Measure lens resolution with a lower resolution sensor?

Beatsy · Post by **Beatsy** » Tue Mar 19, 2019 2:27 pm

I have a few "100 megapixel ready" camera lenses. That is, the manufacturers (Zeiss and Sony) claim they can resolve detail onto each and every pixel of a 100 megapixel FF sensor (which doesn't exist - yet).

The highest resolution sensor I have is 42 megapixels (Sony A7rii). Can I test if the lenses live up to these claims using this camera? If so, how and with what other equipment?

Thanks

rjlittlefield · Post by **rjlittlefield** » Tue Mar 19, 2019 4:53 pm

Maybe look at the aerial image with a microscope objective?

Like at http://www.photomacrography.net/forum/v ... 164#101164 .

Best if you have a bellows with shift, to capture the off-axis parts.

--Rik

Lou Jost · Post by **Lou Jost** » Tue Mar 19, 2019 4:59 pm

Usa 2x teleconverter and see if that outresolves the sensor. If so, you confirm the claim. If not, then you won't know whether to blame the lens or the TC.

Beatsy · Post by **Beatsy** » Tue Mar 19, 2019 5:42 pm

Thanks Rik. I vaguely remembered that thread, but didn't know what to search for. Ta for the link.

And thanks Lou. a 2x TC was my first thought (I have one), but then I thought it's own performance would skew the results. It didn't occur to me to use that test as a "gate". I'll definitely try that first. If I'm left with the second outcome (not sure if TC or lens is sub-par) then I can move on to Rik's solution.

A project for tomorrow (along with finding/making an objectively measurable target of some sort).

Cheers

ray_parkhurst · Post by **ray_parkhurst** » Tue Mar 19, 2019 8:56 pm

Which 100MP-ready lenses are we talking about? I guess we're talking 3um pixels on FF.

enricosavazzi · Post by **enricosavazzi** » Tue Mar 19, 2019 11:26 pm

Just for testing (as opposed to saving usable images at 100 Mpixel resolution), you could use a small sensor with small pixels and shift it sideways (leaving the lens immobile) to test in the periphery of the image circle.

In practice, this would mean using a micro 4/3 camera, which has pixel size around 3-3.5 micron. An HD webcam would give you even smaller pixels.

The depth of the sensor within the lens mount flange puts a limit to the amount of shift possible without shadowing the sensor. Still, a substantial amount of shift is possible with most Micro 4/3 cameras.

Beatsy · Post by **Beatsy** » Wed Mar 20, 2019 2:42 am

ray_parkhurst wrote:Which 100MP-ready lenses are we talking about? I guess we're talking 3um pixels on FF.

All three of the Zeiss Otuses and some(?) Sony GM lenses (only based on a comment from a Sony engineer that GM lenses were 100mpix-ready).

Yes, just a smidge under 3 microns.

Lou Jost · Post by **Lou Jost** » Wed Mar 20, 2019 5:13 am

Enrico that's a good idea. Yes, you can pretty much capture a FF image circle with an MFT sensor by shifting.

ray_parkhurst · Post by **ray_parkhurst** » Wed Mar 20, 2019 1:44 pm

Beatsy wrote:
ray_parkhurst wrote:Which 100MP-ready lenses are we talking about? I guess we're talking 3um pixels on FF.
All three of the Zeiss Otuses and some(?) Sony GM lenses (only based on a comment from a Sony engineer that GM lenses were 100mpix-ready).

Yes, just a smidge under 3 microns.

Assuming 3um pixels, that means ~170lp/mm. I think that can be resolved at f8, so aren't most lenses "100MP-ready", assuming they are diffraction-limited at f8 or larger?

Beatsy · Post by **Beatsy** » Wed Mar 20, 2019 2:12 pm

ray_parkhurst wrote: Assuming 3um pixels, that means ~170lp/mm. I think that can be resolved at f8, so aren't most lenses "100MP-ready", assuming they are diffraction-limited at f8 or larger?

I'm not convinced "most lenses" are that capable, but accept I could be wrong. High quality (and high ticket) primes maybe, but even then I'm not convinced it would be most.

The best I've seen reported for an Otus lens was around 120 lpmm (forget which) but that was on a 54mp FF sensor which is likely the limiting factor. That's why I've become intrigued with testing the "absolute best" a lens can resolve in practice.

I got sidetracked today but will dive into some tests based on the suggestions above ASAP.

wpl · Post by **wpl** » Wed Mar 20, 2019 2:41 pm

Don't slanted-edge MTF measurements extend well past the Nyquist frequency? I've done some measurements like these without really understanding the theory. I think the magic is that information from many pixels is used to determine the resolution (actually contrast for given spatial frequency) in a given region of the detector.

rjlittlefield · Post by **rjlittlefield** » Wed Mar 20, 2019 3:11 pm

Beatsy wrote:
ray_parkhurst wrote: Assuming 3um pixels, that means ~170lp/mm. I think that can be resolved at f8, so aren't most lenses "100MP-ready", assuming they are diffraction-limited at f8 or larger?
I'm not convinced "most lenses" are that capable, but accept I could be wrong. High quality (and high ticket) primes maybe, but even then I'm not convinced it would be most.

Consider http://www.photomacrography.net/forum/v ... 540#163540 . That's a 50 mm f/2.8 Schneider-Kreuznach Componon S at f/4 and 1X (effective f/8 ), shown to out-resolve a 36 MP Nikon D800E by about 2X on each axis.

I'd expect any lens that's decent at f/8 to have similar characteristics. If it's decent at f/5.6 even more megapixels will be required to capture all the detail that the lens is capable of forming.

around 120 lpmm (forget which) but that was on a 54mp FF sensor which is likely the limiting factor

Yeah, that sounds about right. 54 MP on FF is 4 microns per pixel, so 8 microns per cycle = 125 cycles per mm.

--Rik