Axial Lights II (Ghostbuster?)

[rjlittlefield]

in my case with the slanted glass the light would be travelling through the same amount of glass regardless of where it hits along its surface - assuming the light is hitting it a 45 degree angle relative to the surface.
...
I'm likely missing something obvious as I have a long way to go before being able to keep up with many members here when it comes to the behaviour of light and optics.

The underlying issue is that the lens accepts a cone of light from every point on the subject. That means it is seeing light that travels through the glass at different angles. When the glass is slanted with respect to the optical axis, the light on the "high" side of the cone spends more time inside the glass than the light on the "low" side of the cone.

Here is a diagram to illustrate. This is a ridiculously wide aperture compared to what you'll be using, but hopefully it gets the point across. Compare the lengths of the bold lines. They represent the amount of glass to be traversed by light on the "high" and "low" sides of the cone. It's that difference in time inside the glass that causes the aberration.

--Rik

[kaleun96]

in my case with the slanted glass the light would be travelling through the same amount of glass regardless of where it hits along its surface - assuming the light is hitting it a 45 degree angle relative to the surface.
...
I'm likely missing something obvious as I have a long way to go before being able to keep up with many members here when it comes to the behaviour of light and optics.

The underlying issue is that the lens accepts a cone of light from every point on the subject. That means it is seeing light that travels through the glass at different angles. When the glass is slanted with respect to the optical axis, the light on the "high" side of the cone spends more time inside the glass than the light on the "low" side of the cone.

Here is a diagram to illustrate. This is a ridiculously wide aperture compared to what you'll be using, but hopefully it gets the point across. Compare the lengths of the bold lines. They represent the amount of glass to be traversed by light on the "high" and "low" sides of the cone. It's that difference in time inside the glass that causes the aberration.


2023-01-04_14-05-22.png


--Rik

Thanks for that explanation and diagram, made everything perfectly clear! I probably should have been able to have made that deduction given what you had said previously about glass thickness not affecting this type of aberration in infinity sections. It's easy to see why that is the case by imagining parallel rays in your diagram above.

[mjkzz]

in my case with the slanted glass the light would be travelling through the same amount of glass regardless of where it hits along its surface - assuming the light is hitting it a 45 degree angle relative to the surface.
...
I'm likely missing something obvious as I have a long way to go before being able to keep up with many members here when it comes to the behaviour of light and optics.

The underlying issue is that the lens accepts a cone of light from every point on the subject. That means it is seeing light that travels through the glass at different angles. When the glass is slanted with respect to the optical axis, the light on the "high" side of the cone spends more time inside the glass than the light on the "low" side of the cone.

Here is a diagram to illustrate. This is a ridiculously wide aperture compared to what you'll be using, but hopefully it gets the point across. Compare the lengths of the bold lines. They represent the amount of glass to be traversed by light on the "high" and "low" sides of the cone. It's that difference in time inside the glass that causes the aberration.


2023-01-04_14-05-22.png


--Rik

Nice explanation!

[mjkzz]

Well it's that time of the year again when I decide to revisit axial lighting for no good reason. I've been trying some various setups the past few weeks, partly influenced by this thread that I forgot to follow up on earlier in 2022. Reading through it again, as well as some of the old threads Rik linked, helped me understand a lot about what is going on and my issues with axial lighting.

I tried half mirror and it works great, maybe you can try it, too. Take a look at this thread over coin community

[kaleun96]

Well it's that time of the year again when I decide to revisit axial lighting for no good reason. I've been trying some various setups the past few weeks, partly influenced by this thread that I forgot to follow up on earlier in 2022. Reading through it again, as well as some of the old threads Rik linked, helped me understand a lot about what is going on and my issues with axial lighting.

I tried half mirror and it works great, maybe you can try it, too. Take a look at this thread over coin community

Interesting, seems we have been working in parallel quite closely

I had done some tests with various lens filters, even those as thin as 0.7mm, but none were particularly good. Some had quite bad aberrations, others a strong tint, and none reflected much light - likely due to the double AR coatings. I ended up ordering one of those 80x80x1.1mm 50/50 beamsplitters that look similar to the one you got and the one Ray mentioned. Took a few weeks to arrive but it has proved to be really good in practice. Reflects a tonne of light and seems to be much sharper than the UV filters - just have a stack running now to test.

I had come up with a similar box to yours but the glass inserts from the top. It fits over an existing coin platform setup I've been using. I think your solution might protect the glass better, it's not like I need access to it anyway. I made the two sides of the box dovetailed so I can print new "sides" and replace the existing ones. I need to print a back piece with flocking, currently I'm using flocking behind the box but would be nice to integrate it into the box. I also will reprint the side facing the light source so I have a way of easily attaching diffusers and similar, maybe a magnetic solution will work here.

I liked the setup you attached to the camera lens too. Maybe I might try that solution to see how it works. The benefits that I see are that you can move the axial box out of the way by moving the camera out of the way, rather than moving the camera and then taking the box away to access the coin. It'll also allow for more distance between the mirror and coin, meaning I can potentially do some additional light modifications, such as illuminating the coin's edge with another light source or adding a reflector. The downside is that it requires a more difficult printed solution and the box will move relative to the light source, but a few mm of vertical movement should be fine.

[kaleun96]

Here's the stack test I did today. First photo is shot axially with the 80x80x1.1mm beamsplitter. Second photo is non-axial, using my custom "ring flash" adapter. You can click on these photos to see high resolution versions.

In the comparisons below those, the one on the left is axial and the one on the right is non-axial. N.B. I forgot I had sharpened the axial images when taking these screenshots so they may have that appearance of being sharpened relative to the non-axial images.

Overall, I'm quite pleased with the sharpness of the axial setup, much better than I've achieved previously. There is still some aberrations that are discernible but not always so. I can notice in particularly in the 2nd and 3rd comparison, less so in the first. It seems where you have a sort of finely detailed pattern you can see the aberrations have swirled the detail a bit but this is almost at the level of pixel-peeping. I think you could still do a high quality macro stack at these magnifications and not be too put-off by the reduction in sharpness.

Axial (click here for higher res)

Non-Axial (click here for higher res)

[ray_parkhurst]

Please remind me what lighting method is the non-axial?

[mjkzz]

yes, actually, cheap UV filters work better, those with any kind of coating either have some kind of colour cast or aberration. I did try a piece of glass, 1.0mm thick, supposedly to be K9 optical grade, it is cheaper than half-mirror, and works well, too, just less light.

The one attached to the lens was intended for stack and stitch where when you move subject, normal lighting from side might cause lighting condition change -- lighting and camera view will be different for each segments, and eventually cause problems in stitched image. It is not practical to build a super sized coaxial box, so attaching it to the lens is a good solution. But be aware that it takes away some working distance.

Nice setup, super results!!! One suggestion is to put black material one the other side of mirror where light exits, else, reflection of surroundings will be captured by camera, after all, it is a mirror

[kaleun96]

Please remind me what lighting method is the non-axial?

It's a custom 3D printed "ring flash" for my Godox TT350. I've shared some earlier designs here but since then I've made a few changes, like making the flash adapter be able to swivel 360 degrees independent of the filter thread and I also have one to support two flash heads, placed at 120 degrees from one another (third photo). I have some diffuser attachments for them too but didn't use one for the photos above (visible on the whiteboard in the third photo).

[kaleun96]

yes, actually, cheap UV filters work better, those with any kind of coating either have some kind of colour cast or aberration. I did try a piece of glass, 1.0mm thick, supposedly to be K9 optical grade, it is cheaper than half-mirror, and works well, too, just less light.

The one attached to the lens was intended for stack and stitch where when you move subject, normal lighting from side might cause lighting condition change -- lighting and camera view will be different for each segments, and eventually cause problems in stitched image. It is not practical to build a super sized coaxial box, so attaching it to the lens is a good solution. But be aware that it takes away some working distance.

Nice setup, super results!!!

Thanks! I'll have more of a play around with the setup today before deciding on whether I go with a lens-mount setup. I did buy some 35x35mm beamsplitters but unless there's a noticeable improvement in sharpness with them, I think I'll stick with the 80x80 one. Is there a good formula for calculating how small the mirror could be for a given magnification and distance from the lens? I assume some vignetting calculation exists out there. My custom flash adapters shown in the post above present a similar situation. On one of them, I have a 35mm diameter aperture that extends about 35mm from the front of the lens and this still works for all my coins, with some vignetting at the corners. So perhaps I could get by with just the 35x35mm mirrors.

One suggestion is to put black material one the other side of mirror where light exits, else, reflection of surroundings will be captured by camera, after all, it is a mirror

I need to print a back piece with flocking, currently I'm using flocking behind the box but would be nice to integrate it into the box.

I just printed this backing panel overnight so will see how it fares with some flocking material on it. I was initially worried that having a flat panel, i.e. perpendicular to the axis of the light, will result in too much light still be reflected back off the flocking material but perhaps it will be ok. If not, I'll make the back panel into a "light trap" of sorts, perhaps by angling two walls at a 45 degree angle to each other.

[Pau]

There is a Chinese ebay seller bohr-optics selling plate BS and other interesting optics for low prices, no idea about their quality, BTW
https://www.ebay.com/itm/224868117408?h ... R_bx6afEYQ

[kaleun96]

There is a Chinese ebay seller bohr-optics selling plate BS and other interesting optics for low prices, no idea about their quality, BTW
https://www.ebay.com/itm/224868117408?h ... R_bx6afEYQ

That's the seller I bought from. I got two of the 80x80 1.1mm 50/50 plates. Quality seems good, they arrived well-packaged and only after a few weeks. They were also shipped with tracking.

[Pau]

That's the seller I bought from. I got two of the 80x80 1.1mm 50/50 plates. Quality seems good, they arrived well-packaged and only after a few weeks. They were also shipped with tracking

Good to know for future projects!

[Scarodactyl]

I have seen the ghost. It was in a trinocular olympus sz4045 clone that had been awkwardly converted from 100/0 to 50/50 by glueing the swing arm in place and replacing the first surface mirror with a half mirror. It was fine in most lighting but darkfield showed a significant offset double image.

[mjkzz]

I do not have a formula for view angles, I usually just do a trial and error. but most of the time, you'd be surprised, but still, longer focal length lenses are better as they usually have long working distance and you can go free on designing the attachment (to the lens). one note is that, the smaller the design, the harder to flock the inside, and when light is really intense, you might see the inside wall in your image. so it is better to go as large as you can.

in terms of flocking, I now use Mousou black paint, just be very careful when paint it, best way is to use a spray tool thing that artists use, so you get even paint and do it in out doors -- I think it is based on nano tech, not good to breath it in. also let it dry completely before use else, it can get on the lens and hard to get rid of.

yes, the 80x80 half mirror works well and cheap enough (about 12USD in China), but I would not use it for lens attachment type of setup, too heavy and not necessary -- since it is attached to the lens, it is closer to the tip of viewing cone, so it is not necessary.

one more thing about making attached setup -- if you use fusion 360, you can find ready-made thread library so you can print different type of threads. I still could not access many sites out side China, so I can not post a link, but google "fusion 360 camera threads" or something like that. One library even has Mitutoyo, RMS, and Nikon thread. also be aware of printing tolerance, you might have to expand or shrink the thread a bit in fusion 360 before print, and the worst part is, it is filament dependent, so no fixed amount.