The camera orientation is changed - the entrance pupil don't move

The center of the entrance pupil must not move when the camera is pivoted around the two axes of the panohead. In the above illustrations the panohead settings are well adjusted and the two yellow axes cross each other at the exact place where the entrance pupil is located.

A series of shots in different directions but from the exact same location

Many tutorials about finding the NPP (No Parallax Point) or entrance pupil of a camera lens are available on Internet, like this one [1]. The following one derives from a method described by John Houghton [2].

To follow this method you need one more digital camera (any P&S will do it) and one more tripod (you could borrow one from a friend photographer.) Besides that, this is a rather intuitive method. The most interesting point is it make possible to estimate the residual error : everybody can dream of the sturdiest tripod and seek for the highest precision a panohead can reach, but then both will be heavy, cumbersome and expensive. On the contrary one can try to find the most efficient compromise by comparing the residual error of the different solutions. This depends on the photographer, there is no one fits all solution.

After I adjusted my Nodal Ninja 3 panohead for an old film camera (I picked this antique camera because it have a lens where the entrance pupil is very easy to see) I shot a series of views of the head and camera aimed at different positions (20° upward, 20° downward, etc.) On the resulting images, the yellow crosses are located in the same place (as much as the camera on a tripod I used for the shots didn't moved). As the entrance pupil size is about 1 mm in size, this can be used to appraise the entrance pupil displacement: I found the precision I was able to reach was excellent (see the enlarged images at the bottom of this page.)

The first steps are rather easy but they must be accomplished first.

- not tilted, the camera must point straight down. (on most panohead the lens must be parallel to the arm where the camera is fastened, in the above example knob 2 is to be used for that.) Use a level as shown in the picture, or use a hotshoe bubble level for this.

- Then adjust so that the rotation point of the tripod is centered. Use knob 1 (in the above example) for this. Call this "top to bottom" adjustment--this adjusts along the portrait side of the image.

Some panoheads (e.g. Manfrotto 303SPH) have another adjustment to move the camera left or right in order to compensate for an off-center tripod mounting hole. Call this one "left to right".

Start by lining up the center focus point on the rotation point of the tripod (there is usually a screw or a knob there).

If you want more accuracy, take one image, then rotate the head 180 degrees (yaw). Put your hand or some other reference mark in the frame so you know this is the second picture and take another image. Use your camera's playback zoom control (don't zoom the lens) until the image is about life-size. By toggling between the two images, you can see how much they are off. Note that because the images are life-sized, you move the panohead about as much as the amount shown in the LCD.

To check these two axises, bring these images into photoshop, and make them two layers of the same picture. Make the top one 50% transparent. The rotation point should line up exactly. Note-use the screw or knob outline, the lines you draw on the screw may not be precise.

The final adjustment is "fore and aft" controlled by moving the camera along the arm it is attached to. This can be done in several ways, one of which is detailed below. Note that you must get the left to right and top to bottom adjustments as close to perfect as possible before starting in on the fore and aft adjustment.

The remaining part of this paragraph deals with the third step, this is the less easy among the three steps.

The diaphragm of the Leicaflex 21 mm lens is such that the lozenge shaped hole in its middle was easy to locate and to photography: to take the above images an Olympus E-330 on a tripod was fine but in most cases the easiest way to locate the entrance pupil is to place the camera against a light source: the entry pupil is then the small bright disc you can see. If the camera have one, actuating the depth of field preview button will change its size (you can then adjust the aperture to get a convenient size.) On the Olympus rear screen, using the enlarged Live View B mode (up to 10x), it was easy to adjust the camera position forward or backward until the entrance pupil didn't moved anymore on the rear screen while the panohead was rotated.

Watching the rear screen works perfectly when the observing camera is a compact or a bridge camera. If you are using a DSLR for that, you will end taking some shots to be examined on the rear screen using the enlarged playback mode to get a closer view.

- for the tested camera entry pupil being brightened, placing a lamp or a window behind this camera is enough (if DOF preview is available on the tested camera, you could use it to adjust the size of the "hole".)

- watching the moves of the tested camera entry pupil on the rear screen of the observing camera is easy. When using a DSLR for that, taking a shot and displaying it on this rear screen is OK (but slower.) In both situation you should zoom to get a closer view.

- to appraise the amount of the residual error, placing a ruler near the camera is the way to go (try to place it at the same distance from the shooting camera as the NPP.)

The principle could be seen as a reciprocity rule : for example when, as seen from a 6 feet distant point, the no parallax point to be verified is moving by 1 mm then the parallax error for an object located at 6 feet from the camera will be equal to 1 mm (the proof is easy, the two implied triangles are equal but we will skip the math.)

To verify the panohead settings for a 50 mm or equivalent lens I suggest to make a series of shots of the camera and corresponding lens as set on the panohead :

- rotating the head by 20 degrees from one shot to the next

- placing in the view a graduated ruler at the same distance than the NPP.

When this is done it's easy to appraise on the resulting images the entrance pupil displacement from one shot to the next. This is the residual parallax error corresponding to the panohead settings for this lens. Using the following approximate values and knowing the nearest focused object in the panorama, one can guess the parallax error amount in pixels for this panorama (a 1 pixel error is not visible, a 10 pixels error is clearly noticeable, a 100 pixels error is difficult to compensate for.)

27 degrees is the horizontal FOV for a 50 mm lens and a 35 mm camera in portrait orientation so that one pixel (on most DSLR) corresponds roughly to:

- 1 mm for a 15 feet distance

- .2 mm for a 3 feet distance

4.5 degrees is the horizontal FOV for a 300 mm lens and a 35 mm camera in portrait orientation so that one pixel (on most DSLR) corresponds roughly to:

- 10 mm for a 1000 feet distance

- 1 mm for a 100 feet distance

- 0.1 mm for a 10 feet distance

For a fisheye giving a rectangular image covering the whole sensor, a 5 feet distance (the ground next to the tripod), a 100 degrees horizontal FOV, a 60 degrees rotation between shots (6 views for 360 degrees) should correspond to a 1 mm residual error (100°/360°) x (2 pi x 5 feet) / 2300 pix

The above values suggest that the needed precision is strongly related to the kind of panorama the panohead and the lens are used for, to the number of pixels in the final result and to the distance of the nearest focused objects. For a high-resolution mosaic some problems could arise for a tree distant of 100 feet or more. The same error would be absolutely unnoticeable for 360 x 180 panorama and a fisheye lens (this error would then correspond to much less than one pixel on the panorama.)

This could look inaccurate because a well known fact is that a panohead is needed for QTVR and the right answer could be that a 4 inch error (a common error when the camera is handheld) is too much in a room while a precision of 1 mm (one percent of 4 inches) is useless. As far as I know such a precision for the NPP of a fisheye lens is meaningless.

At my own risk, I will say that when evaluated in millimeters, the parallax error don't depend on the lens nor on the object distance (a one inch error at the lens level cause a one inch error for an object located next to the lens and for a distant star.) What is varying, and there are huge variations, is the angle under which this error is seen by the camera.

- the smaller the lens FOV (the longer the focal length) the more visible the errors

- the higher the sensor pixel number the more visible the errors

- the less distant the nearest focused objects the more visible the errors

- the larger the depth of field the more visible the errors

When do you need a panohead ?

The truth is: I don't know.

Being able to measure the residual error will be very useful to refine this answer, I believe. It would be interesting to photography and measure the NPP displacement for a given camera/lens pair when using a strong or a light tripod, when using a cable release or not, when using a monopod or when the camera is hand-held. Obviously, the results would depend on the ground and the wind and would depend on the photographer skill and many other factors.

This could be an heavy task but should be useful for selecting the right stuff or nothing at all when a hand-held camera is the only possible solution or is enough. Knowing the cost of a pound of panohead, to find the weak link in a pano gear chain would be nice, too.

The following are zoomed-in views of the panohead. Yellow lines (corresponding to the panohead axes) are linked to the camera used to shoot, the red dots were placed in the center of lens diaphragm (click on the image to see this red dot).

Feel free to verify my calculations, pleez correct ze misspellations and ze gramar and let you discuss of this subject on the forum!