PDA

View Full Version : How do you test the "sharpness" of a lens?



Cor
9-Jun-2010, 05:32
In many discussions or ads you read : this lens is sharper than....

I wonder how people come to this conclusion. Recently I doubted the sharpness of my 120mm/f6.8 Angulon a bit in the sense that a relative small enlargement (8x10) of a negative did not give me the sharpness I was expecting. So I decided to run a crude test to see if it was an operator error and not the 120 Angulon.

So I set up a little test scene in my back garden, with some garden furniture, mounted my 4x5 TechIII on a tripod and shot the scene on EFKE PL100 (not the sharpest film in my hands, but that does not matter, I use it for both lenses) at f16.

As a kind of bench mark I used my most modern lens a recently acquired 90mm/f8 SuperAngulon MC, also at f16. Moving in a bit closer to get the same scene on film (distance to the focusing point now about 2 meters).

Processed the negatives (shot at 50ASA) stand developed in PyrocatMC, put in my enlarger, as high as possible (for a 50*60 cm print). Made some prints (13*18 cm) from various parts of the scene (150 CompononS at f11).

Against my expectation the 120mm Angulon outperformed the 90SA, and yielded the sharpest/crispiest prints..

So much for a test. Drawbacks are: I should have shot more film, use another lens as a benchmark (the 90SA I recently obtained, but I did shoot a couple of 4X5 FP4 negatives which look under the focusing scope plenty sharp to me at 5* enlargement)


So how do you tackle such tests?

thanks,

Best,

Cor

Ash
9-Jun-2010, 06:29
best way is to use a chart/poster with all the lines & shapes. can't remember the proper name

You focus on the static object, then see how well the lens resolves detail at various apertures. because of the lines/shapes on the poster you can see when each lens bottoms out

rdenney
9-Jun-2010, 06:30
There are a couple of things to remember.

One is that large-format lenses were mostly not intended to be used at wider apertures. Even f/16 is a larger aperture than is usually recommended for Super Angulons, and especially for Angulons. Try again with an aperture no larger than f/22. Realize, though, that getting even performance across the field might also introduce some visible diffraction, and the result is a balance between the need for a smaller aperture to control aberrations and a larger aperture to minimize diffraction.

Another is that the wide-angle designs of old, both the double-anastigmat design of the Angulon and the original reversed-telephoto designs for smaller-format SLR cameas, we optimized for sharpness in the middle of the frame, based on my own experiences. They are softer in the corners, where greater magnification is required to maintain rectilinear projection.

The Super Angulon, which is really two reversed-telephoto lenses placed in opposition around the aperture, was a variation on the Biogon concept. That design was optimized for wide coverage, and even performance across the field.

I have some very old Flektogon and Curtagon reversed-telephoto lenses in smaller formats that are extremely sharp in the middle of the frame, but soft at the edges. But even my 50's-era Super Angulon 121/8 is uniformly sharp across the frame. That was really an advance in design.

The next thing to remember is that the capabilities of the lens can only be determined exactly on the plane of sharp focus. One cannot evaluate lens sharpness and also depend on depth of field. So, when evaluating lens performance with a three-dimensional scene, do your evaluating for things that are precisely in focus, not for things that merely seem sharp because the aperture is small. When comparing different focal lengths, the effects of depth of field can really foul up a test like this, especially when comparing the center to the corners. I have some wide-angle lenses that suffer from field curvature and look terrible when evaluated using test charts, but in three-dimensional scenes may perform just fine because the shape of the scene conforms to the curved field better than a flat test target.

Then, there is the issue of magnification. Lenses are optimized for a certain distance to the subject, and using it for some very different distance will reveal that fact. Stopping down helps.

Finally, make sure you are evaluating sharpness at the print size you intend to make. It's not fair to an otherwise completely serviceable lens to judge it using an actual pixels display on the computer screen, after scanning it at a zillion SPI, when the screen image corresponds to a print size five times bigger than you intend to print. Make some real prints (or at least fragments of prints) at your intended print size.

Lens tests are fun in their own right and I've done a lot of them. But the way I really judge a lens for actual photography is that I use it in making pictures. If the pictures it makes are sharp enough, and if I like the way they are rendered, I keep using it. If they aren't, and if I'm sure it's not a correctable problem with my technique, then I put it back on the shelf.

Rick "keeping things in perspective" Denney

Cor
9-Jun-2010, 06:53
Thanks for the elaborate answer, Rick !

I too was thinking that f16 was perhaps to big, I did check the point I focused, and noticed declining sharpness at the edges.

I also agree with your practical advise in the last paragraph, I can not even process 50*60 cm prints in my darkroom, only 40*50 max.

But I was curious how other people approach this,

thanks again,

Best,

Cor

Ps I did make actual smaller test prints of the max. enlargement, as evaluation tools

carverlux
9-Jun-2010, 07:51
Mr. Denney has pointed out some very important objective observations about LF lenses and the way to properly test them. I found his tests on bokeh particularly enlightening - Mr. Denney, thank you.

I would also add that production variation and sample testing varied to very significant degrees between lenses, batches, factories and brands. It is commonplace for us to expect that products perform to their stated specs, especially when the products we use daily, such as the PC, the cell phone, or the car seldom if ever veer far from working every day and do not surprise with failures to a significant degree very often.

This assumption cannot be so easily made in the times when many of these lenses were built.

First, lens designs from these times were not always done with production realities in mind, which resulted in significant variations between one lens and the next, even within the same batch. Missed tolerances are cumulative and if each of 100 components were only off by 1% of their spec, the cumulative result is a product that only meets 36.7% of this same spec. If there are 8 lens elements in 4 groups in a lens like the SA or 6 elements in the Angulon, you can count how many mechanical and optical tolerances will accumulate to result in a final product that has to ressemble the original design. The Angulon has a better chance.

Modern optical design include a discipline called "tolerancing" which modify otherwise difficult-to-produce designs into ones that are much more tolerant of slight variations. This is one of the major reasons why lenses made in the last 10 to 15 years by competent and conscientious brands are consistently better performers than their predecessors, at least in a technical, objective sense.

Second, optical production measurement techniques (or available test equipment within any particular lens factory) all the way through even the early 1990's did not always reveal all the production sample errors and required significant manual intervention to ensure consistent quality. This is a result of both inadequate measurement technology or test equipment investments as well as meeting production quotas. If the final product HAD to be good, like those used on satellites or for the military, it WILL be built right at all costs, and many times, this required re-building the whole lens over again and again until it meets spec. You can well imagine that this is simply not practical or economical for a commercial product and is a major reason why very few if any lens manufacturers actually ship an actual test chart with their lenses. In contrast, satellite, military or semiconductor lenses are required to be tested periodically at third-party specialty houses to certify their compliance before being rendered fit for service.

The moral of the story on this point is that you have to be lucky. If you are into vintage lenses, then you must be prepared to test the actual sample to see what you get and whether it satisfies you. Setting your expectations based on a brand or a datasheet can be disappointing.

Third, evaluating a lens should be done with performance objectives as close to your final application as possible. Shooting test charts can be a lot of fun but is usually not very conclusive for a photographer even if it is satisfying to the closet engineer in some of us.

The tests you conducted included many more elements than just the Angulon or SA. You did not mention whether you focused wide open then stopped down, or you focused at f/16. It included an enlarging lens. You tested at 2 different magnifcation ratios: one for the taking lens, one for the enlarging lens. You enlarged a portion of the picture to 8x10 but you did not mention if you had a test target that would be consistent from picture to picture and is representative of the subject matter you'd like to photograph. Most important is that these are YOUR conditions and it would be very helpful if you can standardize on these parameters so variations in different lenses can be traced to the lens and not a variation in one of these parameters.

Under the principle of cumulative errors, you can see that you can get lucky and get a lens where everything lined up, or you can be unlucky and conclude that a lens is off when it may be errors showing up in the other parts within the test chain that had nothing to do with the lens per se but the testing conditions and test element combinations.

So - lots can be done to see if you have a gem or a dud. Hopefully, some of these points may help you with your expectations as well as your execution.

Enjoy!
Carver

Atul Mohidekar
9-Jun-2010, 08:29
....

So I set up a little test scene in my back garden, with some garden furniture, mounted my 4x5 TechIII on a tripod and shot the scene on EFKE PL100 (not the sharpest film in my hands, but that does not matter, I use it for both lenses) at f16.

...

Cor

You could buy a test chart or download, print this free ISO 12233 test chart on a high-resolution printer and use it as the target:
http://www.graphics.cornell.edu/~westin/misc/ISO_12233-reschart.pdf


// Atul

Cor
9-Jun-2010, 08:32
Thank you too Carver for elaboration and food for thought..so there are no short cuts here, I go back to the more empirical routine..use the lens shoot the scenes I normally , and make prints as I normally do..;-)..

And after sufficient practice try to decide if I like the lens or not.

I also brought this up to see how people arrive at there statement: this lens is much sharper that that lens, a statement which should be taken with a grain of salt in most cases, as I already assumed..;-)

Best,

Cor

Btw I focused wide open, stopped down but did not re-focus. No real test target either, just the wire mesh of the same garden chair

Andrew O'Neill
9-Jun-2010, 09:53
By putting the lens on my camera, exposing a piece of film, and making a print. There is no better test.

Ron Marshall
9-Jun-2010, 10:14
These test results may interest you:

http://www.hevanet.com/cperez/testing.html

They tested at f11, f16 and f22.

Arne Croell
9-Jun-2010, 11:42
The best would be to use an MTF measurement setup like the Zeiss K8, but since this is usually not an option, evaluation of negatives is it. However, as has been pointed put by Rick and Carver, 3D objects are problematic. Plus, one has to keep in mind that you are always testing your whole system, from film holder to development procedures. So if you want to get any meaningful data about one component of this system like the lens, you need to be very diligent in keeping everything else as constant as possible. Possible error sources range from film holder problems to ground glass alignment, parallelism of the object, lens , and film planes, to development temperature and agitation.

Let me describe my present setup that I use if I feel the urge to test, but realize that it has some remaining problems*, too.

As test object I chose the Norman Koren lens test chart (http://www.normankoren.com/Tutorials/MTF5.html) which allows a judgement of performance at two contrast levels, 50% MTF, and 10% MTF. It is different from a pure bar chart, and gives slightly worse results in lp/mm for the 10% MTF than a bar chart. The text at the linked web site is worth reading to understand the concept. I have printed out 7 of these charts, and they are mounted on a flat door indoors, one in the center, and the remaining 6 in pairs for sagittal and tangential orientation in 2 corners and at 1 intermediate position. Lighting is by 2000W of video lighting. I photograph these on 4x5 TMAX 100 at a magnification of 1:20. The numbers on Koren's charts are calibrated for 1:50, but 1:20 works for the distances I can achieve, so I have to divide the numbers by 2.5. This means the maximum resolution I can detect is 80lp/mm, but that is already above what most LF lenses can do, and above the theoretical resolution at f/22. I actually have only encountered two LF lenses that reached that number (at f/11-16) in the center. Also, 1:20 is the magnification that is used for the MTF charts of many LF lenses like Rodenstock's Apo-Sironar N and W.
The film holders I use for tests are the Sinar Precision 4x5 holders employing a pressure plate and full metal construction; the camera is a Linhof Super Color, which is a super stable monorail, on a Gitzo 1325 (not extended) with a Linhof 3663 head. To ensure parallelism, I use a combination of the zig-align ZV1 setup (http://www.zig-align.com/zig-align/View_Cameras.html) and the Versalab Parallel laser alignment tool(http://www.versalab.com/server/photo/products/parallel.htm). To start I measure the approximate distance between lens and object plane for the 1:20 magnification. Then I check the size of the individual test chart images on the ground glass with a caliper and adjust the distance until I have the right value, then I center the image. Then I remove the lens, put the zig-align 4x5 main mirror in the camera back and hold the Versalab Laser module onto the door surface at the image center, pointing to the mirror. I adjust the lateral tripod position and angle and the back standard until the laser reflects straight back onto the module. Now the film and object plane are parallel. Then I mount the zig-align secondary mirror onto the lens standard and adjust the lens standard until it is parallel to the film plane. I focus on the ground glass using loupes between 6x and 15x (I have a Bosscreen which allows the use of such high magnifications) and take images from wide open to f/22 in full stop increments. I use a "rubber ball" release with a long tube to reduce vibrations to a minimum. I do not refocus at other openings, so any focus difference from stopping down will show. I develop the TMX negatives in TMAX RS 1+9 (from concentrate) to N contrast in a Jobo CPP-2 at 75F and 50rpm. After drying, the negatives are examined in a microscope at 50-200x magnification, and the lp/mm cutoff values for 50% and 10% MTF are noted for each individual chart, 14 values altogether, or 70 numbers for an f/5.6 lens (14x5 different apertures). Tedious, but occasionally you learn something about your lenses.

*e.g. focusing errors, variability of different shutter speeds, shutter vibrations at different speeds, the use of a Bosscreen, the subjective assessment of the contrast cutoff values (that's why having lots of numbers for one lens helps, improves the statistics)

Mark Sawyer
9-Jun-2010, 12:50
These test results may interest you:

http://www.hevanet.com/cperez/testing.html

They tested at f11, f16 and f22.

One of the most interesting things about that test is that there is about as much variation between two samples of the same modern lenses (Nikon 300-M) as there is between completely different lenses of similar focal length.

ic-racer
9-Jun-2010, 13:10
I have some wide-angle lenses that suffer from field curvature and look terrible when evaluated using test charts, but in three-dimensional scenes may perform just fine because the shape of the scene conforms to the curved field better than a flat test target.

A little OT, but have you experimented with lens spacing. I posted in a thread a few years back about my experiments with lens spacing and field flatness at the far edges of the image circle.

I agree with you about not knowing how to interpret the edges in a standard test. What a field photographer would want is to know about the edges is the sharpest image the lens can produce and a second number indicating how far from the flat plane the image is produced.

Nathan Potter
9-Jun-2010, 14:24
Arne gets into the nitty/gritty of careful lens testing and shows how complicated it can get. I used a similar technique some years ago in a laboratory environment for macro lenses and normal 4X5 lenses. In my case I was interested in removing some of the variables that have been mentioned above by others. I used reflective resolution targets in a vertical orientation but did not employ film. The lens was mounted on an optical bench in front of a fine ground glass at the lens focal plane. A horizontally mounted microscope was focused on the ground glass using long focal length objectives. The objectives were shimmed to be parfocal. The lens image on the ground glass was located and roughly brought into focus by racking the ground glass. Then the microscope was focused on the ground glass image of the resolution target and the ground glass then removed so that the aerial image of the resolution target could be viewed and critically refocused. The photo column of the microscope was equipped with a camera and photosensor and could be translated precisely in X and Y direction orthogonal to the optical axis. Thus the resolution bars could be scanned for period and contrast by connecting the photosensor to an X-Y recorder. Crude by todays instrumentation but effective as a microdensitometer. The whole thing really tedious to go through.

Well, just another way to remove some of the variables from lens resolution tests.

Nate Potter, Austin TX.

rdenney
9-Jun-2010, 15:04
A little OT, but have you experimented with lens spacing. I posted in a thread a few years back about my experiments with lens spacing and field flatness at the far edges of the image circle.

The lens in question is a retrofocus wide for the Pentacon Six mount. It's the 45mm Mir 26, which had a reputation for soft corners. I found that the corners were sharp, if you could get them in focus at the same time as the subject. Worked in some situations and not in others.

These lenses were an inconsistent mess. Mine wasn't that bad, but others would buy two or three and mix and match elements, and even the rotation of the elements, to get the best performance. Nasty stuff.

Rick "who loved the color rendition of that lens, however" Denney

engl
9-Jun-2010, 17:55
I only own a grand total of two lenses. I compared the sharpness of these two by removing the back of my LF camera and putting a lens-less DSLR there, effectively acting as a digital back (with rather small area). It eliminates many variables regarding film choice and developing materials/method, minimizes focus errors (live view) and there is no risk of spoiling the comparison by errors in printing/scanning the film. It is rather quick to set up, gives information about sharpness/contrast/color of the lens, and gives instant feedback.

The conclusion is that my old Optar 135/4.7 is lacking in contrast, soft at big apertures, kind of sharp at F16-F22 (although way into diffraction on the small high-density digital sensor). My Fujinon NW-125/5.6 MC surprised me by being as sharp and contrasty as lenses built for 135/APS-C format. More contrast at all apertures, at F5.6-F11 it resolved considerably more detail at the same distance than the Optar, in spite of being a wider lens.

Ideally Id compare them by shooting a ton of pictures and printing them big though :)

John Bowen
9-Jun-2010, 18:24
The late Fred Picker used to describe his "sharpness" test. He would photograph bare tree branches with no leaves and no wind. If the prints looked like tree branches and not steel wool, then the lens was sharp enough for anything he was likely to throw at it.

Cor
10-Jun-2010, 00:15
Thanks again for all the extensive feedback and point of views/approaches, I appreciate it!

Best,

Cor

Dennis
10-Jun-2010, 07:15
If you just want to know how people personally do it...
I make lens tests in my basement because I know the concrete floor is steady and then I have a place marked on the floor that I know if I aim at a particular spot on the wall at a certain height I am squared up. I have on the wall a test chart, a page of newspaper, a dart board and some tape. A few holes in the wall from darts.
I light the test spot with incandescent bulbs and my exposure is always rather long.
It isn't super scientific but it is what I always do and it gives me a relative idea of lenses compared to my other lenses.
Dennis

E. von Hoegh
10-Jun-2010, 11:15
I stick some newspaper to the wall and (with large format lenses) shoot it at f:22. Then I look at the negative under a dissecting microscope, 40x. That, believe it or not, tells me all I need to know. If I don't feel like wasting film, I use a star test. Google that. It's very informative.

cowanw
10-Jun-2010, 12:58
For a different perspective, how do I test the sharpness of a lens?
I take a portrait of a woman of a certain age and if the lens is not soft enough I try to pawn it off on an unsuspecting newbie.
Regards
Bill

Ernest Purdum
11-Jun-2010, 17:04
One popular test target is the USAF. Many years ago, there was a "mystery" paint job on the bottom of a U-2, newspapers asking what it could be. It was the USAF test pattern. Some good thinker had decided it would be easier to conduct tests from the ground up than from the U-2 down.

Ernest Purdum
11-Jun-2010, 17:09
One popular test target is the USAF. Many years ago, there was a "mystery" paint job on the bottom of a U-2, newspapers asking what it could be. It was the USAF test pattern. Some good thinker had decided it would be more convenient to conduct tests from the ground up than from the U-2 down.