Continuing with tests of my new scanning rig, i found the following:
With a 100mm lens on APS-C, the raster of an iPad backlight registered when the scan was taken at f/11, but not at f/8. Imaging ratio was about 1:2, which corresponds to a 645 repro on full frame.
Example at 300%
Conversion, straight out of NLP, of a 1970s negative scanned with “violet” (lavender) backlight.
From a colour point of view, NLP does a great job, considering how far this backlight is away from the recommended white, high CRI light. My point is not colour though. I try to minimise digital noise by pushing the backlight’s R and B channels up to the level of the R channel response to the orange mask. I think that the app proposed by @VladS above could help in the process.
I usually convert at very low pre-saturation and “Basic” colour model, here it was “Frontier”.