^Shame that that wikipedia article has not one photographic example of artifacts related to the point spread function. I doubt that it's ever that clear cut that you can identify that as the cause of something like the artifact in FPF's image, but that's just my uneducated thinking.
There's a Zeiss? article around the 'net somewhere that has some examples of what different lenses' PSF looks like, actual pictures of the weird-looking blobs resulting from a single point input. I'm not sure where it is but I might go look for it. They can be pretty strange-looking.
One thing that I just realized though, is that they can look like what happens when you're trying to burn something with a magnifying glass and sunlight. If you tilt the lens a bit wrong, the resulting image on the ground or whatever has a certain look to it, like a point that kind of "explodes" out on one side.
My main point is that lenses can and do a lot more than just blur things smoothly. Another thing we've all seen: bokeh rings. While this is what happens to part of the image that is OOF, it's similar to what might be happening at a micro level for in-focus things. Lenses that produce "ring" bokeh are taking energy which should be spread out evenly, but are actually adding a type of high-frequency information by making the edges of the disks higher intensity.
Best not to confound the discussion with differences in both format and resolution (Mp). My view is that larger formats deliver a 'lazy' resolution pattern to image data that is entirely satisfying down to 100%. I see better tonal transitions in FF than in smaller format images, reminiscent of the effect of film formats and indeed MF digital.
Higher Mp images seem disadvantaged in per pixel comparisons with lower Mp images, for any given angle of view and scene - more pixels on an image object demands more processing than larger pixels covering the same object. Maybe more image space is taken up with between-pixel spaces, given equivalents of the dimensions of these spaces in between for all sensors? A speculation, I have no idea why.
Regarding the philosophy of sharpness and photography, the very last people I would listen to are the fossils from a bygone age, and these tales from the crypt serve only to inflate their importance while sneering their condescencion at 'lesser' photographers with an interest in technical issues - of course there is a wide gradient of this balance in the population - of art 'versus' technique.
Of course the activity has moved away from their view as we now need focus attention on camera settings, post processing s/ware, as well as printers/paper, etc. All make a huge difference to the output.
Note that HCB and the other luminaries of the past almost all used the best gear they could get their hands on, the hypocrites.
Taylor, you can find it at the beginning of the seminal work by Nasse: How to Read MTF Charts Part I:
OK, so I've heard this "better tonal transitions in bigger formats" over and over and was never able to come up with a concept to understand it. Does anyone have a mildly technical explanation where this is supposed to come from? If two sensors of different format have similar pixel pitch (like the Sony 16MP sensors and the D800), where do the "better tonal transitions" come from, and do they disappear when the D800 image is cropped to 16MP? Do I also get "better tonal transitions" when stitching several frames with the smaller sensor?
I get the fact that larger pixels (or more pixels on a larger area that get downsampled later) produce an overall better result, but my understanding of "tonal transitions" is that it is a local effect (small areas compared to the overall sensor), and I don't see how sensor size could play a role here.
dcjs wrote:
OK, so I've heard this "better tonal transitions in bigger formats" over and over and was never able to come up with a concept to understand it. Does anyone have a mildly technical explanation where this is supposed to come from? If two sensors of different format have similar pixel pitch (like the Sony 16MP sensors and the D800), where do the "better tonal transitions" come from, and do they disappear when the D800 image is cropped to 16MP? Do I also get "better tonal transitions" when stitching several frames with the smaller sensor?
I get the fact that larger pixels (or more pixels on a larger area that get downsampled later) produce an overall better result, but my understanding of "tonal transitions" is that it is a local effect (small areas compared to the overall sensor), and I don't see how sensor size could play a role here.
I don't know the answer to your question, myself. If I were to guess though, I'd think that "tonal transitions" would be a combination of dynamic range and color separation. The larger pixels *could* effect the dynamic range, while I'd think the CFA would be more responsible for the color separation. Either one without the other could create, visually, a small posterization effect.
Scroll down about 2/3 and read "PIXEL SIZE: NOISE LEVELS & DYNAMIC RANGE"
Or just the text without the graphic aids and reference links here:
Cambridge Colour wrote:
"Larger sensors generally also have larger pixels (although this is not always the case), which give them the potential to produce lower image noise and have a higher dynamic range. Dynamic range describes the range of tones which a sensor can capture below when a pixel becomes completely white, but yet above when texture is indiscernible from background noise (near black). Since larger pixels have a greater volume — and thus a greater range of photon capacity — these generally have a higher dynamic range.
Further, larger pixels receive a greater flux of photons during a given exposure time (at the same f-stop), so their light signal is much stronger. For a given amount of background noise, this produces a higher signal to noise ratio — and thus a smoother looking photo.
This is not always the case however, because the amount of background noise also depends on sensor manufacturing process and how efficiently the camera extracts tonal information from each pixel (without introducing additional noise). In general though, the above trend holds true. Another aspect to consider is that even if two sensors have the same apparent noise when viewed at 100%, the sensor with the higher pixel count will produce a cleaner looking final print. This is because the noise gets enlarged less for the higher pixel count sensor (for a given print size), therefore this noise has a higher frequency and thus appears finer grained." ...Show more →
ーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーー philip_pj wrote:
Regarding the philosophy of sharpness and photography, the very last people I would listen to are the fossils from a bygone age, and these tales from the crypt serve only to inflate their importance while sneering their condescencion at 'lesser' photographers with an interest in technical issues...
dcjs wrote:
OK, so I've heard this "better tonal transitions in bigger formats" over and over and was never able to come up with a concept to understand it. Does anyone have a mildly technical explanation where this is supposed to come from? If two sensors of different format have similar pixel pitch (like the Sony 16MP sensors and the D800), where do the "better tonal transitions" come from, and do they disappear when the D800 image is cropped to 16MP? Do I also get "better tonal transitions" when stitching several frames with the smaller sensor?
I guess the thing is that your example is an extreme one, D800 is the first FF that comes close to the pixel density of an APS-C. And, since the MP count is vastly different a similar sized output image would for the D800 use many more "photon-buckets" to combine a point in the final image than the APS-C.
For the more normal FF cameras, the MP counts of APS-C are closing (even surpassing) while the sensor sizes stay the same. Hence instead of more "photon-buckets for the FF", they will simply be larger and therefor gather more light. And the bigger sized pixels gives other benefits which is more tied to handling and lens issues.
The big and bold thing with the D800 from my perspective, is that it house the potential for enormous sized and detailed prints (/or provides ample material for cropping), but it takes some effort to harvest it, not just point and shoot.
I totally get the fact that larger pixels, in principle, can have higher dynamic range and less noise. However, this just supports my suspicion that the "better tonal transition" thing is erroneously attributed to larger sensors just because, on average, they tend to have larger pixels. As I have pointed out with the 16MP APSC/36MP FF-example, this is not always the case, and so far I see nothing to convince me that this whole thing is not just another confusion of correlation and causation.
As you can see, the moire stops in patches where there's an interruption to the pattern.
I was also probably using a B+W Top Linear MRC Polarizer.
Could someone put arrows or markings in that image to point out the moire? I see a noisy facade with what I believe is chroma noise strangely patterned in some adjacent fields between the windows. Looks like the sensor was not able to harvest enough light there - but light wise this is not a very challenging situation, no?
Moire to me is something else, natural (high-frequency) repetitive patterns yielding new un-natural patterns which have no origin in the original scene.
wfrank wrote:
Could someone put arrows or markings in that image to point out the moire?
Sure, but
I see a noisy facade with what I believe is chroma noise strangely patterned in some adjacent fields between the windows.
you pinpoint it yourself. It is no noise, but moiré, which is a special case of aliasing.
Moire to me is something else, natural (high-frequency) repetitive patterns yielding new un-natural patterns which have no origin in the original scene.
Moiré is a low-frequency pattern that has its origin in a high-frequency pattern in the original scene.
Cambridge Colour wrote:
"Larger sensors generally also have larger pixels..
dcjs wrote:
I totally get the fact that larger pixels, in principle, can have higher dynamic range and less noise. However, this just supports my suspicion that the "better tonal transition" thing is erroneously attributed to larger sensors just because, on average, they tend to have larger pixels. As I have pointed out with the 16MP APSC/36MP FF-example, this is not always the case, and so far I see nothing to convince me that this whole thing is not just another confusion of correlation and causation.
Yup, part of the reason I linked that particular explanation is that he says "This is not always the case" repeatedly.
I haven't checked out the official spec-sheet (if one even exists) or any bench-tests on the D800 yet but I would naturally assume the IQ to be a downgrade over a 24mpx FF sensor and probably even over the D700. In my thinking the only reason to go with a FF is DR. I don't even think it's better in the FL equivalency arena - as I like long lenses and to me 20mm on a 135 format is too wide to work with - unless it's fisheye curved for funky creativity. And I also dig being able to use 135 format lens's center area sharpness advantage! I never even see "bad corners" unless the lens is a really poor DX or a µ4/3 native.
I personally want less noise, more DR, and less diffraction limitation!!! More pixels kicks each one of those things right in the balls! So I'm not really understanding why camera companies would want to go over 24mpx on a FF or over 12mpx on a µ4/3. It would make MUCH more sense to me if they added all the other improvements and enhancements but left the sensors at 12 for µ4/3, about 16mpx for APS-C and 24 for FF. <shrug> I could even understand them wanting to go lower like 8,12,16mpx respectively! I'd be happy as a clam with GH2 speed and ability yet with an 8pmx sensor! Continuous rate would double, the buffer would offload faster, there would be more shots per buffer, I'd get more DR, less noise, and less diffraction limitation! And the only downside would be that crops would be limited to 25% of the image space as opposed to 16% as the GH2 is now.
Bifurcator wrote:
I personally want less noise, more DR, and less diffraction limitation!!! More pixels kicks each one of those things right in the balls! So I'm not really understanding why camera companies would want to go over 24mpx on a FF or over 12mpx on a µ4/3. It would make MUCH more sense to me if they added all the other improvements and enhancements but left the sensors at 12 for µ4/3, about 16mpx for APS-C and 24 for FF. <shrug>
Well, the funny thing is that the D800 has better DR than almost all MF backs and the ability to push exposure without noise and with good colours is insane from what I've seen.
People have been bitching about the "megapixel race" in very much the same way when the move from 8MP to 12MP sensors came, including the fear of diffraction, even though this fear of high pixel density sensors being worse because of diffraction is a misconception very similar to the "tonal transition" thing. High pixel densities don't mean that you will get hit by diffraction earlier, it just means that low pixel densities have lower resolution even before diffraction kicks in. They were worse all the time without you noticing it (resolution-wise, that is).
My impression is that the general acceptance of reasonable pixel densities lags behind the available sensors by a factor of about 1,5 at all times. When the 21MP Canon 5d2 came out, people said "16MP would have been enough", now that the 36MP D800 is out, the acceptable "reasonable" limit for FF is 24MP.
Btw, Your 12MP MFT sensor has almost twice the pixel density compared to a 24MP FF sensor, why is 12MP OK for MFT, but 48MP is too much for F?
Bifurcator wrote:
I could even understand them wanting to go lower like 8,12,16mpx respectively! I'd be happy as a clam with GH2 speed and ability yet with an 8pmx sensor! Continuous rate would double, the buffer would offload faster, there would be more shots per buffer, I'd get more DR, less noise, and less diffraction limitation! And the only downside would be that crops would be limited to 25% of the image space as opposed to 16% as the GH2 is now.
and yet you can't have enough sharpness?
less diffraction limitation means less resolution by the way. hitting the diffraction limit means you are achieving the maximum resolution theoretically possible at a given aperture size. since you can never have enough sharpness you should want sensors with enough pixels so that they demonstrate that your super sharp lens is in fact diffraction limited by f/2.8 or whatever it happens to be.
Bifurcator wrote:
I personally want less noise, more DR, and less diffraction limitation!!! More pixels kicks each one of those things right in the balls! So I'm not really understanding why camera companies would want to go over 24mpx on a FF or over 12mpx on a µ4/3. It would make MUCH more sense to me if they added all the other improvements and enhancements but left the sensors at 12 for µ4/3, about 16mpx for APS-C and 24 for FF. <shrug>
dcjs wrote:
Well, the funny thing is that the D800 has better DR than almost all MF backs and the ability to push exposure without noise and with good colours is insane from what I've seen.
OK, but isn't that due to other improvements? Like the OM-D has better DR over all other µ4/3 16mpx sensors yet it's the same size and pixel density as all the others. Why? Better supporting components, better engine, and maybe even some slight cheating.
People have been bitching about the "megapixel race" in very much the same way when the move from 8MP to 12MP sensors came, including the fear of diffraction, even though this fear of high pixel density sensors being worse because of diffraction is a misconception very similar to the "tonal transition" thing. High pixel densities don't mean that you will get hit by diffraction earlier, it just means that low pixel densities have lower resolution even before diffraction kicks in. They were worse all the time without you noticing it (resolution-wise, that is).
Not sure I follow that. I mean is it or isn't it true that f/32 looks like crap on a 12mpx µ4/3 while f/22 looks to be the same level of crappyness on a 16mpx µ4/3? The tiny (and it's SUPER tiny!) bit of resolution increase you get at 16mpx doesn't even begin to recover the diffraction losses - this I've tested and have read "experts" demonstrating it as well.
My impression is that the general acceptance of reasonable pixel densities lags behind the available sensors by a factor of about 1,5 at all times. When the 21MP Canon 5d2 came out, people said "16MP would have been enough", now that the 36MP D800 is out, the acceptable "reasonable" limit for FF is 24MP.
I see your point. I'm still thinking 16mpx on FF is better than 36. In fact I think 12mpx would be awesome! Of course given that they also add in all of the other enhancements they currently have in place. I betcha the D800 with 12mpx could do an ISO 4 or 5 times what it can now before noise became an issue. And with continuous rates of 24 to 30 fps. Screw the video codecs!!!! etc etc.
Btw, Your 12MP MFT sensor has almost twice the pixel density compared to a 24MP FF sensor, why is 12MP OK for MFT, but 48MP is too much for F?
But I don't really think it is. There is a trade off. 6mpx is too low as there is zero room for cropping and prints actually do start coming apart above A4. 8mpx is probably about right. I'll still think the same things myself when chips are doing 200mpx. It seems to me that it's the other (non-resolution) enhancements which are employed generationally which allow the higher pixel densities to be acceptable each go-around. Add those same improvements (and maybe something better than Bayer Color Filter Arrays for more clarity) and I think the general state of digital photography would be better than it currently is. Of course all I know is what I read and what is produced from my limited tests with my own cameras. So I could be totally off.
ーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーーー Bifurcator wrote:
I could even understand them wanting to go lower like 8,12,16mpx respectively! I'd be happy as a clam with GH2 speed and ability yet with an 8pmx sensor! Continuous rate would double, the buffer would offload faster, there would be more shots per buffer, I'd get more DR, less noise, and less diffraction limitation! And the only downside would be that crops would be limited to 25% of the image space as opposed to 16% as the GH2 is now.
sebboh wrote:
and yet you can't have enough sharpness?
That's a quote from Ed tho, right?
less diffraction limitation means less resolution by the way. hitting the diffraction limit means you are achieving the maximum resolution theoretically possible at a given aperture size. since you can never have enough sharpness you should want sensors with enough pixels so that they demonstrate that your super sharp lens is in fact diffraction limited by f/2.8 or whatever it happens to be.
Then how would I get DOF? I don't wish to be shooting mostly at under f/8 ya know... I guess no one does. While I don't think deep DOF shots are generally very useful in this gear forum I love shooting at f/32 and deeper - it's wonderful!
Bifurcator wrote:
Then how would I get DOF? I don't wish to be shooting mostly at under f/8 ya know... I guess no one does. While I don't think deep DOF shots are generally very useful in this gear forum I love shooting at f/32 and deeper - it's wonderful!
increasing pixel density doesn't increase the effects of diffraction, it just allows you to see the effects of diffraction that you did not notice before because your sensor couldn't resolve the difference in diffraction. increasing pixel density won't take away detail at f/32, it will just decrease per pixel sharpness. at f/32 a 36mp image will look better printed at 8 x 10 than the same image taken with a 12mp sensor of the same size.
sebboh wrote:
and yet you can't have enough sharpness?
less diffraction limitation means less resolution by the way. hitting the diffraction limit means you are achieving the maximum resolution theoretically possible at a given aperture size. since you can never have enough sharpness you should want sensors with enough pixels so that they demonstrate that your super sharp lens is in fact diffraction limited by f/2.8 or whatever it happens to be.
Bifurcator wrote:
Then how would I get DOF? I don't wish to be shooting mostly at under f/8 ya know... I guess no one does. While I don't think deep DOF shots are generally very useful in this gear forum I love shooting at f/32 and deeper - it's wonderful!
sebboh wrote:
increasing pixel density doesn't increase the effects of diffraction, it just allows you to see the effects of diffraction that you did not notice before because your sensor couldn't resolve the difference in diffraction. increasing pixel density won't take away detail at f/32, it will just decrease per pixel sharpness. at f/32 a 36mp image will look better printed at 8 x 10 than the same image taken with a 12mp sensor of the same size.
But you're already way way WAY past the line frequency of any ink-jets I know of at 12mpx for a full 8x10 page so I don't see how this is even possible. Got any proof? Like some scans of the two? And then of course human vision is limited to about one arc minute (1/60°) at any distance - so that's like 200ppi on the high outside or about 3.5mpx at only 10 inches away from the page. (yeah, that is three point five and not thirty five megapixels.) So that assertion kinda flies in a lot of faces.
Bifurcator wrote:
But you're already way way WAY past the line frequency of any ink-jets I know of at 12mpx for a full 8x10 page so I don't see how this is even possible. Got any proof? Like some scans of the two? And then of course human vision is limited to about one arc minute (1/60°) at any distance - so that's like 200ppi on the high outside or about 3.5mpx at only 10 inches away from the page. (yeah, that is three point five and not thirty five megapixels.) So that assertion kinda flies in a lot of faces.
indeed, at 8x10 the difference will be very hard to see and come mainly from smoother transitions and better/more downsizing based enhancements. i just picked the size arbitrarily, the point is that even at f/32 the higher mp image will look better at any given size. there is no viewing size where diffraction hurts the higher mp sensor compared to the lower mp sensor.
