1951 USAF resolution target
The 1951 USAF resolution test chart, or target, contains subsets of three lines of gradually decreasing thickness. The lines are subdivided into groups and sets. Each set consists of a subset of three vertical lines and a subset of three horizontal lines. Sets within a group are arranged in a column (with the exception of the first set in each odd-numbered group, like group 0). Groups are identified by a number at the top of their column. Sets are identified by a number at the right or left of the set.
The line density in lppm (line pairs per mm) in a specific group and element of the target is given by
where R is the line density, G the group number and E the element number. The optical system being tested is said to have this resolution if the three lines in the element are perceived as distinct from each other. The test image itself can be magnified as necessary to distinctly see the lines, albeit without post-processing. A measure of subjective interpretation is of course implicit in visual judgments. Automated testing with this target is also possible.
The target contains gradually smaller groups as one approaches the center of the target. At least in a target of good quality, this allows a single target to be used for testing a broad range of resolutions and magnifications. Cheaply made targets, or targets printed on an office printer, may however be limited in their range of resolution.
The absolute sizes of lines in each group and set are standardized, which means that group 1 set 3, for example, has the same absolute size in all properly made 1951 USAF targets. Groups coarser than the standard group 0 (zero) are identified by negative numbers. Therefore, as long as the group numbers are visible in a test image, they provide enough information to compute the magnification of the image at the size being viewed. It is still necessary to know the sensor size to compute the on-sensor magnification of the test image. In practice, this means that you may end up with meaningless group and element numbers if you print out a 1951 USAF target from a downloaded document.
One of the first specifications to look for when choosing a 1951 USAF target is the largest and smallest groups present on the pattern, e.g. from -2 to +7. Especially in high resolution targets, very fine elements can be present, but only a given group and element are guaranteed to still be within specifications. For example, a target may include all elements up to group 7 element 6, but the manufacturer may only guarantee that elements up to group 7 element 3 are within specifications.
Typically, these test patterns are made by
The thickness of the chrome must be sufficient to block all light, but not excessive because the latter would invite undercutting during the etching process. Etching is also critical in duration, temperature, agitation of the solution and freshness of the chemicals. Any scratches or "holes" in the photoresist will of course result in defects of the pattern. Adhesion of the chrome to the glass is important to make the target durable. Sputtering is superior to electroplating to guarantee a chrome layer of good quality.
A target can be chrome on transparent background ("positive" pattern), or vice versa ("negative" pattern). Targets printed on photographic paper are also available. A modified 1951 USAF target type type with multiple chrome-on-glass targets of different optical densities is available for automated measurements of contrast and dynamic range. Multiple 1951 USAF targets are often arranged in a wheel pattern, with a target in the center surrounded by a circle of eight more targets.
The sizes of elements of the 1951 USAF target are set by the MIL-STD-150A standard (subsequent notes and change notices were also released). For more information, see the Wikipedia article on the 1951 USAF resolution test chart.
Using a chrome-on-glass target
According to general descriptions, resolution targets of this type are meant to be illuminated either from the front (especially positive ones) or backlit (especially negative ones). In my experience, front near-axial illumination can easily cause light to reflect off the chrome and into the lens (see figure below, top row). This can result in bizarre side effects related to coherent illumination, especially if the light source subtends a small angle. Light reflecting from the chrome surface may strongly reduce the utilized aperture of the lens and cause odd effects, like a visual sideways shifting or pinching of the pattern lines when the light source is moved. Another effect is an observed inability to focus with precision on detail of the pattern, even though the lens is known to provide the needed resolution for accurate focusing. Just to be sure, using diffused transmitted illumination (and eliminating incident illumination from ambient light) is the safest way to avoid these problems.
The chrome surface must face toward the lens, because with many lenses used in photomacrography the glass substrate may introduce optical aberrations if the pattern is imaged through the glass.
No-name 1951 USAF target from China
The resolution target shown above is currently available from many eBay sellers based in China. There is no brand visible anywhere, but the target offered by these sellers seems to always be the same, so it is probably produced by a single factory. This is the cheapest chrome-on-glass 1951 USAF target I am aware of. However, it is not especially cheap, considering that for a similar price one can often find an equivalent second-hand 1951 USAF target from known brand names like Thorlabs.
The target is packaged in a PTFE plastic box. The glass substrate rests on small plastic feet on the inside of the box, which prevent the pattern from touching the plastic. The box does not rest completely flat against a surface when opened, and the target therefore must be extracted from the box for use.
There is a wide chrome plated edge around the glass substrate for handling the target. This lessens the risk of accidentally touching the pattern with one's fingers, which should be carefully avoided. It is a good idea to wear clean cotton gloves when handling this target.
Cleaning the pattern surface involves the risk of lifting the smaller chrome details off the substrate, and should be avoided at all times.
This target combines one positive and one negative pattern on the same substrate, each starting from group 0 and ending with group 7. In addition, there are also a positive and a negative area, each containing nine patterns that contain groups 4 to 7. The instruction sheet says that these extra patterns can be used in case the corresponding central area of the groups 0-6 pattern is damaged.
Typically, high-quality targets of this type are only positive or only negative. One reason is probably that the large amount of transparent substrate in the positive pattern invites flare and lower contrast in the adjacent negative pattern. Another may have to do with yield and quality control: the more targets one crams on a single substrate, the more likely it becomes that a defect will make it necessary to discard the whole substrate. It becomes likewise more likely that damage to one target during use will force the whole substrate to be discarded and replaced. While combining positive and negative patterns on the same substrate might seem a good way to reduce the cost of the target, this practice involves multiple types of negative side effects.
As always, the proof is in the pudding. I used for this test a good lens (Nikon Printing Nikkor 105 mm A), known to provide excellent resolution and contrast at 1x. The lens was mounted via adapters (devoid of optics) to an Olympus E-M1 Mark II Micro 4/3 camera with 20 Mpixel sensor. Groups 0 to 4 look sharp, with straight edges. There are some irregularities, including occasional "holes" in dark (chrome) areas as well as chrome spots on areas that are supposed to be transparent. Both are indicative of poor manufacturing methods and loose quality checks, but one could live with these defects for infrequent testing and qualitative (visual) evaluations of image resolution. It can already be seen, however, that something is wrong with the numbering of the sets in group 5, as well as the lines and numbering in group 6.
To decide whether this is caused by the lens or target, I switched to a higher magnification with the Leitz Photar 25 mm f/2 (stopped down to f/2.8). The numbering in group 5 (left side of picture) is visibly "faded" because of undercutting during the etching process. The corners of the lines in this group are visibly rounded. They are not usable for precision measurements, but the lines may still be useful for visual evaluation. There is no really readable numbering in group 6. The group 6 lines (center) are too "faded" and irregular to be of any use. They are clearly undercut by etching and/or partly abraded away by whatever brushing or swabbing was used to remove the photoresist. A faint ghost of part of group 7 is visible (at the immediate right of the black square).
At least in my specimen of the target, the additional replicas of groups 4-7 are all as bad as the one in the center of the larger groups 0-7 pattern. None of them has usable group 6-7 lines.
As a comparison, the above figure shows in the middle row a detail (a calibration scale) of a photomask made by Compugraphics International for Intel in 2005. The finest lines in this target (along the left margin of the pattern) are approximately equal to group 7 set 1 in a 1951 USAF target. This detail is close to the top edge of the frame, and there is a little lateral chromatic aberration visible at the top of the lines.
The bottom row shows a detail from a microscope calibration ruler, with ticks spaced 100 μm. The line thickness roughly corresponds to group 6 set 6 in a 1951 USAF target (albeit in the calibration ruler the spacing between lines is far higher). Note that the ends of the lines in the microscope ruler are square, and the line thickness uniform in spite of the twice higher magnification, not "frayed" like in the Chinese 1951 USAF target. It is therefore entirely possible to obtain far better results than the Chinese target with much cheaper non-standard targets (see also below).
This Chinese 1951 USAF resolution target is at the limit of its usability for testing reasonably good lenses at 1x on current (20 Mpixel) Micro 4/3 cameras (but insufficient for any serious pixel-peeping). Performance may be marginally better on full frame sensors with relatively large (5 μm) pixels. Testing at higher magnifications than 1x with this target on Micro 4/3 is out of the question.
It is instead feasible to use this target in the close-up range, for example at 0.1x (1:10) on Micro 4/3 and 0.2x (1:5) on full frame, and lower magnifications. Testing at up to 0.5x (1:2) might be possible with full frame sensors.
Other implementations of the 1951 USAF target
So far, I have not personally tested 1951 USAF targets of better quality (this might change in view of the results of the current test). The specifications of a Thorlabs equivalent target (dead link) with groups -2 to +7 include a line spacing tolerance of ±1µm and a line width tolerance of ±0.5µm, while the Chinese target has a tolerance of tens of µm. Thorlabs also makes a smaller 1951 USAF target with groups 2 to 7, for testing resolutions on the subject side up to 228 lp/mm. Edmund Optics additionally has high-resolution 1951 USAF targets with groups -2 to +9, for testing resolutions up to 645 lp/mm (the finest set is group 9 element 3). So-called extreme resolution 1951 USAF targets are available, and include e.g. groups 4 to 11. One specific implementation (dead link) guarantees up to group 11 element 3, and includes group 11 elements 4-6 without guarantees.
Standard targets for testing high resolutions are very expensive, with the best targets as expensive as top-of-the-line lenses and delicate enough to almost require clean-room handling. The need for cheaper alternatives for non-professional lens testing in macrophotography and photomacrography is therefore obvious. Both biological and technological subjects offer patterns suitable for testing (e.g., this post on photomacrography.net discusses the use of silicon wafers discarded during IC production, often available as curiosities at reasonable prices). Current LSI wafers are too large for easy handling and contain detail far too small to be useful to microscopists, but legacy wafers are smaller and more useful. Generations of microscopists have used diatoms as resolution targets with transmission microscopes.
While there is no scarcity of suitable subjects, their main problem is that they do not allow easily repeatable results for comparing among tests made with different targets (which is indeed the reason why standard targets were developed). Virtually all species of microscopic organisms can grow to different sizes under different environmental conditions. Their surface patterns are likewise variable under environmental influences.
IC wafers have been made in a myriad of types, and obtaining a significant number of duplicates may be problematic, especially as separate purchases done at different times. Individual packaged ICs can be purchased in large numbers but are embedded in resin and difficult to extract. Unpackaged IC dies are ordinarily sold only in large batches and, being tested and working devices, can be relatively expensive. ICs are also periodically redesigned to take advantage of improved and cheaper manufacturing methods, so there are no guarantees of continued availability.
In my search for suitable resolution targets for my personal use, I purchased a few legacy photomasks used in the manufacture of ICs, which are less common than discarded silicon wafers. These photomasks are optically reduced for projection onto wafers and stepped multiple times across a wafer, and therefore contain details larger than those observed on wafers. Photomasks are also similar to resolution targets in that they consist of chrome-on-glass patterns. Some are "clear field" masks (equivalent to positive targets), other "dark field" (equivalent to negative targets). Photomasks designed for temporary use during the development of new ICs (and then discarded) may contain additional test and alignment marks, as well as multiple types of devices, and are especially interesting for the present purpose.
All photomasks I purchased turned out to be useful as resolution targets in one way or another, and at a range of magnifications.
Additional items that I found useful include small alignment targets, similar in construction to photomasks but containing only a combination of reticles and concentric circles. They appear to be designed for optical equipment alignment. A single lot I purchased contained a couple dozen targets, both positive and negative and with different grid/circle spacings, on 20 by 20 mm glass substrates. They are among the most useful of my repurposed resolution targets.
Microscope calibration slides are also useful as medium-resolution targets and available at reasonable cost, although the range of available patterns (mostly, graduated rulers and concentric circles) is somewhat limited. CD and DVD disks are potentially useful for higher magnification, but their patterns must be exposed by opening up the two halves that constitute a disk. This is necessary because most of the lenses used in photomacrography are not optically corrected for shooting through glass or transparent plastic.
The main problem in using alignment targets and calibration rulers as resolution targets is that most do not provide alternating light and dark lines of equal thickness. While it is still feasible to visually compare a few images of a specific pattern detail taken with different lenses and/or in different conditions, and to choose the "best" one, these repurposed targets do not allow a quantitative evaluation of resolution.
This 1951 USAF resolution target made in China is usable, as long as you need to use only the groups 0 to 4. Group 5 is of inferior quality, but may still be used for qualitative assessments. Group 6 is useless. Only a faint ghost of part of group 7 is visible.
This target is at its usability limits when testing any reasonably good macro lens at 1x. If you need a better 1951 USAF pattern, you must be prepared to spend more, or to use a non-standard target.
If you need repeatable result, e.g. to directly compare your test results with tests published on the Internet, then you should use a standard resolution target, and the 1951 USAF is one of the most often used targets. If you can do with a non-standard target, on the other hand, the surplus market offers plenty of cheaper alternatives.