The lenses we use

Have questions about the equipment used for macro- or micro- photography? Post those questions in this forum.

Moderators: rjlittlefield, ChrisR, Chris S., Pau

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

Post by ray_parkhurst »

Lou Jost wrote:Diffraction is not an all-or-none thing, and there is no such thing as diffraction-free pixels. Diffraction is linear with respect to aperture setting, and there is some diffraction blurring even at apertures that are normally considered to be diffraction-free. The more pixels you can put under a subject feature, the better the image, though we do soon each a point of diminishing returns.
Yes, sampling all the way to the limit is certainly an exercise in diminishing returns, hence my question about "validity". Personally, I would be of the opinion that MTF50 is a good limit for "validity", such that 100% crops still appear reasonably sharp. Beyond that and the amount of information being captured doesn't seem like it would add much. I don't know how to quantify that concept in any better way, though.

Perhaps a discussion of "output" is a way to start. What is the output of the image capture? I would guess that someone doing large prints would like to see reasonable pixel-level detail, perhaps to the level of MTF50 as I proposed above. Certainly a lot less than this and the image would be considered "soft", even though in reality it may contain more information in the overall image. The only way to see this may be to step back from the print significantly so the eye can down-sample and integrate that information into a whole. This is analogous to down-sizing the image, but if you must downsize in order to see the image clearly, does it really make "valid" sense?

Lou Jost
Posts: 5945
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Post by Lou Jost »

The idea of a 100% crop is not really useful when we are comparing different formats. We should be talking about a crop of a certain fraction of the FOV.

But of course lens choice depends on the desired output. In these digital days, though, we can publish interactive photos that a user can explore at leisure by zooming in. I think these would be an interesting goal for my scientific publications of new orchid species in journals which have online editions (as most do nowadays). Print size is no longer limiting. I will be pushing for this in my next publication, though I don't know what the editors will think of the idea.

Edit: My plan is to use a 5x or 10x wafer photolithography lens, NA around 0.20 or 0.28, using blue or violet laser light, and using a FF camera to capture stacked samples from a single aerial image around 100mm across, then stitching them for the final image.
Last edited by Lou Jost on Thu Aug 02, 2018 5:36 pm, edited 1 time in total.

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

Post by ray_parkhurst »

rjlittlefield wrote: Regarding pixel count, I'm hoping it's obvious that if you need N pixels on the subject (or on a sensor at 1:1 magnification), then you also need N pixels on any other sensor that is going to capture the same information.
No, this was not obvious when I wrote the reply (hence the 1:1 comment) but I do understand it now. Same number of "pixels" on both subject and sensor, but the sensor pixel size and effective aperture changes with magnification.
rjlittlefield wrote: I have no idea how you're using the word "valid".

As for achievable, that depends on how large an FOV and how much resolution you need. Obviously 24mm x 36mm by 325 cycles/mm is pretty challenging. But is it out of reach? My guess is that if you used a microscope to look at the optical image of your PN105 wide open, you'd see a level of finest detail right in line with these numbers.
Yeah, very imprecise word.

And indeed the 105PN would come close to meeting the numbers, though technically "not quite" since it is not diffraction-limited at f2.8.

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

Post by ray_parkhurst »

Lou Jost wrote:The idea of a 100% crop is not really useful when we are comparing different formats. We should be talking about a crop of a certain fraction of the FOV.
I just can't agree with this. 100% crops are a common tool for evaluation, regardless of format.
Lou Jost wrote: But of course lens choice depends on the desired output. In these digital days, though, we can publish interactive photos that a user can explore at leisure by zooming in. I think these would be an interesting goal for my scientific publications of new orchid species in journals which have online editions (as most do nowadays). Print size is no longer limiting. I will be pushing for this in my next publication, though I don't know what the editors will think of the idea.

Edit: My plan is to use a 5x or 10x wafer photolithography lens, NA around 0.20 or 0.28, using blue or violet laser light, and using a FF camera to capture stacked samples from a single aerial image around 100mm across, then stitching them for the final image.
Sounds like the GigaMacro approach, except in monochrome to get higher resolution.

Lou Jost
Posts: 5945
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Post by Lou Jost »

Ray, sure, if we are comparing lenses on the same sensors, it makes good sense to use a 100% crop. and sure, it is always nice to know that a lens looks sharp at 100% on a given sensor.

But what if you are comparing an image on a small sensor with an image of the same FOV on a much larger sensor? Even if the larger image is blurry at 100%, and even if the small image is sharp at 100%, it might still be true that the image on the larger sensor has more detail. The fairest way to compare them is to compare the same fraction of the FOV on both images. That is also how the output will be judged visually by an observer, if both are printed to the same size.

Even for two equally large sensors with different size pixels, it is helpful to compare the same fraction of the FOV. A 3Mp camera will nearly always looks sharp at 100% while a 50Mp camera may not. That doesn't mean the 3mp camera-lens combo is better than the 50Mp combo.

Does the GigaMacro approach use a single aerial image or a lens that they move around?

Lou Jost
Posts: 5945
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Post by Lou Jost »

Ray, it looks like GigaMacro's approach is fundamentally different from mine. I take multiple samples of a single large aerial image, while they seem to use ordinary lenses with small aerial images, moved around to create their mosaic. There are advantages to both approaches.

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

Post by ray_parkhurst »

Lou Jost wrote:Ray, sure, if we are comparing lenses on the same sensors, it makes good sense to use a 100% crop. and sure, it is always nice to know that a lens looks sharp at 100% on a given sensor.

But what if you are comparing an image on a small sensor with an image of the same FOV on a much larger sensor? Even if the larger image is blurry at 100%, and even if the small image is sharp at 100%, it might still be true that the image on the larger sensor has more detail. The fairest way to compare them is to compare the same fraction of the FOV on both images. That is also how the output will be judged visually by an observer, if both are printed to the same size.

Even for two equally large sensors with different size pixels, it is helpful to compare the same fraction of the FOV. A 3Mp camera will nearly always looks sharp at 100% while a 50Mp camera may not. That doesn't mean the 3mp camera-lens combo is better than the 50Mp combo.

Does the GigaMacro approach use a single aerial image or a lens that they move around?
I'm not arguing with the constant FOV comparison concept. However, that concept is most useful in a narrow range of downsize factors. The extra information present in the over-sampled image gets lost upon down-sampling. In Rik's extreme example, the 100% crop will be very blurry. I think a 2x downsize will be less blurry but still blurry, while a 4x downsize will be pretty sharp. If printed at 100%, you'd see a similar effect when backing farther away from the image. But as you do these downsamplings either by downsizing or backing away, information is lost, only a little with the first 2x, more with the 2nd 2x, and a lot with a 3rd 2x. So in the end, how do you decide an FOV for best comparison?

Regarding GigaMacro, last time I checked, they move the camera/lens on an XY platform in order to capture the images for stitching.

Lou Jost
Posts: 5945
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Post by Lou Jost »

Ray, I don't understand your comment about downsizing factors. I am arguing for picking a fraction of the FOV (say 5%) in both images and looking at them at the same physical sizes. Not downsizing. Upsizing at least one of them so both are the same size. Then you'll see which image contains more detail.

Yes, the Gigapixel approach is fundamentally different from mine. I am using the same aerial image for all the stacks. The lens does not move, except in the z-direction for stacking. I think my approach is easier to stitch, especially since my lens is also telecentric.

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

Post by ray_parkhurst »

Lou Jost wrote:Ray, I don't understand your comment about downsizing factors. I am arguing for picking a fraction of the FOV (say 5%) in both images and looking at them at the same physical sizes. Not downsizing. Upsizing at least one of them so both are the same size. Then you'll see which image contains more detail.

Yes, the Gigapixel approach is fundamentally different from mine. I am using the same aerial image for all the stacks. The lens does not move, except in the z-direction for stacking. I think my approach is easier to stitch, especially since my lens is also telecentric.
Hmm, I am not sure if upsizing works very well in the case Rik described, where the 100% detail is very blurry. You'd basically end up upsizing the lower-pixel-count image until it was as blurry as the higher-count, and in Rik's scenario this is nearly MTF0. I think only downsizing the higher-count image makes sense for comparison. Or maybe meet in the middle?

Your approach sounds sort of like what Austrokiwi does. He uses a large format lens, fixed in place, then moves the sensor to create the images for stitching. Seems a good way to go to eliminate some of the perspective issues that mess up stitching.

Lou Jost
Posts: 5945
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Post by Lou Jost »

Ray, what I've been doing in my comparisons between medium format and FF (both with same FOV) is to pick a feature and a size that looks reasonably sharp in the medium format image, and then magnify the FF image to the same size on the monitor, and comparing them.

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

Post by ray_parkhurst »

Lou Jost wrote:Ray, what I've been doing in my comparisons between medium format and FF (both with same FOV) is to pick a feature and a size that looks reasonably sharp in the medium format image, and then magnify the FF image to the same size on the monitor, and comparing them.
I'm curious why you use the upsize method. Everything we've discussed regarding outputs ends up being downsized. Upsizing always results in a reduction in sharpness, so the smaller format will always lose. Seems a built-in bias to the large format.

Lou Jost
Posts: 5945
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Post by Lou Jost »

Since the FOV is the same, either image could look less sharp. I think it is valid for comparing the lens+sensor combo. I agree with you that it is not necessarily a valid comparison between the lenses themselves. To compare lenses rather than lens+sensor combos, I use my MFT sensor on both the FF and medium-format images; the sensor outresolves both lenses.

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

Post by ray_parkhurst »

Lou Jost wrote:Since the FOV is the same, either image could look less sharp. I think it is valid for comparing the lens+sensor combo. I agree with you that it is not necessarily a valid comparison between the lenses themselves. To compare lenses rather than lens+sensor combos, I use my MFT sensor on both the FF and medium-format images; the sensor outresolves both lenses.
I don't understand. Do you mean you use the MFT sensor and then stitch to create FF or MF outputs? I also am curious why you say the sensor outresolves both lenses. This would be the case at higher mag but at lower mag I'd expect most decent lenses to out-resolve most MFT sensors. Or are you talking about pixel-shifting? Please elaborate.

Lou Jost
Posts: 5945
Joined: Fri Sep 04, 2015 7:03 am
Location: Ecuador
Contact:

Post by Lou Jost »

For these tests I don't bother to stitch the images, I am just looking at a single feature. I want to use the highest pixel-pitch sensor that I have (short of pixel-shifting, just because I don't have time). That sensor does out-resolve most medium-format lenses, and many though not all FF lenses.

mjkzz
Posts: 1681
Joined: Wed Jul 01, 2015 3:38 pm
Location: California/Shenzhen
Contact:

Post by mjkzz »

rjlittlefield wrote:
mjkzz wrote:To get diffraction free pixels, maximum MP is 91.25MP where each pixel is diffraction free. without down sampling, and I believe this is what most photographers mean.
I would be interested to know where that number comes from.

Can you state the assumptions and show the calculation?

--Rik
Same assumption as yours. Calculated using:

https://www.cambridgeincolour.com/tutor ... calculator

with "show advanced" and check "Set Circle Of Confusion based on pixels".

If I were a photographer (which I am not) and use that calculator, by using binary search (or whatever is called), I get the limit is about 96MP (vs 91.25MP) as maximum MPs for a f/5.6 aperture on a 24x36 (full frame) sensor.

We are sampling the airy disc imposed by diffraction, we will not get more details than 96MP, we can over sample it, but to show "true" image, we have to down sample it to 96MP.

[Edit]

IMHO, Stepping back a bit, even with your (Rik's) calculation, the diffraction free resolution is 91.25MP, so it really does not matter how many samples we take, we will NOT gain any details than 91.25MP, we are only getting "more accurate" pixel values AFTER down sampling.

[/Edit]

Post Reply Previous topicNext topic