I have tried several techniques (grid etc) for finding the nodal point and I find that my old eyes (66 years) have problems using them. I developed a rail sight that seems to work for me. The rail sight is a C-channel with two screws: one short screw and one long screw. The screws are at each end of the 24 inch C-channel rail. I line the rail sight up with my camera and focus on the short screw while aligning the long screw with the short screw. The long screw sticks up above the short screw. If I rotate the camera clockwise and the long screw moves counter clockwise, then I move the camera forward. Actually, if the long screw moves the same direction as the camera body, then move the camera forward. If the long screw moves the opposite direction of the camera body, then move the camera backward. I have attached three pictures. A side view of the setup, a centered sight view and an off center sight view. Is this correct, or am I missing something?
The problem with your procedure is that you don't have the lens focus set as for a normal panorama shoot. By focusing on the extremely close screw, the position of the entrance pupil is likely to be significantly different. (This depends very much on the particular lens design). Otherwise, your method is not really any different to the usual one where you set up something like a rod close to the camera, viewed against a feature in the background a long way away. I use a piece of sticky tape on a window looking out onto a distant townscape. Alternatively, you could use your improvised sight in the reverse direction to check for movement of the visible entrance pupil as described in http://www.johnhpanos.com/epcalib.htm. After all, you are wanting to get the entrance pupil to be in the same position for two adjacent detent positions. What better than to watch it directly to see if it changes position?
I used the grid method. Our hardware stores sell wire shelves and I used one of them close to the camera and sighted at mid distance objects. Far distance objects probably won't be affected by parallax much (I reasoned).
But I found that looking at the LCD was not really a good way because the LCD was too small to show the small differences that show up in a photograph. So I just took a series of pictures moving the nodal point back 1mm at a time -- starting with my best guess -- over about 10mm then compared them.
Thanks for the info. I did not consider the focus point in my thought process. However, I think you can use the LCD if your camera has the zoom in capability. I do not see any different between zooming in on my camera and zooming in in Photoshop.
I have just bought a Tokina 12-24 lens and have been using John H's excellent tutorial to find the nodal point at 12mm - my variation is to use a grid on a distant wall, lined up with a pin near the lens.
I have been setting the focus at the hyperfocal for the focal length/f-number combination I plan to use in the field - only for the testing I close down the f-stop to make the pin clearer (f22 instead of f8 - would this affect the nodal point ?).
Although I got to zero parallax on the horizontal and vertical centrelines, there are still parallax errors to be seen in the corners - e.g. bottom left corner to bottom right corner. I am wondering if this is the same issue as for fisheyes where the the entrance pupil in a fisheye lens is not a single point - does this also apply for non fisheye ultra-wide-angle lenses ? and will it make stitching difficult ?
Stopping the lens down will not affect the entrance pupil position. If parallax shift is observed in the corners when it has been eliminated at the centre line, then it does imply that the entrance pupil is not a single point. You can, of course, measure the position of the entrance pupil that eliminates the parallax in the corners and see how different it is. If there's parallax in the corners then stitching errors are likely if you have things very close to the camera in those areas.
Thanks for the advice John H and for confirming I'm not doing anything stupid (yet !). Will do a bit more testing at other focal lengths/focus points to see if there is a better 'sweet spot' .