by Roger N. Clark
All images, text and data on this site are copyrighted.
They may not be used except by written permission from Roger N. Clark.
All rights reserved.
How good are lenses? Are prime lenses really better than zooms? Do teleconverters degrade sharpness? On this page I'll show experiments that illustrate relative lens sharpness that can answer these questions. Because there are many lenses, only a few are shown to give examples.
Below is the test chart. It was printed at 300 ppi and imaged with several lenses but normalizing the distance to the target so that it appears the same size in the frame. This allows "relative image sharpness" to be evaluated. For my tests, the distance to the target in feet = focal length in mm /10. For example, for a 500 mm focal length, the distance is set to 50 feet; for 100 mm focal length, it is set to 10 feet.
Figure 1. Roger's resolution test chart. You have my permission to download the above test chart and test your personal lenses (non-commercial applications only). Set the ppi to 300 and print on high quality photo paper, then the lines at 1 pixel (labeled 1 px on the figure) are separated by 1/150 inch.
All tests used a Canon 10D camera at ISO 100, highest quality jpeg image mode, on a sturdy gitzo carbon-fiber tripod and bogen head. Because I am interested in using the long lenses for wildlife, I thought I would use my normal imaging strategy for the 500 mm lens on a tripod, image stabilization on, and frame using my finger on the shutter (no mirror lockup or cable release). Shutter speeds were generally above 1/1000 second so camera shake was not an issue. For all other lenses, the tests were done with image stabilization off, mirror lockup, and a cable release. The 500 mm f/4 L IS lens performs well with image stabilization on when on the tripod, and compensates for vibrations at those extreme magnifications. I get consistently better results with the 500 mm lens with the IS on.
The Canon sensor has 7.4 microns per pixel in case you want to compute results yourself. The images are shown as they came out of the camera with default factory settings except the images have had the intensity leveled to the same brightness (this reduces apparent differences to show resolution changes only).
The first series uses a 500 mm prime lens and teleconverters. The results are in Figure 2, below. The image with the 2x teleconverter is nearly identical to the one without a teleconverter, showing essentially no degradation in image sharpness. The image on the right for 1.4x teleconverter is close but slightly degraded but by only about 5 percent I have learned that this is due to slight camera shake. At f/4 with the 1.4 teleconverter, the combination is as sharp as the prime lens with no teleconverter. In general, the test shows that if you want to zoom in on a subject, this teleconverter combination works very well. Note the 2x teleconverter magnifies by about 2.1x, not 2x.
The resolution, when given in arc-seconds is arc-seconds per line pair and is
derived from the equation:
arctan(resolved_pixel#/(150*distance in inches))*3600
assuming the arctan is in degrees. The line pair labeled 1 px is separated by 1/150 inch, so that is the source of the 150 in the equation. Example: 500mm f/5.6 resolves the 4 px at 50 feet, so arctan (4/(150*50*12)*3600 = 9.16. Since the 4 px is resolved pretty well, but not the 3, I rounded down a little to 8, and added the approximate sign, ~.
What modulation transfer does the resolution indicate? The 500 mm tests that resolve the 4px line pair is at about 50% MTF. Modulation transfer is the response (DN light - DN dark) / (ideal light - ideal dark) where DN is the data number in the image. One can see that as the lines get closer together, the black line becomes gray and the light in between also becomes more gray. For example, the 500 mm f/4 at f/4 with the 1.4x TC (left side image in Figure 2) has an MTF near 100% for the 8 px bars, degrading to 85% at 5px, 49% at 4px, 11% at 3 px and 0 at 2 px.
NOTE: I have completed other lenses, and will continue adding results to this page as time permits.
First published August, 2003.
Last updated January 28, 2005