Pixel Perfect wrote:
This is so funny - it's shit stirring at it's best
But what size prints do you want. Those hyperfocal distances are based on a 8x10" print, hardly what I'd call a large print. Make 16x20" print and use those numbers and you'll be disappointed.
Ah, ok, I did not know that they based their numbers on small prints! thanks for the heads up.
n0b0 wrote:
If you wanna test for diffraction, try shooting high magnification macro, you'll see the effect better.
Here's a couple of examples. f/4 and f/5.6 f/8 and f/11
These were shots using natural light so the shutter speed might've exaggerated the effect. I'll do another test later with flash.
Last note, diffraction is not so bad if you don't crop your photos, but if you do, you're in trouble.
I don't care what word one would pick to describe the difference. But the difference demonstrated in these examples are pretty significant to me. And the word significant is not an exaggeration at all in this case.
ejmartin wrote:
It's not the size of the photosite that matters for resolution, it's their spacing.
I don't know how a topic on diffraction got turned into a topic of resolution. You are right in that spatial resolution is dependent on the distance between the centroid from one pixel to its adjacent pixel. However, it is entirely possible to have two different sensors with the same resolution but different photosite sizes whereby one sensor is susceptable to diffraction while the other is not.
After some more thought into this idea, I must backtrack a bit and retract my statement that the "gapless microlens" sensor will be able to detect less diffraction effects. Infact it's quite the opposite by the following reasoning:
Picture a light ray that passes through a lens. Due to a small aperture, diffraction occurs in that this ray of light is no longer a parallel beam but a very narrow cone. That cone of light will have a certain diameter when it hits the sensor (this is known as the Airy disc). Not taking into account the use of anti-aliasing filters, if the diameter of the airy disc is greater than half of the pixel width and the cone of light is centered over the middle of the photosite, the edges of the Airy disc will spill over onto the next pixel. This is what causes blurring and is what we know as the effect of diffraction. Now, if the photosite in the adjacent pixel is very small, the edge of the Airy disc can spill over to this adjacent pixel without being detected. In this case, theoretically speaking diffraction can be calculated and physically speaking, diffraction actually do occur. However, from a user point of view, no diffraction will actually be detected in the image.
thw2 wrote:
OK. But this thread is addressed to a large crowd. Very few people shoot super large magnification macro, let alone own the MPE-65. Besides, as has been correctly pointed out already, diffraction has nothing to do with pixel density. High pixel density merely goes to show what has already happened, without affecting diffraction in any way.
I shouldn't supply any info about how the lens behave with diffraction because there are only a few people who shoot macro at high magnification or own the MP-E? What are you trying to say here? minority doesn't count?
Why are you telling me about the pixel density stuff? I don't remember talking about that at all here. What have you been smokin mate?
beewee wrote:
Just something to ponder... diffraction is a function of the size of the light sensitive area for each tiny sensor (photosite) that make up a pixel and not pixel size which is a function of how many pixels per given area..
No it isn't.
Diffraction is purely a product of the wavelength of light and the size of the aperture used.
The effects of diffraction will however be seen better with a higher resolution recording medium - film or sensor - and/or at higher image magnifications - on screen or print.
beewee wrote:
> However, it is entirely possible to have two different sensors with the same resolution but different photosite sizes whereby one sensor is susceptable to diffraction while the other is not.
That seems entirely wrong to me. Diffraction is a resolution effect. Sensor resolution is set by the distance between each pixel, not the area of the light well. Diffraction will affect a sensor with small light wells and large interpixel gaps no differently than it affects a sensor with the same pixel density ('pitch') with large microlenses and no gaps.
I may be out of my depth on this brilliant thread (thanks, Brainiac), but please tell me where I am wrong. Diffraction, as far as I know is purely optical phenomenon, a form of optical distortion. So, if we compare 2 sensors where resolution is the only difference (no size or technology differences), then the higher-resolution sensor should actually show more diffraction, simply because its higher resolution picks up distortion, as well as "clean" images, better.
This however should not be construed as the lesser-definition image being better. It would be the same as saying that a "noiseless" image is better, when actually noise suppression has wiped out detail as well as noise..
philber wrote:
I may be out of my depth on this brilliant thread (thanks, Brainiac), but please tell me where I am wrong. Diffraction, as far as I know is purely optical phenomenon, a form of optical distortion. So, if we compare 2 sensors where resolution is the only difference (no size or technology differences), then the higher-resolution sensor should actually show more diffraction, simply because its higher resolution picks up distortion, as well as "clean" images, better.
This however should not be construed as the lesser-definition image being better. It would be the same as saying that a "noiseless" image is better, when actually noise suppression has wiped out detail as well as noise.....Show more →
You're not out of your depth at all, you nailed it. To summarize: Diffration=optical phenomenon... Resolution/detail=good.
brainiac wrote:
Diffraction will affect a sensor with small light wells and large interpixel gaps no differently than it affects a sensor with the same pixel density ('pitch') with large microlenses and no gaps.
I don't agree. The more diffraction there is, the larger the airy disk (diffraction pattern) is. Thus, for the same pixels density, pixels with larger gaps are less diffraction-sensitive since a larger airy disk (more diffraction) is needed to cover more than 1 pixel (in average).
Etadam wrote:
I don't agree. The more diffraction there is, the larger the airy disk (diffraction pattern) is. Thus, for the same pixels density, pixels with larger gaps are less diffraction-sensitive since a larger airy disk (more diffraction) is needed to cover more than 1 pixel (in average).
Yes, more diffraction means a larger Airy disk. But what is the effect of blurring the image over the size of the Airy disk? It is to reduce resolution by the amount of the blurring. Resolution is governed by the pixel spacing, not the size of the light-sensitive area in each photosite.
Etadam wrote:
I don't agree. The more diffraction there is, the larger the airy disk (diffraction pattern) is. Thus, for the same pixels density, pixels with larger gaps are less diffraction-sensitive since a larger airy disk (more diffraction) is needed to cover more than 1 pixel (in average).
OK - I see what you mean. But in fact, if diffraction is limiting sharpness, then the lesser contrast between the larger pair of adjacent sensels is entirely recovered by sharpening. In other words, sharpening can make the larger sensel results look exactly like the smaller sensel results, CONSERVATIVELY. There is no gain in resolution by using smaller sensor sites at diffraction limiting apertures, only a mathematically predictable gain in contrast between adjacent sensels. So with a little extra sharpening, the sensor with large sensels can resolve just as much as the sensor with smaller sensels (and better in real life, because there will probably be less noise).
At apertures where diffraction isn't limiting, then the sensor with larger sensels will be MORE accurate, since the gaps in the sensor with smaller sensels could MISS pictorial information that the larger sensel would take into account. The smaller sensels will be less accurate when dealing with detail resolved at the sub-pixel scale. This is a separate issue from noise. A pattern of red and green stripes each one third of a pixel wide might easily look red or green, partly, or wholly, on the small sensel sensor (moire), whereas the gapless sensor would record the image more truly as something nearer a consistent brown. Even disregarding noise, false colour abounds in our files, and moire affects lurk along every edge. This ought to be worse when small light wells account for large pixels, i.e. when gaps between microlenses are relatively large.
this has drifted away from optics to sampling theory... and I think really, given the random nature of real photos that the difference between larger and smaller sensels will average out and it won't make a huge difference with respect to defraction either way. You can devise thought experiments that could prove either side...
Without looking at the other comments I compared by cutting and pasting and aligning on different layers so I could compare one-to-one by turning off layers. I equalized the brightness with levels adj. layers so it wouldn't influence my perception.
It was obvious from the size difference they were in pairs and in each pair 1/2, 3/4, and 5/6 the even numbered image had better overall sharpness and contrast. So I guessed they were the shots taken at f/5.6. I ranked them 1, 6 and 4 in terms of overall IQ preference. Comparing the odd numbered frames I ranked them 1, 3, and 5.
The even numbered shots had more CA than the odd (esp. noticable in trash on roof) and the odd numbered shots had less contrast in the shadows and more random chromatic noise. 1/2 and 3/4 where more or less equal in that regard, but there was more of a difference in 5/6.
Most lenses resolve best at 2-3 stops smaller than wide open, so all other things being equal it would make sense to take advantage of that if shooting a static subject on a tripod. But for average hand-held shooting camera movement and subject movement will probably degrade image sharpness far more than diffraction. A more useful test would be to take a lens and shoot the same scene through the entire range of f/stops so the rate of degradation in IQ could be seen.
mh2000 wrote:
this has drifted away from optics to sampling theory... and I think really, given the random nature of real photos that the difference between larger and smaller sensels will average out and it won't make a huge difference with respect to defraction either way. You can devise thought experiments that could prove either side...
I agree. I don't think braniacs last assertion can be a general result.
Pixel Perfect wrote:
I agree. I don't think braniacs last assertion can be a general result.
When diffraction is limiting resolution, there will be a fairly simple gradient of colour/contrast between any sensel and its neighbour in any direction. The consequence is that a larger sensel will produce slightly less contrast between each pixel and its neighbours in each direction. USM reverses this process by amplifying contrast between each pixel and each of its neighbours. It can do this because light striking a sensor can be regarded as obeying the principle of superposition. The lessening of contrast due to increased sensel size is reversible because it is predictable by pure geometry. The only difference you will see between equally spaced smaller and larger sensels will be a slight lessening of interpixel contrast which is entirely undoable mathematically, e.g. by USM. You either see that this is true, or you don't. I don't think any purpose is served by discussing it here since in reality it will be a very marginal effect. But it is an interesting feature of the sensel size/gap relationship that sensors with narrower gaps probably require slightly more sharpening in order to look identical to output from an otherwise similar sensor with larger gaps.
Chuck Gardner's suggestion of shooting through the entire range of f stops brings up an issue that hasn't been discussed directly and explicitly:
Diffraction increases gradually as the aperture is stopped down, but some lens aberrations decrease gradually at the same time.
The optimal aperture for a lens (aside from any need for depth of field) will be the point where increasing diffraction negates any advantage in control of coma, CA, etc. One reason for saying that a lens is "diffraction limited" at a particular stop is that the resolution is not going to increase past that point. However, in some cases closing down one more stop or so will not decrease the resolution - the diffraction and aberration variables will essentially balance each other in that range. So there may be no measurable performance difference for a certain lens between 5.6 and 8, and maybe between 8 and 11 for another.
Currently, the closest thing to a "high pixel density" body I have is a 1DIII. Pixel peeping with an 85L, the degradation between f/11 and f/16 is quite subtle to me.
"The 50D's higher resolution allows you to see detail at f/16 that you can't see on a 40D file shot at f/5.6. Yes, I know, you can see the loss of sharpness as you hit f/16 but regardless, even at this aperture the higher resolution of the 50D carries the day."
"The 50D's higher resolution allows you to see detail at f/16 that you can't see on a 40D file shot at f/5.6. Yes, I know, you can see the loss of sharpness as you hit f/16 but regardless, even at this aperture the higher resolution of the 50D carries the day."
If the 50D at f/16 is showing more detail than the 40D at f/5.6 based on those shots, I need a double corneal transplant. It's such a small difference it's not even worth mentioning.
