Frans van den Bergh

Hi Gray-Box, Thanks for sharing the raw images, they are quite interesting! I agree that the edge quality is not as good as one would hope. I know that solder masks are made using different processes: some are laser cut (likely the worst edge quality), others are etched (likely much better edge quality). Regardless, their intended application probably does not depend on edge quality but rather feature size accuracy. I recommend using the good old razor blade as a reference for a given magnification,...

Hi! I have a slight preference for the lensgrid chart because the edges are better aligned with radial and tangential directions, provided you position the center of our lens on the center of the chart, of course. And the lensgrid has some fine-tuning to try and avoid the worst case edge orientations (like exactly 45 degrees). This is more of a purist thing, though. A lot of people end up combining the radial and tangential measurements anyway. On the other hand, as long as the edges are "mostly...

Just to add to that, the only feature that requires the circular fiducials is the chart orientation estimation feature. The circular features are not needed if you perform your target alignment manually.

Hi Gray-Box, You can focus on using only the trapezoidal features. My recommendation is to use the mtf_generate_test_chart program included with MTF Mapper to produce a suitable .svg file, then use something like Inkscape to simply delete all the circular fiducials. You can even generate your own target design, as long as the edges of your squares (or somewhat trapezoidal shapes) are oriented roughly along the radial and tangential lines. Just test your chart design using a laser printed copy before...

Hi Chris, Thanks for reporting this. I only have a copy of Windows 10, so my guess is that this is indeed some rendering bug in the OpenGL code when running on Windows 11. A bit of a tangent, but I have also seen rendering issues on Linux when using Wayland. It is indeed a sad day when we can no longer rely on OpenGL to work correctly across different OS versions. It may very well be a bug on MTF Mapper's side that just happened to not show up on Windows 7 and 10? Perhaps one solution is to upgrade...

fixed pathologically bad lensgrid curves using clamping

Hi! I have managed to open and process your image fairly easily. I would recommend to try setting the "Threshold" parameter to 0.4 as shown in my attached screenshot; this seem to be enough to allow the whole image to be processed using the usual "File->Open" method. After that I expanded the tree-view triangle in the "Data set" list on the right, and clicked on the "annotated" list entry. Then I clicked on the cyan-coloured text that says "0.10" on the leftmost edge of the square, which caused the...

Hi Ryan, I think the absolute minimum number of fiducials that must be visible is 5 (not counting the two central markers), if I remember correctly. But with only 5 fiducials you are unlikely to have a good estimate of the lens radial distortion parameters, so this might only work with lenses with very low distortion. I would say a reasonable number is around 10, but again this depends on their distribution across the image; measuring pitch or yaw accurately depends on having fiducials near the edges...