Observations.
It appeared that every 6 frames or so - on the average some of the rows of pixels were missing from the image. 6 or 10 rows (a guess only).
The images were not blurred or stretched in other parts as would be the case with the film being in motion when image.

As part of our algorithms we do change frame sizes after moving the film to a new frames. This allows us to guarantee the first 5mpixel frame we get occurred after film stopped moving. The intermediate frames are much smaller.

If we toss out the first frame back and take the second there is no problem, however scanning takes twice as long which is a big problem.

My guess as to the cause.
o Is it possible that not all images returned are from the same scan of the sensor?
Long ago when I first got the unit I noticed a shearing in the image if I waved the camera around while viewing live video.

o Is it possible that when we switch frame sizes that there is a possibility that the first frame back is actually a combination of two scans and the buffers for the two frames do not line up in memory?

o Although I am guessing we are the only ones using your controller for scanning film I have to believe if this is the cases it would play havoc with anyone using image stabilization algorithms etc.