Lens Resolution: Are My Measurement Results Bogus?
I have read claims on the internet that printed test charts are nearly worthless for use in measuring lens resolution. I have also read that a sharp razor blade or utility knife blade can be used to get really accurate resolution measurements. Which claim is true? Both? Neither?
I use the MTFMapper program to analyze lens resolution. NASA has used the MTFMapper program to analyze lenses that they sent to Mars onboard their Rover Perseverance. I don’t think that this program is providing bad results. The software can be obtained from here.
You should be very skeptical of internet sites that don’t tell you how they arrived at their resolution numbers.
A sample resolution test chart
MTFMapper provides printable files, which I used to make my test chart targets. This same software provides a way to use things such as back-lit razor blades or utility knives for resolution targets. The target edges being measured are all on a slant; the measurement mathematics doesn’t like edges that are vertical, 45 degrees, or horizontal.
I believe that MTFMapper uses LibRaw to decode raw files, which uses zero sharpening (which would increase resolution measurements). For raw formats that LibRaw doesn’t support, I use the Adobe DNGConverter to make DNG raw files; these files also have zero sharpening applied. The downside to the DNG converter program is that it strips out some exif data, such as focus distance.
For sites that use other camera photo file formats, particularly jpeg, resolution measurement results are worthless. Just about all of these formats add some level of sharpening. Depending upon the amount of sharpening, you can make the resolution measurements as high as you wish.
I use a printed test chart that measures 40 inches by 56 inches. The chart is printed at 1200 dpi on heavy-weight paper with a fairly glossy finish. The chart is dry-mounted and placed into a frame to keep it perfectly flat, and I temporarily mount a mirror to the center of it using magnets to align my camera and get it perfectly parallel to the camera sensor. The chart is clamped into position to eliminate any movement, and the camera is on a heavy tripod. I use either a wired shutter release or a self-timer, using either live view or a mirrorless camera to eliminate vibrations.
Chart lighting needs to be even, and it’s best to keep illumination levels above at least EV 10. Surprisingly, the ISO value has little effect upon resolution measurements; it’s still best to keep the ISO low.
A sample setup to use a quality blade edge for a resolution target
I conducted some tests to compare resolution measurements using my chart and a pristine utility knife blade. I can’t prove resolution results in an absolute sense, but I can at least compare results from two entirely different test methods.
When I first started doing resolution testing, I tried using small (11” X 17”) printed charts, both with inkjet and laser. I also tried matte/glossy/satin surfaces and single/double weight papers. I determined that laser prints weren’t quite as good as inkjet, and that satin-like surfaces worked best. Small charts are poor for testing lenses at realistic shooting distances (you want to fill the frame with the chart if possible). I’m forced to use laser prints for infrared testing; my inkjet ink is invisible in infrared!
I always struggled to accurately align the test chart to the camera until I began attaching a mirror to the chart surface (using powerful magnets). When you can see your own reflection looking through the viewfinder and the lens center reflection is in the middle of the viewfinder, you’re perfectly aligned. Rotation is easy; just align the chart edge to the viewfinder edge. The MTFMapper program will change the color of the resolution measurements to yellow for edges that end up at a poor angle.
Camera (Nikon Z8) mounted on an accurate sliding linear rail
Accurate focus is an absolute requirement to get the best resolution measurements. It’s possible, using contrast-detect focus or a mirrorless camera with autofocus, to get reasonably good focus. I can get a bit better focus using low-sensitivity focus-peaking and image magnification. It’s necessary to take several shots, with focus set both in front and behind the target before re-focusing. You need to pick the sharpest results from the many test shots, and make sure the focus is done at the shooting aperture
The best way to get optimal focus is to use a linear rail and move the camera in very small (1 mm or so) increments starting in front of the subject correct focus and taking shots until you’re behind the subject focus plane. Again, pick the sharpest result (highest measured resolution).
Select where to take the measurement
The photo above shows how to pick where to take a resolution measurement along the knife blade. Choose a location and orientation to match a similar edge in the resolution test chart. Selecting a different section of the blade will probably give slightly different measurements, because the MTFMapper program is supremely sensitive to edges.
The illumination behind the blade only needs to be even in the selected region of interest that is being measured. You want to do this in a dark place, to maximize the silhouette contrast.
Note that every position and orientation in the camera’s field of view will likely give a different resolution reading. Life and physics isn’t as simple as what is portrayed at most photography websites. A single resolution number is nearly meaningless (as is a single center/edge/corner number). Also keep in mind that different camera sensor resolutions will give different answers as well, because the resolution measurement is actually a combination of the lens and the camera sensor.
Blade placed near to chart target edge in viewfinder
As shown above, I placed the blade edge in a similar location to a chart measurement that I was interested in comparing. I tried to select the section of the blade edge in MTFMapper that would roughly match the length of the chart target edge (about half of the blade edge length).
Blade placement relative to test chart placement
As you can see above, I have drawn in roughly where I placed the blade in the camera viewfinder that I measured in red. The little cyan numbers are the MTF50 resolution measurements on every target edge in the chart, as calculated by MTFMapper.
A potential upside to using a blade edge is that you can focus on it wherever you place it in the frame. For lenses with field curvature, this will probably get you a higher resolution measurement than simply using a chart that you probably just focused in the frame center.
A downside to using the blade is that you get a single measurement from your photograph, versus over 700 measurements by using a chart like that shown above.
I decided to use my Nikkor 24-120mm f/4 S lens on a Nikon Z9 camera to compare resolution test results between my printed chart and the blade.
The lens was zoomed to 34.5mm (the lens barrel marking was 35mm). The MTFMapper program was configured to provide resolution measurements in units of MTF50 lines pairs per millimeter. I like to use these measurement units, since you can get the same answer using any size of camera sensor.
MTFMapper measurement of blade edge: 34.5mm, f/4
Pretty comparable measurements between the blade edge and the chart!
Although none of what I have written can absolutely prove that I’m getting correct lens resolution measurements from my printed test chart, I think it shows that the measurements are at least in pretty close agreement between these two very different test methods.
A site that I use to compare by own lens resolution results against the same lens models is Lenstip, found here.
They also measure resolution in units of MTF50 lp/mm, and our results are typically very comparable (when using similar camera sensors). Lens sample variation is a real thing, so you should never expect to see the exact same results between any two lenses.