Investigating Sensor Misalignment

P. Fahrenfort, January 2009.



Why I wrote this page

When I bought a Nikon 70-200VR lens two years ago for use with my D200 I experienced lots of 'issues', which I blamed on a loose lens element, probably the VR unit. The complaints were a skewed plane of focus, visible in some shots but not in others and always at a different angle, random front focus or back focus and sometimes an overall 'murky' look.

I went through three of these lenses over a period of 1.5 years that all had the same 'issues', but the techs at Nikon could not find anything, no matter how much I harassed them. In the end I shied away from the 70-200VR and later sold my D200, because I was disappointed in it and used my D80 most of the time. Then recently I bought a 105 VR for use with my D300 and suddenly all the issues were back like never before!

To make a long story short, after methodically comparing my D300, D80, 105VR, 105MF and 28-105 I finally cornered the problem and found it to be sensor misalignment. In hindsight when I look back over my shots it seems both cameras started acting weird after being aligned at the same repair facility. The main reason to quickly create this page was to help the techs judge if my assumptions are right and to help them locate the source of the misalignment problem. I've never seen this information before and was not aware of the potential problem. Because the actual problem caused a lot of avoidable trouble for Nikon, the repair facility, the shop where I bought my lenses and myself I decided to dedicate some time to present my findings as clearly and readable as possible for others and for future reference. I will adapt this page as new insights emerge.

I focused the D300 shots using live view and Contrast AF. All the NEFs are in the Project Directory. English is not my native language.





Theory

A normal plane of focus is parallel to the sensor plane and the lens plane.




According to the Scheimpflug principle an intersection point S exists for the sensor plane, the lens plane and the plane of focus when either the lens or the sensor is tilted. Normally this principle is used in view cameras to enhance the sharpness distribution over the subject.




When the sensor of a DSLR is misaligned the same effect exists, only unintentional so the plane of focus intersects with the subject in an unexpected manner.




Because the distance between the lens and the sensor only changes a little when focusing point S is fixed, so the plane of focus rotates around it when the focus is changed. Note that when the focus distance is short the depth of field is very shallow but then the effect is almost absent. When the focus distance is long the effect is in full bloom but then the depth of field is wide. This effectively hides the effect to an extent but the sharpness will still visually decline from the center towards the edges.




The angled plane of focus causes unexpected changes in the sharpness distribution over subjects, but also invokes seemingly random front and back focus when the off-center auto focus sensors are used.







Estimating the location of point S

To get the location of point S I pinned the packing of a background roll to a tripod and put two empty beer cans on the edges. I then rotated the contraption and focused with live view until I got both beer cans more or less in focus. The rotation was about 15 degrees at 7.2 meters, so I estimated point S around 30-40 meters left of the camera.




To check this outcome I selected a place where I estimated the distance to the edge of the canal at the other side to be 30-40 meters away. Then I chose three reference points, A, B and C.




This is the *predicted* situation from above for focus point A. The purple line is used later in the story.




Once more the overview of the situation.




I now focus with live view on points A, B and C, take each image and compare them.

First points A of all photos. To be totally clear, in the left image below I focused on the car in point A, in the middle image I focused on point B in the center, which is not visible in the crop, and in the right image I focused on the building in point C at the far right of the image, which is also not visible. Because the sharpness remains (almost) the same this shows that the focal plane travels from point A to C.

If you look at the houses behind the car you can see them get sharper from the left to the right image. It looks like point S is further away than 30 meters, maybe 50, so the plane of focus is not really accurately along the border of the canal as predicted, but close enough to show the effect, at least in my opinion. An actual plane of focus is not a straight line by the way.




Points C of all photos.




You can indeed see that the focus only shifts a little, due to the inaccuracy of the location of point S, while point C is 900 meters further away than point A. This proves to me the Scheimpflug principle is actively shifting the plane of focus.



Finding the extremes

Now I turn over the D300 180 degrees so it's upside down. This reverses the sensor misalignment so the plane of focus now intersects with the canal border at an oblique angle. The purple line is the plane of focus of the upside down D300. Again the chart of focus point A with the *predicted* planes of focus.




Again the shot overview for easy reference.



Below all three images are focused on the car in point A.

The top image is the D300 in normal orientation, the bottom image the upside down D300. Remeber that the building image from the normal D300 is not as sharp as it could be, because point S was inaccurate. Anyway, when you take this photo focused on the car you expect it to look like the D80 image in the middle, but when you use the misaligned D300 the building in the distance can semi-randomly look like anything between the top or the bottom image, only depending on the angle of the subject relative to the optical axis. Image 00038223 is particularly interesting because a flock of birds is flying in the otherwise invisible plane of focus of the turned over D300.





Why is it so hard to detect?

So far it looks rather obvious what happens with a misaligned sensor but now let's look at a less extreme case, focusing at point B. Below is the *predicted* chart when focus is at point B. Note that the two planes of focus of the two D300 photos are much closer together now.




Again the shot overview.




The focus point B is not visible in the images below. The D80 seems back focusing a bit and the estimation of point S was not really accurate. If everything was done accurately the shown parts of the top photo should be sharp, for they're both in the plane of focus, the D80 image parts should be a little less in focus, for point A is in front of the plane of focus and point C is behind it. The bottom image from the upside down D300 should be the least sharp for the plane of focus is at an oblique angle. As the D80 and top D300 building image are close in sharpness the effect is already hiding in the focusing inaccuracies while I shoot @ f/2.8.

This shows nicely why the effect is so hard to find if you're not aware of it. In most cases it tends to dissolve in the inaccuracies that are an unavoidable part of photography and then suddenly the dreaded effect rears it's ugly head to ruin a photo while you can't figure out for the life of you what went wrong.





In the example below the focus point is shifted horizontally but remains at the same distance from the camera. The effect on the foreground is that the '6' gets unsharp in the right image due to front focus as the plane of focus is shifted a little towards the camera. There's also a change in the background blur. If you see this comparison the effect is clearly visible, but if you do not suspect Scheimpflug at work it's very hard to understand the cause of the sudden front focus while other images were fine, especially when shooting handheld. The amount of front focus changes with the focusing distance but the depth of field changes as well.




When you look at a more distant scene the effect gets really hard to find. This is the overview.




Below you can see the focus points in the inserts on the left. The top image from D300 was focused all the way to the left of the image on the bridge so the plane of focus is forced to extend straight into the image, probably going through the center of the horizon. The middle photo is the D80 focused at the center of the bridge. The bottom image is the D300 focused at the center of the bridge like the D80, but now the plane of focus is diagonally across the image in the foreground.

Differences are there, but could you tell they're caused by a misaligned sensor if you didn't know and saw only one of these images? Plus that the aperture is now wide open at f/2.8.







Murky?

Another thing I seem to notice is a 'murkiness' in the images of the D300 that was not there before. This is dangerous ground and I'm mainly guessing here, because I know you can't make this comparison using different type cameras. But a D80 (on the right) is all I got at the moment so I did it anyway. I also focused at different positions for I could not get a good focus lock with the D80 on the utter left. The D300 is focused as far to the left as possible to get the plane of focus extending into the image pointing at the horizon. I also had to process the images a bit differently in the converter, so in the comparison below you're looking at my perception rather than at a real test.

There seems to be more contrast in the D80 image, the D300 seems, well, 'murky' and also seems to have lost some resolution. The right area of the D300 image below (left) has no actual focused parts, for the plane of focus is pointing toward the middle of the image. It seems logical the image quality is less in that situation for this part of the image is completely in the extended depth of field without anything really sharp.





Examples from D200

I shoot a lot and delete a lot, but I found an untouched directory of a short holiday in the summer of 2007 where I used my now sold D200, the 12-24 f/4 I still own, a sold 17-55 f/2.8 and a 70-200VR f/2.8 loaner lens from Nikon. The Nikon techs know that lens very well. In my opinion the next images show that this camera had a misaligned sensor as well. The images are taken from f/4 to f/8 so you have to be ready to do some serious pixel peeping, but the effect is clear.


I placed green dots in different zones on these two images taken @ 120mm f/4. The dots mark the sharpest areas if you look from right to left over the image. The collection of the dots shows the plane of focus.


Click here to see the full horizontal image
Click here to see the full vertical image


In the next image, shot @ 200mm f/5.6, I marked (relatively) in focus areas green and some out of focus areas red. Then I tried to visualize a chunk of the plane of focus.


Click here to see the full image


Both images below are taken with my 12-24mm lens @ 12mm f/5.6. The image with the car was taken in France in 2007 with the D200, the right image was shot in Italy in 2008 with the D80. The scenes are almost equal if you only think about the depth of field. Now look at the sharpness in the crops below. It's obvious the image with the car is shot using the Scheimpflug principle.


Click here to see the full image of the car
Click here to see the full image of the statue




This D80 reference image was also shot @ 12mm f/5.6. I stepped back and focused on the fountain. Both the statue and the white church are unsharp, as they should be. The image with the car cannot exist without Scheimpflug.


Click here to see the full image


At this point I remebered the gallery I made from this holiday with my D200. It contains 5 images, of which 4 are shot exactly like the image with the car above. The fifth is shot with the 17-55 @ f/13. It is strange, for I lost my interest a bit in the wide angle effect; I usually shoot in the 'normal' range. Now I start to see what happened and understand why I chose these images. They're the only ones without 'issues'.




The image below, 12mm f/8, is from the same scene as the gallery image with the tree, above. I focused on the bail in front. I projected the apparent plane of focus and some unsharp areas.


Click here to see the full image


Finally two images from the 17-55, both shot @ f/5.6 but one at 17mm and the other at 55mm 45 seconds later. Especially the 17mm image shows a very strange plane of focus, for everything around it is unsharp. The edges of the ship are clearly out of focus. Using 17mm, f/5.6 and focused at almost infinity (the ship) that's not logical. In my opinion the complicatedly constructed zoom shows so-called 'spherical abbarations' directly outside the plane of focus, for it's not designed for this use.


Click here to see the full image @ 17mm
Click here to see the full image @ 55mm



How to detect the problem

The best way seems to focus manually at 20 to 50 meter distance, take a shot, rotate the camera 180 degrees (very easy with the 70-200VR in the lens mount), take a second shot without touching the focus ring and compare. The shots should be equally (mis)focused with a well aligned sensor and a well centered lens.

Another way is to shoot a field with grass or something where the detail tends to dissolve easily in to the out of focus blur. When both the camera and the field are horizontal the sharp area should be a straight line from left to right. In the image below (00033668.nef) the yellow lines mark the expected 'sharp' area while the red lines show the actual 'sharp' area. Note that I'm reasonably straight in front of the house and the canal edge in the front (look at the wooden boarding). I'm pointing downwards which makes the effect more pronounced I guess. The foreground and background are progressing in blurriness in opposite directions (only visible on the original).





Conclusion

After finishing this page it seems to me as if the x-axis and z-axis were swapped when these sensors were aligned. What is frightening is that one was misaligned two years ago, the other a few month ago. Nobody, not me nor the techs at Nikon recognized the problem for what it is, sensor misalignment. Very strange and I hope the Nikon techs will find the solution quickly.

I'm very relieved I found the problem that haunted me and I hope that this page will help people to recognize or exclude sensor misalignment in an early stage. The photos above are mostly awful and blurry images, what I don't like because I want to take beautiful and sharp photos. So I want to conclude with a better image, from the 70-200VR. Once finances allow I'll surely buy one again, it's an amazing lens.



Continued in