Just MTF Charts

Just the Cinema MTF Charts: Zeiss Cine Lenses

By
Published June 16, 2019
Tags: , , ,
Zeiss Cine MTF Tests

Zeiss makes a lot of cinema lenses, there are a lot of graphs in this post. Please remember for some, like the Supreme Primes and some of the CP.3 lenses we didn’t get much access to, the MTF graph is the average of just a few copies (the number of copies is identified on the graph). This will be the longest one of these posts, there are a lot of Zeiss Cinema lenses.

A Quick How to on Reading MTF Charts

If 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 Lines

The 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.

 

CP.2 Standard Primes

CP.2 15mm T2.9

Lensrentals.com, 2019

CP.2 18mm T3.6

CP.2 21mm T2.9

Lensrentals.com, 2019

CP.2 25mm T2.1

Lensrentals.com, 2019

CP.2 28mm T2.1

Lensrentals.com, 2019

CP.2 35mm T2.1

Lensrentals.com, 2019

CP.2 50mm T2.1

Lensrentals.com, 2019

CP.2 85mm T2.1

Lensrentals.com, 2019

CP.2 100mm T2.1

Lensrentals.com, 2019

CP.2 135mm T2.1

Lensrentals.com, 2019

 

CP.2 Super Speed Primes

CP.2 35mm T1.5 SS

Lensrentals.com, 2019

CP.2 50mm T1.5 SS

Lensrentals.com, 2019

CP.2 85mm T1.5 SS

Lensrentals.com, 2019

 

CP.3 Primes

CP.3 XD 15mm T2.9

The original graph was incorrect. This is a corrected graph.

Lensrentals.com, 2019

CP.3 XD 18mm T2.9

The original graph was incorrect. This is a corrected graph.

Lensrentals.com, 2019

CP.3 XD 21mm T2.9

Lensrentals.com, 2019

CP.3 XD 28mm T2.1

The original graph was incorrect. This is a corrected graph.

Lensrentals.com, 2019

CP.3 XD 35mm T2.1

Lensrentals.com, 2019

CP.3 XD 50mm T2.1

Lensrentals.com, 2019

CP.3 XD 85mm T2.1

Lensrentals.com, 2019

CP.3 XD 100mm T2.1

The original graph was incorrect. This is a corrected graph.

Lensrentals.com, 2019

 

Zeiss Supreme Primes

Supreme Prime 25mm T1.5

Lensrentals.com, 2019

Supreme Prime 29mm T1.5

Lensrentals.com, 2019

Supreme Prime 35mm T1.5

Lensrentals.com, 2019

Supreme Prime 50mm T1.5

Lensrentals.com, 2019

Supreme Prime 65mm T1.5

The original graph was incorrect. This is a corrected graph.

Lensrentals.com, 2019

Supreme Prime 85mm T1.5

Lensrentals.com, 2019

Supreme Prime 100mm T1.5

The original graph was incorrect. This is a corrected graph.

Lensrentals.com, 2019

 

Roger Cicala, Aaron Closz, and Brandon Dube

Lensrentals.com

June, 2019

Author: Roger Cicala

I’m Roger and I am the founder of Lensrentals.com. 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.

Tweet
Posted in Just MTF Charts
← 
Newer Comments  →
  • Carleton Foxx

    With its 50lp/mm at almost .8, the CP.2 15mm T2.9 seems to have insane resolution.
    Is this the lens to use when I’m doing a wide shot of a scene with lots of detail (like a Rose Parade float composed of millions of glued-on flower petals or a battlefield scene with dozens of actors all wearing intricately decorated costumes)?

  • I had a similar issue with 10mm vs 20mm results with one of the Tokinas. I’m traveling with almost no internet at the moment, I’ll recheck and correct these when I get back this week.

  • Max Manzan

    I have to agree with Shen Hong.
    The wrong naming doesn’t help, but the apparently randomly mixed up frame heights (20mm vs 10mm) where Supreme Primes all cover a picture circle diamter of 46mm, well, it’s very confusing. And some results are puzzling to say the least: Supreme Primes 50mm vs 65mm vs 85mm vs 100mm.

  • Someone

    Yes, this is strange. I guess we would need a clarification from Roger about the correctness of the axis marks.

  • Shen Hong

    I did miss the axes marks, thank you for remingding!

    However, according to my knowledge and experience with many of the lenses listed, in this particular post, every single one of these is suitable for 135 Full Frame shooting, with minimum image circle 43mm. You may find the data from ZEISS official website and technical data sheet showing the same fact. The Supreme Prime line is actually marked as having 46mm image circle, while many of these FF cine lenses are capable of shooting still on 44×33 frame (like Supreme Prime 100 and CP.2 135). Therefore the marking of 10mm doesn’t really make sense either…I guess it is just another typo or software problem.

  • Someone

    One thing you seem to miss is the fact that some lenses here are Super35, while others are 24×36. (Compare horizontal axes, sometimes it goes to 10mm and sometimes to 20)

  • Shen Hong

    Thank you for the reply!

    I completely understand your point and I agree with you that a good cine lens doesn’t necessarily have a good MTF performance, though my comment was referring to resolution only. Nevertheless, the difference between SP 65/100 and SP50/85 can still not be justified.

    Obviously the 65 and 100 are among the sharpest in the line, they don’t seems to have a source from photo lens for rehousing, not like the case of CP.2 which has almost identical curves with some lenses from Classic SLR line. For 100, it may come from Otus, but 65 is most likely a new design and it outperformance the 50/85 extremely much. This one case cannot be explain by your theory.

    BTW, wish you good luck with the naming…and thanks for your work!

  • iKonOkLasT

    A line pair is one black line and one white line adjacent to each other.

  • THe naming issue is an ongoing problem I have with new / rarely tested lenses. The names aren’t in the graphing programs database and I and supposed to rename them by hand, but failed in this case. I’ve had too much going on.

    As to “look terribly bad” I think, perhaps, you don’t consider that photo lenses are designed for best resolution, cinema lenses for other things. The highest resolution cinema lenses are generally the rehoused, recently made, photo lenses. In the case of both Canon and Zeiss, the Cine version of a lens is often a previous generation compared to the current photo version.

    Classic, well regarded cinema lenses are rarely high-resolution. Why should they be? Until 4k and up became common resolution was of little importance.

  • Shen Hong

    Am I the only one got confused with the different titles of the Supreme Primes charts?

    Some of the charts like the SP65 and SP100 looks totally fine but some like SP50 or SP85 have titles of CP.2 on the charts.

    More bizzarre, these SP/CP.2 charts looks terribly bad regarding a cine lens of this price, worse than Otus and even Milvus, and they do not match the ones of CP.2 either, they don’t match with any chart. Where are they coming from? Which are the real charts for SP? And what is the real performance of SP?

  • JordanViray

    Ah yeah got it. The MTF formula I use is doesn’t use square root but is obviously equivalent. Wavelength and f-number ended up cancelling out which makes some sense intuitively but I had to work it out to see for myself. Sure enough, it did and Rayleigh gives MTF8.94. Here’s my working for anyone interested.

    https://uploads.disquscdn.com/images/108773067d0fd8fa73502f0c7973b02cb1b5c8703d916271ce1ac16cbb7f2d06.jpg

  • Brandon Dube

    MTF9 — 1.22 lambda F# is the airy radius (= rayleigh resolution criteria). Diffraction cutoff is 1/(lambda F#). => 1/(1.22*lambda*F#) (=> evaluate the big long nasty diffraction limited MTF formula with the arccosine and the sqrt in it at this frequency).

  • JordanViray

    The original image I used for Washington had a contrast range from about 40 to 244 or 80% contrast. Since this was a noiseless image, the details could be recovered as can be seen in my followup post earlier. Naturally in a real world situation, the noise floor makes such a recovery dicey whereas a higher MTF gives me better chances.

    I don’t know GIMP well enough to know how to do the modeling you described, unfortunately. As for MTF (almost zero) I get why maximum resolution is of interest as a hobby astronomer, microscopist, and as someone who follows semiconductor manufacturing, so don’t get me wrong, I am not saying low MTF is irrelevant. But it isn’t as useful for my application in photography where 1) noise is an important factor, and, 2) most of the big lens test sites use MTF50 which is helpful when comparing lenses. Yes, the resolution data Imatest produces and places like Lenstip and Photozone/Opticallimits publish isn’t nearly as valuable as the actual lens MTF data Lens Rentals puts out. But it is still valuable as a shopper.

    Interesting that MTF9 corresponds to Rayleigh so I might try deriving that myself as an exercise.

    From a photography standpoint, all else being equal, “more MTF more good” is indeed true, at least in my case. I tend to purchase lenses for the long term knowing that future sensors will be able to wring out more resolution than current ones can. To quote Roger re the Sony 135mm GM “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.’” Will it resolve 90MP with as good contrast as it can 43? No, but it will do it better than any other 135mm he’s tested so that is important to me.

  • Brandon Dube

    Throw in some noise and watch it disappear, though.

  • Brandon Dube

    Regarding the “MTF” of a scene — the text has contrast of something like (black) on (white) = 100%. The details in GW have much lower contrast, perhaps just 5% of their own. In the frequency domain, the imaging model is a multiplication, so you would have 5% object * 5% MTF => 0.25% contrast, which is probably “below perception.” This does not mean the information is destroyed — with no noise it is still there, but with a noise floor you may not be able to “recover” it by stretching it back out (increasing contrast). This subject is called information theory these days.

    For modeling different MTFs in GIMP — if there is a convolution option (“kernel blur”) that is what you would do, given a PSF associated with a given MTF.

    MTF (almost zero) has a lot of historical context coming from physics. MTF9 is the rayleigh limit and comes straight out of physics, MTF15 is from the sparrow distance, and MTF20 from a third which I forget the name of. MTF50 has no analog in this way (that I know of) and is the fabrication of Imatest. This isn’t to say any of these are the ideal cutoff for your application — indeed that depends what you are trying to do. In a simple understanding, “more MTF more good” but this does nothing for setting a requirement. For that, I think that for photography a context of “I know what pictures from lenses A-F look like, and know what their MTFs look like” lets you pick lenses based on MTF curves (and a few other data points…) without a rigorous understanding of the physics behind it all, which is maybe ideal.

  • JordanViray

    Some qualifications I did not include: in a real photograph, the low frequency detail would show higher contrast but I have no idea how I would fully model MTF in GIMP.

    I did wonder whether reducing the output luminance range and converting to PNG would have completely eliminated the gradients which should show up as banding when the luminance was re-normalized or whether the gradients were still there but just imperceptible to me. So I took the image posted and applied unsharp masking and adjusted the levels and indeed the gradients were still there. There is banding present but I think that might not be the case had I been working with the original at full precision.

    https://uploads.disquscdn.com/images/ee8f9baec75a83849d987fd324fc53a6dc3b35ca36ff12e2ce4a882758524c3b.png

    This presents a rosier picture for MTF5 but in a regular photograph, in order to extract that detail, the software would need to identify frequencies in different areas of the image and adjust contrast accordingly. Sharpening filters do a decent job of this, perhaps inherently. But with MTF50, the original values are much closer to what they should be with an attendant reduction in risk of sharpening artifacts, e.g. haloes, or, more importantly for me, increased noise. For a studio shooter, that might be a nonissue but I need all the SNR I can get.

  • DrJon

    I think you want 16:9 not 3:2 to compare pixel counts…
    30*4*36 x 30*4*24 -> 30*4*36 x 30*4*20.5 = 10.6MP

  • JordanViray

    Thanks for the link and discussion. Even the 1% is easily seen for me so I wonder where the roughly 15% MTF I see bandied about as “barely resolvable figure to humans” comes from, e.g. Mr. Jam’s post above.

    https://uploads.disquscdn.com/images/bc0fdcb26b2393ec21d4ff602d05fe06f664d2c67afe56b9b36a35f7548a30ab.png

    However, here is an image comparing MTF50 and MTF5 using a photographic example to illustrate why MTF5 is not as useful for me versus MTF50. High contrast detail like the text is readable with MTF5, albeit not comfortably, but the gradations in a subject like the folds in George Washington’s clothes or the waves in his hair are basically gone.

    It’s possible that I did not remap the tones properly but I basically changed the selection levels output in GIMP 2.8.16 to 116/128 and 64/192 for MTF5 and MTF50 respectively and saved it as a PNG.

    It might be a different story if I had used more bits and was viewing on a 10-bit HDR monitor but that’s the general idea.

    It would be great if the admins could give you elevated Disqus privileges since I, and no doubt many others, consider you a really valuable contributor.

  • MisterWU

    Great test this a DP heaven.
    And nice pinch and zoom enable mobile web site.

  • Athanasius Kirchner

    Wow, that’s pretty amazing. 1% contrast is indeed visible, quite so already.

  • Brandon Dube

    You can certainly see 5% contrast… you probably can see 0.5% contrast without noise. The filter will eat any link I post, so go to imgur / a / 61HLs and see for yourself how good the human eye is…

  • Roger, thanks for this ongoing series! Do you know why the CP.3 21mm lens performs so differently than the CP.2 version? There are a couple of other lenses like that. Were those lenses redesigned?

  • JordanViray

    If we are looking for the physical limit and not caring about contrast, yes, MTF5 is better. MTF5 doesn’t matter to me as a photographer because I can’t see it even if the computer can. Sure, software can enhance contrast there but I think we can agree that we want the curves as close to diffraction limited as possible.

    Even though the method for determining lens resolution is far better at lensrentals, I still think Imatest has merit. Well, at least until lens rentals has tested every lens I am interested in (which it isn’t too far off of).

  • Brandon Dube

    Let’s not conflate popularity with quality. And IMO, MTF5 is better than MTF50. 50% contrast is still a huge amount and is far away from any sort of limit.

  • JordanViray

    Yeah it is arbitrary which is why I stated that it “depends on what you consider an acceptable MTF value”. Although arbitrary, MTF50 is more useful than something like MTF5 for photographic applications. And given the widespread use of Imatest on lens testing sites, although it is a systemwide MTF measurement, it’s useful for comparison purposes.

  • Brandon Dube

    MTF50 is an arbitrary cutoff determined by Imatest and is uncorrelated to any idea of imaging limits.

  • Les

    It’s moving images, not microfilm. Contrast at lower frequencies is probably more important than the finest high frequency detail.

  • Mr.JAM

    Line pair(s) means 2 black lines that can distinguish from each other. Line pair in diigtal should be 4 pixels, because at least you need 2 white pixels and 2 black pixels to distinguish each other. So, theorerically 30LP/mm in full frame means 30*4*36 x 30*4*24 pixels, roughly 12.44MP. So it is denser than DCI 4K(8.8MP). 50LP/mm is 34.56MP and it is roughly same as 8K.

  • JordanViray

    Depends on what you consider an acceptable MTF value. Most of them are at or higher than MTF50 in the center and the performance should be much better stopped down.

  • If I am not mistaken, the resolution of a 4K full-frame camera is 4096/36mm/2 pixels per line pair or roughly 50 lp/mm. Most of the lenses you tested don’t seem like they are up to 4K, let alone 8K.

←  
Newer Comments  →
Follow on Feedly