Finding the aperture range which gives the sharpest images

[thereisnobeginning]

I had heard that every lens has a sweet spot in its aperture range which results in sharp pictures. I was curious about how this range is determined.

A search revealed a number of sites which described the procedure,. Essentially, one took multiple pictures, at different apertures, of the same object from the minimum distance at which the lens focuses.

Having some time to spend, I decided to see if I could find this range in the lenses that I have.

A picture of the setup that I used is appended. The camera is the Nikon D5600, with the AF-P 18-55 G VR kit lens that it came with. It is tethered to a laptop running Darktable under Linux. All operations are done from the laptop. The lens is placed at the minimum focus distance (in this case about 28cm) from the object, which is a large book supported in an open position. I took pictures at all the apertures at both 18mm and 55mm. All pictures are in RAW format. The procedure was repeated for the AF-P 70-300G ED VR lens, whose minimum focus distance is 110cm.

I found that, from the subjective viewpoint of my jaundiced eye, the sweet ranges are as follows:

For the 18-55: At 18mm : f6.3 to f11. At 55mm : f8 to f16

Can I conclude that f8 to f11 will give me the sharpest images at all focal lengths for this lens?

For the 70-300: At 70mm: f7 to f14. At 300mm: f10 to f20.

Similarly, can I conclude that f10 to f14 will give me the sharpest images at all focal lengths for this lens?

Is it necessary to repeat this at different object distances as well?

Is this in line with what is typical of these lenses? I would be grateful for comments from those with more experience in this, as well as suggestions which can improve the observations.

I realize that this setup is just too crude to give reliable results. However, it did give me some results, reliable or no.

Many thanks in advance.


The D5600 is tethered to the laptop via the USB cable, and pictures are taken remotely using Darktable.

 

[anotherMike]

There's a lot to cover, yet I don't have the time, so my simple (for now) answer is as follows:

Resolution of a lens is limited by two things in general:

1. Aberrations which reduce the resolution from the ideal
   1. These occur mostly within the first two F stops starting at wide open
      1. Meaning as you stop down past wide open, things improve
      2. But some aberrations never go away as you stop down
2. Diffraction, which is a physical limitation to resolution that can't be gotten around.
   1. It depends on wavelength of light, but it's the same - not a brand thing, a laws of physics.
   2. In a very general sense when we get around F/11 the odds are that diffraction is going to start contributing to resolution reduction more than lens performance will be. By F/16, most definitely diffraction will be impacting things. Critical users of high resolution bodies (d8xx series) will often note the start of diffraction impacting thing just before F/8 even.

Lens performance can also vary by focus distance. As such, your test of resolution at minimum focus distance is not only incredibly unrealistic in terms of real world shooting, but a tough one. ALL lenses that we buy as consumers have trade-offs inherent in the design. There is no 100% perfect lens. Doesn't matter what brand, everything has trade-offs. The better the lens, the less the trade-off. One of the trade-offs, and sometimes even an *intentional* design point, is to have a lens be sharper across frame at distance (where this is needed, say, for landscape), while in closer ranges for portrait, not as much, because the need for across-frame-resolution isn't as great and in many cases, the need for "max" resolution in close ranges isn't likely to be attained. Some lenses will be incredible at distance, very good in portrait range, but only "average-good" at minimum focus distance, because the designer accepted that as a trade off because the odds are the lens won't be primarily used at minimum focus distance.

Hope that helps.

-m

 

[bjn70]

You could try this at distances you expect to use the most. The problem will be finding a suitable target.

but yes a crude rule of thumb is that f5.6 to f11 will generally be sharpest.

 

[just Tony]

I had heard that every lens has a sweet spot in its aperture range which results in sharp pictures. I was curious about how this range is determined.

A search revealed a number of sites which described the procedure,. Essentially, one took multiple pictures, at different apertures, of the same object from the minimum distance at which the lens focuses.

The only time you would tend to expect a test report to include lens performance results at the minimum focusing distance will be for macro lenses. And by that I mean specialized lenses for macro work, not d-everything zoom lenses that have a somewhat "macro" near limit.

A common distance for lab lens testing is something on the order of 50X the focal length of the lens. That is a compromise made in part by practical considerations such as the available working space for the testing. And yes that would imply that zoom lenses are tested at a different distance for each focal length.

As to the optimal aperture there is a generic rule of thumb that will be fairly reliable for you: go 2 stops down from the widest opening. It won't always be the exact peak performance but it won't be falling very far short compared to the peak either. The curves of performance vs aperture tend to be somewhat plateau shaped with only minor drop off near the top.

Having some time to spend, I decided to see if I could find this range in the lenses that I have.

For the 70-300: At 70mm: f7 to f14. At 300mm: f10 to f20.

That result for 300mm was surprisingly biased towards tiny apertures, where diffraction should surely be making a bid for dominance. Then it made sense when I saw how close your camera was to the test target. That lens is far from the distance that the 300mm position was optimized for, so there were some extreme problems that the extreme stopping down was addressing.

Try your 300mm test again at a greater distance such as 20-50 feet and I would expect the optimal range to be more like f/8-f/11.

Be sure to do all you can to kill vibration including the Exposure Delay function if it exists in that camera body.

Take some time to consider what distances will dominate during your use of the lenses. In some cases you might want to get to know them at portrait distances, or at extreme distances you may use with scenic shots. If possible test at those distances. Of course you won't be making test charts 1000 ft wide, so available fine detail in the scene may have to suffice.

The D5600 is tethered to the laptop via the USB cable, and pictures are taken remotely using Darktable.

--
Render unto Digital, that which is Digital's,
and unto Analog, that which is Analog's

 

[jeffnles1]

F8 and be there.

Most of the lenses I have and the ones I've rented are sharpest from 1 stop to about 2 stops from wide open. Some of the exotic primes (600mm F4, etc) are just as sharp wide open as stopped down.

 

[Selex]

All good advice above.

Don't forget review sites that test lenses in a much more scientific way, like DxOmark and Lenstip and even some older lenses on here on DPR. They will report the sweet spot for the lens resolution. (Although DPR gave up on tested lenses, so you won't find the absolute latest).

Absent reviews the rule of 2 stops down from the widest aperture works well.

If you want to get fancy then here is a modification.

For expensive high quality lenses and good primes, it is probably at around 2 stops almost on the dot. Take a look at how Sigma 1.4 primes behave, they are often sharpest between 2 and 3.2.

For "consumer" average quality lenses the sweet spot is probably at 3 stops down due to higher degree of uncorrected aberrations.

I am not familiar with your lenses, so can't comment on those details.

Lastly, and this is coming from others, not myself, but resolution/sharpness is often degraded by other factor besides the lens and sometimes much more so. So if you need sharper images the lens is just one parameter to pay attention to. An amazing lens shot at a slow shutter speed with give soft images due to camera shake. The lens might be outresolving your sensor, so upgrading to a higher pixel density body is another way to go up in resolution.

 

[mbecke]

This is a very good and important thread. I am not a professional photographer, or anything close. However, before I purchase a new lens, i usually read the reviews, review the resolution charts, and attempt to determine at what aperture settings the lens will likely produce its sharpest images (or provide the best bokeh). So when I actually purchase the lens, I already have some idea as to how I am going to use this new "tool" -- and it is a new tool in every sense of the word. I am not going to go into further detail here as there are much more competent and experienced photographers who can elaborate much, much better than I. But I look forward to reading the comments that are posted relative to this important issue. Thank you, OP, for your original post.

 

[Mole Delta]

Many years ago a picture editor told me that "there are no bad lenses at f8" . At the time I only had krap kit and tried to stick to f8 and I really benefited. Now I own a lot of high end stuff that is wonderful wide open but still find myself gravitating towards f8, using wider apertures only if the light conditions call for it or if I am looking for a particular effect.

 

[labalaba]

Just look this up at a site that has tested it eg:http://www.opticallimits.com/

Your results with the 70-300 are suspicious, diffraction should be degrading resolution at these apertures. Most likely your lens was not focused optimally and the results you obtained reflect greater depth of field at small apertures.

 

[thereisnobeginning]

Many thanks to everyone for their considered and thoughtful answers.

I never thought for a moment that the procedure that I used was in any way proper enough to give credible results. What was surprising is that it gave results in line with what was to be expected.

On second thoughts, perhaps not that surprising?

Interestingly, the (relatively) new Nikon AF-P DX lenses do not seem to have the publicly available data that the older established lenses have.

For example, the site opticallimits.com has tested and published data on the AF-S DX 18-55 lens, but not the AF-P version (these two are optically different). As for the 70-300 DX, it has no data, even the AF-S version, which is surprising given its popularity.

On DxoMark I could not locate results for these lenses either, though tests on the the AF-S versions are published.

@labalaba commented that "Your results with the 70-300 are suspicious, diffraction should be degrading resolution at these apertures. Most likely your lens was not focused optimally and the results you obtained reflect greater depth of field at small apertures."

That could be the case, but I was using the camera's auto-focus, and the image did look focussed under live view when zoomed in. Also, the aperture range for the 70-300 is f4.5-f22 at 70mm, and f6.3-f32 at 300mm. So the results seem to reflect the accepted rule of two stops away from maximum aperture.

@EliotKramer commented that "Lastly, and this is coming from others, not myself, but resolution/sharpness is often degraded by other factor besides the lens and sometimes much more so. So if you need sharper images the lens is just one parameter to pay attention to. An amazing lens shot at a slow shutter speed with give soft images due to camera shake. The lens might be outresolving your sensor, so upgrading to a higher pixel density body is another way to go up in resolution."

That could also be the case, though unlikely, since the camera was mounted on a tripod, and triggered remotely without touching it. In addition, there was a one second delay before the picture was taken. Upgrading to a higher pixel density sensor is not within my economic radius at the moment. Oh, well.

@just Tony suggested using a larger distance, 50 times the focal length. This is certanly doable indoors with the 18-55. For the 70-300, it would mean going outdoors, and the risk of introducing more uncontrollable parameters. Lets see, it is something to put on the to-do agenda, time and space permitting. Many thanks for your comments.

@anotherMike, it is obvious that your remarks have years of experience behind them. I will definitely repeat the tests at greater distances.

Thanks again, and regards.

 

[anotherMike]

This is what I didn't have time to write the first time I responded. This might be too much, or too detailed; in that case, my apologies. Perhaps you'll get some use out of it, or at least it will be food for thought.

Sharpness:

We have to talk about this one a bit deeper. Lens "sharpness" is not and can't be measured or defined by a single number, score, or grade. Yes, some sites attempt to do just that, because the consumer in todays ATM/fast food/instant gratification mindset wants to simplify. With lenses, it doesn't work that way. So that means you have to toss out a lot of the so called "lens test sites" that grade in single scores or give one resolution number. Let me rephrase that - you don't toss those out, but don't over weight those numbers in the grand scheme of things, particularly if you shoot in longer distance ranges, which don't correlate too well with the test chart tests done in closer ranges.

So sharpness is generally broken down into resolution - how fine a set of detail can we see, at a particular level of contrast (difference between black and white). The finer the detail, the less the contrast. The gold standard for measuring this contrast/resolution component is optical bench MTF, which, sadly, none of the flat test chart sites use. In this, if we were presented all of the information measured, we would see a graph, of contrast vertically, and resolution in lp/mm (line pairs per millimeter) horizontally for various apertures of the lens. The graph would slope downwards: at the lower resolution values, we'd have a lot of contrast, and at some point, there wouldn't be enough contrast to be able to discern any difference between two very closely spaced objects (high resolution). When you see a manufacturers MTF graph for a lens, you'll see a different sort of graph: contrast vertically, but position in the frame (center to edge to corner) horizontally, and four traces, two for one resolution, two for another. What we have to talk about is sometimes phrased "coarse structures", "medium structures" and "fine structures", or, in lp/mm terms, increasing in resolution. A 10lp/mm is a coarse structure, a 20 might be getting into a medium, and 40 would a fine structure. Many of us might argue the exact numbers to define each, but it's a general idea.

In terms of "Image Quality", which is really what we seek, studies in the 1950's (I believe I have the decade correct) by some folks, notably Otto Schade from RCA, postulated that image quality was related to the *area* under the MTF graph in the first way I described it (contrast vertically, resolution horizontally). This tells us something important: It means that any *one* point along the graph is *not* the way to tell the image quality of a lens. Yet when you read a opticalimits test, or any imatest based system, they all report at one frequency, one spot on the curve. Yes, some folks might argue that "MTF50" is a suitable substitute - I (and others) would disagree. So this is why I believe we have to be careful and not get into a habit of over-relying on "lens test sites" that report single numbers from test charts. Some sites report "grades", but those are often averaged grades whose weighting of the grade we don't know, so they don't always correlate to what we actually get. Lens testing is far more complex than single numbers and grades. Wish it weren't so, but it is what it is.

Structures - let's get back into that: coarse structures represent what we kind of think of as general contrast. Medium structures represent generally the detail and sharpness. Fine structures are actually tricky to always achieve - we'll see why in the next section. Human behavior tends to over-weight the very fine structure performance of a lens, yet, as we'll see in a bit, it's difficult to achieve, and based upon those early studies, it's not the sole determination point of image quality.

Things that reduce sharpness

There are many, many things that conspire against us to remove potential resolution fro the system. Reality is, you, I, and most everyone reading this are contributing to reducing resolution. Consider this: You'll never achieve the finest resolution the lens is capable of unless you are PERFECTLY focused (and I mean perfect - and autofocus isn't perfect), and both camera and subject are PERFECTLY still, and you're shooting at a low ISO, because any noise will also reduce sharpness. So, pondering that, think about what you shoot, how others shoot, and you'll see that unless you are a studio photographer precisely focusing and using strobe to stop action, a landscape photographer on an expensive tripod and ballhead, or a macro (close up) photographer, all practicing great craft, you'll never be able to realize "the sharpest" that the lens might be capable of delivering. Note that this ties into structures as well: A music industry photographer shooting at high ISO, wide open, will never "see" (attain) fine structure detail, but a lens that has excellent coarse and medium structure performance wide open will absolutely be of benefit. But the lens that does this might not have tremendous "lens test site" numbers. So we have to tie what we shoot and perhaps, roughly stated, what area of lens performance matters to us, together.

Getting Real

So, we now have to think about how critical we each are. Some folks have very high standards, and likely print large, and maybe sell in galleries. Some folks don't print at all, or print small, and don't care about getting "the best", and importantly, a while bunch of people with views that lie between these two poles. So, we have to get real about what are standards truly are (and be willing to accept another person might have higher, or lower standards and that this is ok), AND think about whether situations we shoot allow us to realize the best the lens can offer.

Add those two things into the mix, and then one can start to decide whether to obsess over sharpness and getting the best fine structure detail out of lenses. We all want reasonably sharp lenses, but I sometimes sense that many folks who don't need to be obsessed over sharpness gravitate too much towards having to own "the sharpest lens, as tested by lens test site YYY", which doesn't make sense. Of course, this is a lens forum, and a number of us here ARE striving for technical excellence in lens performance. But one must get real what their own goals are. So many things in our daily shooting take away from sharpness, so chasing sharpness when one truly doesn't need it is not the wisest move.

-m

 

[labalaba]

That could be the case, but I was using the camera's auto-focus, and the image did look focussed under live view when zoomed in. Also, the aperture range for the 70-300 is f4.5-f22 at 70mm, and f6.3-f32 at 300mm. So the results seem to reflect the accepted rule of two stops away from maximum aperture.

Does this 'rule' have any optical basis? My understanding was that it usually reflects design/marketing compromises. Lenses can be made that are sharper closer to maximum aperture but this requires either 1) better correction of aberrations, or 2) artificially limiting the maximum aperture. #1 especially is expensive and the decision to or not to design and manufacture to such specifications is a decision about what the market is.

If you use better lenses then stopping down for sharpness ceases to be such an issue and aperture choice is based on other considerations

 

[anotherMike]

Getting a bit technical here, but...

The predominant aberrations that reduce resolution wide open are ones that go away reasonably quickly as you stop down (the tricky part is NOT all aberrations go away as we stop down, disproving the myth that all lenses are equal at F/8 when they are not), and that two stops down is about the point where they start to diminish. Sometimes lenses are designed, on purpose, to be a bit under corrected in spherical aberration, in order to help the look of the bokeh and OOF transitions; and given that this aberration is one that goes way as we stop down, such correction choice can give the lens a specific "look" wide open that isn't tack sharp but has a nice look for people, and then when stopped down, is sharper for more critical tasks. Lenses we buy are always trade-offs, if you want "perfection", or as close as we can get to that impossible goal, you're probably looking at 40 grand panavision primo 70 cinema lenses. Not something that the masses can afford.

But in general, you're right - if you are willing to pay enough, you can get lenses that are quite good wide open, particularly in the easier-to-design mid telephotos. The wide angles are much tougher; even the very best wide angle (which I deem as the 25/1.4 Zeiss Milvus) needs to be stopped down a couple stops to get the edges and corners in order.

-m

 

[labalaba]

@labalaba commented that "Your results with the 70-300 are suspicious, diffraction should be degrading resolution at these apertures. Most likely your lens was not focused optimally and the results you obtained reflect greater depth of field at small apertures."

That could be the case, but I was using the camera's auto-focus, and the image did look focussed under live view when zoomed in. Also, the aperture range for the 70-300 is f4.5-f22 at 70mm, and f6.3-f32 at 300mm. So the results seem to reflect the accepted rule of two stops away from maximum aperture.

There is really no way that resolution should be maintained on that sensor beyond f11. If you saw no deterioration due to diffraction going beyond that aperture, resolution may have been limited by something else (such as focus).

 

[calson]

It is also dependent on the camera. Images from higher resolution cameras exhibit more diffraction at apertures smaller than f/8.

 

[labalaba]

It is also dependent on the camera. Images from higher resolution cameras exhibit more diffraction at apertures smaller than f/8.

The amount of diffraction is the same, but the higher resolution sensor will detect it more readily

 

[thereisnobeginning]

So the results seem to reflect the accepted rule of two stops away from maximum aperture.

Does this 'rule' have any optical basis? My understanding was that it usually reflects design/marketing compromises.

You are right, there is no such 'rule'. Unfortunately my choice of words did not express what I really meant, which was 'accepted rule of thumb'.

Regards.

 

[thereisnobeginning]

I was using the camera's auto-focus, and the image did look focussed under live view when zoomed in.

There is really no way that resolution should be maintained on that sensor beyond f11. If you saw no deterioration due to diffraction going beyond that aperture, resolution may have been limited by something else (such as focus).

As mentioned above, I was using the camera auto-focus all the time. The evaluation of the images was visual, so of course it is subjective.

It would be nice to get a second opinion, though. The RAW files are available.

Regards.

 

[thereisnobeginning]

It is also dependent on the camera. Images from higher resolution cameras exhibit more diffraction at apertures smaller than f/8.

The amount of diffraction is the same, but the higher resolution sensor will detect it more readily

That is correct. The diffraction would be governed by the wavelength and aperture width, irrespective of what the sensor resolution is.

A sensor with a higher resolution would show the diffraction pattern more clearly, relative to a lower resolution sensor.

Regards.

 

[thereisnobeginning]

@anotherMike, it is a delight to read your thoughtful, informative and in some ways philosophical post.

Please consider writing a book.

Regards.