How do I test the registration?

Registration tests are tricky and have to be exact, so Richard Garbutt wrote this (with compositing in mind):

35mm Kinor 35H with 75mm lens
Bruce Taylor's Kinor 35H with 75mm
It is also useful to shoot a grid, preferably of white lines on a black background [like this chart from Panavision]; use a clip of only about 50-60 feet. Before you start the camera rolling, remove the lens, open the shutter, and mark the frame outline with a fine felt-tipped pen. Use the inching knob and repeat this for several frames. Replace the lens and shoot the grid. It's useful to put a little tag on the grid, listing the date, camera serial number, and anything else you might want to be sure to remember, like, say, the framing rate. If you plan to use the camera at rates other than 24 fps, it's great to do a separate clip for, say, 4 fps, 12 fps, 18 fps, 24 fps, 32 fps, 48 fps... you get the idea. For all these tests, the camera must be mounted on a tripod that is ROCK solid.

Once you've shot your clips, (which you've carefully labelled on the mags as well, so you don't confuse them) retire to the darkroom, rewind to the beginning (if you don't have darkroom rewinds, this is why keeping the clip to 50 feet or so will save you frustration) and relace each clip in the same mag it was shot in. Be sure you haven't rethreaded farther in than your marked frames.

Back in the light, dock your first mag back on the camera. Remove the lens, and use the inching knob to bring the first marked frame into the gate. If it doesn't line up with your first aperture marks, undock the mag, release the perf off its indexing pin, and move it so the aperture will line up. Replace the mag. Move the chart on a diagonal a distance roughly 1/3 the distance between your grid lines. Remove your ID tag from the chart - you only need to expose that info once.

Konvas 2M with 17ep crystal sync motor
Konvas 2M with crystal sync motor
Now, double-expose the shifted grid onto your original image at the same fps rate as your original exposure. Repeat for the other mags shot at other frame rates.

Have the lab process it all. Now, view the ORIGINAL NEG on a projector with no masking aperture plates in its gate. Do NOT look at a print - this just adds the printer as a possible variable.

Without aperture plates in the projector, you should see the sprocket holes projected as well. Wander up to the screen and stare hard at them: they should not move, relative to the screen. If they do, there is a problem with steadiness in the projector movement. It's also a great time to squint at the upper and lower edges of the perfs to see that your camera is not putting any little nicks in them, which would indicate a worn camera pulldown claw.

If the perfs are nice and steady, then it's time to look at the grid. If you do not see the double-exposed grids moving with respect to the perfs or each other, shout Hallelujah, because that means your camera is fine. If the grids move with respect to each other, your camera is exhibiting weave, and needs servicing - possibly the movement (which is why it's good to check with more than one mag) or mag tension too tight. If both grids move in sync with respect to the screen, it's another clue the projector movement needs work.

modified 35mm Kinor 5000
Aranda Group's Kinor 5000
BEWARE: I've seen inexperienced folk loftily explain that you can "just rethread the camera test and shoot the second exposure tail-to-head," avoiding that pain-in-the-neck rewinding in the dark process. Wrong. Wrongwrongwrong. If you do this, one exposure will have the camera using one edge of the perfs for pulldown; the second exposure will use the opposite perf edge for pulldown, and the test is not valid.

This registration test is absolutely essential if you're hoping to do compositing work. And if you're going to do comp work, you should be using pin-registered camera gear.