RE-THINKING RESOLUTION

[Bill McCrystyn]

I had always been taught the smaller the aperture the better the resolution. Set for f/8 or f/11 is always better than f/2.8 or f/4. I have been told now, by people I respect, that a lower ISO, 100-200 and a more wide open aperture is the way to go. Sweet 16 at say ISO100 1/800 f/5.6 or 1/1600 f/4 if necessary is a better setting.

I assume there must be math behind this to prove it out but it is beyond me. While the top lenses we use are very good wide or near wide open it seems contrary to me that a wide open lens can have the resolution of f/8-f/11.

Can anyone speak on this?

[John Chardine]

I don't think there's much mystery here Bill. Noise contributes to apparent sharpness or "resolution" of an image so the lower the better. In regard to f-stops, I think it's fair to say that many modern, fast, high-quality lenses are designed by the optical engineers to work at their best wide open so although you might get a small improvement in sharpness by stopping down, it's not a big. The other factor is that high-res sensors have small sensor sites packed close together and apparently you then can see a degradation in image quality due to diffraction effects and small apertures. BPN member Roger Clark has the best web page on the planet on this and other digital photography technical subjects:

http://www.clarkvision.com/imagedetail/index.html

and

http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary/

[Bill McCrystyn]

Thank you John. Roger has commented on this before but in regard Dynamic Range improvement, but I wondered, assumed, that lens aperture settings at a more wide open posture would negate or overwealm any resolution benefit. Reading it again from Charles Glazier got my attention. I appreciate the links and will read then carefully. I think poor Roger gave up on me about it.

[Kerry Perkins]

Bill, times have changed and digital has re-written some of the rules. Modern professional lenses are designed to be very sharp at their wider apertures, and due to diffraction issues already discussed you will actually find a degradation at higher f/stops. For example, the Canon EF70-200mm f/2.8 L IS lens is sharpest at its wider apertures and it is not even recommended to shoot with it at f/8 or beyond. Not at all like the old days of "f/8 uber alles". =:o)

[Bill McCrystyn]

Thanks Kerry. Reading some of Rogers stuff http://www.clarkvision.com/imagedeta...mance.summary/ (http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary/) that John refered has gone a long way. Now I feel like I have a much better understanding of what's going on. Also what I didn't realize was how far these top lens have come and just how well they perform. Apparently allot of work has been done to overcome diffraction.

[Jason Franke]

I would wager that a very, very large portion of the “pro lenses are sharp at wide apertures” comes from trying to provide sensors with enough resolution not trying to overcome diffraction. That coupled with the idea that you’re paying for a fast lens, it better be sharp at wide apertures other wise it’s just a waist of money.

Either way, sharp and fast are very difficult to achieve in a lens (look at the Canon 70-200s, the f/4s are sharper than the f/2.8s even when the 2.8 is stopped down to f/4), and require significant design resources behind them. We are seeing much better designs now, I think by and large, because we have much more computational power and better computer models to put behind optimizing lens shape and placement to maximize image quality.

As for diffraction, it hasn’t changed, it can’t, it’s the behavior of light not what’s recording it. What has changed is our ability to look at things much more closely, possibly too closely.
<o></o>
Ultimately, the real concern is whether the airy disk a given aperture produces is smaller or larger than the circle of confusion that we see in a given print. If it’s smaller, diffraction isn’t an issue anyway, if not then it may be. In other words, can we actually see the effects of diffraction in the final print?
<o></o>
In any case, I wouldn’t worry about diffraction all that much if at all. I certainly wouldn’t limit myself to a given max aperture unless I knew from previous experience that I would have problems there. The decrease in acutance can, to some degree, be recovered with sharpening and contrast corrections in post processing, so long as diffraction hasn’t killed all the detail.
<o></o>
Either way you cut it, knowing your largest intended print size and viewing distances, I think is far more important than anything else, as that will set your maximum acceptable aperture far more than anything else. As the image gets smaller (so long as the viewing distance doesn’t change disproportionately) it should tend to appear sharper. The web is really bad, or good for doing that. For example, based on what I can tell, an uncropped image from a 1Ds-II or 5D-II presented here would need to have been stopped down past f/25 for softness due to diffraction to noticeable. I just threw together a test*, and I’m hard pressed to tell which images are shot at which aperture with out looking at the metadata, at least on my monitor.

At least that's my understanding and I'm sicking by it. :p
<o></o><o></o>


* I used a 40D with 100-400 @ 100mm shooting a bar code on a receipt, from about 8 feet away. Images were taken at f/5.6, f/8 f/11, f/16, f/22 and f/32. When down sampled to BPN’s max resolution I’m hard pressed to see any difference between them, at 100% the image shot at f/32 there are differences, but not a whole lot. At any aperture smaller than f/32 I’m hard pressed to tell what was shot at what with out looking at the metadata. I’ve posted the images I generated for this quick and dirty test at my blog, here (http://www.pointsinfocus.com/2009/01/31/quick-diffraction-test/). I would recommend doing a similar set of images and playing with them yourself (especially printing them), just to see how things work with your specific camera/processing workflow and output size.

[John Chardine]

I would wager that a very, very large portion of the “pro lenses are sharp at wide apertures” comes from trying to provide sensors with enough resolution not trying to overcome diffraction.
snip.

Not really Jason when you consider that many of the fast, pro lenses available today were there when film was the medium, in one form or another.

[Jason Franke]

Not really Jason when you consider that many of the fast, pro lenses available today were there when film was the medium, in one form or another.

If you're saying that the lens have been around in one form or another, that's not useful at all. A 16-35mm f/2.8L II isn't just a 16-35 f/2.8L with a II painted on the barrel, they are completely different designs. Though yes, there are fast pro-grade lenses that haven't been replaced/updated recently.

If you're saying that lenses were designed with diffraction limits being the primary motivation for increasing resolving power (i.e. a lot of work to overcome diffraction), even though film was the medium in common use; well I can certainly see the argument there. But increasing resolving power can just as easily be argued for on it's own merits. Certainly a higher resolving lens is beneficial in dealing with diffraction, but I'm not sold on it [diffraction] being the primary motivator for increasing resolving power (let alone how big of an issue diffraction really is out side of the test bench). At least that's what I was trying to say however poorly worded it ended up being.

More precisely, I should have phrased the statement as, "I would wager a very large portion of the desire to increase the sharpness of pro lenses at wide apertures is simply to have a sharper lens at wide apertures, and not because of diffraction at smaller apertures". Either way, I'm not a lens designer, unfortunately, so I don't know what the exact criteria and motivations are; which is why I phrased the statement as a guess not a statement of fact. If there is someone out there, who can authoritatively say that yes, diffraction at narrower apertures is the overriding reason for designing towards increased sharpness at wider apertures, I'd love to know.

[John Chardine]

Jason- I'm thinking for example of the classic bird lens for many years- the Canon 500 f4.5 L. Not the same lens as the current 500/4 but apparently sharpest wide-open just the same. There must be lots of other examples but maybe they are mostly at the tele end where I imagine it is much easier to do this than at the wide-angle end. It's not a mystery- you design fast, long lenses, charge the price of a car for them, and they end up being used at all the big sports venues you care to mention, wide open with fast shutter speeds and relatively low ISOs of film. If you do this as a lens company such as Canon or Nikon you had better make sure the lens performs optimally wide open. Again thinking mainly of super-teles.

I would agree with your rephrase.

[Bill McCrystyn]

I think no one argues that strides have been made in lens resolution particuarly at the wide end, but I am willing to bet (not my new lens) that my thinking has to change and understand in the light of digital sensors advancement and A/D and amp performance after read has more to do with a result of better S/N (hence res) than tweeking the lens. In a way some things haven't changed. Low ISO (ASA ;)) still rules the day and bigger badder lenses are just iceing on the cake.

[Roger Clark]

Hello All,
(Bill, I didn't give up on you; I went to Tanzania--another GREAT trip. This time great leopards,
one image is up so far on my home page.)

Regarding diffraction, I agree with a lot of the discussion so far. The higher quality lenses do better wide open than lower cost lenses in general, but few lenses are diffraction limited wide open. What people usually notice is in the better lenses is only a little improvement as one stops down, at least with medium to long telephotos. With wider angle lenses (less than about 100 mm) stopping down to f/8 can help a lot, especially in off-axis details, even on high end lenses.
Most lenses have designs that date before the digital era (~2000) began.

My 500 f/4, for example is noticeably sharper if I stop down 1/3 stop from wide open, but f/8 is pretty similar to f/4 (diffraction is kicking in a little). So every lens is different and when you get picky, it probably is every copy of the same model lens may be a little different. I know for example, my 100-400 L is not as sharp at 400 as a friend's lens of the same model.

Sometimes I find images a little soft and there are several factors: e.g. the trade between ISO and
available light, f/stop (trading lens sharpness and depth of field), subject motion, camera movement, and autofocus accuracy. It is not always obvious which to trade, especially in the heat of action when you have a fraction of a second to make decisions as well as compose and follow the action.

I must say I have a new appreciation for the sharpness of my 500 f/4 L IS lens. I took my 1D Mark II and 5D Mark II to Tanzania. I figured on the 1DII for action and the 5DII for scenics. After a couple of days into the trip and comparing images, it became obvious the 5DII was so superior that it became my primary camera for both action and scenics, and the 1DII as backup when I didn't have time to change lenses. (By the way not a single dust spot on either sensor the whole trip, despite dry and dusty and very frequent lens changes.)

The 5D2 failed at perfect AF more often on action than I believe the 1D2 would have done, but I found the 1D2 had a slight error in focus so I would reject about the same number of images from each camera in the end. But 21 megapixels is so awesome and the focus was spot on down to the pixel level on so many images (probably >90%) that I really got a new respect for the 500 f/4. Note the leopard image on my home page was down 1/3 stop from wide open.

Having said all that, and probably rambling as I am very jet-lagged, a couple of references I have on diffraction. See figure 8 at:
http://www.clarkvision.com/imagedetail/does.pixel.size.matter
It shows, for example, that at 5-micron sampling (in the range of many current DSLRs), the MTF, or contrast is zero at about f/8. The blur filter mitigates this problem so you won't notice much softening until lower than f/8. So don't hesitate to use f/8. I use f/8 to f/11 for landscapes all the time. Most lenses I've tested for landscape work (20 to 150 mm) tend to have peak performance around f/8 to f/11. And if I need the depth of field, I'll certainly use f/16, or slower and not be concerned about diffraction.

Roger

[Don Lacy]

Bill, times have changed and digital has re-written some of the rules. Modern professional lenses are designed to be very sharp at their wider apertures, and due to diffraction issues already discussed you will actually find a degradation at higher f/stops. For example, the Canon EF70-200mm f/2.8 L IS lens is sharpest at its wider apertures and it is not even recommended to shoot with it at f/8 or beyond. Not at all like the old days of "f/8 uber alles". =:o)
An image made with a modern digital camera at f/8 our higher will still have more detail then an image made with film diffraction is a property of light and has nothing to do with the sensor. One should shot at the aperture needed to achieve the image they wish to create with the knowledge that diffraction exist and that they might be giving up some resolution for DOF.

[Bill McCrystyn]

Thanks Roger. Congradulations on your new toy. 21 MP ought to take care of it. Your cats are cool.

Is my thinking right that a 21MP full sized sensor will achieve the "equivalent" of about 1.75 stops of resolution compared to a 12MP (D300) 1/2 sensor size version? If so, does a 12MP D300 fall 1 stop short on resolution of a 12MP D700?

[Roger Clark]

Is my thinking right that a 21MP full sized sensor will achieve the "equivalent" of about 1.75 stops of resolution compared to a 12MP (D300) 1/2 sensor size version? If so, does a 12MP D300 fall 1 stop short on resolution of a 12MP D700?

Hi Bill,
I am not quite sure what you mean by "stops of resolution."

The way I look at resolution is to consider the sample interval, or more precisely the pixel pitch.

The pixel pitch of the camera in question are:
D300: 5.5 microns
5D2: 6.4 microns
So in terms of resolving the finest detail on a subject, the D300 would do a little better with a given focal length lens.

Another way to look at resolution would be pixels/picture height. The 5D2 is 3744 pixels high, and the D300 is 2848 pixels. So given the same field of view, the 5D2 would produce a higher resolution image by a little. In practice, the actual difference would depend on the lens.

Roger

[Bill McCrystyn]

Hmmm, I'm probably not expressing it correctly. It was of interest to me at the time because I was considering going from an 8MP D20 to the 12MP D300 so as to better print (resolve) a 12x18 and give me better crop room. Surely (all things being equal) more MP's must equate to better resolution.

[Jim Fenton]

I've got to beg to differ relative to fast leneses being at their best wide open.

If you look at MTF charts, very rarely is this ever the case no matter whose glass you are shooting.

Typically, stopped down a couple is where they all hit their highest ratings.

[Jay Gould]

"the Canon EF70-200mm f/2.8 L IS lens is sharpest at its wider apertures and it is not even recommended to shoot with it at f/8 or beyond."

Kerry, where can I read about this specifically and more on this topic in general nonsuper technical terms? Thanks

[Don Lacy]

"the Canon EF70-200mm f/2.8 L IS lens is sharpest at its wider apertures and it is not even recommended to shoot with it at f/8 or beyond."

Kerry, where can I read about this specifically and more on this topic in general nonsuper technical terms? Thanks
Jay the above statement is uniformed, while it is true you lose resolution due to diffraction at the larger apertures and you should be aware of that as you determine your exposure settings the 70-200 or any other lens does not become useless above f/8. If you need f/16 or higher to get enough DOF then you should shoot at f/16.