Fig 1. Olympus M.Zuiko Digital ED 60mm f/2.8 macro lens mounted on an Olympus OM-D E-M5 Mark II camera
This four-third macro lens was mounted on an Olympus OM-D E-M5 Mark II camera. The camera normally produces 16 MP raw files, and has a 'high resolution mode' in which it can synthesize 64 MP raw files. Results for each mode are presented below.
As per the LoCAte Focus Analyser procedure, the test target was a carbon-soot-coated straight-edge razor blade inclined (with respect to the optical axis) by 6.5° and tilted (with respect to the camera frame) by ~5°, elevated about 5mm above an out-of-focus white paper background. The target was illuminated obliquely by a single LED lamp. At 1:1, the field of view was measured as 17.1 mm. The camera's white balance was set to Auto. The raw ('ORF') files were processed with 'dcraw' to produce pgm files used by LoCAte.
This macro lens has a maximum reproduction ratio of 1:1. At 1:1, it has an effective f-number (working f-number) of f/2.8 * (1 + 1) = f/5.6.
The numeric aperture (NA) ≅ 1 / 2feffective = 1 / (2 * 5.6) = 0.089, which corresponds to an Raleigh Criterion resolution in green light of 0.61 * 0.55 / 0.089 = 3.8 μm.
At the closest focus distance (1:1), there was about 8.6 cm of working distance between the end of the lens and the object in focus. The diameter of the outer lens glass is approximately 3.5 cm. The lens length does not vary with focus distance (ie., the lens does not change length with focus or zoom).
Below are LoCAte analyses for selected aperture settings, ranging from f/2.8 to f/22 at the camera's normal resolution (16 MP). Recall that the target is an incline edge and that the curves show the degree of edge spread along the target (the less edge spread, the sharper the focus). In all cases, the lens is focused on a point near the middle of the target; the only change is the f/stop. The depth of field increases (indicated by a broadening of the curves) as the aperture decreases. At f/22, the entire length of the inclined target is nearly equally focused, but with less sharpness than at lower f/stops.
The peaked feature that emerges at f/16 is due to a white speck of dust on the target. The waviness of the smaller aperture curves is due to textural variations (unnoticed at the time) in the white background becoming significant as the depth of field increases.
The small grey tick marks on horizontal line within each chart indicate plus/minus 100 μm of focal distance. The values for the minima are listed in the text below each chart. For example, at f/2.8, the red minimum spread occurs 139 μm closer to the camera than the blue minimum. See edge spread chart for more information about how to interpret the charts.
In high resolution mode, the camera uses its image stabilizer system to shift its 16 MP sensor a half-pixel-width in eight directions to collect information used to synthesize a higher resolution 64 MP image. The resulting raw file is 9260 x 6912 pixels (twice the dimensions, four times as many pixels). With a field of view of 17.1 mm, that's 1.85 μm/px.
The imaging in high resolution mode is especially sensitive to vibrations. The camera was operated in its anti-shock mode with an 8-second delay.
Below are LoCAte analyses for selected aperture settings, ranging from f/2.8 to f/8 at the camera's high resolution (64 MP).
(The camera does not permit f/stops greater than f/8 in high resolution mode.)
The resolution of the lens + camera is nearly constant, at 12.5 μm, from f/2.8 to f4.5, while depth of field steadily increases.
The green and blue colour channels are well-aligned, but the red channel comes to focus further away from the camera, by about 150 μm. The width of the 'bottoms' of the curves indicate that the edge had a similar degree of spread (focus) over a distance of about 125 μm, ie., a depth of field of about 125 μm. However, assessed visually, the point of best focus and least colour fringing (chromatic aberration) is where the three colour curves nearly intersect.
Figure 9 plots edge spread (pixels) at the point of best focus against f/stop, in normal and high resolution modes, for the green channel.
Fig 9. Edge spread at best focus vs f/stop, for normal and high resolution modes of the camera (green channel).
Both normal and high resolution mode saw minimum edge spreads that were greater than 2 pixels, suggesting that the lens was not able to deliver more detail to the sensor than the sensor could record, but close.
The image (and thus the edge spread) landing on the sensor is surely the same in normal and high resolution modes, yet the two modes have different apparent resolutions (causing the gap between the two curves in figure 9). The high resolution image has twice the pixel dimensions (suggesting twice the resolution), but the LoCAte results above suggest that the actual improvement in resolution is 1.45x, not 2x. Image dimensions of 6750 x 3400 (a 34MP image) would have yielded the expected resolution (6750 = 4640 * 1.45; 4640 is the pixel width of the normal resolution image).
Here's an interesting analysis of the Olympus E-M5 high resolution mode by 'Jack of AlmaPhoto'.
This camera and lens can automatically take a series of photos at incrementing focus distances. The photographer can set the number of photos and the size of the focus increment (in multiples of a fixed focus increment unit). Focus bracketing starts at a given focus and then focuses outward, away from the camera. Because of the focus increment, each successive image has a slightly larger FOV. Photoshop can compensate for that during alignment of the images, prior to focus stacking.
Figure 11 shows a LoCAte analysis of a focus-stacked composite of the seven images of an inclined razor blade. CombineZP was used to stack the images. It selects the best-focused parts of each image to create a composite view of the inclined razor edge, which was then analysed by LoCAte, producing an analysis of edge spread (focus) along the composite image of the edge. From this we can find the approximate size of the lens focus-bracketing increment. The FOV is approximate, so the z-values are also approximate, but nevertheless: (1792 - 382) / 5 = 286 μm for a 5-step increment, so a single focus increment is approximately 286 / 5 = 57 μm.
Fig 10. LoCAte analysis of a composite image composed of a series of focus-bracketed images of an inclined razor blade edge. The focus increment is 5 units; each unit is thus about 57 μm. f/3.5
The lens shows good consistency across its field of view, with the edges being only slightly less focused than the center.
Below is a LoCAte analysis of a camera-produced jpg image that accompanied one of the raw images from figure 10. The camera was set to its sharpening level 0 (normal).
Fig 13. LoCAte analysis of a JPG image produced by the camera, accompanying one of the raw images of figure 10. f/2.8, sharpening level 0 (normal). Sharpening overshoots are visible in the edge profile (right chart).
Sharpening artifacts (overshoots) are clearly visible in the edge profile chart (right side of figure 13). Evidence of sharpening includes overshoots near the edge and odd shape in the red channel edge spread curve (left side of figure 13).
Sharpening has reduced the edge spread in the vicinity of best-focus and increased the depth of field. The red channel is now better aligned with the blue and green channels in the sharpened region (beyond which the red channel is unaligned, as in the raw file). I didn't test whether the camera's sharpening algorithm works as well for content other than high-contrast straight-edges, eg., for high spatial frequency content of an image of random curves (eg., as described here, FFT analysis).
For performing LoCAte (and MTF) analysis, this demonstrates the importance of using raw files that haven't been sharpened. (Or, conversely, cautions against relying too heavily upon edge spread analyses, since edges are so sensitive and susceptible to image processing.)
Figure 14 is a photo of an integrated circuit, lit by a ring LED light, taken at 1/100 s, f/4.5 with the macro lens at 1:1 and the camera in high-resolution mode, with a putative pixel pitch of 1.85 μm/px (as discussed above, some of that may be 'empty magnification'). The teeth in the comb-like feature in the image at the right are spaced by approximately 11 microns.
Fig 14. Left: Photo of a portion of an integrated circuit wafer, 1:1, lit by a ring LED light. Right: Near-center piece of the left-hand image, at 100%, cropped. The teeth of comb-like feature are spaced ~11 microns.
Generally, the camera and lens were a joy to work with and provided excellent macro images with little effort, thanks to their ability to perform focus bracketing. For example, here's a 3D view of sand grains produced using this camera and lens.
Comments or suggestions