Diffraction and 21 MP, 1DsMk III and 5D Mk II

[images_in_light_north_west]

I am very interested to hear from people that actually have a 1Ds Mk III, I have heard and read much, I don't

need to see more numbers I want to here real world experience from owners/users. Has 21 MP put you up against a

wall when stopped down with an sMkIII and where ? f11 ? with very good conditions can you go lower say f16 with

acceptable results, I am most interested in Landscapes and don't want to blow all the cash to find 21MP is to

much for landscape work, I use all L glass and am thinking of getting a couple of prime wide angle's. Thanks in

advance for your replies. Ross

[bobatkins]

"...can you go lower say f16 with acceptable results..."

 

Yes.

[zml]

f/9 is risk free, but, you are not likely to see any effects of diffraction in real life even at f/16. Yes, I'm an actual user of 1Ds3 (even though I don't do straight landscapes.)

[obakesan]

Ross

<P>

I found that rather than spend big buck$ on snazzy and intricate L zooms that I got great results from three

simple lenses TS-E 24, 45 and 90. Especially if you have full frame (I used to used these on film where there was

only full frame) and do landscapes you'll find the tilt particularly advantageous. Try looking at the larger

version of this test shot I took

<P>

<a href=" title="tilt advantage in DoF by obakesan, on

Flickr"><img src="http://farm4.static.flickr.com/3204/2869751916_736bfac8cc.jpg" width="500" height="328"

alt="tilt advantage in DoF" /></a>

<P>

the DoF goes from left to right back to front and this was with the 90mm @ f5.6

<P>

HTH

[alanrusso]

Sorry - what's the relationship between the number of pixels and diffraction? I thought this was more a function of the lens and aperture.

[scott_ferris]

Hi Ross, I got you thinking didn't I?

 

Sorry, but my original comment was of the, where do we go from here, type. 21MP work fine at f22 on the 36X24mm sensor, it is when we go higher in MP's that there will be no IQ increase, the limits will be diffraction and lens resolution.

 

Best advice for you, don't believe anybody or the numbers, rent a camera for the week, most places will give you some credit towards a purchase, use your lenses in the situations you would use your camera and then judge the IQ for yourself. But believe me the IQ from the next generation cameras will not be higher, at normal range ISO, than the latest 5D MkII or the 1DS MkIII.

[savagesax]

As I've repeated several times I think the 1DS mark 3 is so good I sold my hasselblad gear. The turning point was viewing huge enlargements, up to 40X60, both nature and wedding pics. The Canon "L" glass works and the colors are better then any other Canon camera. If you set the 1DS mark 3 correctly you can expect to print landscapes very close to medium format Velvia 50 film, which has been my favorite landscape color positive film for at least 15 years.

 

As far as glass, I've been very happy with results from F4 to F-22. It depends on what you are shooting and the length of the lenses.

[davin_george]

Got a 1DSIII myself and I haven't seen any issues at all with it, and I've used it a f/22 on my 24-70/2.8 and she's been absolutely beautiful! I also use only L series bar the Canon 100mm Macro and the MPE-65 and wouldn't have it any other way. I'm disappointed in Canon effectively devaluing my 1DSIII but having checked the 5DII I wouldn't give it up. Personally I wouldn't give up my 1DSIII and can't recommend it highly enough, however I do have to wonder whether Canon has a 1DSMKIV on the horizon? And if so what they're doing to do for us MKIIIs?

[scott_ferris]

No Alan, this is the page that I started Ross thinking about all this

 

http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm

 

 

It is a function of pixel size to aperture. and the distance from the diaphram to the sensor.

[bobatkins]

You might also want to take a look at this page which discusses both the theory and shows some image samples - http://www.bobatkins.com/photography/technical/diffraction.html

 

Basically with full frame you're pretty much fine at f16. For 35mm you'll always get the highest resolution from a lens when it's somewhere around f5.6 to f8, at least in the center. Some lenses may need stopping down more to improve edge and corner quality.

 

The pixel count really doesn't matter, it's the format size. So whether you shoot film, a 21MP camera or a 40MP camera, if they are all 24x36mm it really doesn't matter as far as stopping down goes if you are pinting the same sized prints

[oscar_van_der_velde]

<A HREF="http://web.canon.jp/imaging/eosd/eos5dm2/html/sample2_e.html">http://web.canon.jp/imaging/eosd/eos5dm2/html/sample2_e.html</A> shows a nice landscape sample image taken at f/7.1. If you view it large you will notice the low contrast areas hardly contain any detail, somehow. Diffraction should not be an issue at this aperture, so maybe the sensor outresolved the L lens?

[awindsor]

Remember that there is only one plane of focus. Everything else is out of focus to a greater or lower degree. At f7.1 the softness you are observing is most likely due to depth of field considerations.

[images_in_light_north_west]

I saw that and wondered why Canon would show that shot, Lanscape at f7.1 when it probably s/b f11, I have some landscapes shot at f7.1 but they didnt have that kind of depth. Ross

[images_in_light_north_west]

Thanks all for your help, I am considering the new Distagon T* 2.8/21 and the 5D MkII, I am really looking forward to

this camera. Ross

[g dan mitchell]

The notion that increasing the number of photosites somehow compromises performance in terms of diffraction blur is

common...

 

... and precisely wrong, backwards actually.

 

If you shoot a scene with some lens (let's say a 50mm f/1.4 EF lens) and some aperture (let's say f/11) on a camera

with lower photosite density (e.g. - the old 5D) and a camera with higher photosite density (e.g. 5D II) the image

projected on the sensors will have _exactly_ the same amount of diffraction. Make a print from the two at any size at

which you would realistically want to make a print from the 5D and the blur from diffraction will be identical.

 

(It would be accurate to say that the camera with more photosites can capture a more accurate image of the diffraction,

but this does _not_ increase the amount of diffraction.)

 

However, if you open up the lens a bit on the two cameras there is the _potential_ (depending on the lens quality and

your shooting technique) that the camera with greater number for photosites can produce a marginally _better_ image at

the larger aperture. This is a Good Thing.

 

So, if you move from one camera in a given format to another in the same format that has more photosites, there is NO

DETERIORATION of the image quality due to diffraction and there could be some improvement at larger apertures.

 

Dan

[scott_ferris]

Hi G Dan Mitchell,

 

Well it depends if you read and understand the links really, if you don't then you are going to live under a misconception.

 

Your bracketed fourth paragraph sums it up perfectly, you are just resolving the same blur more accurately! So there is a finite number of pixels above which you get no effective IQ increase, it seems that given the sensor size and the f stops we are using we have pretty much got there at 20+ MP. Certainly when they start pushing 30+ MP then all you are doing is using up storage space not increasing IQ even at wide f stops.

[._kaa]

<p>I tend to think of it this way. There are three limits to what resolution you can get out of your camera/lens

combination: (1) sensor resolution; (2) lens resolution; (3) diffraction. These limits are not all-or-nothing and

they interact, but as a useful oversimplification your resolution is constrained by <b>the first limit you run

into</b>.

 

<p>Long time ago sensors were crude. The first limit you hit was the number of sensor pixels and that determined all.

 

<p>Then sensors improved. If you didn't have good lenses, you could hit the lens limit before the sensor limit.

Thus good lenses started to matter more.

 

<p>Then sensors improved more. With a good lens and a closed down aperture you can now hit the diffraction limit

before you encounter the sensor limit or the lens limit. Somebody mentioned that one could shoot 1DsMkIII at f/22

-- of course one could. But you won't get 21 megapixels of resolving power because you hit the diffraction limit

before you hit the sensor limit and it's the first limit that matters.

 

<p>Kaa

[zafar1]

Pictures speak louder than calculations

 

http://www.flickr.com/photos/pizzo/2744127241/sizes/o/ (@f/13

[g dan mitchell]

"Your bracketed fourth paragraph sums it up perfectly, you are just resolving the same blur more accurately! So there is a

finite number of pixels above which you get no effective IQ increase, it seems that given the sensor size and the f stops we

are using we have pretty much got there at 20+ MP. Certainly when they start pushing 30+ MP then all you are doing is using

up storage space not increasing IQ even at wide f stops."

 

My point about "resolving the same blur more accurately" is only relevant to the question of resolution at a given aperture

where blur is the limiting factor.

 

As I wrote, there is the "potential" for better system resolution with the camera with a larger number of photosites if a) you use

an excellent lens, b) you select a larger aperture, c) you make a large enough print such that the difference might be visible.

 

It is also good to keep in mind that equal sensor and lens resolution is essentially an impossibility. The only other alternatives

are a) lens outresolves sensor, and b) sensor outresolves lens. I would prefer a sensor system that has some headroom for

the best lens I will use in the best circumstances in which I'll use it. For that reason alone I would ultimately want a sensor

that can outresolve my lenses.

 

None of this matters much if you don't make fairly large prints, use excellent lenses, and use careful technique.

 

Dan

[scott_ferris]

Isn't that what we were talking about?

 

I agree with you 100% and the physics shows that the 5D MkII is effectively limited by blur at f16 and above.

 

It raises an interesting thought for the development of the next pro series cameras. Until now the studio,

greater DOF, people have had the higher pixel counts, but the people who could use them are the wide open sports

shooters. Does this lead us towards a future of high pixel count and high frame rate and very expensive, and a

lower pixel count lower frame rate and nearly as expensive, or will Canon have to take the 1DS up to medium

format to increase IQ and give us a 28MP FF 10fps digital 1VHS, I could only hope!

[oscar_van_der_velde]

G Dan Mitchell wrote "So, if you move from one camera in a given format to another in the same format that has more photosites, there is NO DETERIORATION of the image quality due to diffraction and there could be some improvement at larger apertures."

 

Diffraction works like a blur with a fixed metric radius (circle of confusion) which depends on f-stop. So it will have the same effect looking at the entire image which is meant by G Dan. At 1.6 crop you magnify this blur more than at full frame. If you have a higher pixel density sensor for the same format you are digging deeper into the blur, so that pixel to pixel contrast is lower and the advantage over the less dense sensor can become very small. This becomes worse when the diffraction circle becomes larger than the space between the sensor sites.

 

So, while your image as a whole doesn't get worse by diffraction, your extra pixels on the same sensor size do not necessarily give extra detail. The same goes for depth of field: a certain f-stop may render everything in focus in the image on both sensors, but the higher the amount of MP the narrower the region that looks truly sharp at the pixel level and the larger the fraction of pixels carrying no additional detail.

[g dan mitchell]

"So, while your image as a whole doesn't get worse by diffraction, your extra pixels on the same sensor size do not

necessarily give extra detail."

 

Right. Unless you use a very good lens that can provide more resolution at a larger aperture then the one at which

diffraction blur becomes an issue. Here the added photosites _may_ allow you to take better advantage of large apertures

on such lenses - with no loss in image quality at the smaller aperture.

 

Dan

[mark_blum]

I am a film macro photographer doing research to go digital. I am aware diffraction and how it's effects relate to

photosite size on the sensor. But I also understand that this is just calculations and real world photography is

something else. I don't have a digital SLR yet, so can't do any tests. So I am asking how small an aperture can be

used to still get "acceptable quality".

 

(What is acceptable depends on the proposed use...let's just say 18" x 27" print for common reference. How small

an aperture will allow it to remain sharp and contrasty enough? Assume contrast can be bumped up with USM, at a

cost, but if the detail is not there, no amount of USM will sharpen it.)

 

Just by the numbers, even with a Canon EOS 5D (12.7MP) and its huge 8.2 micron photosites, the smallest optimal

f/stop is f/13. But of course, macro photographers often shoot much smaller effective apertures than this (for

example, due to extension of the lens from the sensor). I am considering the 5D Mark II with 6.4 micron photosites.

My camera design (a special purpose application), would ideally use these lens and aperture combinations.

 

75 mm, M = 0.94, D = 145 mm, f/64 desired, f/45 acceptable.

 

90 mm, M = 0.59, D = 235 mm, f/45 desired, f/32 acceptable.

 

105 mm, M = 0.33, D = 420 mm, f/32 desired, f/27 acceptable.

 

M = Magnification in my application

 

D = Distance from front of lens in my application.

 

Can I get away with these apertures or am I dreaming? I know this depends on other variables beside photosite

size, including the lens, magnification, vibration, etc. But any general or specific observations you can offer will be

welcome.

 

Mark