Will I get more resolution/information?

[ray_parkhurst]

Is it really x vs y difference, or could it be relative to the angle of the light? What happens if you swap from landscape to portrait orientation?

OK, I tried it, and it's still more sensitive when moved in sensor Y.

[Lou Jost]

Hmmm, that's surprising.

What do you see when you look at this coin under a microscope, using just your eyes? Except for this last result showing a difference between x and y translations, it seems like these are "real" colors...

[ray_parkhurst]

Hmmm, that's surprising.

What do you see when you look at this coin under a microscope, using just your eyes? Except for this last result showing a difference between x and y translations, it seems like these are "real" colors...

I don't see such colors in a stereoscopic view. I tried shooting an afocal earlier but my phone gave me problems. There are colors of course, but the "red/blue pattern" is missing.

I may try shooting an untoned silver coin to eliminate the toned copper color background.

[rjlittlefield]

We had a long discussion about false color, back in 2013. See the discussion at http://www.photomacrography.net/forum/v ... hp?t=19582 .

I don't think we ever did fully resolve what physics is producing the colors, but to my eye one of the most important posts in that thread is back on page 6, at http://www.photomacrography.net/forum/v ... hp?t=19582 , where it is written that

From Goodman's "Introduction to Fourier Optics", 3rd ed, p.135, through a relationship of angular size of the source, entrance pupil and angular sprectrum of the subject, an incoherent optical system can behave mostly like a coherent system. The relation for coherence is when angular size of the source is much less than the angular size of the entrance pupil. By inverse, the system is incoherent when the angular size of the source is greater than or equal to the angular size of the entrance pupil plus the angular size of the angular spectrum of the object. All angular diameters are as measured from the object plane. "In between" cases are considered to be "partially coherent".

So, quick summary, my own guess is that the colors are mostly coming from speckle patterns, enabled by partially coherent illumination caused by having an illumination angle that is small compared to the angular sizes of the entrance pupil and the reflections off the subject.

--Rik

[ChrisR]

So, if you use 4 pixels instead of one, it "averages out"??

[Lou Jost]

We can test this. I'll use my milk diffusers which completely remove coherence. If the colors go away, it is probably due to some kind of speckle pattern.

[ray_parkhurst]

We had a long discussion about false color, back in 2013. See the discussion at http://www.photomacrography.net/forum/v ... hp?t=19582 .

I don't think we ever did fully resolve what physics is producing the colors, but to my eye one of the most important posts in that thread is back on page 6, at http://www.photomacrography.net/forum/v ... hp?t=19582 , where it is written that

From Goodman's "Introduction to Fourier Optics", 3rd ed, p.135, through a relationship of angular size of the source, entrance pupil and angular sprectrum of the subject, an incoherent optical system can behave mostly like a coherent system. The relation for coherence is when angular size of the source is much less than the angular size of the entrance pupil. By inverse, the system is incoherent when the angular size of the source is greater than or equal to the angular size of the entrance pupil plus the angular size of the angular spectrum of the object. All angular diameters are as measured from the object plane. "In between" cases are considered to be "partially coherent".

So, quick summary, my own guess is that the colors are mostly coming from speckle patterns, enabled by partially coherent illumination caused by having an illumination angle that is small compared to the angular sizes of the entrance pupil and the reflections off the subject.

--Rik

I do remember that discussion, and have utlized significant diffusion ever since in order to minimize the effects. The light I used in these photos is ~50% of a ringlight, which is a fairly diffuse source. That said, I'm fairly sure you're right that this is the culprit, but I guess I can't diffuse enough to make it go away at reasonable apertures while still maintaining the "look" I am after in the final photo.

But this still does not explain the systematic nature of the speckles, and the sensor pitch level interaction, does it? Why is it more prominent in one dimension than the other? I guess there are additional variables that are interacting.

[rjlittlefield]

The light I used in these photos is ~50% of a ringlight

Can you put dimensions on that? What is the diameter of the ring, and the thickness of the illuminated band, and how far away from the subject is it?

--Rik

[Lou Jost]

Is it a COB ringlight?

[ray_parkhurst]

The light I used in these photos is ~50% of a ringlight

Can you put dimensions on that? What is the diameter of the ring, and the thickness of the illuminated band, and how far away from the subject is it?

--Rik

Here's a drawing of the illuminator. It's indeed based on a 50mm COB ringlight, with added diffusion and masking. The light is suspended ~100mm above the coin.

Edited to add: I can add more diffusion by increasing the width of the diffuser to the outside diameter.

Edit 2: I went ahead and greatly increased the diffusion by making the diffuser in front of the COB 10x wider. I still see the same effect in the aperture sweep, ie sparkles at f4.7, decreasing at f5.6, and mostly gone at f6.7.

[rjlittlefield]

Thanks for the additional info. Now I'm a little curious how far in front of the ring light the diffuser is located, to have some idea how that "10X wider" is actually being utilized. It's the usual problem: if the diffuser is close to the ring light then it spreads the light out farther but doesn't make the light source any bigger. But I'm not extremely curious, because I've never seen a careful analysis of half-ring illumination and I have no idea how to do one myself.

Anyway, all of that bears on the possibility that you're getting colored speckles in the optical image, due to partially coherent illumination.

The other possibility, that you're getting colored speckles due to Bayer filter mosaic, is also very much in the game. Your trio of images at http://www.photomacrography.net/forum/v ... 881#244881 , showing that the colors shift as you slightly move the camera, strikes me as supporting Bayer filter effects.

I suppose that it's again time to point out that naked Bayer filter sensors cannot possibly work the way you'd like. Imagine a star field, sharply focused and positioned so that each star hits only one photosite. Then the sensor provides no information about what color each star actually is. With less pathological cases the effect is not so strong, but the principle is the same. At http://www.photomacrography.net/forum/v ... 537#163537 , you'll see a colorless target imaged a couple of different ways, and the way that puts pixel-sized details onto a naked Bayer sensor has obvious false colors. The general problem is completely unavoidable --whatever pixel-level colors appear in the developed images have been imputed by assumption, not actually measured. If there aren't any strange colors, then that too has been imputed by assumption, and you were lucky enough that the assumptions+processing matched reality.

If you want to tease apart the question of whether the colors are present in the optical image, or are added because of sensor+processing effects, then I can strongly recommend the approach of using two-stage optics, using something like a 10X objective in the second stage to expand the image so that each optically resolved element spreads across many pixels.

--Rik

[Lou Jost]

Theoretically, Pentax/Sony pixel-shifting should completely remove the Bayer-induced problems without causing new problems. Olympus/Panasonic pixel shifting should remove it too but might, because of its greater complexity, cause new problems.

I should be able to do my tests of both types of shifts (with completely incoherent light by milk diffusers) this weekend.

[ray_parkhurst]

Thanks for the additional info. Now I'm a little curious how far in front of the ring light the diffuser is located, to have some idea how that "10X wider" is actually being utilized. It's the usual problem: if the diffuser is close to the ring light then it spreads the light out farther but doesn't make the light source any bigger. But I'm not extremely curious, because I've never seen a careful analysis of half-ring illumination and I have no idea how to do one myself.

Anyway, all of that bears on the possibility that you're getting colored speckles in the optical image, due to partially coherent illumination.

The other possibility, that you're getting colored speckles due to Bayer filter mosaic, is also very much in the game. Your trio of images at http://www.photomacrography.net/forum/v ... 881#244881 , showing that the colors shift as you slightly move the camera, strikes me as supporting Bayer filter effects.

I suppose that it's again time to point out that naked Bayer filter sensors cannot possibly work the way you'd like. Imagine a star field, sharply focused and positioned so that each star hits only one photosite. Then the sensor provides no information about what color each star actually is. With less pathological cases the effect is not so strong, but the principle is the same. At http://www.photomacrography.net/forum/v ... 537#163537 , you'll see a colorless target imaged a couple of different ways, and the way that puts pixel-sized details onto a naked Bayer sensor has obvious false colors. The general problem is completely unavoidable --whatever pixel-level colors appear in the developed images have been imputed by assumption, not actually measured. If there aren't any strange colors, then that too has been imputed by assumption, and you were lucky enough that the assumptions+processing matched reality.

If you want to tease apart the question of whether the colors are present in the optical image, or are added because of sensor+processing effects, then I can strongly recommend the approach of using two-stage optics, using something like a 10X objective in the second stage to expand the image so that each optically resolved element spreads across many pixels.

--Rik

I'm assuming these issues are all part of the Bayer demosaicing, and not actually present in the optical image. The color shifts with small movements are near proof of this, though there could be some tertiary effect as well I don't know about.

Now I'm curious if these effects are present to the same extent in all Bayer sensor cameras. It seems they are there in the A7Riii, both in my own shots as well as those posted by Robert. The Sony software mitigates this through relatively heavy chroma noise suppression, but that's not such a great method IMO. Will I need to stop down to f14 or smaller (effective) to get rid of these false colors? That's not a huge problem for higher mags, so much of my work will be fine. I guess it's just another factor to be careful of.

[ray_parkhurst]

Theoretically, Pentax/Sony pixel-shifting should completely remove the Bayer-induced problems without causing new problems. Olympus/Panasonic pixel shifting should remove it too but might, because of its greater complexity, cause new problems.

I should be able to do my tests of both types of shifts (with completely incoherent light by milk diffusers) this weekend.

Unfortunately, theory does not seem to hold, at least for the Sony.

[Lou Jost]

I'm not sure about that. The Sony/Pentax shifting has given excellent results in most trials, including Robert's very recent tests. I'm not sure we understand the origin of the colors you are finding.

Edit: I see the odd colors in your blow-up of Robert's test in both the shifted and unshifted versions.....not sure what to make of that.