Now, getting back to the issue of how many pixels are needed, and looking at some of those implicit and explicit assumptions...
As Charles Krebs noted, by using a 90 mm tube lens, you've pushed the 10X objective down to 4.5X .
On the camera side, at 4.5X the optics will be working at an effective f-number of 4.5/(2*0.28 ) = f/8.
Going back to those formulas I mentioned
earlier, this gives a cutoff frequency around 226 cycles/mm at 550 nm, requiring Nyquist minimum sampling of 452 pixels/mm but more realistically 1.5-2 times higher to avoid excessive contrast loss due to issues with pixel alignment, anti-aliasing filters, and such. Call it 2*452 = 904 to catch everything. Your results at 840 pixels/mm, 648 pixels/mm, and 181 pixels/mm seem pretty consistent with theory.
They are also consistent with other experiment. I am thinking specifically of the thread titled
Pixels for use at 4-5X on an APS-sized sensor, where at
http://www.photomacrography.net/forum/v ... 164#101164 there is a direct visual demonstration followed by a simple calculation to end up with the result that at f/11 a reasonable number is 47 megapixels needed. Scale that by a factor of 2 to account for the difference between f/8 and f/11, to get a number of 94 megapixels, and this seems pretty similar to your statement that "
An APS-C sensor should have 120 megapixels, not 10 or 20."
That number, say 120 megapixels, will apply whenever using the same objective (NA 0.28 ) to view the same size subject area. It's in that latter issue that things get a little sticky.
If you have an object that's 3 mm across, you can spread it across 5000 pixels on the bridge camera sensor using just 2x optics. Or you can spread it across 5000 pixels on an APS-C sensor, but you'll need 8x optics.
True, but in both cases you'll be needing the same NA on the subject side.
The price difference between these two options is very large. It may be useful to be aware of this cheap easy alternative.
Perhaps, but I'm not convinced about exactly how large "very" is. When I look at prices, I'm seeing the Sony Cyber-shot DSC-HX400V at 20.4MP for about $430. Meanwhile a Canon T5 at Adorama, with 18-55 mm kit lens, is $399. To make that work with the Mitutoyo objective at rated 10X magnification would require something like a 200 mm telephoto; I'm thinking the Canon EF-S 55-250 would work OK, add $179 for that. Or maybe even the body + 18-55 + 55-300, currently offered at
http://www.adorama.com/ICAT5K1A.html for $499.
This doesn't strike me as a huge difference, particularly since we've implicitly assumed that there's a Mitutoyo 10X NA 0.28 sitting out front -- list price $885 for that.
I don't intend any of this to argue that a bridge camera is a bad idea. However, given the historical problem of vignetting with zooms at much less than max focal length, combined with your observation that
"There was also vignetting at middle and lower focal lengths.", I think some caution is called for.
I have no problem believing that using a bridge camera is a quick and easy way of capturing the
center of the Mitutoyo's field -- much quicker and easier than sticking a long zoom on APS-C. But for capturing the whole field, or even as much of the field as would be typically captured by an APS-C setup, I'm still very concerned about the vignetting issue.
Comparative images
at same field width would be very helpful for addressing that concern.
--Rik