Smaller pixels yield better MTF, even with diffraction

[Daniel Browning]

This discussion about diffraction has been moved from the original thread because it didn't really relate to the topic.



Diffraction limit is diffraction limit - there is basically insignificant or no improvement in MTF by reducing pixel size below 6 microns when shooting at f/5.6 and beyond. Laws of physics are unavoidable.


Diffraction is a physical law, but it's not a brick wall; the effect is gradual. Again, even after diffraction has started to affect the image, further MTF improvement is possible. You don't have to take my word for it, since diffraction is something we can measure. I'll volunteer to take the burden of proof (you can go first next time).

For proof, the easiest comparison would be the 50D at f/5.6 (4.7 micron) with the 40D (5.7 micron), but, unfortunately, there are no suitable tests on the web yet. (By the way, I was mistaken about the 50D, it's only 4.6K, not 5K.)

So until then, we'll have to use test data for other cameras. The best I can find is only for full frame cameras, so we'll have to change the experiment to compensate. The area of Still35 is larger than Super35 (1.6X field of view), which is about one and 1/3 stop, so an equivalent test would be f/9 instead of your f/5.6. We'll test a variety of focal ratios from f/5.6 to f/16.

I'll use the data from Bryan Carnathan's the-digital-picture.com (http://www.the-digital-picture.com/Reviews/ISO-12233-Sample-Crops.aspx?Lens=245&Camera=453&Sample=0&FLI=0&API=6&LensComp=245&CameraComp=356&SampleComp=0&FLIComp=0&APIComp=6). This comparison is the 5D (12 MP) with the 1Ds Mark 3 (21 MP) using the EF 200mm f/2.8. The 5D has a much weaker AA filter, relative to the pixel size, than the 40D, so that will skew the results in favor of your position. Furthermore, although the same raw conversion software (DPP) and settings were used for each camera, Canon might be using a different debayer algorithm behind the scenes for different camera models (I don't know). That factor could be isolated with dcraw.c, if desired.

I've taken the center crops, converted them for the same output size (i.e. so the crop would look the same as if you printed them both at the same size). Click on the thumbnails below to see the image (the thumbnails themselves, of course, are not intended for comparison, they're just links).

Below: The 5D and 1Ds Mark III at f/5.6. There is no effect at all from diffraction in either camera. The debayer artifacts (green and color patterns) are a natural result of the weakness of the anti-alias filter. As expected, the 1Ds Mark III, with over 50% more pixels, has higher resolution. I'm only including the f/5.6 focal ratio to establish a baseline of how much improvement is possible when there is no diffraction at all. Be sure that you can see the difference at f/5.6 before commenting on the other focal ratios.

http://thebrownings.name/photo/diffraction/65-5d-f5.6.jpg (http://thebrownings.name/photo/diffraction/500-5d-f5.6.jpg)

http://thebrownings.name/photo/diffraction/65-1dsm3-f5.6.jpg (http://thebrownings.name/photo/diffraction/500-1dsm3-f5.6.jpg)

Below: The 5D and 1Ds Mark III at f/8.0. Diffraction is beginning to have a very slight effect here, which is noticeable on the 1Ds because the 5D's 8.2 micron pixels add too much of their own blur for the diffraction to be visible. It is softening the very highest frequency of detail.

http://thebrownings.name/photo/diffraction/65-5d-f8.0.jpg (http://thebrownings.name/photo/diffraction/500-5d-f8.0.jpg)

http://thebrownings.name/photo/diffraction/65-1dsm3-f8.0.jpg (http://thebrownings.name/photo/diffraction/500-1dsm3-f8.0.jpg)

[reduser.net limits posts to only four pictures, so the rest of this post is continued below.]

[Daniel Browning]

Below: The 5D and 1Ds Mark III at f/11.0. Now diffraction is very obvious, even in the 5D. But it's plain that the 6.4 micron pixels resolve more detail.

http://thebrownings.name/photo/diffraction/65-5d-f11.0.jpg (http://thebrownings.name/photo/diffraction/500-5d-f11.0.jpg)

http://thebrownings.name/photo/diffraction/65-1dsm3-f11.0.jpg (http://thebrownings.name/photo/diffraction/500-1dsm3-f11.0.jpg)

Below: The 5D and 1Ds Mark III at f/16.0. This focal ratio results in a *lot* of diffraction, as you can see. However, you can still see that the 21 MP provides more detail than the 12 MP. The difference isn't as large as f/5.6, above, but it's there.

http://thebrownings.name/photo/diffraction/65-5d-f16.0.jpg (http://thebrownings.name/photo/diffraction/500-5d-f16.0.jpg)

http://thebrownings.name/photo/diffraction/65-1dsm3-f16.0.jpg (http://thebrownings.name/photo/diffraction/500-1dsm3-f16.0.jpg)

Furthermore, in all the cases above, the higher megapixel camera provided more contrast (MTF) in addition to the increased resolution.

[Pawel Achtel]

Thanks Daniel for the test, greatly appreciated!

I am still trying to digest the results. What is the magnification ratio in every of those test pictures? We need to know that in order to quantify the effect and calculate resolving power [lp/mm] at, say 10% MTF (easiest). Either that, or run the images through Imatest and calculate MTFs at some chosen frequency.

At a glance, it looks like our opinions are of glass being half empty or half full. :biggrin: You say: "Huge" improvement by reducing pixel size, I say "insignificant".

I'd like to put a figure, an estimate on an improvement from, say reducing 6 micron pixels down to, say 3 microns at f/5.6. As we do not have 3 micron pixel chip to compare, we can scale it up and use your results with f/16 and larger pixels, of course.

Let me also chew over the results bit more before drawing conclusions :clown2:

[Pawel Achtel]

OK, assuming the same magnification in both 5D and 1Ds (I assumed the magnification was 1x), here is what my eye seas as resolving power:

f/5.6

36.5 (5D) - 37 (1Ds) [lp/mm]

f/8

36.5 (5D) - 36.5 (1Ds) [lp/mm]

f/11

27 (5D) - 29 (1Ds) [lp/mm]

f/16

26 (5D) - 29 (1Ds) [lp/mm]

So, my point is that the diffraction caused about 30% of resolution loss by going from f/5.6 to f/11, while the benefit of twice the number of pixels resulted in less than 7% improvement.

Thus, I stand by my original statement that reducing pixel size gives insignificant improvement over the loss of resolution caused by diffraction.

In other words, you have doubled the data size and only improved the resolution by less than 10%. Hardly "huge" improvement. Where you lose "huge" 30% of resolution by stepping down from f/5.6 to f/11. This loss due to diffraction limit will, of course, start sooner (like f/4.0) for pixel sizes smaller than 5-6 microns and, it is much more siginificant (namely ~300% more significant in this example) than the benefit of using twice the resolution.

[Daniel Browning]

Thanks Daniel for the test, greatly appreciated!


You're welcome!



I am still trying to digest the results. What is the magnification ratio in every of those test pictures? We need to know that in order to quantify the effect and calculate resolving power [lp/mm] at, say 10% MTF (easiest).



The website says: (http://www.the-digital-picture.com/Help/ISO-12233.aspx)

"The chart I am using is a very expensive, finely printed variation of the ISO 12233 standard that extends resolution to 4,000 lines per picture height (l/ph)"

So the big "28" marker is 2,800 l/ph, 36 is 3,600 l/ph, and 4,000 l/ph at the very left. You can convert these to lp/mm. Both cameras are 36x24mm, so one "ph" is 24mm.



At a glance, it looks like our opinions are of glass being half empty or half full. :biggrin: You say: "Huge" improvement by reducing pixel size, I say "insignificant".


Toe-may-to, toe-mah-to. :) Even at f/5.6 (no diffraction), the difference between the 12 MP and 21 MP would be called insignificant by some. But to me, it is huge.

[Daniel Browning]

f/5.6

36.5 (5D) - 37 (1Ds) [lp/mm]


Are you saying that you only see a difference of 0.5 at f/5.6?

That focal ratio is not affected by diffraction at all, so something must be amiss in my test (or your interpretation) if that is the case.

The 1Ds Mark III has 29% more vertical resolution than the 5D, so we should be seeing that, at least.

[Pawel Achtel]

Are you saying that you only see a difference of 0.5 at f/5.6?

That focal ratio is not affected by diffraction at all, so something must be amiss in my test (or your interpretation) if that is the case.

The 1Ds Mark III has 29% more vertical resolution than the 5D, so we should be seeing that, at least.

Yes, and this is expected (and unfortunate for our test). Essentially you maxed out the lens's resolving power at f/5.6. Here the lens is limited by the aberrations and impurities of the glass, not by diffraction. Thus, there is hardly any difference between 5D and 1Ds at f/5.6 using that lens. This does not mean 1Ds would not resolve significantly more with a sharper lens. There is no surprise here, but unfortunate for our test because the maximum resolving power of both cameras was probably not reached at f/5.6.

This draws me to another conclusion that, in order to further resolving power of the combined system, one has to seek very sharp and very expensive lenses that perform better at apertures wider than f/5.6. If we were to use 5k+ S35 sensors, the only way to see some benefit would be at f/4 and wider with some wicked sharp glass.

Glad I invested in Master Primes - there is a chance they may not get obsolete within a couple of years :biggrin:

[Daniel Browning]

Yes, and this is expected (and unfortunate for our test). Essentially you maxed out the lens's resolving power at f/5.6. Here the lens is limited by the aberrations and impurities of the glass, not by diffraction. Thus, there is hardly any difference between 5D and 1Ds at f/5.6 using that lens.

When I compare the EF 200mm f/2.0 L IS (http://www.the-digital-picture.com/Reviews/ISO-12233-Sample-Crops.aspx?Lens=458&Camera=453&Sample=0&FLIComp=0&APIComp=3&LensComp=245&CameraComp=453&SampleComp=0&FLI=0&API=4) with the f/2.8, I don't see a large difference in resolution, which leads me to believe it is the camera, not the lens, that is limiting resolution.

In any case, my point in asking you the question is that I don't interpret the above graphics the same way you do. To my eyes, the 21 MP camera has about 24% higher lp/mm at f/5.6: I would say that "25" on the 5D corresponds to about "31" on the 1Dsm3.

Maybe it would help to look at an area of the chart with slightly lower detail frequency.

[Pawel Achtel]

Yeah, you may be right, now that I look at f/5.6 again, it looks like 26 (5D) and 30-32 (1Ds) - really hard to interpret where the lines start to blur :blink:

I guess, at this resolution, the MTF of the lens would be not higher than 20%, the MTF of the sensor would be not higher than 20%, thus combined MTF would be only 4%, or simply put: bugger all :biggrin:

To see the effect of the resolution clearly, it'd be nice to have a lens with MTF greater than 80% at the sensor resolution limit. I don't think such glass exists.