Sony FE 135mm f1.8 GM Early MTF Results

As is sometimes the case, I got access to a couple of pre-release copies of the new Sony FE 135mm f1.8 GM lens. Of course, if I get access, it gets MTF bench tested. I mounted the first one, sipped my coffee and then lost my mind and started shouting various expletives, enough to bring Aaron running in from the other room to see what I’d broken.

I hadn’t broken anything; I just saw MTF curves higher than anything I’d ever seen in a normal-range lens. (Lenses like 400mm f/2.8 super telephotos, are about this high. But those are super telephotos. And f/2.8.)

Anyway, I tested the two copies we had and sent a subtle note of congratulations to some friends at Sony. The note turned into a video conference with one of the designers of the lens and some phone calls that went like “you can write up those two copies” and “no, I only write up 10-copy sets”. This turned into Sony giving me to access to 8 more copies and permission release the test results early.

Sony 135mm GM Rentals

So, this write up is my usual MTF post; 10 new-from-box copies tested and averaged. They are Sony’s own copies, however, not the usual lenses we’ve bought off the shelf. I’ll repeat the test in 6 weeks when we get our own copies, but I have no reason to think it will be different. And just to be clear, Sony didn’t hover over me or approve my results; they’ll see this blog post for the first time exactly when you do.

A Bit About the Lens

I was permitted to share a bit of the background I was given on this lens; it has some new features. The paired linear motors moving two separate focusing groups haven’t been done before. There have been some attempts at paired focusing groups in zooms, but not in primes, and the pairs have generally been one ring and one linear motor. Also, these are new linear motors (depending on how you count, 4th generation) that are much more powerful and robust than earlier ones. This is the same motor design used in the Sony FE 400mm f/2.8 GM OSS lens, which we showed you in the teardown of that lens.

If I understand correctly, this focusing system allows the 135mm f1.8 GM to execute up to 60 AF instructions per second. {Correction: I misunderstood this part during the teleconference. What was said was that the A9 can give 60 instructions per second, and that this and the 400 f/2.8 come closest to keeping up with that.} That is faster than anything else Sony has made and does it to a higher degree of accuracy than they’ve achieved before.

Optically, the lens has what Sony’s engineers call the largest ‘extreme’ aspheric element ever made, and it’s up in the front of the lens, which they say helps both sharpness and bokeh. I think ‘extreme’ aspheric may be more of a marketing, than an optical, term. But what was very clear is they have (and I saw micrographs to demonstrate it) been able to polish this aspheric to a smoother degree than has been possible, reducing or eliminating any onion-skin bokeh.

Sharpness Testing Sony 135mm GM

There were more features, like the 11-blade aperture and the aluminum-magnesium composite chassis (the same material used in the Sony 400 f2.8 again). I’m not trying to make this into a lens review; it’s just my report of MTF tests. But I wanted to let you know that I was really impressed by the discussions I had with Sony engineers. As many of you who follow this blog know, ‘impressed’ has not always been my opinion of Sony’s lenses. But I’m impressed this time.

MTF Results

Let’s make this simple and straightforward. In the center, that’s the highest MTF I’ve seen on a non-supertelephoto lens. The highest. Let’s put particular emphasis on the purple line, which is 50 lp/mm. That’s a higher frequency than any manufacturer tests (that we know of), appropriate for fine detail on the highest resolution cameras. We would consider an MTF of 0.5 at 50 lp/mm to be very acceptable. This is hugely better, nearly 0.8 in the center. We’ve never seen that kind of resolution before., 2018


The MTF drops away from the center, of course, but even at the very edges, the readings are still quite high.

Let’s compare it to the Sigma 135mm f1.8, which until today was the sharpest 135mm we had tested. In the outer 1/2 of the image they’re pretty even, but in the center half, the Sony GM is dramatically better, especially at higher resolutions., 2018


I’ll also throw up a comparison with the Zeiss 135mm Batis, which is really excellent, although not wide-aperture. The Batis has a considerable advantage since it’s being tested at f/2.8. Even at f/1.8, though, the Sony 135mm GM is clearly better in the center half of the image. 2018


But Wait! There’s More!

Aaron brought up that this was the highest center resolution either of us remembered seeing on standard testing, with 50 lp/mm reaching a ridiculous 0.78 MTF. We have, in the past, tested lenses at a higher frequency for ultra-high resolution sensors (150 megapixels). We found that a lot of lenses that were really good at standard frequencies died quickly at higher frequencies.

So we tested the 135mm GM up to 100 lp/mm, something we don’t normally do., 2019

These results are insanely good. At 100 lp/mm the Sony 135mm f/1.8 GM has a higher MTF than most excellent primes do at 50 lp / mm. If you don’t speak MTF, basically that means this lens can resolve fine details that would be a blur on excellent lenses.

Back when we were doing that ultra-high resolution testing we tested all the lenses stopped down to f/2.8 or f/4; there was no way to get the kind of resolution our client needed otherwise. So we tried the 100 lp test at f/2.8. Honestly, I thought the resolution wouldn’t go up all that much. As is so often the case, I thought wrong., 2018

No lens we’ve ever tested has resolved 100 lp/mm this well at any aperture. One other lens was close, but I can’t tell you the name of it. We were under such strict nondisclosure that we never referred to it by name. It was just referred to as ‘the lens in question’ and was a huge prototype. But even that lens wasn’t quite this good at 100 lp/mm.

What does this mean for you? Well, in a couple of years if you are shooting a 90-megapixel camera, this lens will be the one that wrings the most detail out of that sensor. Right now it looks at your 43 megapixels and goes, “that’s cute.”


This has been an MTF test. It has only been an MTF test. If it had been an actual lens review, I would have 762 images showing you pretty models, dramatic landscapes, and bokeh examples. Lens reviewers will do that in a while; be patient.

But as far as the test goes, the results are pretty simple. This is the sharpest lens we’ve tested. Period. (At last count, that’s out of 300+ lenses tested.)


Roger Cicala and Aaron Closz

March, 2019

Author: Roger Cicala

I’m Roger and I am the founder of Hailed as one of the optic nerds here, I enjoy shooting collimated light through 30X microscope objectives in my spare time. When I do take real pictures I like using something different: a Medium format, or Pentax K1, or a Sony RX1R.

Posted in Equipment
  • sahil brar

    Roger I appreciate what you do for the community and rarely disagree with your findings but I dont think that the little more centre sharpness should make you declare this as the sharpest whereas the Sigma 135 lens performs way better overall across most of the frame. That clearly qualifies as a superior optic overall. Also the sigma samples and sony samples were not acquired the same way and that makes me somewhat doubt because I kind of adament on the Sigma being the overall winner by a definite margin while keeping its superior rendering still aside.

  • joel richards

    Were you ever able to test the copy variance for the 135GM? I’m curious how much the resolution varied? Sony’s earlier lenses were sometimes great . . . but only if you got the rare “good” copy.

  • Just wondering if this lens is still the sharpest, detail resolving lens, in Sony platform, or in any platform? Also would it be possible to review Fuji GF lenses? I love the blogs that you do here.

  • Franz Graphstill

    What were the results six weeks later, @roger_cicala:disqus ? You mention planning to re-test when you got your shipment. I am guessing that they came out about the same?


    Excellent article. Thanks a lot for your MTF charts and for your explanations of them. I’ve learned a lot!

  • Franz Graphstill

    Well, we’ve tried it on the next higher resolution now – the 60Mpixel A7R IV, and the 135 GM is still quietly thinking to itself “that’s cute”.

  • Gerard R

    Hi Roger, did you all get an L Mount made yet?

  • chriswilliams

    Old Post, New Question.
    So, this lens has been on “backorder” for months now. Will the “second batch” be as good as the first? (Is the delay even a manufacturing thing?)

    Really wondering if they continue to build top quality even after the first round has sold out…

  • Roger Cicala

    Martin, I think that’s true for most fields of science. In this case we have a lot of errors to consider: Each lens is done at 4 rotations and there is variation in each rotation. For each rotation there is also variation from one side compared to the other. Then there is variation for the 10 different copies for each of the above.

    We can’t treat it as 80 measurement points since some, but not all of the measurements are related, being same lens. Although, interestingly intra- lens variation is sometimes greater than inter- lens variation off axis.

    We’ve had numerous discussions both with optical science people and statistics people on the best way to analyze said variation. I would guess we’ve had about 6 or 7 adamant opinions, none of which agreed, and another dozen ‘well all of them tell you something opinions about those. Monte Carlo analysis is the accepted method in the optical field, but then no one has analyzed this much data per lens, nor has anyone really done such analysis at high frequencies (variation increases the higher you go).

    We came up with a method in house and refined and changed it over the years, but I can’t say that it’s the ‘right’ analysis since smarter people than me don’t agree about it so I stopped publishing it since this is the internet and people love to overreact and abuse things.

    Here’s a variance graph that we use in-house. The area is not actually +/- 1.5 SDs (our calculation includes SD, but modifies it slightly) but it’s in that ballpark.

  • Martin Huisman

    In my field of science, when we test ‘something’ multiple times we plot data-points with their respective errors.
    Say, standard deviation (SD) or standard error of the mean (SEM).
    I’d be very keen to see -in general whenever you test multiple copies- you plot data this way, so we can visually assess in one view how much variation there is accross lenses of the same model and/or between measurements (lp/mm for instance, since usually errors increase along with increase of precision).

  • Franz Graphstill

    I have my copy now. It’s awe-inspiring. It really rewards manual focus. It produces razor sharp images wide open.

  • GuyWith

    You’re forgetting about handholding’s best friend—the strobe. Good studio strobes give you a motion-stopping 1/12,000 of a second flash duration. If you want to gang a bunch of speedlights you can get as fast as 1/40,000.
    Regardless, if you are shooting portraits you have to use a tripod because it breaks the flow if you stop everything to adjust lights when you drift out of place. The tech is there; might as well use it.

  • John Draper

    They are available now.

Follow on Feedly