Just MTF Charts: Nikon Prime Lenses
We’ll continue our just MTF chart series of posts with the Nikon brand prime lenses. For those of you who have been asking, The Digital Picture is going to add all our MTF graphs to their comparison tool as we release them. That will let you do logical comparisons like ‘is the Sony 135mm f/1.8 GM sharper at the edges than the Sigma 14mm f/1.8‘ to prove your brand is the absolute best in online forum wars.
Three quick answers to questions that I’m sure will be asked:
- Z lenses testing requires sacrificing a camera and lens for electronics, machining a new mount for the bench, and wiring everything together. It’s a long and expensive process and we won’t have Z-lens test results for at least months.
- No, we don’t test stopped down. Lensrentals is willing to pay the cost for 10-copy tests wide open to establish standards for QA testing (which is always done wide open) and we share them. If you really want stop-down results, a 10 lens series costs $5,500. Email me and I’ll tell you where to send your check.
- The 24mm f1.8 isn’t missing because I forgot. I didn’t have enough results to post a 10-copy average.
A Quick How to on Reading MTF ChartsIf you’re new here, you’ll see we have a scientific methodology to our approach, and use MTF charts to measure lens resolution and sharpness. All of our MTF charts test ten of the same lenses, and then we average out the results. MTF (or (or Modulation Transfer Function) Charts measure the optical potential of a lens by plotting the contrast and resolution of the lens from the center to the outer corners of the frame. An MTF chart has two axis, the y-axis (vertical) and the x-axis (horizontal). The y-axis (vertical) measures how accurately the lens reproduces the object (sharpness), where 1.0 would be the theoretical “perfect lens”. The x-axis (horizontal) measures the distance from the center of a lens to the edges (measured in millimeters where 0mm represents the center, and 20mm represents the corner point). Generally, a lens has the greatest theoretical sharpness in the center, with the sharpness being reduced in the corners. Tangential & Sagittal LinesThe graph then plots two sets of five different ranges. These sets are broken down into Tangential lines (solid lines on our graphs) and Sagittal (dotted lines on our graphs). Sagittal lines are a pattern where the lines are oriented parallel to a line through the center of the image. Tangential (or Meridonial) lines are tested where the lines are aligned perpendicular to a line through the center of the image.
From there, the Sagittal and Tangential tests are done in 5 sets, started at 10 lines per millimeter (lp/mm), all the way up to 50 lines per millimeter (lp/mm). To put this in layman’s terms, the higher lp/mm measure how well the lens resolves fine detail. So, higher MTF is better than lower, and less separation of the sagittal and tangential lines are better than a lot of separation. Please keep in mind this is a simple introduction to MTF charts, for a more scientific explanation, feel free to read this article. |
Nikon MTF Charts
Nikkor 20mm f/1.8G ED
Nikkor 24mm f1.4 G ED
Nikkor 28mm f1.8G
Nikkor 28mm f1.4E ED (by special request)
We didn’t put this in originally because I didn’t have 10 copies to test. Here’s the results for 5 copies, it’s all I could get to.
Nikkor 35mm f1.4G
Nikkor 35mm f1.8G ED
Nikkor 50mm f1.4G
Nikkor 50mm f1.8G
Nikkor 58mm f1.4G
Nikkor 85mm f1.4G
Nikkor 85mm f1.8G
Nikkor 105mm f1.4E ED
Nikkor 105mm f2.8G IF-ED VR Micro
For a look at all the Just MTF Articles we’ve done so far, be sure to check them out here.
Roger Cicala, Aaron Closz, and Brandon Dube
Lensrentals.com
April, 2019
Addendum: The Digital Picture is hosting the MTF charts on their comparison tool, putting them up as we publish them here. It’s a great way to compare two lenses for the one or two of you who like to do that. The Canon and Zeiss primes are already up, the Sigma should be added by tomorrow.
56 Comments
denneboom ·
Is there a limiting factor for lenses on the bench (size maybe?) because I dont see tele primes on this list.
Roger Cicala ·
Yes. At the 200mm f2 is the max size we’re supposed to test.
GuyWith ·
Based on what I’ve read here before, to you guys the term “supposed to” is less of an impassible limit and more of a challenge to your ingenuity.
Roger Cicala ·
Guy, we’ve done up to 400mm f/2.8. But there’s always the question in my mind of ‘are those absolutely accurate’? And let’s just pretend that I did, say, Canon 400 f/2.8 Mk II and III and Sony GM. They’re all going to be excellent and the difference splitting hairs.
And the fanboys will split those hairs, magnify them like they were looking under an electron microscope, and I’ll spend days and days on forums going “No, that’s not what I said at all. The test results don’t show that.” I’m tired of that.
As I said the other day, “Fanboys are why we can’t have nice tests”.
But seriously, let’s say lens A is a tiny bit better than lens B. But B is heavier and longer. Could it be the increased weight, way over the theoretical limit of the bench, is allowing some vibration that makes the results lower? Is it possible the front element is so close to the collimator that it has some effect? I won’t know for sure.
Andreas Werle ·
Thanks for sharing the data with us, Roger.
Another question. In case of the 24mm 1.4 ED there is another “strange” behaviour of the MFT-curve. The contrast measurements for 40/50 lp/mm in tangential orientation follow an “up and down” path, decreasing at 4mm image height, then increasing and 6 mm and decreasing again. At the edge of the image circle the contrast does again increase slightly. Is there any explanation for this?
Greetings Andy
Roger Cicala ·
Hi Andy. That’s most likely field curvature; when the tangential field curves more out of focus than the sagittal, you get lower MTF readings. There are also some higher order aberrations that will have that behavior.
GuyWith ·
Amid the terror and tears of the day Uncle Sam demands his annual pound of flesh, along comes our Uncle Roger with a delightful surprise.
Thank you for beaming down this ray of sunshine, you have made zTax Day almost bearable.
Marcello Mura ·
If i where you, I will test try a kickstarter project to found a stopped down lens test, so we will see how much people really want thoose data! 😀 😀 😀
David Bateman ·
I think Roger could easily collect over 1 million. Thats how the tv show, video game high school collected money to be made. Now that depicted future is a reality with Overwatch leagues.
The issue is does Roger want to test the full range on thousands of lenses?
Brandon Dube ·
A million isn’t even a third of the funding it would take…
Roger Cicala ·
Marcello, unlike most kickstarter fund-ups, I’d actually feel like I had to deliver on promises made if I went that route, so it’s a big commitment.
To give a rough idea, though, to do just f/2.8 and f/4 on 10 copies each for just the prime lenses would take 276 testing days, and the cost (my cost, not retail) would be close to $200,000.
Last year one of the largest review sites entered into discussions about taking over the MTF project. They’re smart guys and after we talked about everything they basically said there was no way they could afford to do it.
P. Jasinski ·
The kickstarter idea does not really sound that abstract. I would most definititely back that up, as this might spare many people a lot of trips to camera shops attempting to test their lenses on shop displays, small AF targets and random camera store items 😀
If you count the time (or man-hours), I’d rather pay someone who actually knows what they’re doing 🙂
A 100$? cound me in.
Marcello Mura ·
Thank you for your reply Roger. The information you provide was very interesting and worth asking. I fully understand that commit to a single project for 276 days is not the best for businnes.
Ayoh ·
Why is it so expensive?
Roger Cicala ·
Ayoh,
Start with the optical bench is a $200,000 piece of equipment that requires about $20,000 a year in maintenance and upkeep. Add in $50,000 worth of mounts for various lenses and another $60,000 in custom software, rent and utilities on a building to keep it in and the bare breakeven cost to run it is about $50 an hour (it doesn’t last forever, and I’m not doing what a business would do and depreciating it over 3 years and adding in the interest on the loan to buy it — hence my cost only, not what I’d actually charge which is way more).
Plus it takes a skilled tech to run it, which is another $40 an hour (they don’t make $40, but employment costs, benefits, and taxes put that as a bare minimum). So $100 an hour ($720 a day) to break even; 276 days estimated to complete the testing = $198,200.
But that’s unrealistically low. During that same 276 days we’re supposed to be doing testing to fulfill contracts, etc. I’d lose way more income than that estimate if we were to just do stop-down testing. And before you get all excited about how rich optical testers could get, Olaf optical testing (which owns the machine and all) has never made a profit and never paid me a dime.
Roger
Ayoh ·
Thank you Roger for the very candid reply, much appreciated.
Based on comments you have made at other times it seems you are pretty confident that your test system is much superior to that of lens manufacturers. Why do you think this is the case? it seems that for someone like Canon/Sony/Sigma/Nikon the costs above would be trivial compared to the cost committed to in lens manufacturing and cost associated with failure to control quality. It seems it would not be a question that they can’t do it rather than that they feel they don’t have to. Give the engineering expertise in those companies they could probably easily design and build their own OLAF-grade testing system if they committed to it. After all if comparatively small lens rental shop with a few guys and interns can do it surely a large scale lens design and manufacturing house can. Why would they not do that?
Ayoh ·
Thank you Roger for the very candid reply, much appreciated.
Based on comments you have made at other times it seems you are pretty confident that your test system is much superior to that of lens manufacturers. Why do you think this is the case? it seems that for someone like Canon/Sony/Sigma/Nikon the costs above would be trivial compared to the cost committed to in lens manufacturing and cost associated with failure to control quality. It seems it would not be a question that they can’t do it rather than that they feel they don’t have to. Give the engineering expertise in those companies they could probably easily design and build their own OLAF-grade testing system if they committed to it. After all if comparatively small lens rental shop with a few guys and interns can do it surely a large scale lens design and manufacturing house can. Why would they not do that themselves?
Roger Cicala ·
Ayoh,
That is, indeed, the question. The answer is that some do. We have had engineers from several manufacturers come over and exchange methodologies and data. Others choose not to and some are testing exactly the same way they did in the 60s.
They all could, of course, the expense would be trivial for them. Why they don’t varies. In a couple of instances we’ve had a company’s engineers come over and they were very excited and couldn’t wait to use the techniques. The company’s managers absolutely denied there was any need for such testing, the way they were doing it was just fine, and refused to invest any money in testing.
These are very large companies with different departments wanting different things. Marketing departments don’t want test results released, they can’t possibly be as good as the computer generated theoretical MTF tests. Accounting departments don’t want to ever spend money. Some optical engineers love bad metrology (lens testing) because if there design looks great on computer but doesn’t manufacture well, bad metrology won’t reveal that. And most large companies have a lot of resistance to any idea of ‘you’ve been doing it wrong’.
My favorite example is a company who has zero interest in improving their in-house testing and has told us our testing is of no value. Yet their R&D department pays us to do testing of developmental products for them.
Before you ask, I can’t name name’s; if they have had interactions with us we’ve signed non-disclosures.
Ayoh ·
Thanks Roger, that is great information. Seems some companies have very short-sighted management. Manufacturing variation data is very useful for robust design. They don’t have to publicise the variation, or do anything about it but it still is much better to have that information than not
Ayoh ·
Thank you Roger for the very candid reply, much appreciated.
Based on comments you have made at other times it seems you are pretty confident that your test system is much superior to that of lens manufacturers. Why do you think this is the case? it seems that for someone like Canon/Sony/Sigma/Nikon the costs above would be trivial compared to the cost committed to in lens manufacturing and cost associated with failure to control quality. It seems it would not be a question that they can't do it rather than that they feel they don't have to. Give the engineering expertise in those companies they could probably easily design and build their own OLAF-grade testing system if they committed to it. After all if comparatively small lens rental shop with a few guys and interns can do it surely a large scale lens design and manufacturing house can. Why would they not do that themselves?
David Bateman ·
Roger that 24mm f0.14 lens looks crazy. Didn’t know Nikon made one. Or you may want to fix that title.
Roger Cicala ·
Thank you, David. Fixed that.
Roger Cicala ·
Thank you, David. Fixed that.
MassimoTava ·
Roger, would you prefer a MTF chart that has perfectly overlapping lines at lower lp/mm vs a lens that has higher lp/mm but Tangential & Sagittal Lines diverge wildly.
Roger Cicala ·
It’s not an absolute, but I’d sure trade a bit of MTF for some off-axis sag-tan separation improvement. Depends on what I’m doing, though. I’m looking for a very different MTF in an architectural lens than a sports lens.
Carleton Foxx ·
Can someone answer the magnitude question?
What percentage of betterness is represented by each horizontal line in the chart?
For instance if the 50 lp/mm of Lens A is at 0.4 and Lens B is at 0.5 does that mean Lens B is 10 percent better? Twice as good?
A related question: How much of a change is significant? Does the lp/mm have to jump a full 0.1 to say that a lens is sincerely better? Or is 0.01 a noticeable improvement?
Brandon Dube ·
A vertical bar is an increase of 10% (relative to 1.0… it would be +50% to go from .2 to .3) contrast at a particular scale. What is significant is not straightforward to answer. If you take pictures of a clear blue sky, the MTF is irrelevant because the object has no detail. If you take pictures of very fine patterns and textures, you have a great sensitivity to what MTF is measuring.
0.01 is way below the perceptible threshold. In general, as a very rough rule of thumb two lenses which differ by 0.1 at 40lp/mm are noticeably different. Two that differ that much at 20lp/mm are night and day.
Andreas Werle ·
This is realy interesting, thanks for the question and the answer! So, the contrast for 20lp/mm is most relevant for the perception of the image quality. This is an important aspect and may be helpful for the interpretation of the data.
Roger Cicala ·
I don’t want to speak for Brandon, but I’m pretty sure that’s NOT what he meant, it’s a huge oversimplification.
Put another way, there’s rarely a 10% difference between similar lenses at 20 lp/mm. If we test a dozen or two copies of a prime lens, almost all would be within 2-3% and one that’s 10% lower is almost certainly a ‘bad copy’.
If we test the same two dozen lenses at 40 lp/mm they’d probably be scattered in a range of at least 10%. If you evaluate images with fine detail critically you would see that some resolve better than others. That’s getting into what we consider sample variation and is inevitable.
This is why manufacturer’s, who generally test at a 20 lp/mm equivalent, often say a lens is fine while a photographer, who resolves detail at 40 or 50 lp will state ‘it’s obviously softer on one side than the other’.
Andreas Werle ·
Sorry for that.
Of course the statement, that “Two that differ that much …” refers to copy variance, not as a guide to compare different lenses (50/1.4 from Brand A compared to 50/1.4 from Brand B).
I understand, that this would be an oversimplification of your Data and i respect, that you want to avoid it. Peace! 🙂
Greetings Andy
Brandon Dube ·
No… MTF is just a thing. Whether lens A and B are the same model or different models is not relevant. If lens A and lens B differ at 20lp/mm by 0.1, one is going to look substantially better than the other.
Brandon Dube ·
Please don’t turn a very rough rule of thumb into some kind of statement about the relevance to perceptual image quality. I did not say 20 lp/mm was more or less important than anything else. It is simply the case that unless your lens is dogshit, the MTF is going to be high at very low spatial frequencies (10, 20 lp/mm). Since the range of ‘typical’ values at those frequencies is quite compressed (say, 0.7 – 0.1 instead of 0.1 – 0.8), the sensitivity is a lot higher.
Doug McEwen ·
Roger, thanks again for being willing to publish this treasure of information for all these lenses. The only thing I wish for is that you would resume publishing the variance data, because for me it was just as important as the MTF data. If I lived in the US I would only buy lenses by renting them first from Lens Rentals, and then keeping the ones that were reasonably close to what reviews and test sites such as yours indicate they should be. But I live in Canada, so I have to purchase them at retail and hope for the best. I would rather buy a lens with a bit worse MTF but good variance, than the other way around which becomes somewhat of a lottery. You indicated in a past post that you stopped publishing the variance data because it was being abused. I was very sorry to hear that – I thought it was good for the whole industry that the variance data be published, because it gave manufacturers incentive to improve that aspect of the lenses, and obviously, customers won too.
Roger Cicala ·
Doug, I may come back to doing it, but the variance data has a bit of an issue that makes me uncomfortable. There are some lenses where every copy has a soft side or corner. The variance formula misses this and says that lens has OK variance because math says ‘normal’ is a soft side or corner; they all have that, so low variance.
What we probably need is a two-fold ‘variation in center sharpness between copies, variation in edges within a copy’ kind of thing. But I don’t have the math to make that kind of thing.
Kenneth Tai ·
28mm F1.4E please
Gerard Roulssen ·
A fantastic effort, much appreciated!
Any chance of 28/1.4E entering the party …?
Roger Cicala ·
Unfortunately that lens hasn’t been very popular and we don’t have 10 copies.
Matti6950 . ·
If the Sigma blasts tne nikon 28mm F1.4 in astigmatism and corner sharpness the choice is easy for me. To bad no comparison yet (though i’m already pretty sure i won’t get the nikon, as it seems optimized for close and not infinity.
Roger Cicala ·
I could maybe get a 5 copy set done on the Nikon. While I don’t like doing less than 10 copies, I guess it would be better than no data at all.
Matti6950 . ·
Thanks a lot Roger 🙂 Hmm choice between Sigma art and nikon harder now.
Roger Cicala ·
OK, I did the 5-copy run I was able to do and posted the results above.
Samuel Chia ·
Roger, I recall Brian from The-Digital-Picture used to have your MTF graphs and for quite a few lenses he had stopped-down measurements as well. Certainly this means you’ve measured lenses stopped down in the past, contrary to what you mentioned. I now can’t find those MTF graphs on the-digital-picture anymore, which is a pity.
Brandon Dube ·
I did some (~ Canon and Zeiss primes) in 2015, and we’ve done something like < 10 other lenses since.
Roger Cicala ·
As Brandon said, we do not have the complete set. Also, in those cases we didn’t do 10 copies, we did one copy at different apertures. And got 7,312 fanpeople screaming not fair, that wasn’t a good copy of the lens, they knew it was better than that, we were picking an awful copy to do because we hate their brand, etc. etc.
So I stopped releasing the single copy stopped down data, although there is some still out there.
bdbender4 ·
Ah, good to know, I also wondered about this.
I also was curious about Nikon Z mount lenses, but am grateful for and not surprised at the explanation.
As for the rest, it’s the same as it ever was. Thank you!
Ernest Green ·
I assume all these tests are done wide open. Can/does a typical prime lens (or any lens for that matter) easily score at or above a “5” on the 50 line pair test when stopped down? Say to 2.8, F4, F5.6. In other words do all modern quality lenses perform similarly at 5.6 when it comes to the 50 line pair? I’d like to see just one example of a stopped down lens test.
Roger Cicala ·
They all get better, and most primes will resolve 0.5 at 50 lp in at least the center 1/2 of the image by f/4. They aren’t all equal by any means, especially in the outer 1/3, and wide-angles generally struggle more to get there.
Newer design (last 5 years or so) high-quality f/2.8 zooms will get there in the center by f5.6 but few of them will reach 0.5 at 50 lp in the outer 1/3.
Couple of f5.6 examples, a Sigma 85 Art gets there, a Canon 85mm f1.2 doesn’t quite but it’s close, a Canon 70-200 f2.8 isn’t close away from center. Also a good example of why I’m always laughing when people tell me their zoom is ‘as good as a prime’.
https://uploads.disquscdn.com/images/301e96d0545a4cc9d08a040a35801bf3d2a2fe1b95cfff7d9d42341e47039c0a.png https://uploads.disquscdn.com/images/d5bd9429ad9b39c44f7c55f00c48b1eab85590b1fa031ea4786b70e6efb3b713.png https://uploads.disquscdn.com/images/67d043c7b3c79f2197f0dd7b189433266fb593ee85f8e954fdba7263498398ca.png
Mattr ·
Interesting how differently the two 85mm primes respond to stopping down (when compared to your wide open results). The main improvement with the Sigma Art seems to be better sagittal resolution in the outer parts of the image circle. Thanks for posting!
BTW, I understand that this is a lot of work/money. But even single copy stopped down results are informative.
Rokoko ·
Thank you Roger!
Chris in NH ·
How about the 200mm Micro F4 IF-ED? It’s still legendary for macro use, astonishing for a 20+ year old design, and I believe the oldest lens still active in Nikon’s lineup.
Andre Yew ·
Thanks for adding the 28/1.4E results Roger! It looks like the black sheep in the family even compared to the 104/1.4E, and if you’d hid the label and told me that was a recent Sigma Art, I’d totally believe you. Many of the Nikons’ results look like they’re a few generations old, which is ironic since they were the ones to push high-resolution DSLRs. (FWIW, I use many of the f/1.8 and f/1.4 Nikon primes including the 28/1.4E on a D850 and am happy with them.)
Lance Blackburn ·
Thank you very much for your efforts, Roger. Always much appreciated. It would be good to have a comparison to the original Nikon published theoretical MTFs side by side.
Patrick Cligny ·
Thank you Roger for your work on the 28mm/1.4E which confirm the FTM given by Nikon for 10 and 30 lines pairs (https://imaging.nikon.com/lineup/lens/f-mount/singlefocal/wide/af-s_28mmf14e_ed/spec.htm).
Nikon did a real work on 105 and 28mm/1.4. But the Sigma pressure, on performance and price is terrific…
Naveen Kailas ·
The Nikon 28/1.4 ED appears to be the sharpest lens of the bunch, and may be the sharpest lens under 50mm in any of your recent runs of MTF tests. Very surprising.
Steven D. Keirstead ·
It’s a really good lens.
Duckie ·
What happened with the Z mount? Is there any update on progress (or lack thereof) for mirrorless Z mount testing? For all I know the disclaimer “we won’t have Z-lens test results for at least months” could be years old. I figured the reason could be as simple as no plan to make a Z mount (or RF mount), because there is no economically viable reason to invest money into testing yet. E.g. not enough bad lenses that need optical realignment or service, so that for now it is easier and cheaper to simply replace bad lenses.