Argh! AF ... no issue, sensor tilted!

Well, I've struggled with soft images and thought maybe it was me. I've taken far far too many test shots (perhaps 1000) and I'm now convinced I have NO AF issue.

When doing yet another focus test (this time in the kitchen as there's better light!), I noticed that in this case, I have an impossible focus (see below). The top row is sharpest at the back corner, and the bottom row is sharpest at the front corner. Even though I'm slightly high of target, this cannot be! So I checked other images, and ditto, to much lesser extents. So, there you go, AF is working, but the sensor must be tilted (or 10+ lenses of mine are tilted - I think not!).

So, off it goes to Fuji and probably a replacement camera I would think, since this isn't going to be straight forward to fix.... that's assuming they concur with my findings.

Excal

PS: Sorry the image is a bit low res, but I think you can see the issue.

excal said:

Well, I've struggled with soft images and thought maybe it was me.
I've taken far far too many test shots (perhaps 1000) and I'm now
convinced I have NO AF issue.

When doing yet another focus test (this time in the kitchen as
there's better light!), I noticed that in this case, I have an
impossible focus (see below). The top row is sharpest at the back
corner, and the bottom row is sharpest at the front corner. Even
though I'm slightly high of target, this cannot be! So I checked
other images, and ditto, to much lesser extents. So, there you go,
AF is working, but the sensor must be tilted (or 10+ lenses of mine
are tilted - I think not!).

So, off it goes to Fuji and probably a replacement camera I would
think, since this isn't going to be straight forward to fix....
that's assuming they concur with my findings.

Excal

PS: Sorry the image is a bit low res, but I think you can see the
issue.

Click to expand...

But your test is not definitive. In order to make such a test you would need a grid or test chart that will impact on the ccd full size with a measured guarantee that the ccd and subject are in absolute paralell.

A good way to do this is to find a house with horizontal siding, drop a couple of ribbons vertically and carefully lay a line perpendicular to the house, stick the camera there with a level and try again. This will almost always show a fault either with curvature of field or something else.

You are not going to diagnose your problem with any kind of makeshift test, the errors you are trying to identify are very small and will require asolute accuracy in testing.
Melvin

It looks like the results I would expect. You are testing for flatness of focus but you have many intersecting planes. The camera is too high above the test target and the sensor plane is placed to intersect the target plane on two axis. That is why you have two points of focus on your test card. The target distance from the focal point is the same at two distinct points, where ideally you want the planes (sensor, lens, and target) to be parallel to each other. When working with large format cameras you can control the plane of sharp focus by adjusting the lens tilt to intersect the film and subject planes, but you don't have the individual controls of a view camera on the S2.
Ross

If he's having the same problem with all his lenses then he may have a lens mount that is not parallel to the sensor plane. That would make all his lenses "tilt" in the same direction. When I worked on cameras years ago, I was taught to be very careful to intall the shims that were behind every lens mount in the same position I found them. If you made a mistake it takes very special equipment (used by the factory) to figure out where the shims need to go. Usually the shims were in the form of very thin washers with different numbers of them behind all but one of the lens mount ring's retaining screws. With this said, I do agree that I would not come to this conclusion from the picture that was posted here.

ross den otter said:

It looks like the results I would expect. You are testing for
flatness of focus but you have many intersecting planes. The camera
is too high above the test target and the sensor plane is placed to
intersect the target plane on two axis. That is why you have two
points of focus on your test card. The target distance from the
focal point is the same at two distinct points, where ideally you
want the planes (sensor, lens, and target) to be parallel to each
other. When working with large format cameras you can control the
plane of sharp focus by adjusting the lens tilt to intersect the
film and subject planes, but you don't have the individual controls
of a view camera on the S2.
Ross

Click to expand...

--
Mike K.

Friends are God's life preservers

Maybe you are right but the test is unconclusive.

Photographing a test chart on the wall on "autofocus" and "manual" with the help of the focussing dot in the viewfinder would be better.

By lack of a good test chart, I use a folded out news paper with type everywhere (loosely) taped to the wall.

I set up my camera by taping a flat piece of mirror on the wall and aim the camera in such a way that the lens of my reflection in the mirror is exactly in the middle of my camera thus guaranteeing a perfect parrallel to the wall view point.

I use a small but flat make-up mirror. It goes with me when I have copywork to do. It is similar to the very expensive mirror setup that Hasselblad makes.

The tests should show any misallignment and tests the focus mechanism for accuracy.

It is best for the tests to shoot the largest tif size rather than RAW (according to Fuji service) and use a sharpening of "normal".

They say that with the sharpening applied it is easy to see where out of focus areas begin as there is only sharpening in sharp detail.

I did my tests on RAW (converted unsharpened) and sharpened at 500/0.3/0 and again on 200/0.3/0 in Photoshop.
That seemed to have worked for me.

If the CCD is tilted, the top and bottom may be out of focus but a strip somewhere between the two should be sharp from left to right. It will be easy to see if you shoot with your sharpest lens at a large lens opening to limit depth of field.
Rinus of Calgary

A quick update on this...

Thinking about this in work today, I can't explain what I saw with a tilted sensor. In fact, aside from maybe not being perfectly square to the target, I can't explain it at all, so that must be it - my angle to the target in that one shot.

Testing again tonight, I have now worked out the problem, and yes, my camera is faulty. I feel much better about this now that I know what the problem is. My results are below (it's taken many many hours to get to the bottom of this and I'm very very confident now that I'm right about the result).

LENSES CONTAINING NO MOTORS:
Sigma 105mm f2.8 (macro) AFD = Perfect!
Sigma 24mm f2.8 AF = Back focus problem (2-3")
Nikon 50mm f1.4 AFD = Back focus problem (2-3")
Sigma 28-70mm f2.8 AFD = Back focus problem (2-3")
Sigma 100-300mm f4.5-6.7 AFD = Back focus problem (2-3")

Sigma 28-105mm f4-5.6 AFD = Inconclusive - but it's a cheapy rubbish lens anyway!

LENSES CONTAINING INTERNAL OPTICS MOTORS:
Sigma 70-200mm f2.8 AFS = Perfect!
Sigma 50-500mm f4-6.3 AFS = Perfect!
Nikon 18-35mm f3.5-f4.5 IF-ED = Perfect!

MANUAL FOCUS:
Pretty good... well, most of the time.

Interesting isn't it. I can't explain the 105 Macro, except it has a wierd long focus thread, so perhaps it's making the AF system life easier. Possibly less motor driving power required, so maybe with more strain on the motors with the other normal lenses, the AF system is starved of juice? And they had that "Err" power failure dead camera return problem over Xmas didn't they?! Hmmm....

But it does seem to indicate the camera is at fault, not the lenses, so on Monday it's going back to Fuji. Since the sensor and view-finder are aligned right, it should be possible to fix this with a new AF module (a CAM900 I seem to recall?)

Excal

Oh, and if you didn't work it out from my last post...

1) If the lens has motors internal to it (ie. IF and AFS lenses) it focuses fine
2) If the lens has no motors in it (ie. AF and AFD), it back focuses 2-3"
3) The view-finder for manual focus is aligned fine

And note, I did have at least one of each type to test (AF, AFD, IF and AFS).

Excal