After ordering the internal base plate and front plate for the lens adapter machined in aluminium the camera can be finally assembled and the sensor to lens distance is easier to control.

An nvme SSD was installed above the interface board just 1mm above the sensor assembly for storing images.
The next version of the interface board got 1.5k pulldowns on each data line to clean up transitions and reduce the risk of glitches.

One big challenge is aligning the distances between the 2 sensors because if the second camera should help with focusing it must be on the same focus plane as the main sensor. Several different mounts have been printed and many more are to come until it is good enough.

Also the FFC routing and protection during movements is tricky.

The next big challenge is pixel shift correction and saving bitmaps directly from the application while capturing the image without external postprocessing.

As previously noted depending on the resolution mode the sensor has 3, 6 or 12 different active lines with different offsets to each other making an image appear blurry/shifted/jagged without correction.

For the correction a cyclic line buffer is used to avoid storing the full image in ram and instead the offsets are applied after every scanned line. It might be possible to speed this up further using parallel threads...

The big milestone is finally exceeding the resolution of commercial digital cameras by capturing a 21700x10000 200MP image:

And a zoomed section of the "Input C" of the frequency counter:

Note that there is still slight chromatic abberation and that every second line is slightly darker as the sub-lines seem to have a slightly different sensitivity.