Some reversed lenses should be shot through thick glass

[Charles Krebs]

So I will try using transmitted light and I'll pile 4mm of cover slips on the wing.

Lou, you have brought up a very interesting conversation. Definitely seems worthy of more investigation. But I must say that 4mm of cover-slips (about 23 cover-slips, 46 uncoated air/glass surfaces) seems like a real recipe for a disastrous image. There are "optical windows" from places like Edmund and Thorlabs that would seem a much more logical approach. They may not be the exact thickness of the glass over the sensor, but there are some reasonably priced ones that are 3mm thick.

[Lou Jost]

Hi Charles, yes, that was a bad idea. Lots of veiling glare. It wasn't an attempt at a final solution but just a way to quickly test whether this effect is important at the apertures I work at. I'm glad to know about those optical windows! Thanks for the tip.

I gave up on the cover slip test and am just using stacked filters for these tests. They are being stacked right now in Zerene and I'll post them tomorrow. Looks like the effect is indeed significant for MFT lenses.

[Lou Jost]

RoyalWinchester recently posted some tests using his f/1 Repro-Nikkor. This lens is unusual in that the exit pupil is actually at the image plane. It was designed for film. At f/1 it was not very sharp and ChrisR expressed surprise that a lens so specialized showed considerable CA wide open.

http://www.photomacrography.net/forum/v ... epronikkor

Perhaps the issue I have described here is the reason for the problems of this lens wide open. This is the absolute worst case mismatch- the sensor-to-exit pupil distance is 0 (should be >>50mm to avoid problems, see graphs at the pages I've linked to above) and the aperture is super-fast (EA=2, again in the danger zone according to the graphs). The lens expects no glass between the rear element and the image plane, but there is a 2mm piece of glass there.

I would suggest that instead of a Repro-Nikkor, we should use a similar specialty lens that is designed for digital use. The Scanner Nikkor of the latest-model Coolscans comes to mind, 14 elements, 7 or 8 of them made of ED glass. One side of this is designed to image a piece of film, and the other side is designed to put the image on an electronic sensor with, presumably, a filter stack of some kind on it. There was a little bit of info on this lens here:

http://www.photomacrography.net/forum/v ... ner+nikkor

Focal length around 80mm like a repro-Nikkor (85mm). Covers medium forma subjects. No idea of the m.

I have one of those old film scanners waiting for me in the US next time I go there, and I will dissect it and report back the actual data (focal length, aperture, working distance, and thickness of the sensor filter pack). This may be Nikon's modernized digital version of the Repro-Nikkor. If so, it can sometimes be bought "cheap" (ie, cheap relative to the Repro-Nikkor which sells for around $2000). There is one right now on eBay (comes with scanner attached!) for $750:

http://www.ebay.com/p/Nikon-Super-CoolS ... r/66779621

[Lou Jost]

Here are my test results which show that the absence of an expected sensor filter stack does indeed need to be taken into account when reversing fast MFT lenses. Red/cyan fringing is reduced considerably when shooting through two ordinary 46mm filters of unknown thickness, versus shooting through air like people usually do. However, in the fast Pana-Leica 25mm f/1.4 MFT lens that I was testing (mounted in reverse on a 75mm f/1.8 MFT lens), there are other aberrations including a violet/green fringing, that are also important and that are not affected by adding filters.

To keep lighting constant between test series I lit the wing from below, which also created terrible color fringes between the bright background and the dark scales.

So this is a comparison between terrible versus merely bad....but it shows the effect I wrote about. This is at 200%. Left is shooting through air, right was shooting through two filters. Note the cyan and red fringing diminishes, but the violet/green fringing is from some other lens aberration and is unchanged.

[rjlittlefield]

I've mostly been lurking on this thread, but be assured that I find it very interesting.

I had no idea that the filter stack was so thick in MFT format. 4 mm!

Yes, definitely an issue at wide aperture.

Red/cyan fringing is reduced considerably when shooting through two ordinary 46mm filters of unknown thickness, versus shooting through air like people usually do. However, in the fast Pana-Leica 25mm f/1.4 MFT lens that I was testing (mounted in reverse on a 75mm f/1.8 MFT lens), there are other aberrations including a violet/green fringing, that are also important and that are not affected by adding filters.

My guess is that the red/cyan fringing is a simple prism effect, caused by dispersion of rays that strike the filter stack at an angle. If that's correct, then stopping down will not have much effect on it.

I don't have a good feel for the violet/green fringing. Have you checked to see if longitudinal CA is in play there?

... f/1 Repro-Nikkor. This lens is unusual in that the exit pupil is actually at the image plane.

I am curious where this description came from. It sounds scrambled. The exit pupil is the apparent position of the aperture, the place where light apparently comes from to strike the sensor. If that were to be at the image plane, then light would have quite some difficulty making it to the corners of the sensor.

The description at http://redbook-jp.com/redbook-e/record/rep03.html says that "Another outstanding feature of the lens is that the entrance and exit pupils are located far behind the lens, which makes the principal rays (rays passing through the center of each pupil) run almost parallel to the optical axis." Noting that the lens is symmetric, I take that description to mean that the entrance pupil is far behind the lens and the exit pupil is far in front of it, both consistent with principal rays running almost parallel to the optical axis on both sides.

--Rik

[Lou Jost]

Thanks for helping us think about this effect, Rik.

"My guess is that the red/cyan fringing is a simple prism effect, caused by dispersion of rays that strike the filter stack at an angle. If that's correct, then stopping down will not have much effect on it."

I don't understand this comment. The red/cyan fringe is on the image WITHOUT the filter stack. It goes away when I add the filters.

I don't know where the violet/green fringing comes from. Is there a way to figure it out?

"I am curious where this description came from. It sounds scrambled. The exit pupil is the apparent position of the aperture, the place where light apparently comes from to strike the sensor. If that were to be at the image plane, then light would have quite some difficulty making it to the corners of the sensor."

My impression might be wrong, but several writers have explained that this lens is an excellent relay lens. I think I got the pupil locations from an Italian writer whose web page I had to google-translate (hence easily garbled), but I don't have it at hand.

[rjlittlefield]

"My guess is that the red/cyan fringing is a simple prism effect, caused by dispersion of rays that strike the filter stack at an angle. If that's correct, then stopping down will not have much effect on it."

I don't understand this comment. The red/cyan fringe is on the image WITHOUT the filter stack. It goes away when I add the filters.

I'm thinking the lens will be designed to compensate for prismatic dispersion introduced by the slab. That means the lens will essentially have negative prismatic dispersion of its own. Take away the slab that it was designed to compensate for, and the dispersion in the lens appears.

--Rik

[Lou Jost]

So the only way to get rid of that would be to shoot through glass, then?

Does the glass have to be close to the subject or can it be anywhere in the optical path?

I suppose that water or alcohol of sufficient depth would also work? This problem could be a blessing in disguise for those of us who have to shoot specimens in liquid, or, as Pau said, only partly in jest, this might be great for shooting through aquarium glass.

[Lou Jost]

Rik, here is what the Italian article says about the positions of the entrance and exit pupils of the Repro-Nikkor:

"Finally, the Repro-Nikkor 85mm adopts a perfectly symmetrical pattern in which the pupils of entry and exit H 'and H are external to the core of lenses and coincide with the plane of focus on the subject and with the film plane: it is therefore an authentic relay-lens which intercepts the two married in the most telecentric as possible, and theoretically would be a perfect target for even the "crabby" of current 24x36mm digital sensors; This particular feature makes it usable as a relay-lens secondary cascade to a primary goal or as combined with fluorescent optical devices such as CRT (cathode ray tubes) or medical fluoroscopes"


Edited to add: " the input and output pupils coincide with the plane front fire and with the film plane, intercepting the image and recording it on film with light pencils absolutely parallel to each other and perpendicular to the planes of conjugated: we are in the presence of the "perfect telecentric" and considering the extreme characteristics of resolution and contrast can say that this "relay lens" transfers the image almost as if it were a copy in contact. "

https://www.nital.it/nikkor-historical/ ... or85mm.php

[Lou Jost]

Rik, I've tried to see where the violet/green fringe is coming from. I used the same reversed Oly 50mmf/2 macro lens but replaced the 75mm lens with another Oly 50 f/2 macro lens. This does not show any violet/green fringe. I'll post those results in a separate thread. This is my "poor man's Repro-Nikkor" since it is a symmetric fixed-magnification 1:1 lens with an aperture of f/1.

[rjlittlefield]

Does the glass have to be close to the subject or can it be anywhere in the optical path?

As far as I know, theory says that if the slab is perfectly flat and smooth then it can be anywhere in the optical path. This is because the deviation of every ray depends only the angle of the ray, not where it intersects the glass. But if the slab has some long-scale variations in thickness, then closer to the subject is better because the cones of light passing through the glass are narrower so they see smaller variations. As a matter of practice, I suspect that the best of those "optical windows" (e.g. Edmund Stock #47-834, 4mm thick with flatness lambda/10) could go anywhere, but the lower class ones (e.g. Stock #32-740, flatness 4-6 lambda) would be better near the subject.

I suppose that water or alcohol of sufficient depth would also work? This problem could be a blessing in disguise for those of us who have to shoot specimens in liquid, or, as Pau said, only partly in jest, this might be great for shooting through aquarium glass.

If Pau was jesting, I missed that aspect. I took his smilies to indicate an idea that he wasn't sure about. Yes, the lens and liquid effects are at least in the right directions to compensate. Whether the compensation could ever perfect would depend on the exact dispersions. Practical tests required...

here is what the Italian article says...

Thank you for the reference. I don't understand Italian, so I cannot say where the problem lies, but surely the English explanation is scrambled. I believe the actual situation is that the lens is bilaterally telecentric. That is both technically feasible and consistent with all the rest of the explanation, except that the entrance and exit pupils are actually at +-infinity, not in the planes of focus.

Perhaps the issue I have described here is the reason for the problems of this lens wide open. This is the absolute worst case mismatch- the sensor-to-exit pupil distance is 0 (should be >>50mm to avoid problems, see graphs at the pages I've linked to above)

What the graphs really show is that problems get worse when the ray cones are more slanted with respect to the glass. The Repro-Nikkor would be the absolute best case, where all the ray cones will be perpendicular to the glass, hence the website's (translated by Bing) comment that "theoretically a perfect lens telecentric for even the most "grumpy" of current digital sensors 24x36mm". The sensor-to-exit pupil distance is not 0, it is infinity.

--Rik

[Lou Jost]

Thanks, I should have realized that the telecentricity should make it robust against filters in the optical path.

I made a "poor man's Repro-Nikkor" and ran my first test a few minutes ago here:
http://www.photomacrography.net/forum/v ... highlight=

I gathered Pau was half-joking, half-serious, but it is a very good idea!!

[Pau]

Lou, I was not joking, I wanted to express a preliminary good idea needed of further confirmation (as per Rik's interpretation)
Knowing the RI of glass and liquid the calculation of the adequate liquid sheet to match the sensor glass effect would not be difficult (but not being good in maths, I don't attempt to do so)

I fact I was specifically thinking in your alcohol immersed orchids, although I didn't say it

[Lou Jost]

It is a great idea Pau and I will use it!!!! One challenge is that we don't know the refractive index of the filter stack. But it will be easy to do the experiment with varying depths of liquid. I suspect that at the apertures we normally use, we won't have to be very exact.

[Lou Jost]

Rik, maybe telecentricity doesn't make a lens escape the problem of filter stacks. I just found this by Brian Caldwell:

"And even a perfectly telecentric design wouldn't help with filter-induced spherical aberration at large apertures, which is the problem Roger and I initially encountered with the boosted Otus without placing the necessary glass plate in the converging beam."

http://www.fredmiranda.com/forum/topic/1299203/1