Optimal sensor pixel density for Mitu 10x

[Lou Jost]

This is a follow-up to my post on bridge cameras as tube lenses.

Here I show how much detail is lost when using a Mitu 10x on a typical APS-C or full-frame sensor because of insufficient pixel density.

I used a Mitu 10x on a 90mm tube lens on three different sensors, each with a different pixel density. The highest density is from the Sony HX400V bridge camera, at 840 pixels/mm. Next the Panasonic FZ300 bridge camera, with a pixel density of 648 pixels/mm. Last comes a Nikon D90 (APS-C) with 181 pixels/mm.

The images of scales are from the same butterfly wing, and the same scale type is shown in each image, at about the same magnification on the screen (a scale is about the same width on the screenshot of each image). Same lighting also, two 1/128 power Yongnuo flashes to stop all movement.



Left to right: Sony HX400V, Panasonic Lumix FZ300, Nikon (APS-C) D90. Straight-from-the-camera jpgs. Aerial image is approx. identical in size and in detail in all three; the difference between images is due to the pixel density of the sensors used to record the images.

The results show that an immense amount of detail, including all the parallel ridges on the scale surfaces, is lost on the APS-C sensor because of its low pixel density. The linear pixel density should be at least 3 or 4 times higher (which is a 9-16 times higher density of pixels per square mm), in order to show the detail that is in the Mitu aerial image. The pixel density is a major limiting factor for image quality for this lens, contrary to what some theoretical discussions had claimed. An APS-C sensor should have 120 megapixels, not 10 or 20.

This result is based on a tube lens of 90mm because that is the longest length that I can arrange easily in all three cameras. (The Panasonic has a lens whose max zoom length is 108mm, and the closest lens I have to that for my Nikon is an excellent 90mm Tokina macro lens, perhaps my sharpest lens. So I standardized everything to 90mm.)

But the result is independent of tube length, for a given object.

Whether this lost detail matters or not depends on the intended application, of course. But it is good to recognize the limitation and choose cameras with higher pixel densities if detail is important to you.

[Charles Krebs]

There's no question that with some of the objectives we are using here you will definitely benefit from greater pixel density than most cameras have today.

But for a photograph you typically use as much of the sensor area as possible to create the final image. The two fixed lens cameras have 1/2.3" sensors that measure 6.17 X 4.55mm, 7.7mm diagonal, and an area of 28 mm^2 (and 20Mp and 12Mp). The Nikon is a 12MP sensor with a size of 23.7 X 15.7mm, diagonal of 28.4mm, and an area of 372 mm^2. If you set up to get the same subject field size filling the frame on each sensor you will need about the same number of M-pixels on the sensor to record about the same amount of detail.

In your example you are providing the same magnification onto each sensor, about 4.5X. The 1/2.3" sensors are recording a subject field size of about 1.4mm X 1mm. The APS-C sensor will be recording a subject field size of about 5.3mm X 3.5mm (and let's hope the corners are still good since we've pushed the objective down to a magnification, and thus have it recording a field size, it was not intended to do.) If you look at a the same 1.4 x 1mm field size that was recorded using all of the smaller sensors pixels (in 28 mm^2), and compare it to the same 6.17 x 4.55mm section on the larger sensor you can see that you are only making use of a small fraction of the pixels on that sensor (about 7.5% of it's area, only about 0.9 Mp).

But if you were to record the same field size (as on the 1/2.3" sensor) onto the larger APS-C sensor using the entire sensor area, then you would need to provide an additional 3.8X magnification to the objective's image. That would "spread out" the image detail over a larger area so you would not need as high a pixel density, and you would be using all of the sensors pixels (although the 12Mp of the D90 sensor is really too low for a 10/0.28, about 20-24 would be much better.)

Your main point is well taken... in many cases we can use many more pixels. In the actual photography we need to keep in mind the magnifications and field sizes if you are going to compare between formats.

[Lou Jost]

Thanks for responding, Charles.

Note that my purpose is not to compare formats. I'm exploring how much information there is in the aerial image of the Mitu.

I think the numbers I've just given let us figure out the pixel density needed to get all the available information out of the aerial image, at any magnification.

For example, if the Mitu good-quality image cone just barely covers the width of a full-frame sensor when used with a 90mm tube lens, then my results show that the sensor needs to be 36*600 = about 21000 pixels wide if you want to record all available information from the aerial image. If you use the lens with a 200mm tube, you are only using half the image cone with that FF sensor, and it should be 10500 pixels across to get all the available information. If you push it to 400mm you are using a quarter of the width of the image cone and the sensor should be 5000 pixels across to get all the information available in that sliver of the image cone.

Does that make sense?

Edit: Note 600 px/mm is at the low end of the pixel density I calculated; maybe 800 px/mm would be slightly better.

[Lou Jost]

The other conclusion I get out of this exercise is that bridge cameras are viable alternatives to larger more expensive cameras for photomacrography.

If you have an object that's 3 mm across, you can spread it across 5000 pixels on the bridge camera sensor using just 2x optics. Or you can spread it across 5000 pixels on an APS-C sensor, but you'll need 8x optics. The price difference between these two options is very large. It may be useful to be aware of this cheap easy alternative.

And many bridge cameras have interesting features that are still rare on larger cameras. The FZ300 for example has a screen that can be rotated and moved all over the place, and it gives the option to use a completely electronic shutter. Nothing moves -- no chance of camera vibrations. (Unfortunately flash is disabled under this option, though external flashes can be triggered manually during the exposure and perform exactly as with mechanical shutters.) And it is light and portable, and it also takes fine pictures of birds and other wildlife with the same tube lens! For backpacking field work I think a bridge camera (especially one that shoots RAW files, like the FZ300) would be a good choice.

Nevertheless I still prefer my Nikon for being less fiddly and having a wider range of lens choices, even though it clearly needs more pixels (and needs to shed the anti-alias filter of course). For objects under bright continuous light, though, I would use the FZ300 because of its completely vibrationless shutter.

[g4lab]

Very informative exposition!

The image quality does not seem to be suffering as a result of the smaller pixels and higher density. I wonder if the wide range zooms on the bridgecams might also introduce some distortions in addition to possible vignetting or light fall off.

[Lou Jost]

g4lab, yes, there seemed to be some strong curvature of field in the Sony, at least at middle focal lengths (100mm, eq to 600mm on a FF sensor; this camera zooms to 1200mm FF equivalent!!). There was also vignetting at middle and lower focal lengths.
Both the Panasonic and Sony performed very well when their zooms were fully extended. In the case of the Panasonic, the camera has an optional hollow tube that fits over the lens and screws into the body. This tube has a 55mm thread in front for front conversion lenses. I mount the heavy microscope objective on this, rather than on the delicate lens itself. It is very very sturdy this way, but the objective is only close to the front lens when fully zoomed out.

[g4lab]

In the case of the Panasonic, the camera has an optional hollow tube that fits over the lens and screws into the body. This tube has a 55mm thread in front for front conversion lenses. I mount the heavy microscope objective on this, rather than on the delicate lens itself. It is very very sturdy this way, but the objective is only close to the front lens when fully zoomed out.

These are the kinds of details that you can only find out by trying it.
Thank you for sharing. Very good idea to use a bridge camera. Saves the trouble of rounding up a tube lens.

[ChrisR]

Did you get reversed perspective at all? I've had that doing this. It can look odd, but often doesn't matter. ( objects further away are bigger)

+1 vote for using a tube adapter on a camera. I taped a Mitutoyo onto a little camera's lens "just temporarily". It worked fine, then something happened and the lens popped off the camera, and never went back on

[Lou Jost]

Chris, I probably wouldn't have noticed or recognized reverse perspective if I had seen it, but from now on I'll be on the watch for it.

Sorry to hear about your lens experience. A Mitu is heavy, and those little lenses on the compact cameras sound like they are driven by little toy motors that strain even without extra burdens. The external tube solves this problem completely. It is more solid than my 35mm gear. I don't know if other brands besides Panasonic offer it, though.

[Lou Jost]

g4lab, I forgot to mention that the bridge camera results seem to have greater depth of field than if I had spread the same exact subject area over a larger APS-C sensor.

[ChrisR]

I wasn't expecting the camera to hold the objective, far from it, but even being very careful wasn't enough.

The tubes you mention are certainly available for many older Canon compacts. Those you can also program with CHDK to run a stack all by themselves . (Have done that with about three).

The DOF riddle is I think bound in how quickly the OOF focus gets to be VERY OOF.
"In Focus" is a calculated band, but when they are the same depth, the OOF gets more OOF quicker on the larger sensor giving the impression that the DOF must be less.
This shows up in a test Rik did. When I can remember a keyword to look for I'll link it!

[Lou Jost]

Thanks, I figured the appearance of extra depth might be an illusion. However, the magnification needed to get a scale to fill the frame on APS-C sensors is very much higher than that needed to get the same scale to fill the frame on small sensors, so I can also imagine it is a real effect.

[ChrisR]

It depends what you "hold still" when doing the comparison. If you pick an aperture for maximum DOF on each sensor,

which is where the diffraction blur = the OOF blur,
for both sensor sizes, (assuming the same number of pixels on each)

then you're no better off with one sensor than the other.

It's tempting to compare say a marked "f8" on both systems, but that doesn't really help - you can use a smaller effective aperture with the bigger sensor than you can with the smaller. The only downside is exposure duration.
It took a lot of effort on Rik's part to convince me of this, a few years ago now. It helps if you believe in mathematics , pictures are always open to some judgement. I have it on a speadsheet but I'd have to strip off all the stuff I've added. Can send when I have a mo...

[Lou Jost]

I actually do believe in mathematics!

Though it is sometimes hard to be sure all the implicit and explicit assumptions in a derivation have been satisfied.

[ChrisR]

Oh certainly, it has to be the right mathematics! I think it was a few months before I gave up looking for something wrong. I'll declutter the SS and let you judge.