Large format macro tests

A forum to ask questions, post setups, and generally discuss anything having to do with photomacrography and photomicroscopy.

Moderators: rjlittlefield, ChrisR, Chris S., Pau

Lou Jost
Posts: 5947
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Large format macro tests

Post by Lou Jost »

I used to do large format semi-macro photography in the old days of film, and I loved it. I lost all of that equipment years ago. But nowadays all the equipment I used to dream about back then is dirt cheap on eBay, since no one wants it. I've accumulated a Horseman 4"x5" view camera, numerous adapters for digital backs, and a few lenses, and now I finally put it all together to see whether it is worth using for low-magnification macro work.

I use a sliding back that accepts a FF digital camera, and this can be moved around ("sensor shifting") to capture portions of the 4" x5" image. These files can then be combined in ICE to make a gigantic digital capture of the view camera's image. In normal photography at infinity, even though any single digital tile is less sharp than the corresponding image taken with a dedicated modern digital lens, the total amount of information in the 4"x5" image is much greater than in an ordinary single FF digital image with the same FOV as the 4"x5" image. I wanted to see if that would be true at magnifications around m=1.

First I tried a Mamiya 6x7 medium format 140mm macro lens. This is a giant bellows lens (made for the old Mamiya RZ series) which does not use a helicoid. It has floating elements like a Macro-Varon which have to be adjusted manually for each magnification. It used to cost thousands but can now be had for under $200. I was surprised to find that at m=1 it covered 4"x5" pretty well. Its best aperture was f/8 and it looked pretty good. When I saw the nice results, I got excited and began to experiment with other lenses. The obvious choice was the 120mm Macro-Symmar.

The 120mm Macro Symmar image circle completely covers 4"x5" at m=1 and its best aperture across that field is f/5.6 plus half a stop. I made a tile of the hindwing and compared it to the corresponding crop of the whole butterfly (same FOV as the 4"x5" image) taken with my Sigma 70mm Art macro lens at f/4 (that excellent lens' sharpest aperture). I used pixel-shifting for both to squeeze out as much resolution as possible. The 4x5 image is vastly superior to the FF image of the same FOV.

Here is the whole FF image; the whole 4"x5" image looks similar but with extra space on the edges because of differences in aspect ration and field orientation:
_2680589.jpg
Here is a 300% crop of the lower right hindwing; on the left is the crop from the FF image and on the right is the 100% crop from the 4"x5" image.
Untitled-1.jpg
As you can see, more pixels are better.

Traditionally no one used large format photography for much higher magnifications because of the need for enormous bellows extensions. Apparently no one knew about coupled lenses in those days. By combining a lens focused at infinity and a reversed lens, we can do large-format macrophotography with no additional bellows extension at all! I've begun to experiment with that, and it looks promising. I'll put the results here in this thread.

lothman
Posts: 959
Joined: Sat Feb 14, 2009 7:00 am
Location: Stuttgart/Germany

Re: Large format macro tests

Post by lothman »

I am not sure whether I understood this. You used the Mamiya lens at 1:1, took a stitched photo of an object bigger than your sensor and compared this to a single shot at a magnification smaller then m=1 :-k

Lou Jost
Posts: 5947
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Re: Large format macro tests

Post by Lou Jost »

No. I'm comparing equal fields of view in single optical images. In one case, the optical image is 5 inches across, made by a large format lens, and in the other case, the same field of view is contained in an optical image 36mm across made with a modern macro lens designed for small sensors.

The image I showed is not from the Mamiya. As the post says, the image is from the Macro Symmar.

Lenses that were designed for large sheets of film generally have lower resolution per mm on the image than lenses designed to cover small sensors. But, when integrated over the whole image, at fixed fields of view, the larger image has more detail. I wanted to now how much more. Turns out to be quite a lot.

By the way, these were taken with the Panasonic S1R.
Last edited by Lou Jost on Wed Sep 02, 2020 7:02 am, edited 1 time in total.

Lou Jost
Posts: 5947
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Re: Large format macro tests

Post by Lou Jost »

I should add that a similar cpomparison at medium format shows that a medium-format image without pixel-shifting is only slightly better (in terms of resolution) than a 35mm image of the same FOV with 8-image pixel shifting.

lothman
Posts: 959
Joined: Sat Feb 14, 2009 7:00 am
Location: Stuttgart/Germany

Re: Large format macro tests

Post by lothman »

Lou Jost wrote:
Wed Sep 02, 2020 5:58 am
No. I'm comparing equal fields of view in single optical images. In one case, the optical image is 5 inches across, made by a large format lens, and in the other case, the same field of view is contained in an optical image 36mm across made with a modern macro lens designed for small sensors.

The image I showed is not from the Mamiya. As the post says, the image is from the Macro Symmar.

Lenses that were designed for large sheets of film generally have lower resolution per mm on the image than lenses designed to cover small sensors. But, when integrated over the whole image, at fixed fields of view, the larger image has more detail. I wanted to now how much more. Turns out to be quite a lot.

By the way, these were taken with the Panasonic S1R.
so you projected a 5 inch object to 5 inch and stacked this 5 inch focal plane with a FF sensor. This is compared this with a single shot where you took the 5 inch object projected direct to FF sensor. Correct?

Lou Jost
Posts: 5947
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Re: Large format macro tests

Post by Lou Jost »

Yes, though not stacked, both are single shots.

ray_parkhurst
Posts: 3416
Joined: Sat Nov 20, 2010 10:40 am
Location: Santa Clara, CA, USA
Contact:

Re: Large format macro tests

Post by ray_parkhurst »

Of course the 300% image is going to look worse than the 100% image.

Lou Jost
Posts: 5947
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Re: Large format macro tests

Post by Lou Jost »

Yes, but that's real loss. Remember we're comparing same FOV on the two images, and that's what counts (unless you have web clients who automaticallly zoom in to 100% without thinking about what they are looking at). If both images were printed at the same size, this is what they would each look like if examined closely.

Lou Jost
Posts: 5947
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Re: Large format macro tests

Post by Lou Jost »

Lest this result seem too obvious, I think we all agree that for m large enough, these two images will be equally blurry, due to empty magnification. I'd like to see where that happens.

ray_parkhurst
Posts: 3416
Joined: Sat Nov 20, 2010 10:40 am
Location: Santa Clara, CA, USA
Contact:

Re: Large format macro tests

Post by ray_parkhurst »

Lou Jost wrote:
Wed Sep 02, 2020 9:05 am
Yes, but that's real loss. Remember we're comparing same FOV on the two images, and that's what counts (unless you have web clients who automaticallly zoom in to 100% without thinking about what they are looking at). If both images were printed at the same size, this is what they would each look like if examined closely.
Yes, of course, but this would be the case no matter what your subject or FOV or lens or magnification is. You mention that at some point the images would appear the same, but I don't think that's true. The 300% will always look worse, except perhaps in the special case of magnification to the point of being just a single color.

Most everyone on the web zooms to 100%, which is why I place such a high premium on high IQ at 100%. Only time I like to zoom further is if the comparison at 100% is slim, and higher magnification is needed to see the subtle differences between IQ, or to better see details that are at pixel level.


Edited to add: I suppose it would be easy to find the point at which a 300% image is as blurry as a 100% (with same FOV) by simply upscaling the images equally to the point of total blur.

rjlittlefield
Site Admin
Posts: 23562
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Re: Large format macro tests

Post by rjlittlefield »

Lest this result seem too obvious
Running some numbers...

It looks to me like the large-format case is imaging at 1X around effective f/13, while the FF case is imaging around 1/3X and effective f/5.3.

Based solely on diffraction, the 1X case should win but only by a little, not the large margin shown here. So something else is going on. Pixel count is a strong candidate.

Ignoring that your subject has lots of blue, and blindly applying Nikon's rule of 2 pixels per cycle at the diffraction limited cutoff frequency, the f/5.3 case would require a pixel size of 1.46 microns. Even smaller pixels would be better, for improved sampling of the optical image. But the physical pixel size on Panasonic DC-S1R is 4.27 microns, nominal 2.14 microns with pixel shifting.

So yes, if that Sigma 70mm Art is diffraction limited at f/4, then the S1R even with pixel shifting does not have nearly enough pixels to capture all the information in the optical image.

So I agree with your point that more pixels are better.

I expect it's also true that a larger sensor is better, since that allows to use a lens with smaller effective aperture = simpler control of aberrations. But that point is not directly addressed here, and would require a lot more work to nail down.

--Rik

Lou Jost
Posts: 5947
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Re: Large format macro tests

Post by Lou Jost »

Most everyone on the web zooms to 100%, which is why I place such a high premium on high IQ at 100%
Yes I know, that's why I mentioned "web clients who automaticallly zoom in to 100%". But it should be obvious that this is not, by itself, a valid indicator of lens or image quality. Otherwise we'd all be happy with 5Mp cameras. Those are perfectly sharp at 100% for almost any lens you put on them. The amount of information contained in the optical image is a better measure. If that is high, you have the option of downsampling so that a 100% crop is sharp, to satisfy the internet crowd, but also gives you the option to get a sharper print or to print larger, using the non-downsapled image. In other words, it is important to know the zoom factor needed to reach the edge of 100% sharpness. Now that would be a good measure of image quality.
The 300% will always look worse, except perhaps in the special case of magnification to the point of being just a single color.
Remember that the 100% image is magnified much more than the 300% image, about 3x more than the 100% image, in roder to cover 4"x5". I think they will eventually be similar as m increases and we reach empty magification. I'm not sure.

Edit: While I was writing, I see Rik worked some numbers. Yes, I agree, the message here is that the best lenses are greatly outresolving even very high Mp sensors, so spreading the image over more pixels, or using pixel shifting, is helpful. Even pixel shifting is not reaching the full potential of some lenses. That means it pays to work with a larger image until the point that diffraction gets too big.

ray_parkhurst
Posts: 3416
Joined: Sat Nov 20, 2010 10:40 am
Location: Santa Clara, CA, USA
Contact:

Re: Large format macro tests

Post by ray_parkhurst »

Lou Jost wrote:
Wed Sep 02, 2020 10:29 am

Yes I know, that's why I mentioned "web clients who automaticallly zoom in to 100%". But it should be obvious that this is not, by itself, a valid indicator of lens or image quality. Otherwise we'd all be happy with 5Mp cameras. Those are perfectly sharp at 100% for almost any lens you put on them.
I sure wish this were true, and I know you did not intend to put it that way. Bayer demosaicing, AA filters, shutter shake and mirror slap all contribute to a wide range of quality coming from most any camera. It is not so easy to get a good looking 100% image, even if your camera is only 5MP.
Remember that the 100% image is magnified much more than the 300% image, about 3x more than the 100% image, in roder to cover 4"x5". I think they will eventually be similar as m increases and we reach empty magification. I'm not sure.
I know you partially addressed this later, but the 100% image is likely not degraded 3x due to the magnification.

Lou Jost
Posts: 5947
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Re: Large format macro tests

Post by Lou Jost »

It is not so easy to get a good looking 100% image, even if your camera is only 5MP.
You're right. But as we've discussed before, the 100% criterion is artificial and makes no sense as a judge of image quality, though it may make sense in practice if your target is users who always zoom in to 100%. Those users, however, are making a mistake.

If the image is sharp at 100%, that only means the lens outresolved the sensor. That's not interesting by itself; just about any decent lens will outresolve a 5mp sensor. And a camera with a 100 Gpixel sensor will be unsharp at 100% for even the best lenses. By this criterion, you would prefer the barely decent lens from the 5Mp camera over a superb lens from the 100 Gpixel sensor.

You generally use great cameras and so your 100% criterion means something real, but only because you are implicitly considering sensor size as well. Given your sensor size, it takes a good lens to outresolve your sensor. My point is that you have to factor in the sensor size somehow. Increasing sensor pixel density should not automatically make the imaging system worse, as it would under your criterion.

ray_parkhurst
Posts: 3416
Joined: Sat Nov 20, 2010 10:40 am
Location: Santa Clara, CA, USA
Contact:

Re: Large format macro tests

Post by ray_parkhurst »

Lou Jost wrote:
Wed Sep 02, 2020 11:50 am
It is not so easy to get a good looking 100% image, even if your camera is only 5MP.
You're right. But as we've discussed before, the 100% criterion is artificial and makes no sense as a judge of image quality, though it may make sense in practice if your target is users who always zoom in to 100%. Those users, however, are making a mistake.
But that is only your opinion Lou. Achieving best resolution and sharpness at 100% is indeed my primary criteria for image quality (aside from the obvious of focus, lighting, composition, and color fidelity). Occasionally I will look at 200-800% when it is difficult to assess the differences between IQ at 100%, but only in a comparison context, not for presentation.

I think the discrepancy here is that we have different ultimate purposes for our images. I don't do large prints. In fact I don't do any printing at all. Every image I take is for consumption on a computer monitor. I share some images on forums, but the typical size is 800x800 pixels. I will occasionally email a full size image to someone, but this is tedious since anything in the 10MB+ range is slow to move around. I do share larger images on EasyZoom website, which can take large images and has functionality for zooming-in, but the max normal zoom is to 100%. They do have a 200% zoom now, but of course image information/quality does not improve when doing this.

Now, it is certainly feasible for me to shoot at higher magnification with larger sensors and higher MP counts, or with MP-expanding pixel-shifting or Super-Resolution techniques, in order to gather more information. I've tried these techniques in order to improve the quality of my images at 100%. The problem is that I would not want to present those images directly since their 100% IQ is generally worse than I get now. I'm still hopeful that non-expanding PS or SR may ultimately work for me, but so far have not found a solution that beats my current system. Going up in magnification, then downsizing to minimize the sensor aberrations, so far is the only sure-fire way I've found to improve 100% IQ, but it's a whole lot of work. That method would need to come with a larger sensor to be practical given the large panoramas.

So is it just large image printing that drives your requirements? Are you sharing your huge digital images somewhere? I'd love to view them if that is the case.

Post Reply Previous topicNext topic