Why I need MPs

[neuroanatomist]

But I certainly don't want to get into an endless argument either as I doubt it would help anyone.

Well, if you believe that the sensor characteristics play no role in determining the effect of diffraction on an image, it definitely would be endless... Can you provide some documentation or evidence that the pixel density of the sensor has no effect on the aperture at which diffraction affects image sharpness?

 

[Canon-F1]

Well, if you believe that the sensor characteristics play no role in determining the effect of diffraction on an image, it definitely would be endless... Can you provide some documentation or evidence that the pixel density of the sensor has no effect on the aperture at which diffraction affects image sharpness?

of course he can not.
but some people would even argue that they can build a spaceship that´s faster then light.

i mean what is so complicated to understand?
everyone with some school education that includes math can look at the wikipedia link and see that what is written there it´s correct.

Cameras

If two objects imaged by a camera are separated by an angle small enough that their Airy disks on the camera detector start overlapping, the objects can not be clearly separated any more in the image, and they start blurring together. Two objects are said to be just resolved when the maximum of the first Airy pattern falls on top of the first minimum of the second Airy pattern (the Rayleigh criterion).
Therefore the smallest angular separation two objects can have before they significantly blur together is given as stated above by

Thus, the ability of the system to resolve detail is limited by the ratio of λ/d. The larger the aperture for a given wavelength, the finer the detail which can be distinguished in the image.
Since θ is small we can approximate this by

where x is the separation of the images of the two objects on the film and f is the distance from the lens to the film. If we take the distance from the lens to the film to be approximately equal to the focal length of the lens, we find

but f/d is the f-number of a lens. A typical setting for use on an overcast day would be f/8.[8] For blue visible light, the wavelength λ is about 420 nanometers.[9] This gives a value for x of about 4 µm. In a digital camera, making the pixels of the image sensor smaller than this would not actually increase image resolution.

 

[Canon-F1]

Perhaps you wrote in haste.

The diffraction from the lens does, of course, produce a slightly fuzzy (unsharp) representation of the image on the sensor. This fuzziness in the focal plane does, by no means depend on anything to do with the sensor.

well you don´t know what we are talking about.

the fuzziness does not depend on the sensor sure.... nobody said it does.
but the RESOLUTION is limited when the airy disc is bigger then the photosites.
and the photosites depend on the sensor.

the bad thing is he gave you TWO links to get some infos.... but you seem to ignore it.

the lens (the lens design) has nothing to do with the resolution limit on sensor level.
it depends on the aperture and the wavelength of the light ONLY.
for this we assume a PERFECT lens.... so no negative effects by lens design (imperfections).

so even in 100 years you will still be bound by this simple rule of physics.
because you can´t build a better lens then a PERFECT lens.

the only thing a REAL lens does it to make the diffraction effect WORSE.
but ultimately at some photosite size you can´t resolve more details even with a physical and mathematical PERFECT lens.

to resolve more details you have to go to smaller wavelengths (smaller then visible light).

when you look at at this link: http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm

you should see why smaller photosites can not resolve more information then the size of the airy disc.

In fact the fuzziness is exactly the same whether there is a sensor there or not e.g. if film were used.

that´s correct but you have to think it through!
no matter if FILM or a DIGITAL SENSOR, when aperture and wavelength combination create an airy disc that´s bigger then a photosite or a grain on film, then you can´t resolve more then that size of the airy disc. because there is not more INFORMATION.

ergo... at some point, independend from lens quality, more megapixel (smaller photosites) on a sensor will not result in better resolution.

Note that you will never see an Airy disk in a real photograph unless you photograph point sources.

you should read the articles again. :
this one sentence shows your lack of understanding.
the first sentences in the wikipedia article should give you an hint:

"In optics, the Airy disk (or Airy disc) and Airy pattern are descriptions of the best focused spot of light that a perfect lens with a circular aperture can make, limited by the diffraction of light.

The most important application of this concept is in cameras and telescopes. Owing to diffraction, the smallest point to which a lens or mirror can focus a beam of light is the size of the Airy disk. Even if one were able to make a perfect lens, there is still a limit to the resolution of an image created by this lens. An optical system in which the resolution is no longer limited by imperfections in the lenses but only by diffraction is said to be diffraction limited."

 

[Denabears]

Can you provide some documentation or evidence that the pixel density of the sensor has no effect on the aperture at which diffraction affects image sharpness?

Let me try this argument, which I hope is based on assumptions that we can agree on. 3 points make the argument (which is by no means original):

1) The sharpness, more precisely the MTF, of the system (camera+lens) can be expressed as the product of exactly 3 factors:
a) MTF curve describing lens aberration excluding diffraction,
b) MTF curve diffraction (not quite a simple as the correct formulae kindly given by Canon-F1 would suggest)
c) sensor MTF.
Note: as usual for MTF curves a, b and c are all functions of spatial frequency (normally declining with frequency).

2) a and b are functions of the aperture (as well as spatial frequency). We all know that factor a usualy improves as the lens is stopped down quite quickly at first, then more slowly (from f/4 or so for most decent lenses).
We also all know that factor b follows the inverse relationship hinted at by Canon-F1's Rayleigh criterion - but that is only one point on the curve. I suspect the most important point is that some (perhaps little) contrast remains to much higher spatial frequencies (and yes, there are nulls where there is no contrast at all, for one color of light at a time).

All I claim is that factor c does not depend on the f/no. (which seems self-evident to me)

3) The product of a x b x c therefore has a maximum determined by max(a x b) and is scaled overall by c. A "perfect" sensor will have c = 1 at all spatial frequencies, and so the MTF is just a x b any lower pixel count sensor (with a smaller value of c independent of f/no.) gives a smaller product at any f/no. (QED)

I believe that anyone who thinks I'm mistaken has to disagree with one of the 3 points (and it would be good to know which), or we have a purely semantic disagreement (most likely due to the 100% viewing when the image gets bigger as the pixels get smaller - that is not the situation I'm describing!).


ps. Seeing another post by Canon-F1. The pixel / Airy disk illustration given by Sean McHugh is seriously misleading I know that many many people have been misled by it. Sampling theory tells quite a different story. It is best to calculate using MTF in the spatial frequency domain, as suggested above.

pps. not even neutrinos go faster than light never mind spaceships.

ppps. there seems to be a bug that prevents me from italicizing one of the "c"s.

 

[Gothmoth]

1) The sharpness, more precisely the MTF, of the system (camera+lens) can be expressed as the product of exactly 3 factors:
a) MTF curve describing lens aberration excluding diffraction,
b) MTF curve diffraction (not quite a simple as the correct formulae kindly given by Canon-F1 would suggest)
c) sensor MTF.

you can forget that argumentation because we already assume a perfect lens.
again, the LENS has NO influence on the SMALLEST POSSIBLE SIZE of an Airy Disc.

a perfect lens can achive this SMALLEST size, a less then perfect lens can only achive a BIGGER airy disc size.

you can´t do better then the physical minimum.

if you go under this minimum with the size of the photosites you will not resolve more detail.
once you are under this minimum you can make the photosites smaller and smaller but you will not get more details. it´s simply not there.

All I claim is that factor c does not depend on the f/no. (which seems self-evident to me

and all i say is that at some point the resolution can not be increased by better lens design nor smaller photosites (because of diffraction).

of course you can resolve more details with smaller photosites when you go to shorter wavelengths. because the size of the airy disc will be smaller with smaller wavelenght.
but that is no real solution for most photographers as we like to photograph visible light not ultraviolett and shorter.

 

[Denabears]

Oh dear, as I feared we are all getting at cross-purposes here, because the assumptions and question are not clear enough (that is even more evident from Gothmoth's perfect lens post that has just popped up).

"In optics, the Airy disk (or Airy disc) and Airy pattern are descriptions of the best focused spot of light that a perfect lens with a circular aperture can make, limited by the diffraction of light.

The most important application of this concept is in cameras and telescopes. Owing to diffraction, the smallest point to which a lens or mirror can focus a beam of light is the size of the Airy disk. Even if one were able to make a perfect lens, there is still a limit to the resolution of an image created by this lens. An optical system in which the resolution is no longer limited by imperfections in the lenses but only by diffraction is said to be diffraction limited."

That's absolutely correct, of course. I did not claim there was no diffraction limit for lenses! - I'm sorry if something I wrote it made you think so.

My claim is that there is no diffraction limit for sensors, so that a higher resolution sensor always gives a better result (however slightly).

Thanks for the links I am already extremely familiar with them (but you were not to know that). Sean McHugh is not, however, the best source as he ignores sampling theory.

If you look carefully at my first post you should see that I effectively say that we need 3 blue pixels to resolve the "Airy disk" not just 1. (If you read between the lines and think what produces the 80 lp/mm figure.)

I am being much more subtle than you give me credit for (but that's cool).

 

[dirtcastle]

What about post-processing?

Do increased megapixels give greater ability to process images?

It seems like people are discussing too many variables here.

PLEASE PEOPLE...

Is there an "all other things equal" way to compare different megapixel counts ???

All these formulas and extra variables mean 90% of people viewing this thread probably don't understand the discussion.

 

[Gothmoth]

so that a higher resolution sensor always gives a better result (however slightly).

and this is only correct if you can make the wavelength of the rays you collect smaller and smaller.

i repeat again:

In a digital camera, making the pixels of the image sensor smaller than this would not actually increase image resolution.

the resolution will not get worse when we ignore noise and other effects... but it will be stagnating at some point.

 

[Denabears]

but that is no real solution for most photographers.

Yes absolutely, perfect lenses are very expensive. That's why I started this whole string of threads off with the 80 lp/mm figure from DXOmark: not theoretical, not nearly the best that can be done, but it still justifies of order 200Mpixels. Diffraction imperfect lens and all.

After all many of us have an 85/1.8 (or an even better lens) and we'd see (a litte) benefit with 200Mpixels at f/5.6.

P.s. another point - compacts with 2 micron pixels are reasonably sharp at f/5.6. That is also about 200Mpixels on full frame (decide yourself).

 

[Denabears]

so that a higher resolution sensor always gives a better result (however slightly).

and this is only correct if you can make the wavelength of the rays you collect smaller and smaller.

i repeat again:

In a digital camera, making the pixels of the image sensor smaller than this would not actually increase image resolution.

No, the wavelength is of course assumed constant in the 3 point argument. Which step don't you agree with?

 

[Denabears]

What about post-processing?

Do increased megapixels give greater ability to process images?

It seems like people are discussing too many variables here.

PLEASE PEOPLE...

Is there an "all other things equal" way to compare different megapixel counts ???

All these formulas and extra variables mean 90% of people viewing this thread probably don't understand the discussion.

That's actually a really good point. Extra pixels help preserve high frequency detail that would be lost when transformations are done e.g. to remove lens distortion, level a horizon etc.

I'm sorry that by my mention of diffraction (which was not even necessary for my argument) I caused the thread to get messy - mea culpa.

Even better: I'll shut up, unless asked a direct question.

 

[Gothmoth]

No, the wavelength is of course assumed constant in the 3 point argument. Which step don't you agree with?

ok.. once more before i go to bed.

my viewpoint is a viewpoint of the physical OPTIMUM.
i don´t care about how good the real lenses are.
for my argumentation i already have a PERFECT lens (because i talk about the OPTIMUM performance, the best theoretical possible performance). it´s a mathematical view on the best possible resolution.

so with a PERFECT lens i have a MINIMUM size of an Airy Disc.

the amount of detail a sensor will resolve will INCREASE when i make the photosites SMALLER.... UNTIL i reach a point when the photosites are SMALLER then the Airy Discs.

how much smaller is open for argumentation.
but you will reach that size at some point (btw: that is also why we need other devices then visible light microscopes. because they are limited by the the wavelength of visible light).

so once we have reached that size we can make the photosites on the sensor smaller and smaller but we will NOT resolve MORE detail. at this point you can increase the megapixels (the theroretical sensor resolution) but the output resolution (real information) will be stagnating.

that is because ot the mathematical formula given in the wikipedia article.

the only way you can resolve more details with smaller photosites at this point is when the wavelength of the rays you collect with your sensor will be smaller/shorter (ultraviolett for example).
because a shorter wavelength creates a smaller airy disc with a given aperture.

 

[CanonLITA]

Is one or more of the following correct?

1. even assuming a perfect lens, there will always be a minimum size of the resolvable light information (due to the way light waves behave when they pass through the lens aperture and are forced to change their direction - i.e. difraction) and therefore no more information would be captured be decreasing the size of the pixel below this physical threshold;

2. the minimum useful size of a pixel (as per 1 above) is so small that nowaday we are pretty far from it and it still makes sense (resolution wise, please disregard cost) to squeeze more than 21mp in a FF sensor;

3. even when the minimum useful size of a pixel (as per 1 above) is reached, squeeze more pixels in the sensor still makes sense (resolution wise, please disregard cost) to (a) better post processing, and/or (b) overcome the limits of a real-world-less-than-perfect lens.

Thank you all for your thoughts.

 

[Gothmoth]

Is one or more of the following correct?

1. even assuming a perfect lens, there will always be a minimum size of the resolvable light information (due to the way light waves behave when they pass through the lens aperture and are forced to change their direction - i.e. difraction) and therefore no more information would be captured be decreasing the size of the pixel below this physical threshold

correct and that is why at some point a visible light microscope is not good enough.
the wavelength of the visible light is limiting the size of the structures we can resolve with a visible light microscope.

2. the minimum useful size of a pixel (as per 1 above) is so small that nowaday we are pretty far from it and it still makes sense (resolution wise, please disregard cost) to squeeze more than 21mp in a FF sensor;

in theory yes (if you use big apertures and not f13 and smaller). (*)
but in the real world we have to keep lens imperfections, noise etc. in mind.
so 21MP is not near the limit but with todays technology i question the usefulness of 36MP or even 45MP FF sensors.

(*) with small apertures you are today limited even with a perfect lens.

3. even when the minimum useful size of a pixel (as per 1 above) is reached, squeeze more pixels in the sensor still makes sense (resolution wise, please disregard cost) to (a) better post processing, and/or (b) overcome the limits of a real-world-less-than-perfect lens.

kind of.
in a perfect world (theoretical world) more photosites would not make things worse.
but we don´t live in a perfect world.

(a) the size of the files will be bigger -> slower to process in camera and on the PC, more storage space needed. but no real gain on information.

(b) in theory yes. but i doubt that you will see it.
i mean look at the 18MP 7D... crap lenses make no visible better images with that camera.

 

[Richard8971]

I can't tell you how many times I see guys out shooting who have high $$$, high MP cameras/L glass who can't shoot their way out of a wet paper bag. It is far better to KNOW your equipment, regardless of what EOS body you use and use it frequently!

I have gotten outstanding photos from my XTi through my 7D. My wife still loves her T1i that we bought new 3 years ago and refuses to upgrade to a "pro" body. I only got my 7D because my 40D was damaged beyond repair, not because I needed/wanted a higher MP camera.

Truth of it is, the higher the MP, the better the glass needs to be to take full advantage of the higher pixel count. Hand shake CAN effect the quality of the photos, regardless of IS or not. Unless you are doing SERIOUS cropping (and if you do, maybe you should buy the lens you really need for the job vs cropping) I really doubt you would ever notice how many MP there were in the PRINTED photo taken from ANY camera.

Guys, the only time you ever really NOTICE the resolution is when the photo is on your monitor and you are scrolling in to each and every pixel. Take a 8-10MP rebel and take a photo. Take the same photo with a 18MP 7D and have 8x 10's made. I doubt you would ever really be able to tell them apart. The 40D vs 50D was a great example of this.

I am still a huge believer that good glass & knowledge of your equipment is far superior to high MP camera bodies. If you are buying a new camera, buy the one that you can AFFORD and best suits your needs, regardless of MP size. (IE, 6.5fps 40D vs 8fps 7D, these are very fast cameras and about a grand difference between the two. BOTH produce outstanding images! You do the math.) If you already have a camera, keep it and learn it WELL! Only upgrade if your current camera cannot keep up with your needs, not wants. But, of course, if you have plenty of money and a budget isn't of concern to you, none of this really applies. I am mainly speaking to the working folk of the world on this point. If you make your living with photography then I would still say the same. Buy what you can afford and learn it!

Will any of this stop the MP race? No, and it shouldn't. Cameras will continue to advance and grow in MP. I think however, that more and more people are starting to see that the QUALITY of the photo (regardless of MP size) is far more important than the SIZE of the photo. Right now, the technology is set to do just that and Canon has led the way with the new 1D X by changing people's perception of needing higher and higher MP!

 

[Richard8971]

I do have to add one point. I would say that the question SHOULD be, not the size of the MP of a sensor but the SIZE OF THE SENSOR!

In other words, FF vs crop sensor. Now, that being said, the original Rebel at 6MP has about the same pixel density of the FF 5D at twice the resolution. Both produce very clean images at similar ISO settings. I'm not talking about cropping, scrolling down to each pixel, blowing up the image... yada, yada, yada. I am simply talking about taking a photo and having a 4 x 6 made!!! I bet you would be hard pressed to tell them apart.

Pixel density has a greater effect on image quality over MP size of the sensor! Unless some breakthrough takes place in CMOS/CCD techology, it will be this way for a long time.

At this point in camera tech, FF produces smoother photos than a crop sensor with the same resolution at higher ISO settings with similar lenses. Just the way it is...

D

 

[dtaylor]

but in the real world we have to keep lens imperfections, noise etc. in mind.
so 21MP is not near the limit but with todays technology i question the usefulness of 36MP or even 45MP FF sensors.

People keep debating the maximum useful FF resolution without considering what we know from APS-C.

The 7D and 5D2 are pretty close in terms of MP (18 and 21). Having a smaller sensor, the 7D is like a crop from a 45 MP FF sensor.

If real world lenses and shooting conditions would render 45 MP of little or no benefit on FF, then they should impact the 7D to the same degree. In other words, even though the 7D and 5D2 are close in terms of MP, their actual performance on resolution charts should be much further apart if the 7D is hobbled by lenses and shooting conditions.

DPReview has tested the resolution of both bodies. They typically stop down to f/8 in these tests. They are:

7D Absolute Extinction (JPEG test)
H 2500 3100
V 2450 3050

5D2 Absolute Extinction (JPEG test)
H 2800 3300
V 2700 3300

Note: I'm using the JPEG test numbers because DPReview didn't do a RAW resolution test on the 5D2. RAW numbers would be a bit higher because ACR is better at resolving fine detail then Canon's JPEG engine / DPP, but the relative difference would remain the same.

The difference in measured resolution is actually a bit less (<12%) then we would expect based on the MP difference (16%). This is real, hard, observed data. If your theory conflicts with it, your theory is wrong, not the real world experience.

The 7D sensor is not limited by the 50mm f/1.4 @ f/8. That's a very good lens at f/8, but it's hardly the only lens that can resolve that much detail, and f/8 is actually just touching on diffraction effects for the 7D's pixel density. This tells us that a 45 MP FF sensor would be quite useful.

How useful? Stitch three 7D frames together overlapped so that you end up with roughly 45 MP. The resulting image will out resolve 6x7 film on an Imacon, and resembles the 645D sample images I've seen in terms of sheer detail and clarity.

Not everyone needs this. Many people would love it. That's why Canon needs a fast shooting/low light FF sensor (1Dx) and a high resolution FF sensor.

 

[dtaylor]

Guys, the only time you ever really NOTICE the resolution is when the photo is on your monitor and you are scrolling in to each and every pixel. Take a 8-10MP rebel and take a photo. Take the same photo with a 18MP 7D and have 8x 10's made. I doubt you would ever really be able to tell them apart.

I make 16x24" and 20x30" prints. The last two portfolios I made were 17x22. I guarantee you I could tell them apart.

 

[Richard8971]

Guys, the only time you ever really NOTICE the resolution is when the photo is on your monitor and you are scrolling in to each and every pixel. Take a 8-10MP rebel and take a photo. Take the same photo with a 18MP 7D and have 8x 10's made. I doubt you would ever really be able to tell them apart.

I make 16x24" and 20x30" prints. The last two portfolios I made were 17x22. I guarantee you I could tell them apart.

??? The statement I made was regarding 8 x10's, not prints that large.

And for the record, I have prints in my studio that I have made from my lowly 10MP XTi as large as 16 x 20 that I would put against ANY camera, regardless of MP. Here are a few that I have made large prints from. Mind you, I had to compress them down a tad to allow the 'CR' server to accept them but trust me, there is NO loss of detail or "crispness" in the full size prints! (The ones I posted are jpeg's at 25% compression!)

 

[archangelrichard]

The OP asked a practical question and got an awful lot of impractical answers (a long with a very few practical ones - like Danabears first response)

Look here, folks. FILM is a monstrously higher resolution (anyone who doubts this, try enlarging a sharp film negative to 48 inch x 60 inch - then try doing so with a digital print ... at 1200 dpi for high quality to match the film as close as you can (my printer does 9600 dpi!) and look at the results)

Remember lenses worked to get that sharpness in film; should be able to achieve the same in digital so the lens is NOT the limiting factor; at least not in the way many have stated here (however chromatic aberrations are different for film over digital so that may cause some issues)

So, to the OP; YES what you say is true and it is possible at some point in time that this will happen - but not with today's technology. In medium format you can go with an 80 MP mamiya / Phase One (645 means 60mm x 45mm sensor size); there can be incremental increases in MP count in 35mm format but you are not going to see anything like 80 MP any time soon - that 80mp sensor is a density of 25.6 MP for FF sensors so we know that this is possible easily and we can see the quality of it; above that there are a lot of guesses of 32 - 36 MP's being possible / practical / pending - see all the rumors on this site

NO we are nowhere near the limits of possible technology as many here would like to think; more is definitely possible; but you could be talking of different materials for sensors, etc. to get much higher; Yes for the purpose the OP mentioned, more resolution / more MP is possible and will help in your work as you need to crop (and ignore the insults you have been getting about reframing, etc. from people who failed to read and comprehend the Original post); it just won;t be that soon