel light" hypothesis. Why have the shorter 500/4 lenses more restrictions than the longer 600/4 models? Makes no sense.

Individual lens design issues perhaps? I don't know.

I'd like to share some new information with you ... I just received the following from a frequent poster at DPreview, Marianne Oelund who, as an electronics engineer, is very well respected among posters and has posted numerous compelling explanations about all matters of camera science. Her explanation and reasoning behind the various limitations that I first inquired about is as follows:

AF systems are full of compromises. One of the main ones is the assumption for the location of the lens exit pupil (i.e., its longitudinal distance from the image sensor). Over the full range of Nikon's AF lens line, keeping the exit pupil location close to the position assumed for the AF system is nearly impossible.
 
Phase detect AF accepts light from pairs of narrow beams coming from the lens exit pupil. These beams are intended to originate from positions +5 and -5 deg away from center (for AF points that work up to f/5.6). You can see how those beams are spaced and shaped here: http://www.dpreview.com/forums/post/54269847 For central AF points, there will never be an alignment issue if the exit pupil moves further away (or closer), however for the far lateral AF points which are looking at the exit pupil from an angle, movement of the exit pupil will shift the beams away from their intended positions.

Now, this is only true for beams which are aimed left/right of center while the beams aimed above/below center will not be affected. The cross-type sensors use both pairs, so they are sensitive to exit pupil distance. Single-direction sensors only use the beams above/below the center of the exit pupil, so will work well even if they are at lateral points - and the beams above/below for the cross-type sensors will also be OK. They can still focus on horizontal lines (parallel to the long edge of the image).

There are also potential vignetting issues for the lateral points, caused either by the limited size of the AF sub-mirror, or the diameter of the lens rear opening. This is not vignetting in the sense that it merely affects the image with darker edges; this is vignetting which completely cuts the light off for one of the AF beams. Users tend to think in terms of AF points losing function because they receive less light, but in reality it's because in those situations the AF points completely lose light from one beam. (end of message)

My takeaway from all of this is (guidance from Ms. Oelund) that focus is still possible at the left/right ends using non-cross sensors and that I should not obsess about it, given the magnificent focus systems of the D5/D500.

[Bjørn Rørslett]

Still, there is a discrepancy as the pupil position of these long lenses aren't that different.

[stenrasmussen]

Maybe these figures can help in the discussion. I've overlain the AF points and how I interpret the actual sensors are arranged when viewed on top of the AF module itself.

[Roland Vink]

There is an obvious conflict with the "parallel light" hypothesis. Why have the shorter 500/4 lenses more restrictions than the longer 600/4 models? Makes no sense.
[...]
Still, there is a discrepancy as the pupil position of these long lenses aren't that different.

That occurred to me while I was writing earlier, I agree it makes no sense given what we know, but it was the best I could think of.
Also, the two AFS 300/4 models are not mentioned, surely the vignetting and/or the pupil position can't be so different from the 500 and 600mm lenses?