01-16-2012, 01:34 PM
Stimulated by [url="http://forum.photozone.de/index.php?/topic/830-do-it-yourself-centering-sanity-check"]Klaus' thread on testing lens-centering[/url], I have been trying to test if my PC-E 24mm f3.5 has the problem of decentering. During the test I found the following problem, which I find confusing: for some reason I rotated the lens by 90 degree to see if the image quality changes. Of course, the shift and tilt knobs were both setted at zero reading (and I confirmed it before and after rotation). But I found that, after the lens was rotated, the image in the live view screen had a shift in the direction in the "vertical" direction-the direction parallel to the short side of the censor (the 24mm side, D700). When the image was magnified to maximum magnification ratio the amount of shift was quite obvious. So I tried to calculate the amount of shift on the censor, surprisingly this amount was somewhat random:
Test 1: short side shift = 0.27mm, long side shift = 0.07mm
Test 2: short side shift = 0.58mm, long side shift = 0.19mm
Test 3: short side shift = 0.38mm, long side shift = 0.07mm
Is this normal for a shift/tilt lens? I can imagine that as you rotate the lens the lens could take a tiny amount of shift or tilt if you did not lock them firmly (and according to the Nikon manual they really can't be locked firmly), but I still wonder a shift like 0.5mm (about 2% of 24mm-the length of the short side of the censor) is too large. It may also be that the tilt/shift scales are not very precise so that althoug they read zero but the true amount of shift/tilt is not zero (indeed I already noticed that the distance scale of the lens suffers large errors).
To make sure that I did my calculation correctly, let me tell you how I calculated the amount of shift on the image censor: I use an image editor to measure the shift in pixels, then multiply it by 24mm (or 36mm for the long side) and divided by 2832 (or 4256 for the long side; the image resolution of D700 is 4256x2832).
Any highlights?
Regards,
Frank
Edit: In test 2 I put the camera in a portrait orientation. It seems that the shift after rotation always predominantly occurs in the short side direction. I don't know if I need do more careful tests. The LV mode is really power-consuming.
Test 1: short side shift = 0.27mm, long side shift = 0.07mm
Test 2: short side shift = 0.58mm, long side shift = 0.19mm
Test 3: short side shift = 0.38mm, long side shift = 0.07mm
Is this normal for a shift/tilt lens? I can imagine that as you rotate the lens the lens could take a tiny amount of shift or tilt if you did not lock them firmly (and according to the Nikon manual they really can't be locked firmly), but I still wonder a shift like 0.5mm (about 2% of 24mm-the length of the short side of the censor) is too large. It may also be that the tilt/shift scales are not very precise so that althoug they read zero but the true amount of shift/tilt is not zero (indeed I already noticed that the distance scale of the lens suffers large errors).
To make sure that I did my calculation correctly, let me tell you how I calculated the amount of shift on the image censor: I use an image editor to measure the shift in pixels, then multiply it by 24mm (or 36mm for the long side) and divided by 2832 (or 4256 for the long side; the image resolution of D700 is 4256x2832).
Any highlights?
Regards,
Frank
Edit: In test 2 I put the camera in a portrait orientation. It seems that the shift after rotation always predominantly occurs in the short side direction. I don't know if I need do more careful tests. The LV mode is really power-consuming.