Measuring LF lens sharpness using USAF 1951 target methodology

[Tin Can]

I am now ready to learn a little more. Long ago I downloaded a USAF 1951 target, printed it about 8x10 and often use it informally to determine focus and DOF with my haphazard method.

Below are links to the download and other links describing aspects of LF lens usage.

What is a DIY method and 'good enough' method of comparing LF lenses using the free USAF 1051 download?

I am certain I have the wrong scale for my usage, but I am also not buying the very expensive Edmond Sci products. Especially as I don't know how to use them.

Even WIKI does a poor job of explanation of practical usage.

Thanks in advance for any input, suggestions for study, links...




https://en.wikipedia.org/wiki/1951_USAF_resolution_test_chart


http://www.edmundoptics.com/test-targets/resolution-test-targets/1951-usaf-glass-slide-resolution-targets/1790/


http://www.largeformatphotography.info/lenses-primer/

[Bob Salomon]

Won't tell you anything unless you are testing process lenses.

[Argentum]

its easier to do a comparative test as opposed to an actual resolution test which requires accuracy in the scale of the target, the distance from the target and very importantly the contrast range from the black to white lines.
Published film manufacturers data quotes 1 and sometimes 2 resolution figures for their film and since you will be capturing your image on film ( I presume ) you are really testing how well any lens and film combination is able to resolve detail. Well the film manufacturers 1st resolution figure is usually at a contrast ratio in the subject of 1:1000 which is 10 stop range. Without knowing what your actual contrast range is between black and white lines and/or being able to set it to 1:1000 means that your measured resolution relates to what? For it to be meaningful you really need to be able use a 1:1000 figure so that it relates to film mnaufacturers quoted figures.
The film manufacturers 2nd quoted figure is at a subject contrast ratio of 1:1.6 and at this contrast ratio obtained resolution may have dropped from 200 lp/mm to 60 lp/mm taking fuji acros as an example. And this figure is a more realistic real world achievable value for local/micro contrast in your real world subjects.
So you see that obtained resolution is not only a function of your lens but also of the lighting contrast ratio and particularly of micro contrast in the subject reproduction.
This makes it a tough task to actually measure resolution in a meaningful way and it is therefore much easier to do compartive tests between lenses and film but in the knowledge that your measured figures may not relate to film and lens manufacturers figures due to your testing regime.

If you just want to know what your lens is able to do then using a newspaper taped to a wall is as almost as good as any other method and doesn't require any special resolution targets. And even better is to do some real world photography using Zeiss suggested method.

see following two Zeiss Camera Lens News Pdfs:

in CLN02 see section titled:
Resolution of Camera Lenses Where are the limits – and why?

in CLN03 see section titled:
How To Determine the Resolution Actually Reached in a Photograph – A Practial Example

This last one is where its really at since what you get on film is what counts.

151501
151502

[Bob Salomon]

Bear in mind, that unless you are going to be photographing flat art then these tests have little to no useful value. If you are photographing real world things then there is depth to your subjects and flat test charts don't test that. That is one reason why lens manufacturers use MTF, distortion, color tests for their lenses. Then there are the variables that using flat art tests introduce: film type, exposure, vibrations, lighting, air conditions, developing, drying, loupe quality, how rested your eyes are, aperture used, magnification ratio, etc.. And one last point. I went to the USAF photo school and spent all the time I was in the USAF as a USAF photographer and jot once did we ever test lenses with a USAF lens chart, or any other type of lens chart, we did use targets to affirm correct focus whenever we used the Robertson process camera, but that was with an automatic focus target system and any gg check was to confirm that the electronic focus setting was correct. And I was in the USAF in the early 60's, well after the chart was produced.

[Argentum]

Agreed and also resolution throughout any real world subject is variable because lighting ratios and focus vary throughout the subject.

But doing these tests is all part of the learning process just so long as you understand that the best most people are able to achieve are comparative tests and NOT actual resolution values although the photographing a fence test does tell you that you are achieving at least a certain resolution value if not the limit.

[Dan Fromm]

Randy, before you waste any time looking at your target, ask yourself what you want to learn from testing. Until you know that there's no point doing the work.

There are differences between acceptance testing, shooting tests to compare lenses' behavior, and formal resolution testing using, e.g., the USAF 1951 target.

Acceptance testing is simple. Will the lens do what I need or not? When I was testing high performance macro lenses I used a 100 ticks/mm Olympus stage micrometer for a target. Simple question, at the magnifications I want to use the lens can it resolve feature 10 microns apart? Yes or no? This sort of testing has no implications for how well the lens will separate closer features or will perform at magnifications not used.

Comparative testing. Any subject with a range of distances between features will do here. Some lenses will separate closer features than others will. If done at distances at which the lenses will probably be used, a good way to separate usable from unusable.

Formal resolution testing with the USAF 1951 target. The USAF 1951s I have are laid out with a 2:3 aspect ratio. Working at a different aspect ratio will require at least two targets carefully attached together. The ones I have are made to be used at 1:10, i.e., with film-to-target distance approximately 11f. Subject to not committing many sins (poor focus, film plane not parallel to target, poor vibration control, ...) a negative examined at high magnification (I use a stereo microscope that goes to 40x) will show how well the lens separates nearby features in a subject, how well it does across the field, how resolution varies with aperture and whether the lens distorts badly (vertical bars tilted in the neg, bad news). The results are useful for learning what stopping down does across the field -- the typical test, so-called, shot posted here has no in-focus detail towards the edges, can't learn much from such shots -- and, in general, whether the lens is much good.

MTF measured across the field and at a range of apertures gives a better sense of what the lens tested can and can't do. Few people have MTF testing machines. I don't. People who advocate MTF or nothing strike me as <negative words deleted>. Yes indeed, if you can do it and interpret it MTF is more informative than raw resolution. So what? Resolution measured with the USAF 1951 chart isn't completely useless.

Bob listed a set of conditions that have to be controlled to get results that can be compared with each other. When I was doing a lot of testing I controlled most of them. Its not rocket science and it isn't particularly difficult.

[Bob Salomon]

One of the biggest problems with photographing charts is that there are so many variables that must be controlled that any results that most people achieve are not reproducible at another time or by someone else.

[Jac@stafford.net]

Dan From: Formal resolution testing with the USAF 1951 target. The USAF 1951s I have are laid out with a 2:3 aspect ratio. Working at a different aspect ratio will require at least two targets carefully attached together.

Indeed. I cannot imaging how photographing a small test chart with the bellows racked out to macro metrics can be helpful for the photography for which the lens is intended.

An aside - the only resolution challenge I have had was trivial. I wanted to copy many full-page photo features I did for newspapers to print on 8x10 paper. Pinning each page on a wall; lit with two 250W floods, using slow film (I think it was Kodak 5240 or some similar very inexpensive film) and a 35mm Nikon F with 105mm/F2.5 lens did a marvelous job. I was stunned, in the good way.
.

[Dan Fromm]

Jac, 1:10 isn't as close to infinity as I'd like but is surely isn't macro.

Bob, yours is the counsel of defeat. Don't try, it can't be done. Or it can't be done well enough. Having tried and having got reproducible results I'm not as sure as you that not trying is best.

[Bob Salomon]

Jac, 1:10 isn't as close to infinity as I'd like but is surely isn't macro.

Bob, yours is the counsel of defeat. Don't try, it can't be done. Or it can't be done well enough. Having tried and having got reproducible results I'm not as sure as you that not trying is best.
Glad you got reproducible results, but the typical tester will not.

[Ken Lee]

You might find it interesting to see how Arne Croell performed his lens tests. See http://www.arnecroell.com/lenstests.pdf

[Nodda Duma]

Hi Randy,

Good luck to you. I think it's great that you're digging into this. If you haven't already, suggest you visit Norman Koren's website at Imatest.com. Note that you'll actually be measuring the Contrast Transfer Function (CTF) with a bar pattern resolution chart such as the USAF 1951.

I don't have a problem philosophically with using a bar target for resolution measurements, as long as one is aware of the limitations. We use USAF 1951 all the time at work for direct view optics (when the imager is the Mk I eyeball).
However digital imaging systems are measured using a slant-edge target and algorithm to determine MTF. It's faster and less subjective.

Critical resolution measurements for objective lenses are measured on an MTF station. By critical measurement I mean when I need to compare lens build performance against the model predictions as part of a design effort or production. We acquired a new MTF station two years ago from Optikos for about the same cost as 3 really nice Ford F-150s. Of course our need justifies the expense.

Anyways, judging from your posts on the site I don't have any doubts that you understand (or will soon understand) all the caveats. Best advice: don't assume that they don't apply. Being anal is key. With that I'll just focus on the meat & potatoes:

For your use I'd avoid the bar targets. Instead, use the slant-edge algorithm method to measure your resolution performance. Imatest, QuickMTF, and ImageJ all provide this capability. If you have MATLAB, you can use the original algorithms written by Peter Burns of Kodak. All of those software packages are based on his algorithms which he released publicly in the late 90s. Google sfrmat3 for the latest. I've used all of those over the years for all sorts of measurements and focusing/alignments of optics.

Once you're making measurements, I suggest you tweak the different aspects of your test setup to get a feel for how they affect your results (lighting, alignment, object distance, etc). Experience is the best teacher here.

Good luck, feel free to ask for advice.

-Jason

[Sal Santamaura]

Won't tell you anything unless you are testing process lenses.I use my Edmunds chart, purchased long ago and mounted on a large sheet of board, as but one element in a "standard scene" to compare general purpose lenses I use for distant landscape subjects. The chart's aspect ratio is irrelevant; it sits 50 feet away in my back yard amid trees, a wood fence, bushes and concrete walkways. The house next door is also a mere six feet further away than the chart.

What the chart provides is a completely stable detailed pattern that can be examined in negatives of it using a magnifier. I never calculate lp/mm, but rather rely on it to objectively show how different lenses of the same or similar focal length compare to each other when used under identical conditions. Conditions, I might add, that are close to those encountered when photographing landscapes in the field.

The ultimate criterion for any lens is "how does it perform when used the way I (any individual) uses lenses." I find the Edmunds chart to be useful in making those judgements. Randy ought consider whether his own use pattern would benefit from the stable resolution reference an Edmunds chart provides.

[Jim Jones]

I generated this more compact and wider range resolution chart than the USAF 1951 chart in GIF format instead of the lossy formats that are often posted online. There are anomalies in the finest lines due to the small file size, but the chart is still valid for comparison tests. Feel free to copy, use, and republish this chart, but only in GIF format and without resizing so the original usefulness will not be compromised.

Oops, apparently the forum converts images to jpeg, so this chart is nearly unusable. I'm too computer illiterate to work around this. The original chart is 4178x5626 Pixels, but only 160kb.

[Argentum]

here's a PDF version of USAF Chart.

151521

If you apply the principles given in the second zeiss lens article above then you should be able to measure actual resolution when chart is placed at a distance from camera.

Note, an inkjet printer isn't capapble of printing the central very fine details of the chart. That doesn't matter, you can just use the outer lines whcih when placed at a suitable diatance from camera will work just fine. You can do the maths to work out the distance you need to get the expected resolution on film from the lines in the chart.

This one may be useful too.
151522