The lenses we use

[Justwalking]

Moritex 0.8X

OK, so then effective f/12.5 on the subject side, f/10 on the sensor at 0.8X .

Do we agree on that much?

--Rik

I can't do nothing with this fixed focus lens. They count that F/10 eff.

[rjlittlefield]

They count that F/10 eff.

Yes, it is effective f/10 on the sensor side.

On the subject side, the f-number can be computed by the standard formula: fnumber = 1/(2*NA).

12.5 = 1/(2*0.04)

The f-number on subject side can also be computed from the sensor-side f-number and the magnification.

In that case the relevant formula is: sensor-side fnumber = subject-side fnumber * magnification

10 = 12.5 * 0.8

Both formulas give the same number, of course.

The lens is f/12.5 on the subject side.

--Rik

[Justwalking]

They count that F/10 eff.

The lens is f/12.5 on the subject side.

--Rik

Well. It is. And? Is it good or bad when F.eff = 10?

[rjlittlefield]

Is it good or bad when F.eff = 10?

Refresh my memory, please...

Exactly what camera are you using? How big are its pixels?

--Rik

[Justwalking]

Is it good or bad when F.eff = 10?

Refresh my memory, please...

Exactly what camera are you using? How big are its pixels?

--Rik

1/2.33" sensor. Crop factor 5.5. Pixels is about 1.33um.

[rjlittlefield]

Is it good or bad when F.eff = 10?

Either or both, depending on what "good or bad" means. The good part is that you have 4 times more pixels than are needed to capture all the detail in that lens's optical image. The bad part is that a sharper lens could capture 4 times more detail using the same sensor.

Standard formulas say that for f/10 on the sensor side, you only need pixels of size 2.75 microns to capture all the detail that is in the image.

Your pixels are half that size, so you are over-sampling a blurred image by about 2X. The actual content of your 16 MP images is about 4 MP.

To fully utilize all 16 MP, you would need a lens with effective f/5 or wider on the sensor side.

At 0.8X, that implies NA 0.08 or wider.

--Rik

[Justwalking]

To fully utilize all 16 MP, you would need a lens with effective f/5 or wider on the sensor side.

At 0.8X, that implies NA 0.08 or wider.

--Rik

Rik, what's mean sharper lens regardless of diffraction?
What is the standart formula?

Well, and what must be F.eff for FF 16MP and magnification X4.4 with FoV 7mm? And how big must be the pixels there?

Well, i know the unswer. FF must have 15um pixels sensor to resolve 16 MP. and lens must take F22 .eff on this magnification.

[rjlittlefield]

what's mean sharper lens regardless of diffraction?

There is no "regardless of diffraction". Everything I have said is about diffraction, and how it limits sharpness.

When I say "sharper lens" here, I mean one with a wider aperture, so that it can resolve finer detail -- a smaller Airy disk.

What is the standart formula?

The simplest formula is just this: pixel size = lambda*fnumber/2

This gives 2 pixels per cycle at the diffraction-limited cutoff frequency. It is the same criterion that is used by Nikon on their web page Matching Camera to Microscope Resolution .

With most listings of formulas, for example HERE, you would have to work through other formulas to get the final result. In this posting I have done the algebra for you, to make the formula as simple as possible.

what must be F.eff for FF 16MP and magnification X4.4 with FoV 7mm? And how big must be the pixels there?

The combination of FF, X4.4, and 7mm FOV is not consistent. Full frame is 36 mm wide, so 4.4X would be FOV 8.2 mm.

Please rework your question so the numbers are consistent.

If you are eventually going to compare numbers between FF and your camera, then please crop the FF sensor to match your camera's aspect ratio (width/height).

--Rik

[Justwalking]

When I say "sharper lens" here, I mean one with a wider aperture, so that it can resolve finer detail -- a smaller Airy disk.

The simplest formula is just this: pixel size = lambda*fnumber/2

This formuls is 0.5CoC - no more. It is not about that all above will be lost in picture and we no need more resolution on sensor.

The combination of FF, X4.4, and 7mm FOV is not consistent. Full frame is 36 mm wide, so 4.4X would be FOV 8.2 mm.
Please rework your question so the numbers are consistent.
If you are eventually going to compare numbers between FF and your camera, then please crop the FF sensor to match your camera's aspect ratio (width/height).

--Rik

I count roughly from stitching the 13mm coin and 6 frames to cover it totally.
I no need to crop FF, but FF must cover same area with larger magnificatin to cover 5.5 times larger sensor for same resolution.
With cropping FF the resolution of the frame give us only 0.52MP instead 16.

[rjlittlefield]

The simplest formula is just this: pixel size = lambda*fnumber/2

This formuls is 0.5CoC - no more. It is not about that all above will be lost in picture and we no need more resolution on sensor.

I do not understand your idea. Please say it again in different words.

--Rik

[ray_parkhurst]

To fully utilize all 16 MP, you would need a lens with effective f/5 or wider on the sensor side.

At 0.8X, that implies NA 0.08 or wider.

Sounds like a 105PN.

[ray_parkhurst]

OK, I modified my lighting, tried a stitch to see if I could meet my goals, and it looks pretty darn good. I uploaded it to EasyZoom, see here:

https://easyzoom.com/image/123994

Here are the stats:

Final image: 12,500x12,500 pixels, 156.25MP, 150MB
# of source images in the panorama: 12, 3 columns x 4 rows, stitched using Image Composite Editor
# of stacked images per source image: 10, in ~50um steps, using Helicon
Magnification: 2.7x using Nikon 3XMM objective
Camera: Canon T2i with AA filter removed

I am pretty happy that I can get this level of documentation detail with this technique, though it is far more work than I originally hoped for. But oh well. On the bright side I am using my T2i and one of my regular lenses, so no investment. I may need to pay for EasyZoom if I do this very much. ICE is free.

[rjlittlefield]

Justwalking, earlier you wrote:

what must be F.eff for FF 16MP and magnification X4.4 with FoV 7mm? And how big must be the pixels there?

Then you edited that posting to say:

Well, i know the unswer. FF must have 15um pixels sensor to resolve 16 MP. and lens must take F22 .eff on this magnification.

This was going in the right direction, but your number is clearly wrong.

With 15 um pixels, an FF sensor would have only 3.84 MP (=(36/0.015)*(24/0.015)). 3.84 MP is quite a bit fewer than 16 MP.

So, recognizing that you have messed this up, let's step through a correct calculation.

Please read carefully, line by line, so that you understand what I am doing at every step.

Suppose we have two sensors. One sensor is 6.16 mm x 4.62 mm (aspect ratio 4:3). The other is 32 mm x 24 mm (FF, cropped to aspect ratio 4:3).

The ratio of sensor sizes is 1 : 5.2 (= 6.16:32 = 4.62:24).

If FOV is the same, and the small sensor is at 0.8X, the big sensor will be at 4.16X (=0.8*5.2).

Let both sensors be 16 MP.

On the small sensor, the pixel size is 1.33 microns (=sqrt(6.16mm * 4.62mm)).

On the large sensor, the pixel size is 6.93 microns (=sqrt(32mm * 24mm)).

The ratio of pixel sizes is 1 : 5.2 (=6.93/1.33). This is the same as the ratio of sensor sizes.

On the small sensor, to utilize 1.34 micron pixels we need effective fnumber = 4.84 (=1.33/0.55*2).

On the large sensor, to utilize 6.93 micron pixels we need effective fnumber = 25.2 (=6.93/0.55*2).

The ratio of effective fnumbers is 1 : 5.2 (=4.84:25.2). This is the same as the ratio of sensor sizes and the ratio of pixel sizes.

Yes, to cover the same FOV, the larger sensor requires more magnification.

But the pixel sizes are larger by the same factor. That allows the effective fnumber to be larger by the same factor.

On the sensor side, the pixel size and effective f-number just scale in proportion to the sensor size.

Meanwhile, on the subject side everything stays the same.

At 0.8X, the small sensor needs NA = 0.082 (=0.8/(2*4.84)).

At 4.16X, the large sensor also needs NA = 0.82 (=4.16/(2*25.2)).

The numbers make it look complicated. Let me put this in a table:

Code: Select all

    Sensor Width   Magnification   FOV    MP   Pixel Size   Effective F-number  Subject-side NA
        6.16mm           0.80     7.7mm   16     1.33 um        f/  4.84             0.082
       32.00mm           4.16     7.7mm   16     6.93 um        f/ 25.2              0.082

Now maybe you can see the pattern.

At same FOV and same MP:
1. on the subject side, NA is not affected by sensor size.
2. on the sensor side, the pixel size and effective f-number just scale in proportion to the sensor size.

--Rik

[rjlittlefield]

OK, I modified my lighting, tried a stitch to see if I could meet my goals, and it looks pretty darn good.

Excellent. I'm glad to see this technique working better for you.

Here are the stats:

Final image: 12,500x12,500 pixels, 156.25MP, 150MB
# of source images in the panorama: 12, 3 columns x 4 rows, stitched using Image Composite Editor
# of stacked images per source image: 10, in ~50um steps, using Helicon
Magnification: 2.7x using Nikon 3XMM objective
Camera: Canon T2i with AA filter removed

I am pretty happy that I can get this level of documentation detail with this technique, though it is far more work than I originally hoped for.

Well, if you want 156 MP, and you can only shoot 18 MP at a time, then it's going to take at least 156/18 = 9 times longer (rounding up). You did it in 3*4 = 12 times longer, which is not very much short of the best you could possibly do. Whether it's worth the trouble is something that you'll have to decide based on your own cost/benefit analysis. Shelling out for a bigger high MP sensor would save you some time, at the cost of some $$.

--Rik

[ray_parkhurst]

OK, I modified my lighting, tried a stitch to see if I could meet my goals, and it looks pretty darn good.

Excellent. I'm glad to see this technique working better for you.
...
Shelling out for a bigger high MP sensor would save you some time, at the cost of some $$.

Since this is the only way I seem to be able to create the zoom-able images I'm looking for, it will have to do.

I am not sure what optics I would use for this on FF. The 3xMM does OK on APS-C but won't do FF. What telecentric 3x objective with at least 0.09 NA covering FF is available?

edited to add:

I learned something interesting from this exercise. I find am less sensitive to un-sharpness due to diffraction than due to sensor aberrations. When shooting at wide enough aperture so that the sensor is limiting sharpness, the image at 100% looks "out of focus" to me. When shooting such that the lens diffraction is limiting sharpness, I don't get the same sense. The 3xMM objective I used for the stitched image is operating around f14.4, so definitely causing diffraction blur on my 4.3um pixels, but it looks "natural". The 105PN operating at f5.8 does not limit the sharpness, so any unsharpness is due to the sensor, and has an unnatural/out of focus quality to my eyes.

It also may be that viewing small features on the lower mag image is more of a strain, and with the features being blurry, it's annoying. With higher magnification, there is less need to look deeply into the image to see the details, so any unsharpness is less annoying.

Bottom line, I expected to downsize the stitched image by 2x to get sharpest pixel detail, but at full size the features are large, and their diffraction blur is not annoying to me.