8K/30 mp+ sensor for the Gh6.....

[pake]

Ask some macro shooters and they will tell you that while degradation from diffraction sets in from f11 it really only becomes a problem from f16. For my underwater photography I regularly go to f16 to eliminate ambient light, because flash sync speed only allows me to go to 1/250s. Global shutter would allow me to open up my aperture alot but even now I don‘t have a problem going to f16. At f22 the image becomes notably soft. This is mostly using the Olympus 12-50mm lens at it‘s macro setting. I assume that lenses that are sharper out of the box (although I‘ve got a very good copy) will be even better at f16 and maybe even at f22, but I didn‘t try it when I had the Olympus 60mm for a few weeks.

I'm one of those. I have plenty of useful macro shots taken at f/16 or even f/18 with the 20MP sensor.

Who cares how the photo looks unprocessed?

Purists? But yeah, I agree 10000%.

 

[DanS]

Who cares how the photo looks unprocessed? No one is ever going to see it. All these people who "hand wave" away the problems of smaller sensors are just realists who understand their use case instead of chasing "IQ" because "physics"...

It's not a sensor size issue, any sensor size can have the issue when you try and cram at large number of ”pixels” into a given area.

BIF for example, Is a use case that can expose diffraction issues. sometimes you have to heavily crop the footage, so you end up much closer to a one-to-one ratio than you might like.

 

[RAH]

Who cares how the photo looks unprocessed? No one is ever going to see it.

Yes, but it gives you a "base" level where you start your editing improvements, so the better the image you start with, the better the potential results. You know this to be true, I think.

 

[Hypilein]

Of course unprocessed files are useful to compare cameras to each other. If I look at a review, I would prefer the images to be mostly unprocessed. However, if I own a camera or test one myself to see if it suits me, I will process the hell out of the files to see how they hold up, because that is the real world after all.

I understand that diffraction is a thing. I can see it when I stop down for my underwater macro shooting and until we get global shutter (or a shutter with about 1/600s sync speed) I will be forced to stop down to eliminate natural light, so it's unavoidable. I think that while diffraction becomes a bigger issue with more pixel density it doesn't really change if you adjust for viewing size. So, while it is a thing, I don't think it's worth a huge discussion.

A lot of my landscape shooting is done at f5.6 which is good enough for plenty DoF. We don't need to stop down as much as the FF guys. If diffraction sets in at f4.5 like it currently does at f8 than really I won't ever be affected for that kind of shooting. UW wide angle I have to go to f8 or f11, but underwater I have much bigger issues that degrade IQ than diffraction anyway.

I guess the only people who really get into diffraction territory are macro shooters and they do so already.

 

[archaeopteryx]

I will be forced to stop down to eliminate natural light, so it's unavoidable.

Does the housing not accommodate an ND filter?

I guess the only people who really get into diffraction territory are macro shooters and they do so already.

If you go with the assumption a lens's decline from peak sharpness indicates the onset of diffraction several μ43 primes (20 1.7, 25 1.8, 45 1.2, 75 1.8) transition to diffraction limited by f/4. And those tests are at 16 MP or below. This holds for centre resolution in some of the f/2.8 zooms as well. Many of the other lenses peak near f/4.

I don't know of any optical aberrations which would contraindicate the above diffraction assumption.

So I would suggest the effective MP curves I posted 2.5 years ago are perhaps on the generous side if one's really trying to maximize information collected by a sensor. By the same token, if one's putting 1-2 MP images online or isn't printing large, very little (if any) information is lost by underutilizing the sensor's resolving ability. Similarly, if one's not focus stacking then more information is collected by stopping down whilst the blur disc in DoF limited areas of a composition remains larger than the Airy disc.

While I agree f/16 and f/22 are where diffraction softening is obvious my experience is it's not hard to spot at f/11 (3.75 μm pixel pitch; 16 MP μ43) if you've practiced a bit or are comparing a couple iterations of optimal radius Richardson-Lucy deconvolution and an uncorrected image. I believe I can also see a difference at f/8 side by side or toggling with a wider aperture but it's small enough I won't make any strong claim there.

 

[Ruairi]

Industrial and surveillance sensors are usually designed with controlled lighting in mind, as with robotics. They aren’t for available light photography. The Super35 studio cams from Panasonic, Canon, Sony, etc. are designed on this principle.

That it offers cinema DCI aspect ratio could be taken as a hint this sensor is designed for creative purposes.

 

[Aristophanes]

That it offers cinema DCI aspect ratio could be taken as a hint this sensor is designed for creative purposes.

Exactly. The sum of all information in the rumour chain is that the higher megapickle sensors appear aimed exclusively at the video market and assume a reliance on predominantly projected scene lighting. It will suffer from visible ISO issues and diffraction limits for available light photography. They are likely to be very costly, possessing less IQ than the current 20 MP sensor for stills.

 

[piggsy]

Ask some macro shooters and they will tell you that while degradation from diffraction sets in from f11 it really only becomes a problem from f16. For my underwater photography I regularly go to f16 to eliminate ambient light, because flash sync speed only allows me to go to 1/250s. Global shutter would allow me to open up my aperture alot but even now I don‘t have a problem going to f16. At f22 the image becomes notably soft. This is mostly using the Olympus 12-50mm lens at it‘s macro setting. I assume that lenses that are sharper out of the box (although I‘ve got a very good copy) will be even better at f16 and maybe even at f22, but I didn‘t try it when I had the Olympus 60mm for a few weeks.

Really I guess the more useful metric is: what exactly is the smallest detail you're trying to capture with the image? Does it slip under the smallest resolvable detail from the whole toolchain of lens > camera > maximum recoverable detail through software reconstruction ?

This is a 4 shot stack at F10 at around 0.6x

Subscribe to see EXIF info for this image (if available)

moth on melaleuca by PIG, on Flickr

This is an 11 shot stack at F9 about 2.25x -

Subscribe to see EXIF info for this image (if available)

cross polarised blue banded bee by PIG, on Flickr

Even though effective aperture here is really huge in equivalent terms (F72 @ macro / 35mm equivalent is getting a bit past the point where we consider diffraction our friend) , the smallest details we're interested in are still resolvable. Possibly there's more detail to be discovered in the grass stalk and the surface of the leaf, and you can see that basically maxing out to a point where every couple of pixels we have quite large differences in colour/luma where maybe we could get better resolution stopped down just a bit at F4 and doing more stacking. But we're not really shooting that, we're shooting the scales on the wings and the edges of compound lenses and hairs.

 

[mfturner]

... snip...

So I would suggest the effective MP curves I posted 2.5 years ago
... snip...

@archaeopteryx, at the end of the 2.5 year old thread, you posted some plots of blur disks at different focal lengths, apertures, etc. Is that calculator available online? I'd like to see what it does a you approach infinity focus, like with the old Harold Merklinger method of infinity focus vs hyperfocal focus to get distant things a bit sharper. My cell phone and the 9mm bcl are good examples, it's hard enough to get either to focus on something useful, so I often slam them to infinity for a shot or two which sometimes is preferable (given the small absolute apertures of either).

 

[archaeopteryx]

Is that calculator available online?

I don't know of any online calculators which chart, or even do more than one calculation at a time, so using one that way seems tedious. I did all of that thread in a couple spreadsheet tabs, for example, since circles of confusion and Airy discs are simple formulae.

(F72 @ macro / 35mm equivalent is getting a bit past the point where we consider diffraction our friend)

I can't reproduce f/72? The DCR-150+250 stack is a 78 mm f/2.1, which provides the front EA ≈ MN = (2.25)2.1 = f/4.7. Since you mention f/9, I presume that's the infinity aperture of the Oly 60. f/9 is enough narrower than f/4.7 the combined EA of the coupled lenses is going to be essentially that of the Oly 60, so presumably in the range of f/18 to f/9 depending where a given frame is in the autofocus bracket's progression. The full frame equivalent is at magnification 2M = 4.5 but the resulting EA also depends on how that's obtained. For example, pushing down a 5x infinity objective is likely to yield a front EA in the vicinity of f/16. Assuming that's the limiting aperture, it's around 4.3 stops faster than f/72.

A corollary is opening up the Oly 60 decreases the extent to which the Raynoxes are stopped down from behind. Since the DCRs are 3/2 formulas they're not well corrected for their speed, so the result may be degradation of image quality due to increase in aberration being larger than decrease in diffraction. My experience suggests infinity f/9 on the Oly 60 may not be too far from optimum but I haven't seen details on how Olympus might automatically adjust exposure to compensate for changes in EA over the bracket. (For Panasonic post-focus I suspect what happens is aperture and ISO are held constant with shutter speed increasing as brackets move to lower magnification.)

What I usually do is measure EA directly and compare image quality on the same stack across a range of front (where applicable) and rear apertures. In cases where the depth of the autofocus bracket noticeably larger than the depth of the composition it's sometimes useful to also vary the position of the composition within the bracketing range. Some of the front lenses I use are noticeably responsive to this, others not much.

 

[piggsy]

The full frame equivalent is at magnification 2M = 4.5 but the resulting EA also depends on how that's obtained. For example, pushing down a 5x infinity objective is likely to yield a front EA in the vicinity of f/16. Assuming that's the limiting aperture, it's around 4.3 stops faster than f/72.

I've just always assumed - macro equivalent aperture at 1x = double indicated. At 2x = same again. Then your m43 crop vs 35mm equivalent = double again.

 

[hoodlum]

It is interesting that Olympus recently published a patent to support AF tracking with a dual/quad pixel sensor. It looks like Olympus believed that they could eventually have access to Canon's Dual/Quad sensor. Maybe Canon is considering making a m43 sensor to help them increase overall sensor production. Competition would be good.

https://asobinet.com/info-patent-oly-dualpixel-af/

 

[mike3996]

I've just always assumed - macro equivalent aperture at 1x = double indicated. At 2x = same again. Then your m43 crop vs 35mm equivalent = double again.

A lens' macro reproduction has nothing to do with crop factors. A 1:2 lens is a 1:2 lens no matter what sized sensor its image is being projected.

 

[BDR-529]

It is interesting that Olympus recently published a patent to support AF tracking with a dual/quad pixel sensor. It looks like Olympus believed that they could eventually have access to Canon's Dual/Quad sensor. Maybe Canon is considering making a m43 sensor to help them increase overall sensor production. Competition would be good.

Canon was the first one to use dual pixel AF in 70D in 2013 but only three years later Samsung launched Galaxy S7 which had dual pixel AF sensor as well.

I don't know who holds essential patents because Canon uses in-house sensors but so did Samsung. S7 had Isocell sensor but S20 family uses mix of Sony (IMX) and Samsung sensors including IMX555 which also supports dual pixel AF.

Sony has quad pixel camera sensor with omnidirectional AF already in production but Canon has also been rumoured to work on one.
https://www.sony-semicon.co.jp/e/products/IS/mobile/2_2_ocl.html

 

[piggsy]

A lens' macro reproduction has nothing to do with crop factors. A 1:2 lens is a 1:2 lens no matter what sized sensor its image is being projected.

Yes I know. The magnifications mentioned in the post aren't doubled. The effective aperture does increase with magnification though.

 

[archaeopteryx]

I've just always assumed - macro equivalent aperture at 1x = double indicated.

That's correct for a lens obtaining magnification by extension as the formula there is EA ≈ (M + 1)N. This is accurate within light meter rounding for all conventional 1x macro lenses I'm aware of as well as for the Canon, Laowa, and Mitakon 1-5x lenses. It appears to also hold for the recent set of 2x macros from Laowa and others. However, the next doubling occurs at M = 3 rather than M = 2.

Coupled lenses have the advantages of not relying on expansion of the image circle as the primary mechanism for obtaining magnification and being one stop faster for sufficiently large magnifications. Microscope designers recognized this in the 1930s, so it's a little odd Canon chose to rely on extension for the MP-E 65 mm and that Mitakon and Laowa retained the approach. EA ≈ M N_front tends to be a good approximation for M ≥ 4 or so. Below that it's typically easiest to measure since one rarely has the design information needed to model the lenses.

An important special case, though, is if the clear aperture of the front lens in a coupled lens exceeds the rear lens's field of view then pair has EA ≈ EA_rear because the front lens isn't the limiting aperture within the pair. Most use of close up lenses occurs under this condition, hence the claims about close up lenses not costing light.

Then your m43 crop vs 35mm equivalent = double again.

I actually can't think of any common case where this occurs. At magnifications around 1x lens configurations tend to result in N being roughly constant, at magnifications around 4x and above there's usually an effort to decrease N (increase the numerical aperture) to offset increases in M. One could certainly construct such situations, though.

A lens' macro reproduction has nothing to do with crop factors.

While correct as to the specific claim made, this isn't really applicable to preceding discussion. Give or take differences in aspect ratio, full frame requires twice the magnification for a given composition to cover the same percentage of sensor area as on m43. Since this is a property of only the sensor sizes it holds for any finite magnification. Whilst other definitions of equivalent magnification can be constructed I would suggest they're typically contortionate and thus rarely useful.

It's widely accepted under the above criteria that m43 offers lower size, cost, and weight of lenses relative to larger formats. Whilst the largest in long glass, reductions occur at all focal lengths. The distinction of macro, photomacrography, and photomicrography in this regard is mainly the relevant maths aren't as widely (mis)understood and the arguments over conventional portrait photography depth of field are irrelevant.