Sony 90 mm f/2.8 G,
a macro lens not designed for macro
In virtually every modern camera system, macro lenses of focal lengths ranging roughly between 70 and 105 mm are the best in terms of high resolution, low geometric distortion, flat field, and superb correction of optical aberrations. Years ago, I decided to complement my Micro 4/3 system with the full-frame Sony Alpha 7II and 7RII. I love Micro 4/3 for its portability and good quality of lenses in the Olympus/OM System M.Zuiko Pro series, and I am aware that their native 20 Mpixel resolution and image quality is sufficient for at least 95% of the uses of digital images. However, the inherent light-collecting advantage and higher native pixel count of full-frame sensors, coupled with the availability of specialty lenses that give their best on full-frame or larger sensors, make full-frame cameras a better choice for specific applications.
The Sony FE system offers a good but slightly antiquated 50 mm f/2.8 macro lens with a noisy AF motor, and a modern 90 mm f/2.8 G macro lens. 50 mm is a bit too short a focal length for macrophotography at 1x. Besides, I already have the CoastalOpt 60 mm f/4 Apo macro lens, which is quite close to 50 mm in focal length. This made the 90 mm my only real alternative among Sony lenses.
Encouraged by a few very positive, and even enthusiastic, reviews of the 90 mm f/2.8 by Ken Rockwell, DxOMark and lensrentals.com, I purchased this lens. A modern macro lens capable of reasonably fast AF is potentially useful also for portrait and close-up photography, in addition to macrophotography in the field. The lens also has built-in optical image stabilization, an AF-stop button, AF-MF switching by sliding the focus ring fore and aft, three-position focus limiter, and communicates with the camera (e.g. the camera switches to high-magnification view while the focus ring is turned). The Sony 90 mm is quite expensive, another factor that often (but not always) suggests a good overall quality.
My tests with this lens at 1x, however, showed a rather mediocre optical performance betwen f/2.8 and f/5.6. Performance is slightly better at f/11, but on high-resolution sensors f/11 already begins to be negatively affected by diffraction. The much cheaper Laowa 100 mm f/2.8 Apo 2x (which, admittedly, has no AF and no electronics) performs much better in the f/2.8-f/5.6 range. Some Tamron and Sigma macro lenses in the same range of focal lengths also perform better than the Sony 90 mm, for a lower price. The legacy Nikon AF Micro Nikkor 105 mm f/2.8 is another not-too-expensive lens that performs optically better than the Sony 90 mm at 1x.
So, how did Ken Rockwell, DxOMark, lensrentals.com and other testers miss the mediocre lens performance of the Sony 90 mm? A likely answer is that these sites do not focus on macrophotography. Ken Rockwell published test images of a group of ceramic tiles, with a field of view probably half a meter or so diagonally. DxoMark does not seem to be equipped for tests in the macrophotography range.
On the other hand, lensrentals.com tested this lens on a test target 6 m away, then repeated the test at 0.6 m. They did not specify whether these measurements are the working distance (i.e. the distance between subject and front of the lens), or focus distance (i.e. the distance between subject and the lens focal plane). Since they mention that at 0.6 m the lens "does not quite reach" 1x, I am led to believe that they are talking about focus distance. At a 0.6 m focus distance (as displayed in the camera with the lens in manual focus mode), the actual working distance is approximately 45 cm and the magnification approximately 0.49x. This can be regarded as already in the macrophotography range.
At both 6 m and 0.6 m, lensrentals.com reported a significantly better MFT 50 for 6 samples of the Sony lens, compared to 6 samples of the Canon 100mm f/2.8. They performed the test on a Sony Alpha 7R (34.7 Mpixel), while my tests are done with a 7RII (42 Mpixel). The different pixel count translates to an only marginally increased resolution of the 7RII, not enough to substantially affect the test. It should also be mentioned that their Sony lens samples were from a 2015 pre-production run, which might be better than average lenses from the mass production line. Unfortunately, I have no direct experience with the Canon lens. According to closeupphotography.com, the Sigma Art 70mm f/2.8 DG Macro Lens (together with four other tested lenses) is better than the Canon lens, and the Canon lens itself is better than the Sony 90 mm, which was rated as "disappointing" at 1x on APS-C. This directly confirms my results with the Sony 90 mm at 1x. Unfortunately, by the time the test on closeupphotography.com was published, I had already owned the Sony lens for more than one year.
On another page of my site, I openly mused whether the Sony 90 mm has some redeeming qualities when used as a close-up lens, rather than in macrophotography. The purpose of the present page is to test whether this is the case, and to help me to decide whether I should keep this lens, or sell it and use only my numerous manual-focus macro lenses on Sony bodies.
I performed the test on a Sony Alpha 7RII (42 Mpixel sensor) with a resolution target originally designed for film scanners, identified on the target itself as film4ever.info and film4ever.digital. Googling these target names reveals several sellers of these resolution targets, both in the US and EU. Different types and sizes of scanning targets are available from these and other makers (e.g. Vlads test target). The part of target used for this test is at the top center in the slide, so nearly in the center of the frame.
Since this is the first time I used this test pattern, I also took a quick series of shots at 1x with the Printing Nikkor 105 mm f/2.8 A, which is my best lens at this magnification. The above figure does show that the pattern is not perfectly perpendicular to the lens axis (the bottom left portion is clearly out of focus). In optimal conditions, this lens can do much better. Nonetheless, the point of this test is to show that the test pattern is sufficiently detailed to test the Sony 90 mm. In particular, the -1 group is completely resolved, and the 0 group is partly resolved at this magnification, so the -1 group (and all lower groups) are certainly adequate for testing at lower magnifications.
If you wonder why the test pattern is not perpendicular to the frame sides, I do this intentionally. This helps to reduce aliasing and sharpening artefacts when sharp edges of the line pattern are imaged at a resolution close to the sensor's Nyquist frequency.
It can be seen that, at both magnifications, resolution at f/2.8 is not very good, and not quite acceptable for a lens of this price. At 0.2x, resolution improves while stopping down, becoming acceptable at f/4 and further slightly improving at f/5.6, while f/8 provides no perceptible advantage over f/5.6. Group -2 is partly resolved. At 0.6x, instead, stopping down from f/2.8 to f/8 does not perceptibly improve the resolution, with remains rather mediocre. Group -1 is partly resolved.
The above image shows that the Sony 90 mm lens at all tested apertures only resolves group -1, element 5 at 0.6x. This is equivalent to 44 lp/mm on the subject side, and therefore 73 lp/mm on the sensor side. The sensor's Nyquist frequency is equivalent to 110 lp/mm, so the sensor outresolves the lens by a significant margin, and the image displayed at 1:1 pixel ratio is perceived as fuzzy. In fact, the test of this lens by lenstips.com gives even poorer results (MTF 50 of only 57 lp/mm at f/2.8, and 71 at f/4).
At 0.2x, the lens at f/2.8 resolves group -2, element 3, or 17 lp/mm on the subject side. This translates to 85 lp/mm on the sensor size. Thus, the lp/mm count on the sensor side is a little higher than at 0.6x, which agrees with my qualitative judgement of a better sharpness. On the other hand, my qualitative impression of a higher sharpness at f/5.6 is not confirmed by the lp/mm count, since at this aperture the lens still resolves 17 lp/mm on the subject side.
This once more confirms that the 90 mm is not quite a good lens in the macro range of magnifications. It also confirms my suspicion that this lens performs better in close-up photography than in the macro range, so here we have the paradox of a macro lens not designed for the macro range around 1x, but rather as a close-up lens. Of course, this lens subjectively performs better on a sensor with a significantly lower pixel count, e.g. 24 Mpixel. However, the clear trend of Sony Alpha cameras is toward higher and higher pixel counts, so this Sony lens, since its introduction in 2015, has not kept up with the Sony cameras that can use this lens. Sony has also failed to provide a more modern macro lens that matches the capabilities of its own cameras.
Alternatively, could it be that Sony, at some point after introducing this lens, discovered that it was too expensive to produce this lens to standards matching its design parameters, and relaxed its acceptance criteria to artificially reduce the number of specimens rejected or sent back for retooling?
I still do not understand why Sony did not go all the way and designed a lens as good as those of many competitors throughout its focusing range. Given the remarkably high price of this lens, its optical performance is not quite acceptable. At most, this lens should sell for one-third to one-half of its actual price. Optically, there are better macro lenses even at this lower price, but the Sony 90 mm has additional features that, at least for some users, could make it worth buying (specifically, if the lens will mostly be used in close-up and product photography rather than macrophotography, and in situations where in-lens image stabilization and AF are important). However, it is definitely overpriced for what it offers.
Shall I keep this lens? Since I already own it and selling it on eBay is a hassle (and eBay gets a significant slice of the price), for now I will keep it. It is my only lens capable on AF on Sony cameras, so after selling it I would have to replace it with another AF macro lens. A few current third-party macro lenses do offer a better optical performance, but their AF is often slow, with lots of hunting back and forth.
Do I recommend this lens?No, not as a macro lens. It is clearly overpriced for what it offers. You should instead look into Tamron or Sigma alternatives, which are optically better and cost a lot less. If you already own this lens and use it mostly as a close-up lens then, like me, you may decide to keep it because of its reasonably good IQ and AF in the close-up range.
As I suspected, the Sony FE 90mm f/2.8 Macro G OSS performs better in close-up photography than in the macrophotography range around 1x. In this lens we have therefore the paradox of a macro lens not designed specifically for macrophotography, or at least a lens designed for sensors with much lower pixel counts than current cameras.