A resolution test of Laowa 2x and Oshiro 2x
(on Micro 4/3)
On an earlier page, I discussed the Laowa 2x and Oshiro 2x lenses, two of the present three models of 60 mm macro lenses capable of focusing from infinity to 2x. I also described several perplexing point, including for example the fact that the user manual and sales materials of the Oshiro lens contains exclusively pictures of the Laowa lens, in spite of the two lenses being distinguishable from their pictures.
I decide to carry out a simple test to compare the resolution of these lenses.
I used for most of this test the ubiquitous 1951 USAF resolution target. The physical target I used is etched in a chome film deposited on a glass substrate. The glass plate contains one positive (black lines on transparent background) and one negative pattern (transparent lines on black background), as well as multiple replicas of the smallest portion of the patterns, in case some of them become damaged or contaminated with use. Below is a photograph of the negative pattern, imaged at f/4 with the CoastalOpt 60 mm Apo and cropped and reduced.
Sets of three lines are arranged in groups (columns with a number on top, and in addition an element at the bottom right belongs to the group on the left side of the target) and elements (two sets of lines, one vertical and one horizontal, with a number at one side). The group 0, for example, contains an element 1 at the bottom right of the pattern, and elements 2 to 6 aligned in a column at the left. Group 1, with elements 1 to 6, is aligned along the right. Group 2 and 3 are smaller versions of groups 0 and 1, respectively, and so on. Larger targets have 0 or a negative number as the largest group, and a given group (group 1 for example) always has the same absolute size, regardless of the size of the target (and therefore it is a wrong practice to e.g. print a double-size copy of a 1951 USAF target found online, without changing all the group numbers accordingly).
The 1951 USAF target used for this test is not a high-quality implementation, and it shows clear limitations in the lines of group 6, which display rounded ends, uneven thickness and other defects when observed with an industrial microscope. Some additional tests were performed using a microscope calibrating ruler of better resolution (see below).
I placed a small LED panel a few cm under the test pattern, and a 3 mm thick TFT panel as a diffuser, 1 cm below the test pattern.
As a comparison lens, I used the CoastalOpt 60 mm Apo f/4. This is a good macro lens with built-in focusing from infinity to 0.5x. It is characterized by absence of both types of chromatic aberration (axial and transversal), and is additionally transparent to, and corrected in, a broad spectrum from NIR to NUV. The latter characteristic is what makes this lens unique, rather than its resolution as a macro lens. LensRentals found the CoastalOpt at 0.2x to have the highest resolution among several of the best macro lenses they tested. This makes it a legitimate term of comparison for tests.
Flare and contrast
To prevent flare and low contrast with the positive 1952 USAF pattern, I had to reduce exposure and prevent overexposure of the transparent parts of the pattern. For this reason, the background is rendered as yellowish/pinkish in the test images, rather than white.
The test pattern I used has positive and negative patterns. The image quality with the negative pattern is decidedly poorer because of flare and excessive contrast, even after underexposing by 1-1.5 stop. The source of the flare is probably the massive amounts of off-axis light that passes through the transparent areas of the target near the target patterns. Masking off these areas would be a fist step to correct this problem, but this would in turn prevent the use of the positive pattern on the same target.
In retrospect, the best solution is probably avoiding targets that combine positive and negative patterns, as well as large transparent areas. While resolution patterns on paper have only a limited contrast range, the contrast of transilluminated chrome-on-glass patterns is orders of magnitude higher, and more than enough to create problems for virtually all imaging lenses.
With the test pattern used in the present test, all three lenses show vulnerability to flare. The CoastalOpt 60 mm is vulnerable to a central flare spot at certain apertures and magnifications (a solution is described here), while the Laowa and Oshiro are sensitive across most of the frame, and the flare manifests as an irregularly-shaped "cloud".
The above test is with the lens fully open. In preliminary test runs, I discovered that short exposure times, coupled with electronic shutter, often produce an evident stripe pattern in the images. I subsequently reduced the illumination intensity, but sometimes the stripe pattern still emerged with a lens fully open, e.g. in the sample at the center of the above figure.
The following samples are taken at f/8, which on Micro 4/3 is about the limit, past which diffraction begins to be noticeable in macrophotography.
The Laowa and Oshiro lenses behave slightly but visibly better in both resolution and contrast at f/8, while the CoastalOpt seems to be very slightly better at f/4. Flare is a problem with all three lenses at f/8, more so than fully open. The Laowa (or at least this specimen of the Laowa) is slightly better in image resolution than the Oshiro, but it is not a substantial difference.
The CoastalOpt is only designed to remain corrected for magnifications up to 0.5x. According to the CoastalOpt literature, it remains "still very good" at 1x with the addition of an extension ring, as done in this test. The test at 2x of this lens, as done for this page, is obviously non-optimal, since this lens should be reversed at magnifications above 1x. Nonetheless, the CoastalOpt performs no worse than the Laowa and Oshiro lenses in the 2x test on a sensor with a linear pixel density of 7,611 ppi (pixels per inch), in spite of the latter two lenses being designed for magnifications of up to 2x and corrected by a floating group up to this magnification.
To put things in perspective, all the above test images, including those with the aperture fully open, would be fully satisfactory when viewed or printed at reasonable sizes, after post-processing with ordinary adjustments (but no sharpening). The following is an example from the Laowa lens, even at its "not so good" fully open aperture of f/2.8. All the group 5 and some of the group 6 elements are resolved if you look at the corresponding crop above.
I mounted all tested lenses on an Olympus E-M1 Mark II Micro 4/3 camera with a 20 Mpixel Bayer sensor. All three lenses are designed for larger sensors, but the Laowa and Oshiro cover only APS-C sensors at infinity. They can be used on full-frame in the macro range. The CoastalOpt covers full-frame at all magnifications. Therefore, on Micro 4/3 a substantial periphery of the image circle is not used. Using a 20 Mpixel Micro 4/3 sensor is roughly equivalent to cropping the center of the image produced by an 80 Mpixel full-frame sensor, i.e, the sensor is likely to outresolve all three lenses.
For the 0.5x test, I focused all three lenses at this magnification, as indicated on their respective magnification scales.
For the 1x test, I used the CoastalOpt lens with a 52.5 mm extension ring and focused it to infinity, in order to provide a magnification approximately equal to the other two lenses when focused at the 1x setting (according to their magnification scale). The actual magnification of the three lenses turns out to be approximately 0.875x.
For the 2x test, the CoastalOpt was focused at 0.5x and additionally mounted on a stack of extension tubes with a total length of 60.5 mm to reach a magnification similar to the two 2x-capable lenses focused at 2x. The actual magnification is approximately 1.8x in all three lenses.
The focal length of the CoastalOpt shortens while focusing closer, because of the presence of a fixed correcting group. Applying to this lens the simplified formulas for macro lenses, with focus set to 0.5x and a 60 mm extension one should expect only a 1.5x magnification. However, these formulas are not accurate if used on lenses that change focal length when refocused. The Laowa and Oshiro lenses also use correcting groups.
A number of conclusions can be reached from the test images:
The 0.5x settings of the Laowa and Oshiro lenses are highly unreliable. The actual magnification is nowhere close to 0.5x, but the Oshiro (about 0.3x) is closer to reality than the Laowa (about 0.2x). Only a short distance (25 mm on the Laowa, 11 mm on the Oshiro) separates this setting from infinity focus, but a better calibration of the magnification scale should be possible nonetheless. In the CoastalOpt, more than half a turn separates 0.5x from infinity, but on the other hand, this lens cannot focus any closer without extension rings. As a consequence, it is nearly impossible to compare the three lenses at this magnification, without recalibrating the magnification scale of the Laowa and Oshiro lenses.
The 1x settings of the Laowa and Oshiro lenses are more close to reality, but still 20% off.
The 2x settings of both lenses, at 1.8x actual magnification, are off by 10%.
Some of the Laowa test crops above show a reddish tint of the black bars at f/2.8 and at the 1x and 2x settings. This is due to axial chromatic aberration, which in these crops becomes visible even with very small changes of focusing (5 μm or less). To confirm the presence of this aberration, I imaged a microscope stage ruler inclined with respect to the normal to the optical axis (below).
This type of aberration is present in both the Laowa and Oshiro lenses, and minor changes in focus turn the prevalent color of details from magenta or red on one side of the focus plane to cyan or green on the opposite side. This aberration is visible in both lenses, but not very large, compared to some other macro lenses. It is more visible at fully open aperture. It is hard to judge whether the Laowa or the Oshiro is better with respect to this aberration. Either one seems to have a slight edge on the other, depending on the magnification at which the test image is displayed and on the white balance (the two above pictures are crops of the same original image). This aberration is absent in the CoastalOpt.
Transversal chromatic aberration has a different effect, and when present is visible in well-focused details as a fringing in a radial direction, of two different colors on opposite sides of a detail, and increases as one approaches the peripheral regions of the image.
Unless the lens is decentered, it is not detectable in the center of the image. This type of chromatic aberration is undetectable in the Laowa and Oshiro lenses on Micro 4/3 (it might become visible on larger sensors). This aberration is also undetectable in the CoastalOpt.
Color and contrast
Both lenses render color and contrast well. The optical formula, of medium complexity for a modern macro lens (9 elements in 7 groups) ensures that the transmission of the lens remains good and spectrally even. At least in some cases, the Oshiro tends to display redder tones than the Laowa, but this is easily corrected in post-processing or with a custom in-camera white balance.
Both the Laowa and Oshiro 60 mm 2x macro lenses perform well enough, and given their competitive prices are an interesting choice where autofocus is not a requirement and a 2x magnification without extension tubes is desirable. Their main weaknesses are (1) the magnification scale around 0.5x is not correct, and (2) there is a moderate amount of axial chromatic aberration and sensitivity to flare. Optical performance is not a factor in choosing between either lens, and other factors, like price and mechanical construction, may guide the choice.