36

I've tried googling this, but have never found a satisfactory answer.

I've heard the term "sweet spot" thrown around by some photographers to mean the f-stop of a lens which results in the highest sharpness that lens can achieve.

A few questions on this:

  1. General photographic knowledge states that the higher the f-stop (the smaller the aperture), the largest depth of field you will achieve. This seems to "suggest" the higher the f-stop, the sharper your image will be (all other factor being equal of course). Does the idea of a "sweet spot" trump this rule? (so, theoretically a f11 could be sharper than f22)

  2. Is the "sweet spot" an optical algorithm that can be applied to any lens, or is it to do with the particularities of the manufacturing of some lens?

  3. Finally, how can I determine the "sweet spot" of my prime lenses?

Note: I know other things come into consideration in sharpness, like ISO, light, glass (lens) etc etc, but please ignore these and assume these are equal for every different lens.

For context, I'm mainly trying to achieve ultimate sharpness in architectural (indoor and outdoor) photography, and urbanscape, where usually I tend to shy away from small depth of field.

mattdm
  • 143,140
  • 52
  • 417
  • 741
andy
  • 1,919
  • 1
  • 18
  • 23

4 Answers4

26

The sweet spot of a lens is probably just as dependent upon the type of image capturing surface used as the lens itself. Both film and digital sensors have a limit of detail they can resolve (although large-format film has the tendency to capture FAR more detail than 35mm or digital sensors at much tighter apertures, around f/22.) Assuming you have a lens with the best resolution imaginable...it will ultimately be limited by the imaging material. This is due to the "diffraction limit" of the film or sensor.

The mechanics behind finding the "sweet spot" of a lens can be fairly complex, as it is very mathematical. To simplify this for consumers, the MTF (modulation transfer function) chart was born as a way to provide clear, mathematically derived information about the sharpness, or resolution, of a lens, film, or sensor. If you are interested in the underlying theory, this article is a good read: Understanding image sharpness.

In simpler terms, assuming you want the maximum clarity for the sensor size and density you are using, for most DSLR image sensors the "sweet spot" of most lenses of decent to high quality is between f/8 and f/11. Entry-level DSLR's, which tend to have smaller sensors with smaller photosites of greater density, are diffraction limited at around f/8 or f/9. Higher-end DSLR's, which tend to have larger sensors with larger photosites and lower density, are diffraction limited around f/11.

Outside of having a really crappy lens that does not have the greatest intrinsic resolution, most lenses can resolve a high degree of fine detail. Most lenses on the market these days have their own MTF chart that can be helpful in knowing the lenses "sweet spot" in and of itself. Most digital cameras have information about when the sensor becomes diffraction limited. Review sites such as DPReview.com, the-digital-picture.com, etc. will also state the apertures at which the sensor becomes diffraction limited for most cameras. I don't do much film myself, so I can't offer you much regarding when various types of film may become diffraction limited.

It should be noted that the diffraction limiting aperture (DLA) is only when diffraction starts affecting quality, but not when it has reached its maximum effect (which is usually several stops beyond the DLA.) Visible image softening from diffraction will usually not be apparent until a couple stops beyond the initial DLA. For sensors of a given size (i.e. APS-C), a higher-density sensor will start revealing diffraction earlier, however the lower-density sensor will be incapable of resolving detail as high as the sensor of greater density. For any given megapixel size (i.e. 18mp), a sensor with larger physical size will usually provide better results. Diffraction affects image quality due to light dispersing beyond a single photosite and affecting others. As larger sensors (i.e. Full-Frame vs. APS-C) have larger photosites, they become diffraction limited at tighter apertures than smaller sensors.

The real trick is finding the overlap between the point of peak sharpness for the lens, and the point at which an image sensor is able to resolve clear detail without visibly softening it due to diffraction. An aperture setting in the overlap area will be the true "sweet spot" of the camera and lens you are using. On the flip side, if depth of field is more important than ultimate sharpness, then a higher aperture may provide a sweet spot more appropriate to your work.

jrista
  • 70,728
  • 15
  • 163
  • 313
  • 3
    dude, you're a frigging genius, thanks! Gimme some time to digest all this info and I'll be back. thanks for a great answer. – andy Aug 30 '10 at 01:30
  • +1 for the scientist. There's a few rule of thumbs kicking around for dSLRs, but I think the changing face of sensors is making it hard for them to stick. – Joanne C Aug 30 '10 at 02:06
  • @jrista - "large-format film has the tendency to resolve FAR more detail than 35mm or digital sensors" - when comparing, say, Velvia 50 in 35mm and 6x9 formats, the resolved line pairs per millimeter is still the same, just there are many more line pairs on the the 6x9 frame. When looking at prints of the same size, 6x9 has more detail, but the underlying resolution is the same. Same goes for digital, if both cropped sensor and medium format sensor have the same pixel density, their theoretical maximum resolution is the same (although more variables influence it). – Karel Aug 30 '10 at 06:39
  • I changed it from "resolve" to "capture", as the key point was their diffraction limit is considerably higher, around f/22. Resolution wasn't really the key point. Which brings up an interesting point about digital sensors...sensors of different sizes with the same pixel size and density will have the same diffraction limit. If we took, say, the new D60, an 18mp APS-C sensor, it gets diffraction limited at the astonishingly low aperture of f/6.8. A full-frame sensor with the same pixel size/density would be limited at the same aperture, which beggs the question...why such high density? ;-) – jrista Aug 30 '10 at 06:53
  • Sorry for the nitpicking. Well, there is a camera company who understands the need for better pixels rather than just more pixels ;-) – Karel Aug 30 '10 at 07:28
  • I'm all for larger, but more closely spaced, photosites myself. If they can increase both density and size, with a moderate pixel count between 20 and 30mp, I think that would be ideal. Modern sensors still have pretty large gaps between photosites, which is somewhat offset by microlenses, but not entirely. – jrista Aug 30 '10 at 15:08
  • @jrista, all else being equal (granted, not always the case), a higher density sensor always capture as much as or more detail than a low density. For example, take two sensors, one limited at f/6.8 and one at f/13. The f/6.8 limited sensor will capture more detail at all apertures below f/6.8 because it is in the sensor limited region. Above f6.8, it is still going to capture more detail because it is the airy disk limited region. Only after f/13 are the sensors equal, and the f/6.8 sensor is never behind. – eruditass Aug 30 '10 at 19:08
  • The actual story is obviously much more complicated. Of course, with more space, there is more room for electronics to lower the black read noise (this has more of an effect on the noise-limited region than say photosite density), increase the quality of micro-lenses, etc. However, only looking at the diffraction limit is misleading, especially when comparing older sensors to newer sensors. – eruditass Aug 30 '10 at 19:15
  • 2
    @jrista, diffraction is an optical phenomenon that is linked to the lens and not the sensor. The physical size of the airy disk that causes the softening is independent of the medium that it is being projected on to. A higher resolution sensor will capture more softening on a per pixel basis, but the overall image will not be softer, as the absolute physical size of the airy disk being captured will be the same on both sensors. – eruditass Aug 30 '10 at 20:30
  • Here's a good article: http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm

    Basically, all this talk about diffraction was meant for choosing an aperture, not for comparing sensors (except perhaps to say smaller sensors suck)

     – eruditass Aug 30 '10 at 20:55
  • As long as you understand my point, which the Cambridge article states in its conclusion and in part 2: diffraction does not make higher density sensors inherently worse and the a higher resolution sensor (say 18mp vs 10mp) will always produce equal or better resolution than the other at the same print sizes and crops, all else equal. This is not to endorse the megapixel race: it is stupid and other sensor enhancements are more important. And no jrista, I don't understand why you would not buy an 18mp over say a 12mp, all else equal (including price and other sensor perf:DR/Noise/color/etc). – eruditass Aug 31 '10 at 01:04
  • I think long conversations like this are best left to StackExchange chat. Makes these threads rather unruly. I would like to continue the discussion, with visual examples if you have any. I think you make an interesting point, but I'm not certain I'm entirely convinced that a new super high density 18mp sensor will always be better than an older lower density 16mp sensor, for example. I'll be logged in here for the next several days: http://chat.meta.stackoverflow.com/rooms/154/photo – jrista Aug 31 '10 at 04:10
  • Sorry, seems the old photo.se room was closed. This one should work: http://chat.meta.stackoverflow.com/rooms/211/photo-se – jrista Aug 31 '10 at 04:34
  • i think this discussion is excellent, and perhaps deserves a question of it's own! which I will craft... if I can – andy Sep 01 '10 at 02:44
  • @jrista nit: raises the question is good; begs the question is bad. (Sorry. Pet peeve.) – luser droog Apr 27 '13 at 06:11
  • @luser: Perhaps. I think the wiki article you linked is inappropriately assuming the use of the phrase "begs the question" means "assuming the initial point" (petitio principii), when in fact in general usage, "begs the question" quite literally means "raises the question". The sole difference is that the former is simply more slangish, and if you wish, "sloppy". Slang or not, I do not believe it is in the average speakers nature to go around saying "assumes the initial point", which makes the notion that "beggs the question" is a direct translation of "petitio principii" a stretch at best. – jrista Apr 27 '13 at 16:17
  • It should also be noted that if I change the phrasing to "raises the question", the circular nature of the statement remains unchanged. Only the phrasing changes, but the impression that I am making one argument to further my original argument remains...so why bother nitpicking the phrasing itself? Better to nitpick the circular argument in general, no? ;P – jrista Apr 27 '13 at 16:21
  • BTW, it should be noted that I've had plenty of time to test Eruditass' statements that higher density never means worse IQ (all else being equal), and I completely agree. – jrista Apr 27 '13 at 16:22
13

With primes I always put a page of text up on the wall, put my camera on a tripod with a remote trigger (the self timer works too) and take a couple of photos at each major f-stop: 2.8, 4, 5.6, 8, 11, 16, 20 and then I compare them for sharpness at the center, edges, and corners. You'll see that there's a range that's sharpest and I use a label maker and print "8-11" to put on the lens itself so I know for each lens.

With zooms it's harder because the sweet spot will change with the focal length, so for a 70-200mm lens you'd want to do it incrementally like 75mm, 100, 125, 150, 200 maybe.

Just keep in mind that even if the text isn't perfectly sharp at any focal length/aperture, we don't generally photograph text and differences in sharpness can be seen with text that you'd never see in a landscape, for instance.

Erica Marshall
  • 1,921
  • 16
  • 13
12

I think "sweet spot" is a rather poorly-defined term in general use, and in fact you'll see some people talk about the sweet-spot of a lens with regard to the sharpest aperture setting, and others talking about the sweet spot of a lens' image circle (e.g. using a full-frame 35mm lens on a cropped-sensor DSLR).

You cannot generalize and say "50mm primes have a sweet spot at f/8". Different lens designs perform differently and will make different trade-offs. So not all lenses of a given type will have the same sweet spot.

With regards to sharpness and aperture, modulation transfer function (MTF) charts will give you a good picture (albeit theoretical), if they're published for the aperture settings you're interested in. But MTF charts can be difficult to find for some lenses, and usually will show just one or two aperture settings.

The empirical way of determining the sweet spot for a lens that you personally own is to take test shots at different apertures, preferably of a flat-field scene with both fine details and high-contrast edges. Then compare the images and draw your conclusions. It may not be clear-cut, depending on what your criteria are. For instance the aperture where the corners sharpen up may be different than the aperture where the center of the images is sharpest. Obviously shooting technique is important for this, so using a tripod with mirror-lockup and a cable release is ideal to remove camera shake as a factor.

While the f/8-f/11 range is generally considered a safe choice, I wouldn't say it's universally true. Higher-quality lenses will already be starting to see the effects of diffraction by f/8, especially on high-resolution camera sensors. For instance, many late-model pro-level lenses will hit their sharpness sweet spot around f/4-f/5.

Philip Ngai
  • 219
  • 1
  • 6
Jeff Kohn
  • 676
  • 3
  • 5
  • "Higher-quality lenses will already be starting to see the affects of diffraction by f/8, especially on high-resolution camera sensors." - diffraction limit is not the property of lens. – Karel Aug 30 '10 at 14:40
  • @Karel, the lower the quality the lens, the higher f-number the "diffraction limit" because the lens-quality will be limiting resolution, not diffraction. Lens-quality goes up with f-number, which is the opposite of diffraction.

    Picture a graph of f-number (x axis) vs resolution capture (y axis). There are 3 curves: diffraction limit, lens-quality, and sensor resolution. Diffraction slopes down, lens-quality slopes up, and sensor resolution is a flat line. The lowest value of the 3 is your resolution at that capture. I think I'll draw this when I get home as it may be helpful.

     – eruditass Aug 30 '10 at 21:13
  • Well, this turned out looking much worse than expected. These are in no way real values, but give an idea of the relationships.
    http://imgur.com/9xtyR.png

    If we follow the high quality lens, it is sensor limited by the medium and low megapixel sensors at 1 and 2. In the high density sensor, it is limited by its own lens quality until f6.3, where diffraction creeps in, but has a higher resolution than with any of the other sensors until f8 with the medium-density sensor and f13 with the low density sensor.

    On the medium quality lens, it starts off lens-limited on all the sensors. By f4

     – eruditass Aug 30 '10 at 23:41
  • That high quality sensor is diffraction limited at f6.3, while the low quality sensor is not diffraction limited until f/13. This is more and more visibile with a higher resolution sensors. – eruditass Aug 30 '10 at 23:43
  • Now to expand on diffraction limiting with regard to megapixels, the medium density sensor does not get diffraction limited until f8. Even though the high-megapixel sensor is diffraction limited around f6.3, it will still capture more detail than the medium-megapixel sensor until f8. This is only with the same-size sensors.

    For those reading the graph: the lens lines all slope up (in reality they are curves generally going up) and the sensor lines are all flat.

     – eruditass Aug 31 '10 at 01:05
  • Karel, the point at which a lens becomes diffraction limited is absolutely a property of the lens, it has to do the resolving power of the lens. For instance the Zeiss 100mm f/2 Makro-Planar is sharpest at f/4, and only goes downhill from there (admittedly very gradually, and even today's 24mp cameras won't show the falloff until you go past f/5.6)

    You may not be able to see the true diffraction limit if the lens out-resolves the camera/sensor you're using, but that doesn't mean the limit isn't there. There's a reason high-end lens testing is done with an optical bench rather than a camera.

     – Jeff Kohn Aug 31 '10 at 06:37
  • Guys, maybe we should have another question and answer about this:)? – Karel Aug 31 '10 at 14:33
  • Regarding diffraction limit being property of lens - please take a look at this calculator: http://www.cambridgeincolour.com/tutorials/diffraction-photography-2.htm. The only inputs are sensor size and amount of pixels, no lens type, focal length etc. Diffraction limit is the same for every lens, given that we've fixed the sensor size and pixel count. Thus diffraction limit is not the property of lens. – Karel Aug 31 '10 at 14:37
  • @Karel, did you read my responses and look at the graph? Yes, diffraction limit is the same for every lens, it is a physical property, but lens of lower quality can mask the diffraction limit. The Cambridge article acknowledges it, but for ease of interpretation and simplification, assumes best-case theoretical scenarios. In your link, notice this phrase: The above values are only theoretical best case scenarios; actual results will also depend on lens characteristics – eruditass Aug 31 '10 at 18:24
  • OK, so now we agree that diffraction limit is not the property of lens, which was my initial statement. I understand perfectly well that the actual perceivable details in the image depend on many other factors. – Karel Aug 31 '10 at 18:45
  • Diffraction is caused by the lens. If we define the diffraction limit as the region where diffraction limits the resolution, a low quality lens will change this limit. If we simplify the model and make the diffraction limit only involve the sensor, then of course the lens won't have an effect. – eruditass Aug 31 '10 at 21:47
  • Karel those calculators only factor sensor resolution because they make the assumption that the lens out-resolves the sensor. If you want to argue that the diffraction limit is the same for the Zeiss 100mm f/2 as it would be for for one of those 28-300mm lenses @ 100mm, we'll just have to agree to disagree. – Jeff Kohn Sep 01 '10 at 07:19
  • I think we should differentiate between lens softness and diffraction limit. The resolution of the system depends on many other factors and I see no reason to take one factor into account and leave out all the others. – Karel Sep 02 '10 at 21:22
4

Regading photography, there are two points that limit the resolution of an image: one is the Depth of field (look at Wikipedia yourself, I'm not allowed to post two links), the other the physical resolution of the lens (Rayleigh criterion of maximum resolution).

A large depth of field is generally obtained with a small aperture (f/11 has a smaller depth of field than f/22), while a large aperture leads to a smaller diffraction limited spot size for these areas of the image that are in focus.

For an ideal picture there are two contradicting goals: large aperture (small f-number) for the points in focus, small aperture (large f-number) for for a large depth of field. Depending on the lens, the film/ccd detector used and what you intend to picture, different settings are optimal.

user1260
  • 71
  • 1