Olympus/OM System macro lenses tested
OM System currently markets three macro lenses, of which two are inherited from Olympus and are so far unchanged in all respects, down to the original branding:
The first to be introduced, in 2012, was the 60 mm, capable of AFautofocus and 1x magnification. The optical scheme uses 13 elements in 10 groups, which is moderately complex for a modern macro lens. It is a good, lightweight and relatively compact macro lens, of average price. The magnification/distance scale, separate from the focus ring and visible through a transparent window, together with a rotary focus limiter dial instead of the more typical slider switch, set it aside from typical macro lenses. A further unusual feature is that the focus limiter has a position to manually force the lens to focus at 1x. This is a premium lens, which means it is positioned between the M. Zuiko Pro series and the M. Zuiko consumer series. It has some features of the Pro series, in particular weather sealing, but lacks other features found in this series.
The 30 mm followed in 2016. At 1.25x, it is one of the earliest AF macro lenses exceeding 1x. Its optical scheme with 7 elements in 6 groups is quite simple for a modern macro lens. It is lightweight and cheap, and lacks multiple features normally associated with macro lenses, most notably a distance/magnification scale and a focus limiter control on the lens barrel. It also lacks weather sealing. Unavoidably, the short focal length results in a short WDworking distance, and the relatively broad front of the barrel makes illumination of the subject problematic in the macro range.
In 2023, OM System introduced the 90 mm f/3.5, which had been displayed on Olympus lens roadmaps for quite a few years, albeit its exact specifications were never publicly announced. Among other things, the focal length was generally expected to be slightly longer, at 100-105 mm. The optical scheme with 18 elements in 13 groups is more complex than average for a modern macro lens.
The 90 mm is the first macro lens in the Pro series, and offers new features like a 2x maximum unaided magnification, the possibility of reaching up to 4x with Olympus teleconverters, and a novel IS system capable of working when hand-held also in the macro range. However, in my opinion the OM System (or possibly the original Olympus) lens designers committed a couple of serious blunders that negatively affect its ergonomics.
All three lenses have digital encoders in their focusing rings, i.e. focus-by-wire is performed even when the the focus ring is manually operated. The 90 mm is alone among these three lenses in having a forward/backward sliding clutch on the focus ring to switch between AF and MFmanual focus. This feature is available in most, but not all, Pro-series lenses.
I tested all three lenses at 1x, since this is the highest magnification shared by all three lenses. It is also a magnification commonly used in photomacrography. Most tests of macro lenses on the web are apparently performed at a much lower magnification, on a test target placed meters away from the camera. This is fine for general-purpose lenses, but when I want to evaluate a macro lens, I am primarily interested in how it performs in the macro range. Some tests were also performed at different magnifications (see the figure captions for details).
A good performance in the close-up and portrait ranges is important for a macro lens, and a good performance at infinity is icing on the cake, but if the lens is only mediocre in photomacrography, it misses its purpose. Make no mistake, I do like the icing on a cake, and I am a bit frustrated when someone else has eaten the icing off my cake, but I am even more disappointed if I get just the icing and no cake. The latter is what you get, for example, with the Sony 90 mm f/2.8 G for FE cameras, which is literally a macro lens not designed for use in the macro range. It performs well when tested on the same test rigs used for general-purpose lenses, but when the magnification approaches 1x its image quality is disappointing for a lens of its price range. Several much cheaper, fully manual current macro lenses provide a far better IQimage quality in the macro range than the Sony 90 mm.
For this test, I used a Thorlabs NBS 1963A resolution target with a finest pattern of 228 lp/mm mounted in the test fixture discussed at this link. Imaged at 1x, the elements on this resolution target are labeled in lp/mm, with no conversion necessary to factor in the image magnification. For all tests I used electronic shutter only, with the camera attached to the stand of a Zeiss measuring microscope and a motorized THK-KR20 rail, with the rail controller used for manual focusing and the ROIregion of interest of the resolution target magnified to the maximum extent (14x) on the camera's LCD screen.
All three lenses were manually focused onto the finest resolvable element of the resolution target, as displayed on the LCD screen of the OM-1 camera at the maximum magnification possible in live view. The lens aperture reported in the figure captions is the nominal aperture, as displayed by the camera.
In this test I did not evaluate the curvature of field. Instead, I refocused when switching between center and corner tests, to record the best possible image resolution at either position.
Since the 90 mm f/3.5 can focus at 1x at two different settings of the focus limiter, as discussed above, I tested this lens at both settings.
30 mm f/3.5 test
The lack of a magnification scale slightly complicates testing this lens at 1x, since the lens significantly exceeds this magnification. A fairly good approximation of 1x magnification can be achieved by trial and error using a ruler as subject (the Micro 4/3 FOVfield of view at 1x is 17.35 mm wide), not changing the magnification afterwards, and focusing by moving the camera + lens assembly with respect to the subject with a micrometric stage, as typical when magnification must not change.
At 1x, this lens barely resolves 128 lp/mm in the center, and 114 lp/mm in the corners.
30 mm test at 1.2x
At approximately 1.2x, the lens barely resolves 161 lp/mm on the subject (134 lp/mm on the sensor) in the center, and 114 lp/mm (95 lp/mm on the sensor) in the corners. The difference in on-sensor resolution between 1x and 1.2x is visible, only slightly lower in the center at 1.2x, and substantially lower in the corners at 1.2x.
30 mm discussion
The disappointing performance of this lens in the corners was a surprise to me. The problems worsen above 1x, as shown in the above test at approximately 1.2x. I mostly use this lens in product and close-up photography because of its very short WD, and in this use I did not notice a corresponding obvious degradation of IQ in the corners. It is possible that I simple overlooked this fault, because with these types of subjects the corners rarely contain significant detail, or (more likely) that this fault is less visible at lower magnification.
A strong coma and uncorrected spherical aberration seems to be some of the causes of this problem. It lessens at f/5.6 and higher, but at these apertures the resolution in the center is already decreasing. At 1.2x, there seems to be also an unusual type of axial chromatic aberration that causes fuzzy orange haloes around dark parts of the subject. Stopping down reduces, but does not completely eliminate, this aberration.
As discussed here, the extremely short WD of this lens in the macro range (around 10 mm at 1.2x when a step-down ring is used as an almost flat lens shade) is an even worse problem than low corner performance, so I don't recomment this lens for macro work in any case. To perform the above test in the corners, I had to remove parts of the fixture containing the resolution target, in order to place the front of the lens close enough to the target.
60 mm f/2.8 test
The maximum magnification of the 60 mm macro is slightly higher than 1x. Instead of attempting a test at exactly 1x, I used images recorded at the maximum lens magnification (apparently close to 1.1x).
At the center of the frame, 1x and f/2.8-f/3.5, resolution barely reaches 144 lp/mm, which is very close to the 149 lp/mm of the Nyquist limit for a 20 Mpixel Micro 4/3 sensor (as discussed here). In the corners, resolution decreases slightly to 128 lp/mm at f/2.8-f/4. Contrast is also lower in the corners. A slight chromatic aliasing is present in the center (suggesting that the lens outresolves the sensor) but not in the corners.
90 mm f/3.5 test
Although this lens focuses to a maximum magnification of 2x, it is possible to limit its focus range to 1x by using the hardware focus limiter. It is less precise to manually focus at 1x by setting the focus limiter to S Auto, pulling back the focus clutch to MF, and turning the focus ring to the 1x position (the focus and magnification scale is only indicative). At nominal 1x as set by the focus limiter, the magnification of this lens seems to be closer to actual 1x than with the other tested lenses.
As mentioned above, the Nyquist limit for a 20 Mpixel Micro 4/3 sensor is 149 lp/mm. At 1x, the actual resolution of the 90 mm on the OM-1 is in practice somehat lower, at 128 lp/mm. Image resolution is very slightly higher in the image center than in the extreme corners, but the difference is quite small. Although some unevenness is visible in the imaged groups of lines at spatial frequencies up to 228 lp/mm, this is not true image resolution but aliasing. A simple way to detect this is counting the number of lines in each group: they are 5 on the resolution target, but only 2 to 4 can be counted in the areas affected by aliasing. The implications of this type of aliasing are further discussed below.
As mentioned above, one of the purposes of this test is to verify whether, at 1x, this lens performs better when its focus limiter is switched to the S Macro setting, compared to the 0.25m-∞ setting. In my earlier tests of this lens at 1x, the MTF of the lens significantly increases at the S Macro setting, especially in the portion of the MTF graph that contains the MTF-70 to MTF-90. Without a significant hands-on experience with MTF tests, however, it is difficult to decipher from the graph exactly in which way the image quality is better at this setting. The above figures help to answer this question in a qualitative way: the main difference between the two focus limiter settings is that image contrast is visibly higher when S Macro is used. Image resolution in lp/mm, on the other hand, at a given lens aperture is largely unaffected by this setting.
The camera reports a lens speed of f/5 when the S Macro setting is engaged, compared to the nominal f/3.5 reported without this setting. A careful comparison of the above test images does show that the image resolution at f/5.6 is very similar at either setting. The images at f/8 are likewise virtually identical at either setting. It can also be noted that, without S Macro, image resolution (albeit not contrast) remains approximately constant between f/3.5 and f/5.6. Therefore, the lens speed advantage of the non-S Macro setting does not translate to a higher image resolution as one should theoretically expect, and in practical use there is no good reason to choose the non-S Macro setting in the magnification range 0.5x-1x. Above 1x, of course, S Macro must be used because this magnification range is otherwise not accessible.
It may also be noticed that the test images show no trace of lateral chromatic aberration or vignetting. Both aberrations, if at all present in the optics, are automatically corrected by the camera, and with this lens, on the OM-1 camera, the correction can be regarded as perfect.
Some of the test images do show a chromatic moiré aberration (mostly orange) at certain spatial frequencies, especially at f/8-f/11. Some aliasing is also visible, especially in the 181 lp/mm group up to f/5.6. It is possible that the software anti-aliasing filter and demosicing algorithm are at least partly responsible for the loss of actual resolution at spatial frequencies around the Nyquist limit, while the lens may in fact outresolve the sensor at f/3.5-f/5.6 (otherwise, no aliasing at spatial frequencies exceeding the Nyquist limit should be observed).
90 mm at 2x and Hi Res
The following test was performed only on the 90 mm. In this test, I imaged the test target first at 2x and normal resolution (20 Mpixel), and subsequently switched to tripod-mode Hi Res (80 Mpixel) with 8-shot sensor shift. In order to make the results visually comparable, I magnified the 20 Mpixel image by a factor of 2, albeit without interpolation (i.e. each original pixel became a square cluster of 4 identical pixels). This unavoidably results in "blocky" image crops, but it is the method that introduces the least post-processing alterations in the magnified image. It is of course possible to obtain visually more pleasing results by using AI-based resolution enhancements in post-processing, but the purpose of this test is to evaluate the lens performance on the OM-1.
At 2x and the native 20 Mpixel resolution, there is virtually no difference at f/5, f/5.6 and f/8. This indicates that the sensor, in this case, is the limiting factor. It is also visible that the 228 lp/mm group (imaged at 114 lp/mm on the sensor) is always resolved at these apertures. This is no surprise, since this resolution is a bit below the Nyquist limit of the sensor. This also implies that the resolution target is too coarse to measure the on-sensor resolution in lp/mm at 2x. 228 lp/mm means the thickness of one of these lines on the target is 2.2 μm (4.4 μm on the sensor).
The 80 Mpixel Hi Res images, on the other hand, are visually much sharper than the preceding set of images. See in particular the texture visible at f/5 on the dust bunny lying on the "181" label. The effect of stopping down from f/5 to f/5.6 is evident, and the additional loss of detail between f/5.6 and f/8 is visible, albeit to a lesser degree. There is no doubt that the lens still delivers more than enough resolution to take advantage of Hi Res at 2x, and that, in this case, the best resolution and contrast are clearly achieved at f/5.
On another page, I tested the 90 mm with teleconverters. The results are good, but in view of the excellent results provided by Hi Res, the latter might be a better alternative than a 2x teleconverter. I have not tried using a teleconverter simultaneously with Hi Res, but I suspect that this combination would exceed the lens capabilities.
Comparison of lenses
The 30 mm displays a reasonable image resolution at 1x and f/3.5-f/4 in the center of the image, but is quite poor in the corners, and gets worse at 1.2x. This is not even the worse of the handicaps of this lens. The extremely low WD is more important in limiting the lens' usability in the macro range.
At 1x, the 60 mm is a good lens. At f/2.8-f/3.5 it has a slight edge in center image resolution over the 90 mm, but in the S Macro setting the 90 mm at 1x has better contrast and slightly better corner resolution.
The 90 mm is very versatile, and optically excellent when the correct focus limiter settings are used. Inexplicably, OM System has left the user in charge of manually choosing which settings to use in the 0.5x-1x interval (without explaining to the user the consequences of this choice), instead of automatically providing the best performance at all magnifications. The focusing range broken into two sections is a related design blunder that could have been avoided.
The three current Olympus/OM System macro lenses for Micro 4/3 differ substantially in capabilities and image quality. The 30 mm is primarily usable as a close-up lens. Its poor image quality in the corners at 1x and above is certainly a handicap, but the very short WD is an even worse handicap. The 60 mm is a good lens, with slightly better image center than the 90 mm and slightly worse corners. The 90 mm has outstanding capabilities, but is penalized by design blunders that worsen its ergonomy.