Unlike the CP.2 lens, the ZE and ZF mount Zeiss 135mm lenses have normal photography housings. The manual focus throw is not nearly as long as with the cinema lens, but it is very smooth and the lens focuses beautifully. With its solid metal housing, the Zeiss weighs in just over 2 pounds compared to 1.65 pounds for the Canon. The Zeiss has a 77mm front element compared to 72mm for the Canon, and 9 aperture blades compared to the Canon’s 8. There’s a bit of price difference, too, with the Zeiss listing for $2,122 and the Canon $989 at the moment.

Imatest Results

We had 8 copies of the ZE 135mm f/2 to test today — not enough to give absolute limits of variation but enough to at least give us a good suggestion. I’ve shown the Imatest MTF 50 results (in Line Pairs / Image Height on a Canon 5D Mk II) at f/2.0 below. As you can see this is a nice, tight grouping of results.

Compared to the average (mean) MTF50 values for the Canon 135mm f/2L, the Zeiss is better wide open across the frame, as shown in the table below. That’s very impressive as the Canon is one of the sharpest lenses around.

	Center MTF 50	Avg MTF 50	Corner MTF 50
Zeiss 135mm f/2	945	840	745
Canon 135mm f/2	800	710	640

As we stop the aperture down, though, the Canon catches up quite quickly. As shown in the graph below, the Zeiss slowly sharpens up steradily through f/5.6 on the Canon 5D II, with the corners reaching their maximum at f/8.

The Canon lens peaks at around the same aperture, but resolution increases to a greater degree as we stop down. By f/5.6 the lenses are virtually equal in resolution.

Summary

The Zeiss 135mm f/2 APO-Sonnar is a superb lens. It has one of the highest resolutions we’ve tested overall and the corners are amazingly good, even wide open. You definitely pay for what you get, though. The Canon 135mm f/2 is a superb lens and while it doesn’t have quite the resolution as the Zeiss wide open, it is less than half the price and autofocuses. (The 135mm f/2 is always on my list of the best value lenses available.)

Possibly in reaction to the Zeiss 135mm hitting the streets, a very widespread rumor has appeared that Sigma will announce a 135mm f/1.8 OS Art Series lens later this year. That’s exciting, but the key word here are ‘rumored’ and ‘announce’. While Sigma is generally fairly quick from announcement to release, that still sounds like a lens that won’t be available until the end of 2013 or early 2014. Assuming the rumors are true.

For video shooters, particularly, this lens is going to be a superb tool. Photographers wanting the very best will be interested, too. The optics are as good as it gets.

Roger Cicala

Lensrentals.com

April, 2013

It looks nice and hefty sitting next to a Canon 135mm f/2.

It looks much better mounted to a Canon 5D Mk III in my hands, don’t you think?

Imatest Results

We tested on the Canon 5D Mk II so we could directly compare it to one of my favorite lenses, the Canon 135mm f/2.0. The Canon is one of the sharpest lenses at f/2.0 we’ve tested. Below are the Imatest MTF50 results at the center, averaged at 13 points over the front surface of the lens, and the average of the 4 corners.

Long live the new King! The new Zeiss 135mm T2.1 CP.2 (and the hopefully soon to be available ZE and ZF f/2.0) have some pretty amazing numbers. Particularly in the corners. We’ve never had anything do better at this aperture.

I don’t have any nice pictures for you, but there are several people who have gotten copies already and have posted some impressive images. Here are a few links but there are lots more:

http://zeissimages.com/showreplies.php?qid=950

http://www.dpreview.com/forums/post/51185931

http://zeissimages.com/standardgallery.php?lenstype=557&showall

Pretty impressive stuff. It’s priced at $2,200. That’s going to make for a tough decision for Canon shooters with the excellent Canon 135mm f/2.0 lens available at less than half that price. I don’t think anyone will ever complain about the Zeiss image quality, though. It’s spectacular.

Roger Cicala
Lensrentals.com
April, 2013

Comparing the Specs

The Zeiss is a bit pricier than the others lenses in this range, and a bit larger. But you get twice as many aperture blades for your money. Not to mention it has significantly more light transmission. Don’t you wish photo lens makers had to use actual transmission (T) instead of theoretical calculations (f)? Looking at the table you kind of see why the camera makers might rather not.

	Zeiss CP.2	Canon IS II	Nikon VR II	Tamron VC	Sigma OS
Price	$19,900.00	$2,200.00	$2,400.00	$1,499.00	$1,249.00
Weight (lb)	6.2	3.3	3.4	3.2	3.15
Length (in)	9.85	7.8	8.1	7.4	7.8
Aperture blades	18	8	9	9	9
Min. Foc Dist. (ft.)	5	3.9	4.6	4.2	4.6
T#	2.9	3.4	3.3	3.2	3.2

Ok, enough of the silliness. The Zeiss lens is clearly an entirely different beast and while we can mount it to our SLRs that’s not what it’s designed for. That extra money and weight go into making it a true cinema lens with long, smooth focus and zoom gearing. It’s also really parfocal, meaning if you focus on something at 70mm and zoom out to 200mm the object is still in focus. None of the photo lenses are (although budget minded cinematographers desperately want them to be).

Just a Little Bit of Handling

I could go on for some time about how accurately it focuses (it does), how smoothly it zooms (totally true) or how it’s not too heavy to hand hold for a while (a complete lie – it weighs almost as much as a Canon 500 f/4 IS II). This is a lens designed from the ground up to be mounted to a set of rails and focused with a geared follow focus system. It’s perfect for that and built as solidly as any cinema lens we carry.

Cinema lenses, as a rule, are designed differently than photo lenses. Photo lenses are about rapid autofocus, which means rear or inner focusing. That in turn means focus breathing, often to the point of massive changes in focal length when you focus closely. Being parfocal is of little importance for a rapidly autofocusing photo lens. When you zoom from 80 to 150mm if the camera can autofocus in a split second, who cares if it’s still in focus after the move? Not to mention the subject might be moving anyway. Being parfocal is very important for a cinema zoom.

We did a quick parfocal check, comparing it with the Canon 70-200 f/2.8 IS II, which is not parfocal (but actually sort of close to it). We simply set the lenses at 70mm and live view focused on the bush in the center with each lens.

Then zoomed to 200mm and took another image.

Here are 100% crops of the bush at 200mm with the Zeiss on the left, Canon on the right.

We did a quick check for focus breathing, too. I won’t repeat the Canon lens, it breathes significantly and the focal length changes as you zoom close. The Zeiss 70-200 did not focus breath significantly from far to near focusing.

Yes, I Had to Run the Numbers

Absolute resolution, historically, has been far more important for a photo lens sitting in front of a high-resolution sensor than a video lens. Even 4K video is about 8 megapixels, not nearly as resolution sensitive as a 36 megapixel SLR. So when we’ve tested video lenses for resolution compared to photo lenses they’ve historically not held up well. Resolution isn’t their primary focus.

But we thought we’d see if the Zeiss could hold its own against the best 70-200 f2.8 photo zoom we have, the Canon 70-200 f/2.8 IS II. Because the Canon is actually shooting at T3.4, we tested the Zeiss wide open (T2.9) and also stopped down slightly to T4. I’m not going to clutter up the tables with the T4 numbers – this lens is as sharp wide open as it is stopped down, with the exception that the corners get just a tiny bit better at T4.

These are Imatest MTF50 results using a Canon 5D II test camera showing point sharpness at the center, average over the entire lens, and average of the 4 corners.

	Center MTF50	Avg. MTF50	Avg. Corner MTF50
Zeiss@ 70mm	990	775	600
Zeiss@ 135mm	915	675	575
Zeiss@ 200mm	815	575	425
Canon@ 70mm	875	755	575
Canon @ 200mm	840	720	525

Well, as you can see from the table, the Zeiss 70-200 T2.9 takes the idea of video lenses being lower resolution and shows that at the right price point, you really do get it all. At 70mm it’s clearly outresolving the Canon 70-200 f/2.8 IS II. The MTF50 decreases steadily at longer zoom lengths, but even at 200mm it’s still as sharp in the center as the Canon, which is the highest resolving 70-200 zoom we’ve tested. And remember the Canon is working at T3.4 wide open, a half stop slower than the Zeiss.

Let’s keep some perspective – if I were a photographer I wouldn’t be spending this kind of money for a 6-pound 70-200mm zoom because it’s sharper at the wide end. And even shooting 6k video I suspect you’d be hard pressed to detect a huge difference in your footage at 70mm. But now you can have a true cinema lens with long focus and zoom throws, properly geared for follow (and zoom) focus, parfocal and without significant breathing that’s as sharp as any photo lens made.

While $20,000 is sticker shock for my photography colleagues, consider a set of three Cooke Panchro primes covering the same focal length at the same aperture costs $22,000 and doesn’t even approach the Zeiss in resolution. Plus Zeiss lenses don’t tend to spit out focus helicoid collars and require a $600 repair every 3 months like Panchros do.

Roger Cicala and Aaron Closz

Lensrentals.com

April, 2012

BTW – I know what you’re thinking. Yes, I do love my job.

Does it improve my composition and technique? No. But knowing this stuff can be helpful. For example, when I want to shoot a landscape at 70mm and f/5.6 will my corners be sharper with my 24-70 f/2.8 or a 70-200 f/2.8? Or which will have less distortion for an architectural shot (since I hate the resolution loss of correcting distortion in post), my 35mm f/1.4 or my 24-70 zoom at 35mm? (Surprisingly, the answer is my zoom.)

This kind of information is easy to find. DxoMark has nice graphs for each lens that show distortion, vignetting, chromatic aberration, and resolution at various focal lengths and apertures for each lens they test. SLRgear.com has a nice pop-up app that shows the resolution across the field of the lens at various apertures and focal lengths.  The Digital Picture has great pop-ups that let you compare two lenses side-by-side for flare, distortion, vignetting and even images of ISO 12233 crops.

A lot of people use those tools when deciding which lens to buy. I use them after I have the lens so I know how to best use it.

 Resolution Maps

One thing that I’ve started using more frequently in post processing is a resolution map of the lens. We all know that every lens has highest resolution in the center and less in the corners. But the pattern of sharpness is different for different lenses.

Some lenses have a high peak of resolution right in the center that quickly drops off. Others maintain significantly high resolution halfway to the corners and then drop like a rock. Others have a rather linear drop-off from the center to the corners.

Just as an example, below are 6 Imatest charts showing MTF50 of 6 different lenses across the field of view. The absolute resolution numbers aren’t important for this demonstration, rather it’s the pattern of how the resolution changes. For each lens, yellow is the highest MTF50, blue is about 1/3 the value of yellow.

Why Does it Matter?

There are a lot of reasons, of course. But one I use a lot is creating sharpening maks for postprocessing. Like a lot of people, I use a masked layer for sharpening, applying less sharpening to the already sharp center of the image, and more sharpening to the softer areas. Instead of just a generic oval, I try to make a mask that mirrors the resolution map of the lens I’m shooting with.

I keep masks as actions for my most commonly used lenses, which speeds up postprocessing considerably. For example, I’d use something like the first mask, below, for images shot with the lens on the upper left above, and the second mask for middle right lens above.

As an example I’ll use two 100% crops from the left edge of this snapshot.

The crop on the left shows what that edge looks like when I sharpened the entire image to give best center sharpness. The crop on the right was when I used a mask to use stronger sharpening, but only at 50% strength in the center of the image. With either technique the center looked the same, but the edges were quite different.

Of course you can simply use an oval mask and adjust it for each image with a bit of trial and error. But I had 500 vacation photos to go through. Since 75% of them were taken with one lens at the same aperture, saving an action with the appropriate sharpening made that quick and easy.

You don’t need Imatest to figure out the sharpness pattern for the lenses you have. A simple photograph of a flat wall or fence with reasonable detail (bricks or unpainted wood are nice) will let you see where each lens starts to soften and by how much. Once you’ve made a good mask for that lens you have it forever. For most lenses, the same mask can be used at different apertures – you simply reduce the strength of the layer if you’ve shot stopped down. For other lenses, though, like my Zeiss 50mm f/1.4, you will need to make masks for different apertures.

Uwe Steinmueller at OutbackPhoto.net and I have been doing a series of articles trying to show how a little gear head knowledge and a little post-processing knowledge compliment each other and help make better images, and this is a great opportunity for that. Uwe’s article and action for corner sharpening, provide a nice photographic demonstration of how sharpening with a mask improves your end result, and a nice script with an adjustable mask.

Roger Cicala

Lensrentals.com

April, 2012

My first impulse was to do Standard Internet Response #1 — give an absolute answer, such as ‘obviously not’, despite having no facts to back that answer up. Then I considered Standard Internet Response #3 — give a useless, but factual, answer like, ‘well, if you have a 70-200 and teleconverter already, that’s certainly adequate’. (I never use Standard Internet Response #2 – the ‘if you’re a good enough photographer it doesn’t matter which you use’ response, nor S.I.R. #4 — ‘Google is your friend’.)

But, since it really is a reasonable question and a lot of people seemed interested, we set up to Imatest the 70-200 f/2.8 VR II / 2X III combination. Please be aware that our longest testing distance is 40 feet, which isn’t ideal for testing 400mm lenses, but it’s the longest we have. (I’m pretty comfortable it’s a longer testing distance than anyone else has, too, except maybe DxO and they aren’t really sharing information about their testing set up). Results may be quite different at 300 feet. I’m not sure which way they’d be different. The 70-200 seems sharper at this distance than it does at infinity, at least that’s what most people say. On the other hand, teleconverters are generally tuned for long distance shooting. So I just don’t know. (BTW – “I don’t know” is not a listed S. I. R.)

Imatest Results

We used an identical setup to the tests we ran last week on the 80-400 AF-S and 80-400 AF lenses to test the 70-200 f/2.8 with 2X combination. The MTF50 results are shown in the table below. The bottom line, from a resolution standpoint, the new 80-400 is clearly better. The previous 80-400 is better than the 70-200 with 2X right in the center, but outside the center the 70-200 with TC is very close.

	Center MTF50	Avg MTF50	Avg. Corner MTF50
Nikon 80-400 AF-S	820	675	480
Nikon 80-400 AF	725	575	410
Nikon 70-200 f/2.8 with 2X	600	560	440

What does it mean? Mostly it means if you’re shooting at 40 feet distance the 70-200 VR II and 2x teleconverter will get you a nice usable image, but not as good as you would get with the 80-400 VR II.

The old 80-400 AF lens is better in the center than the 70-200 VR II combination, although that’s just right at the center. Less than 1/3 of the distance away from the center, the two are even.

I can’t say the results would be the same if the shooting distance was near infinity, and I’m not sure how they’d change. The 70-200 alone is reputed to be a bit less sharp at infinity, though. On the other hand, the teleconverter might well have less of an effect at the longer shooting distance.

Roger Cicala

Lensrentals.com

March, 2013

Or. . . How I Learned to Appreciate Small Aperture Photography

If you read my blog much, you know I’m a resolution fanatic. I test every new lens for resolution. For personal use, I’ll choose the lens with higher resolution over the one with creamy bokeh every time. When choosing a camera, I have a (yes, I’m ashamed to admit it, but it’s true) strong tendency to want the most megapixels. I’m a resoholic.

Being a resoholic, I’ve always been somewhat fanatical about apertures. Whenever possible I shoot with the lens stopped down at least one stop to wring the maximum sharpness out of my lens. But I’m always careful not to stop down too far because I was taught, soon after I picked up a camera, that if you stopped down too far the dreaded diffraction softening would kick in.

With today’s high-pixel density cameras, that meant f/8 was as far as I would ever stop down. My mental map of aperture sharpness was like the ancient maps of the world – past f/8 there was nothing but the notation Here Thar Be Monsters. Or the equivalent label in Latin or Olde English, just because that makes it seem much cooler.

Go to f/11 and the diffraction monster would come and eat the resolution right out of your photographs. The diffraction monster loves to snack on some tasty resolution. When testing I really never checked past f/5.6 or f/8. That’s where the maximum resolution would be. Any further, and, well, you get it by now.

But I knew there were excellent photographers who shot their landscapes and macros at f/11 or even f/16 because they needed the depth of field. I heard rumors of photographers in far off lands who even actually took photographs at f/22. I considered them sort of like those guys who jump off cliffs in batsuits and fly around for a while before pulling their parachute rip cords. It was fascinating to know people did that, but made me a bit queasy. I was certain the survivors would eventually learn the error of their ways.

But lately, some people like Tim Parkin at Onlandscape.com started opening my eyes (by repeatedly beating on my head). They claimed to be shooting at f/16 and even f/22 with high-pixel-density SLRs, carefully postprocessing their images, and getting very nice detailed results. I shook my head sadly at first, hoping they would come to see the light (pun intended). But then I looked at Tim’s recent article The Diffraction Limit and had to admit, their f/22 images didn’t look bad at all.

So I decided it was time to open the closet door and see just how bad the diffraction monster really was.

Preliminaries

Before we get into all of this, let’s remember we’re looking at two simultaneous events when we stop a lens down. I am not going to get into lengthy discussions of Airy Discs, Raleigh Criteria, and other arguments here. You can read about them elsewhere. This is the simple overview of what’s going on.

• 1) As we narrow the aperture (higher f/number) diffraction occurs which causes some loss of resolution.
• 1a) Smaller pixels are affected more than larger pixels since diffraction causes a spread of the point into a disc. A disc of small size might still fit nicely on a large pixel, but might cover two small pixels. The math is mildly complex, it’s not linear, and I’m not going into it more than that.
• 2) As we narrow the aperture, the lens resolution increases.
• 2a) The increase is different for different lenses.
• 2b) The increase may occur at different rates for the center, middle, and corners of a lens.
• 3) Decreasing aperture is sort of a race between these two effects. When we first stop down, the lens sharpening is greater than the diffraction softening. As we stop down further, lens sharpening slows or stops, but diffraction softening continues.

Some Resolution Testing

I’m not one to really believe what I see in online exampless; given enough postprocessing an online jpg can look pretty sharp if the lens was the bottom of a beer bottle. I want at least a side dish of numbers or some comparative crops with my reviews, thank you.

I decided our current Nikon lineup gave me a great opportunity to look at diffraction effects. By shooting the same lenses on a D700, D3x, and D800 I can look at full-frame sensors with 12, 24.5, and 36 megapixel sensors. That gives linear pixel densities of 118, 168, and 204 pixels per mm, respectively.

I decided to use 50mm lenses because we have 3 choices that are quite different in how they behave at various apertures.  The Zeiss 50mm f/1.4 (the schizoid fiftoid) is very soft and dreamy looking wide open, but becomes razor sharp once stopped down to f/5.6 – it’s like two lenses in one. The Zeiss 50mm f/2 Makro planar is quite sharp even wide open, but seems to maximize it’s center resolution by f/4. The Nikon 50mm f/1.4 G is reasonably sharp wide open, but seems to keep getting sharper the more you stop it down.

So I tested all 3 lenses on all 3 bodies at apertures from wide open to completely stopped down in our Imatest lab.

The Effect of Stopping Down on MTF 50

I started right in the middle of my selections: the Nikon D3x with the very predictable Nikon 50mm f/1.4 G lens. Here are the MTF 50 values in line pairs / image height for the center point, weighted average of 13 points, and average of the 4 corner points. Please note that the plotted average is NOT just the average of center and corners, so if the ‘average’ value is near the center, you know the lens stays fairly sharp in the middle regions, while if it’s nearly as low as the corners the lens falls off rapidly away from the center point.

I have to admit I was a bit shocked. Just as expected, the resolution starts to decrease after f/8, but it doesn’t decrease all that much. Even at f/16 the resolution is still quite a bit higher than it was at f/1.4.

The next step was to see how things look with lower and higher pixel density cameras. So I shot the same lens on a D700 and D800.

I was a bit surprised here, too. I had expected the lower pixel density of the D700 would shift the peak resolution a bit, perhaps to f/11, but that wasn’t the case, although the drop after f/8 did seem less severe.

Things weren’t as different as I expected on the D800, either. The center does seem to peak around f/5.6 with the corners peaking at about f/8, which isn’t surprising. The other cameras show only a slight increase in resolution at the center between f/5.6 and f/8 so it makes sense there would be a bit stronger diffraction effect on the D800. I had really expected more than this. The lens still improves in the corners strongly between f/5.6 and f/8 and the improvement is greater than the diffraction softening.

The message I took away, though, is that diffraction softening is real, it occurs where it is supposed to, but it’s really not as severe as I had thought. Even on the D800 resolution is as high, or higher, at f/16 than it was at f/2.8. At f/11 the resolution is as good, or better, than at f/4. And at both f/11 and f/16 resolution is clearly higher than it was wide open. Perhaps the diffraction monster’s teeth aren’t as long and wicked as I thought.

Some Different Lenses

Diffraction softening is fairly constant, but lens sharpening as the aperture decreases is not. Different lenses behave differently. I compared the Zeiss 50mm f/1.4 and 50mm f/2.0 Makro Planar lenses to the Nikon 50mm f/1.4 G we tested above on all three cameras. In the interest of brevity we’ll just show the graphs for the D3x. The variations for the D800 and D700 were similar to these.

Let’s start with wide open performance. At f/1.4 the ZF f/1.4 lens isn’t as sharp as the Nikon 50mm was, while the ZF 50mm f/2 Makro is sharper at f/2 than either of the other lenses at that aperture. The f/2 Makro has reached maximum center sharpness by f/4 and then slowly loses resolution. The lens doesn’t reach maximum corner sharpness until f/8. The ZF 50mm f/1.4 gets maximum center sharpness at f/5.6 and corners again at f/8 on the D3x. (In the graph above, you can see the Nikon reached maximal sharpness at f/8 for both centers and corners.

The pattern was unchanged on the D800, but on the D700 the two Zeiss lenses center sharpness shifted just a bit to the right – moving to f/5.6 for the 50mm f/2 Makro, with corner sharpness remaining peaked at f/8.

So there is some difference in stopped down behavior with different lenses. Before you ask me to go test this or that, the work has largely been done already at sites like SLRgear.com and Photozone – they show center, corner and edge sharpness at various apertures in their lens reviews.

Yes, I Took Pictures

OK, the numbers surprised me a lot, so I went and did what had to be done. I actually took photographs stopped down to f/16 and even f/22.

Here’s one picture I shot at various apertures (this was on a Canon 6D).

What I saw mirrored what the numbers said I would see. Below are some 100% crops of the white gazebo just off center and some tree trunks near the left edge. I also tried something I was told was possible, but hadn’t really believed. I took the obviously diffraction softened f/22 image and did my best to sharpen it in Photoshop. (By best, I mean about 45 minutes testing different combinations of sharpness and contrast enhancement in 3 layers before getting the results shown below as ‘f/22 sharpened’.)

I haven’t tried this kind of sharpening before and was feeling my way along. I’m sure it would be better with practice (I blacked out the large tree-trunk for example and the image is about 1/3 stop darker than when I started) but still I found the results surprisingly acceptable.

One thing I found very interesting is that I could perform what was basically postprocess abuse on the f/22 image to a degree that would have been impossible with one of the other shots. Below, for example, is the f/2.8 image above processed with exactly the same settings I used on the f/22 image. The center crop, particularly, looks like a ‘find edges’ special effect filter.

I don’t mean to suggest that the postprocessed f/22 image is going to be as good as a nice f/5.6 or f/8 image at all. Rather I’m suggesting it can be improved to a larger degree than they can, making up some of the out-of-camera difference between them.

Does this mean f/16 is my new f/5.6? No, not at all. But I think I may become a lot more aggressive about using f/8 and f/11 when I’m trying for a larger depth of field. I even might use f/16 if absolutely needed. I don’t think I’ll be shooting f/22, though. That’s just a step too far for me.

Roger Cicala

Lensrentals.com

March, 2013

An aside: I’ll be going on vacation for 10 days at the end of this week, shooting with my new camera at my new-found apertures. I don’t expect there will be any more blog posts until late March.

When Uwe Steinmueller from The Digital Outback suggested we compare the same lens from both perspectives it sounded like a great idea to me. Uwe would write his assessment of the lens from a photography standpoint, I would write a bit about its testing and measurements.

He suggested we start with the Nikon 24-120 f/4 VR. It’s a lens he really likes for obvious reasons: it’s relatively small, covers a very useful range, and has superb vibration reduction. For him it provides a superb one-lens solution.

I’ve considered it an “OK” lens. I understand the attractions but was less than excited about its distortion, chromatic aberration, and borderline corner resolution. It’s not a bad lens by any means, but not one that made me ‘ooh’ and ‘ah’.

Resolution Numbers

When I first evaluated the 24-120 for use with the D800, I tested it only at it’s extremes (24mm and 120mm) so I went back with a couple of copies and repeated Imatest at several different focal lengths.

First we’ll look at the 24-120′s Imatest MTF 50 at 5 focal lengths. The table below shows MTF50, measured in line pairs/image height for raw files at the center, averaged over the entire lens, and the average of the 4 corners.

	Center	Avg.	Avg corner
24mm	1080	880	460
35mm	1100	885	460
50mm	1040	900	490
85mm	990	845	450
120mm	970	820	420

It’s apparent the lens is quite good in the center, although it’s not quite as sharp at the long end. The corners are OK, but on a D800e I’d really like to see resolution of 500 line pairs / image height and these aren’t quite there.

We’ve repeated them at f/5.6 to see how much benefit the lens gets from closing the aperture a bit.

	Center	Avg.	Avg corner
24mm	1100	920	560
35mm	1100	930	560
50mm	1075	940	575
85mm	1050	920	545
120mm	1020	880	500

The center, which was already quite excellent, doesn’t really improve with stopping down much except a bit at the long end. The corners do sharpen up nicely at f/5.6 and from a pure MTF 50 standpoint I’d call them acceptable at this aperture.

Other Numbers

Like most 5X zooms, there is a fair amount of distortion. The lens has 3.5% barrel distortion at 24mm which zeros out by 35mm. Almost immediately, though, pincushion distortion sets in, reaching 2% at 50mm and staying right there through the end of the zoom range.

There is a fair bit of vignetting at both ends of the zoom range, but it’s not awful by any means. Vignetting isn’t nearly as strong in the central part of the zoom range (30mm-80mm or so). There’s also a good bit of chromatic aberration in the corners throughout the zoom range and even when stopped down.

My Summary

The usefulness of a 24-120mm lens, with excellent vibration reduction and excellent center sharpness, can’t be argued. That makes it the perfect lens for a lot of people.

However, the corners are rather soft at f/4, becoming acceptable at f/5.6. Distortion and chromatic aberration are easy to remove in post processing, obviously, but doing so reduces resolution a bit. So looking at the numbers I wouldn’t recommend this lens for people who need sharp corners and edges, like landscape or architectural photographers.

However, Uwe’s article clearly demonstrates that with careful postprocessing the images shooting the 24-120mm on the D800 are really quite good. I think to some degree this is because the D800 allows such great resolution. Even average corner sharpness can look good dressed up in a D800 outfit. I also think good postprocessing is wringing every bit of resolution out that is there. I suspect out-of-camera jpgs are not quite as good as Uwe’s postprocessed images.

And again, for many people using this lens, the corners aren’t critical for their photographs anyway. They would, I think, they’d be very happy with images right out of the camera.

Roger Cicala

Lensrentals.com

February, 2013

A lot of my posts about lens resolution consist largely of showing the MTF 50 numbers from Imatest or our optical bench. That leads to a lot of questions about what that numerical difference really means, and how much you would notice it in photographs.

My usual response to that has been more numbers: SQF data that shows how large of a print would be required for you to see the difference at close viewing angles. But photographers are visual people and more numbers don’t always seem to answer the question. So I thought we’d try to show what a difference in MTF 50 numbers really looks like.

This is harder than you might think. I’m trying to use your monitor, at 72 dots per inch but with a huge color gamut, to demonstrate the difference you would see in a print at 300 dots per inch but with a smaller color gamut. The most accurate way to do it would be to invite you all over to look at some prints, but that’s not practical.

Today’s Testing

We’ll try something I think will usefully demonstrate what the numbers say and what you can see.

1) I’ll test specific copies of several lenses and give you their MTF50 measured by Imatest.

2) I’ll shoot testing-quality images (measured square, tripod mount, remote release, focus bracketed).

We’ll use ISO 12233 charts with the same lenses and show you 100% crops. It’s certainly the most used, and probably the most sensitive, of the optical charts for testing resolution. This should demonstrate what kind of difference a thorough home tester could see.

Its best feature for our purposes are the 4 crossed resolution charts in each corner, which demonstrate off-axis astigmatism pretty well (horizontal and vertical resolution are obviously different when there’s astigmatism) and let us compare the 4 outer quadrants of the lens.

3) Finally, I’ll send the raw files of some routine photographs taken with the same lenses and camera over to Uwe Steinmueller at Digital Outback Photo. Uwe is an expert post-processor, far better than I, and we’ll see how much difference there is when he’s completed his workflow. (To keep this of reasonable length, Uwe’s article is posted separately HERE.)

One nice side bar: one of the lenses has a slightly soft corner as measured by Imatest, but it’s not obviously decentered. So it will be interesting to see if we can actually detect it outside of Imatest measurebating. I’m guessing not, but I haven’t looked yet, I’m just putting my vote up before I complete the article.

Meet Today’s Contestants

I’m going to play with some Nikon-mount lenses today, mostly because that lets me use the Nikon D800‘s sensor to generate the best resolution we can. We’re going to compare some 35mm lenses: The Sigma 35mm f/1.4; Nikon 35mm f/1.4 G; Nikon 16-35mm f/4 VR at 35mm; and the Nikon 24-120 f/4 VR at 35mm.

Why did I choose these? Well, I’ll tell you.

• The Sigma 35mm f/1.4 is the highest resolving 35mm prime I’ve ever tested.
• The Nikon 35mm f/1.4 G is designed another way with emphasis on things other than wide-open resolution.
• The Nikon 16-35mm f/4 VR is an excellent optic, but is weakest at 35mm where we’ll be testing.
• The  Nikon 24-120mm f/4 VR is another excellent optic, and is at its strongest around 35mm-70mm.

Why did I not choose the Nikon 35mm f/1.8, Zeiss 35mm f/1.4, Rokinon 35mm f/1.4, Schneider Cine-Xenar 35mm, or whatever else you’re going to tell me you wanted to see? Because this isn’t a 35mm lens test. It’s an attempt to use some lenses that I know will generate different MTF 50 charts to see how easily you can actually see (or not see) the difference between them.

I tested one sample of each lens, but, of course, chose them from sets that had been recently lab-tested to be certain they were all reasonably good copies. Even then, I tested a second Nikon 16-35 and Nikon 35mm f/1.4, because fanboys love to scream ‘bad copy’ when their favorite lens doesn’t do quite as well as the others. The second copies were no different, and again, they were selected because they were known to be at least average from a pool of 50 copies.

Let’s Test Some Lenses

First I’ll start with a table showing peak center, weighted average, and average near-corner resolution for each lens. The numbers are MTF 50 in line pairs / image height. I should mention these are not my usual average of 50 copies; these are the specific results for one copy of each lens used in this test. (Each copy was, of course, pretested to make certain it was a good copy compared to our database of known values for each lens.)

Lens	Aperture	Ctr	Lens Avg	Corner Avg
Nikon 16-35mm	f/4	946	715	400
Nikon 16-35mm	f/5.6	969	781	460
Sigma 35mm	f/1.4	973	778	480
Sigma 35mm	f/2.8	1171	961	575
Sigma 35mm	f/4	1218	1016	800
Sigma 35mm	f/5.6	1160	1025	825
Nikon 35mm	f/1.4	740	607	425
Nikon 35mm	f/2.8	1020	830	590
Nikon 35mm	f/4	1065	945	775
Nikon 35mm	f/5.6	1135	985	840
Nikon 24-120mm	f/4	1100	883	460
Nikon 24-120mm	f5.6	1100	930	560

The numbers run together a bit in a large table, so let me summarize a few points:

• Except for the Nikon 35mm f/1.4, all of the lenses are maximally sharp in the center by f/4. The Nikon lenses all sharpen their corners further at f/5.6 compared to f/4,but the Sigma doesn’t, at least not significantly. (See — numbers do have some value for real-world photography.)
• At f/1.4 the Sigma has significantly higher resolution than the Nikon 35mm f/1.4. It’s still a bit higher at f/4 (although I’m not sure that slight difference is significant), and they’re about even at f/5.6.
• The 24-120mm seems to outperform the 16-35mm at 35mm (not shocking, as I mentioned the 16-35 gets a bit weaker right at 35mm).
• At f/4 and f/5.6, the prime lenses are better in the corners than the zooms, but not much different in the center.

It might be simpler to look at one of the charts I generally use for presenting Imatest data. The center MTF 50 is on the horizontal axis, weighted average MTF 50 on the vertical, both in Line Pairs / Image Height. To keep it somewhat simpler I’ve just plotted f/4 (square) and f/5.6 (diamond) data for each lens.

Just looking at center resolution (horizontal axis), I’m fairly certain that at f/4 (squares) I could see a difference between the 16-35 and the Sigma 35mm, but could not see a difference between the Nikon 35mm and the Nikon 24-120mm. I’m not so certain that I could see a difference between the Sigma 35mm and those two Nikon lenses, but think I probably could.

Center Resolution Reality Check

Before I put up more test chart examples, let me graph just center resolution for all of the lenses to simplify things a bit. As always, I know some fanboys are going to spit biscuits about how sharp their Nikon 35mm f/1.4 is wide open. This is a good copy, right at average in test results of the 50+ copies we have in stock. At f/2.8 it’s resolving very well indeed, and its edges and corners are excellent, but it’s not designed to win center resolution championships wide open.

 So let’s look at images of the ISO 12233 chart shots of the lenses shot at f/4. The worst resolves at 945 lp / ih (and worst isn’t a good term, that’s still very, very good resolution). The best resolves at 1220 lp / ih, and the other two just over 1100. Can we see the difference in a test chart shot? Here are 100% crops — to save a little bandwidth I didn’t include the Nikon 35mm f/1.4 image – it’s indistinguishable from the 24-120mm. PLEASE NOTE – the Sigma lens was shot at f/4, just like the other two lenses in this example.

I think this shows pretty well that different numbers (if they’re sufficiently different), are something you can detect with a bit of pixel peeping. But it’s not glaringly obvious.

One thing I suggest you do right now: downsize the image to 50% instead of 100% (which most people suggest as a more reasonable approximation for viewing an actual print). You can probably still see the difference. Now downsize it to 12.5% (about online jpg size). Can you see the difference now? If you can, you need a medication increase.

Now, here are three more center ISO 1223 charts. Before I tell you what they are, see if you can rank them from highest to lowest resolution.

If you said the bottom image is just a bit sharper than the top image, you’ve got good eyes. It is, but just a bit. Both are clearly better than the center image, though, aren’t they?

From top to bottom the crops are: Sigma 35mm at  f/1.4 (973 lp/ih), Nikon 35mm at f/1.4 (740 lp/ih), and Nikon 35mm at f/4 (1065 lp / ih). If you go back to the blue bar graph and look at the numbers, we’re comparing the lowest resolution to a couple of the mid resolution lenses. You can scroll back up and compare these to the others.

So what is my summary at this point? A difference of roughly 20% in MTF 50 in the center of the lens is readily visible in test charts. A difference of 10% is visible but you have to really look closely to see it. When it gets down to someone saying “this lens at 950 line pairs / image height MTF 50 is better than that lens at 900 lp/ih – well, that’s getting silly.

How About a Few Corners?

You might want to scroll back up and look at the computer generated ISO 12233 corner image, just to refresh yourself about what ‘perfect’ would be. With real lenses, we don’t often get close to perfect in the corners. This section will also answer a question I get asked fairly frequently — why don’t I post corner numbers and instead use average lens numbers? The answer is because corners are about a lot more than MTF 50.

Just to refresh, all of our lenses have corner resolution from 400 line pairs / image height wide open.  The Sigma and Nikon 35mm f/1.4 both give near-corner resolution over 800 lp / ih when stopped down. At f/5.6 the 16-35mm and 24-120mm have resolutions of 460 and 560 respectively. So what does that look like? Here are lower left corner images from ISO 12233 shots for a representative sample, and I put the MTF 50 numbers in blue on each sample so you don’t have to look back and forth.

Let’s talk about the real things first, before the ‘gotcha’. I think you can all tell the Nikon 35 at f/5.6 is a bit sharper than the two zooms. My eyes don’t think it’s twice as sharp, which the MTF 50 numbers sort of insinuate it is. I think I can tell the difference between the two zooms, pixel-peeping with this 100% crop. At least I think I can. But it’s a lot closer than I would have expected given the 100 point difference in MTF 50.

But I know what you’re thinking, “What’s up with the 35mm at f/1.4 looking so soft?” If we can hardly tell the difference in the two zooms, shouldn’t the 35mm look about the same as the 16-35 at least?

Well here’s a great example of MTF numbers not telling the whole story. The Nikon 35mm f/1.4G has both longitudinal and lateral chromatic aberration in the corners wide open (look along the edges of the thick square where the lines join it, or the edges of the “5″s). The two f/4 lenses, because they’re f/4, don’t really have these issues so they look better, even though the MTF numbers are roughly the same.

There are other things that affect corners a lot in addition to MTF numbers: coma and other aberrations, distortion (you can fix it in post, but you’ll loose some resolution doing so), vignetting, etc. And finally, I should mention the ISO 12233 chart isn’t quite the corners, although it’s fairly close.

So what is my summary about corners? Well, a big MTF difference is a good thing. Like the numbers suggest, the Nikon 35mm prime at f/5.6 is clearly better than either of the zooms. But in the outer parts of the image, MTF is not the only thing, not by a long shot. A smaller MTF difference, even a 20% (100 lp / ih) difference, may not be a good indicator of  which lens has better corners.

There is another 20% difference in MTF corners, though, that I want to look at. What if the difference is between corners on the same lens? In that case the corners, in theory at least, should have the same vignetting, aberrations, etc.

Demonstrating a Problem

Imatest gives us a lot more numbers than the averages I put in the graphs and that gives us an opportunity for another nice example. The Sigma 35mm f/1.4 we used for this test has an upper left (image) corner that’s a bit softer than the others as shown in the more complete Imatest printout below. (Notice the horizontal / vertical resolutions of 399 / 365 compared to around 500 in the other corners). It’s also a bit softer in the right lower (image) midrange than the other midranges. In other words, it’s very slightly decentered.

The question is whether this slight degree of decentering is really visible? (Trust me, this is slight. A really decentered lens would have areas of resolution under 200 lp/ih and much more astigmatism.) Checking SQF numbers tells me the difference should not be visible in an 8 x 11 print. But is it visible with a bit of pixel peeping?

I don’t think there’s any question the left upper corner softness is noticeable with some pixel-peeping. I’m much less certain how noticeable it will be in a real photograph, particularly given it’s location. But certainly it might be. I did check and the difference between the 4 quadrants was about the same at f/4 and f/5.6 — all were sharper, of course, but there was still a noticeable difference.

This does provide a great demonstration of why I continually harp that checking center resolution is no way to test your lens. This copy of the Sigma lens is outstandingly sharp in the center. Only by comparing the 4 outer quadrants could you possibly tell something is a bit off with this lens. By the way, because I know it will come up in comments, this is something we see commonly in wide-aperture primes of all brands. It’s no more frequent with Sigma than Nikon or Canon.

Summary

There are three kinds of people: those who can count and those who can’t.   – Author Unknown

I’m a geek and spend most of my days testing lens resolution; although mostly to detect copies that are out of sorts. I also test newly released lenses for MTF and resolution, too. Numerical results like Imatest measurements are invaluable for that kind of work. They give us reproducible data that we can use to set standards and compare things.

I think, though, that even just showing images of test charts illustrates some of the limitations these numbers have.

In the center of the lens, where aberrations are minimal, MTF50 correlates pretty well with perceived sharpness as long as we don’t try to split hairs.  A difference of 200 line pairs / image height or roughly 20% is clearly visible. A difference of 100 line pairs / image height or 10% in total resolution is visible but probably requires side-by-side comparison. And that’s in a lab shooting test charts and pixel peeping. I’m interested to see what the difference in photographs is after postprocessing.

In the outer 1/3, though, the numbers are less meaningful. Sure, a big difference in MTF is obvious. But things like vignetting, aberrations, and field curvature may make a much larger difference in how the image looks. It’s nice to know the MTF, but I wouldn’t make a purchase on just that basis, ever.

And, of course, we still haven’t even considered real-world things like exposure, focus accuracy, and vibration. I’ve mentioned before autofocus isn’t nearly accurate enough to make MTF measurements, we require careful live view and even then focus bracket. That and other real-world factors are going to have a huge influence on images above and beyond any MTF differences.

Uwe has a follow up with some actual images shot with theses lenses, properly postprocessed from raw files on his website in the follow up article, Seeing the Numbers with Real World Images.

Then, perhaps, I’ll reevaluate our testing and see if other numbers beside MTF 50 might give more information when we evaluate lenses. That seems to be the standard among those who test lenses, although I’ve not seen good proof of why it’s the best number.

Roger Cicala and Aaron Closz

Lensrentals.com

February, 2013

It’s not  falling that hurts, it’s the sudden stop at the end.   

-Author Unkown

Understandably, people are concerned with whether a new lens they buy is functioning properly and optically within specifications. One thing I harp on is that we also need to watch our lenses over time. Bumps, drops, and normal wear-and-tear can affect a lens optically.

A fair number of pro photographers take advantage of annual ‘clean and check’ services to help keep their lenses in the best possible shape, but other people never find this kind of thing necessary. As usual when there are two different schools of thought, both are correct. Most people don’t ever need to worry about an annual check-up, but it sometimes is worthwhile.

I’m going to use, as an example, some of our data from Canon 70-200 f/2.8 IS II lenses. I choose this lens for a couple of reasons: 1) we have a lot of copies and 2) it’s a high-resolution lens so getting a bit out of sorts may not be apparent at a glance. This lens is NOT chosen because it gets out of adjustment more often than others; it’s about average in that regard.

Let me also make clear that chances are an individual may never see this happen to one of their lenses. I’m always amused when I write about something like this and some fanboy immediately replies ‘my lens never had a problem.’ That’s cool. I’ve got thousands of lenses that never had problems. But about 10% of rental lenses do get out of sorts optically. While rental lenses get abused a lot more than individuals’ lenses, it’s inevitable that this will happen to some lenses in the real word, too.

It’s also fairly easy to fix, although it will usually require a trip to an authorized repair center to get it done.

A Lot of Imatest Data

We test all of our lenses regularly, by photographing test charts, by using Imatest, and more recently by also testing on an optical bench. I’ve summed up the effect of a year’s worth of abuse on slightly over 200 copies of the 70-200 f/2.8 IS II in the graph below. When we put these lenses into service all of them had acceptable results – that is they started the year within the red box.

In the graph, I’ve plotted each copy’s worst results during a year’s heavy rental use. (The graph, for those who aren’t used to my articles, shows MTF 50 in the center of the lens along the horizontal axis, and weighted average MTF 50 on the vertical axis, measured in line pairs / image height from raw images.) I’m showing results at 200mm. The results at 70mm were similar.

It’s pretty obvious that 26 copies (a bit over 10% of all copies) lost their optical edge at some point during a year of heavy use with a lot of shipping tossed in.

How Significant a Problem?

How much difference does it make optically? Looking at these graphed numbers you would probably think not too much. You’d probably be right.

First, if you look at the vertical red line I drew, every copy that was out of the box but to the right of the red line still has normal center sharpness. That’s not too surprising; many types of decentering affect the corners and edges but have little effect on the center. Since many people are using a 70-200 f/2.8 lens for action shots, shooting centered subjects, they may not notice soft edges or corners.

Even many of the other copies might pass routine inspections. A resolution of 700 / 500 line pair per image height is as good as a lot of lenses get on a good day. Unless you were comparing copies of the 70-200 f/2.8 IS II you might not notice a problem with copies that were resolving 700 /550 line pairs per image height. If you took careful test chart images you’d see that there was astigmatism (horizontal and vertical lines aren’t resolved equally) or that one area of the lens didn’t resolve as well as others.

The 4 lowest lenses on the graph had obvious problems that anyone would have noticed, although one of those 4 still remained sharp in the center. The 6 lenses above those but well to the left of the red line all had a soft side or corner that you would have noticed if you checked sides and corners. The others you might, or might not, have thought had a problem if you shot with them extensively.

If It’s Broke, You Can Fix It

If, as I claim, the lenses outside the box aren’t as good as they ought to be, then it should be possible to improve them. Aaron and I performed optical adjustments on each of the ‘bad’ copies and the results are shown in the graph below.

If you look carefully, you’ll see there is one fewer red dot than there were blue dots. One copy is off at Canon because it needed more help than we could give it.

As far as causes among the 25 lenses we adjusted, three had broken collars (the nylon bearings that hold the elements in their helicoid track). The others simply had an element that required recentering, tilting, or spacing.

Many of these elements slide up and down in helicoid tracks during focusing and zooming; wear and tear can eventually move them a tiny bit and rental lenses are heavily used lenses. Other elements may shift a bit with a bump or jar, affecting the optics of the lens. A tilt or decentering of one element by less than 1 millimeter is enough to devastate the optics of a lens.

Just for another look, I’ve put the ‘after adjustment’ values back into the first graph. Now that’s much better!!

A Couple of Thoughts

Does this matter to any of you? It will to at least a few of you at some point in your photographic journey. If you have a lens that seems to not be quite as sharp as it once was, a trip to factory service for a makeover is probably worthwhile. Remember, these changes are fairly subtle, often affecting only the corners or edges, or just increasing astigmatism.

I suspect sometimes people don’t really notice their lens is slightly decentered. They just don’t like the pictures it makes quite as much as they once did and use it less frequently. People may just decide they’ve gotten tired of that focal length or that particular lens without really checking to see if it’s slightly out-of-spec. Lenses with astigmatism will do that, particularly. The images just don’t quite look right but it’s hard to put your finger on why.

I’ll also answer a couple of questions I suspect some people will be asking. The first will be: Can you do this at home? I don’t think so, at least not on a lens like this. Optical adjustment on this lens, for example, requires disassembly, partial reassembly leaving off part of the barrel to expose the adjustable elements, then some careful time with a lens test projector or other optical testing equipment (shooting test charts won’t cut it for this), and finally reassembly.

That may also answer the question I see online every so often, “Why did they charge so much for an optical adjustment”? It takes us 1.5 to 4 hours to adjust one of these lenses. I’m sure the factory service center can do it faster, but it’s still a fair bit of work.

The other question someone will ask is: Can adjust the lowest ‘passing’ lenses to be better and bring them all up to the top scores? To some degree it’s possible. But remember adjusting to be better at 70mm may cause problems at 135mm or 200mm. Adjusting them to be great at 20 feet focusing distance may adversely affect infinity. It’s a balancing act, trying to get the best resolution overall.

A saying from one of my previous careers is really appropriate to lens adjustments — Better is the enemy of good enough. I can’t tell you the number of times we’ve decided that a lens was pretty good, but maybe we could just tweak that upper right corner at 70mm a bit. After we ‘improve’ that corner at 70mm, we spend the next 3 hours desperately trying to fix the center at 200mm, which was ruined by that 70mm corner adjustment.

Roger Cicala and Aaron Closz

Lensrentals.com

February, 2013

A teleconverter spreads out the light leaving the lens so that only the center portion reaches the sensor. The result is the focal length of the lens seems longer (the image is magnified), but at the cost of reducing the amount of light (effective aperture) of the lens. The Speed Booster compresses the light leaving the lens onto a smaller image circle. This makes the focal length seem shorter and actually increases the amount of light reaching the sensor.

The EF to NEX Speed Booster, for example, changes the effective focal length x 0.71, and increases the maximum aperture by 1 stop. A Canon 50mm f/1.2 lens effectively becomes a 35.5mm f/0.9 NEX lens, for example. Videographers all over the internet were singing Hosannah and laying palm leaves along the path of it’s introduction.

I went into my office, shut the door, and sobbed quietly for a while. Why, you ask? I’ll tell you why. About once a day, we get an email saying somthing like, “I just rented a Canon 5D Mk III and shot video of my daughter’s high-school graduation. My footage doesn’t look anything like Vincent LaForet’s. Obviously the camera was defective and I want my money back.”

This adapter was, I thought, going to result in another 50 emails saying, “I just shot video using a generic 50mm f/2.0 lens I bought on eBay with the Speed Booster adapter and Sony NEX VG20 I rented from you. The footage doesn’t look anything like the footage shot with a Zeiss 35mm T1.5 Super Speed shot on a RED Epic. Obviously the equipment is defective and I want my money back.” I know, like I know the sun is going to rise in the east tomorrow, that you don’t put some more glass between a camera and a lens and get a better image.

But marketing hype will be marketing hype and people who want to believe in magic will believe it – and be disappointed when the magic doesn’t happen.

But Then, There Came a Ray of Hope

Then I flipped over to Metabones’ white paper on the Speed Booster and spit coffee. The primary designer of the adapter is Brian Caldwell. If anyone could make optics do magic, he could. He designed, for example, the Coastal Optics UV-VIS-IR Macro lens, an amazing thing that is the gold standard for forensic macro photography. So I read the white paper carefully and it made perfect sense. The White Paper explained how:

1. The Speed Booster introduces zero (none, nada) spherical aberration, even with an f/0.9 output. That’s amazing. The very complete graphics in the White Paper do show it adds a bit of astigmatism and distortion, though.
2. Where teleconverters magnify lens aberration, a focal reducer would reduce aberations basically because it would shrink them.
3. The adapter is physically smaller than a standard, non-optical EF to NEX adapter.
4. Corner illumination is improved.

After reading the white paper, I became convinced that these things were true. And these are all good things.

There was an additional claim made in the white paper, that MTF (modulation transfer function -  acutance and resolution) was also improved. This one I struggled with. To be blunt, I found this section to be, shall we say, selective, in the comparisons made. I was left with the feeling that it might be using some very specific examples to suggest a general conclusion. The section was a bit more carefully worded than other parts of the white paper, and some information in the graphs, that didn’t quite agree with the claim, was downplayed in the text of the section.

I was willing, however, to be convinced that an FX lens mounted to the Speed Booster would have better corner resolution mounted to a Micro 4/3 than the same lens mounted to a full-frame camera – after all, those Micro 4/3 corners are a lot closer to the center of the image. And that it may, or may not, have better corner resolution mounted to an NEX camera compared to a full-frame camera.

So after my research, my impression was this will be, at least a very useful tool. It may be nearly as good as people hope it will be. In other words, it will, like all other imaging gear ever made, follow Roger’s Law of New Product Introduction (pathway A).

Let’s Do Some Testing Boys and Girls!

OK, first and foremost, this is not going to be a bunch of video samples. That’s not what I do. I’m a testing geek that writes words and makes graphs. But for those who have bravely come this far without a picture to ease the heavy burden of reading, let’s have a look at the Metabones Speed Booster.

It’s a nice looking bit of kit – solidly made and well put together. There’s a solid mount for tripods or shoulder mounts underneath. It mounts with a most satisfying thunk and clamps tightly to camera and lens. The optics are close to the surface, though, and some care in handling will be necessary to make sure they don’t scratch.

Let’s Shoot a Few Test Charts

There are a lot of ways to compare lenses with the Speed Booster and not a lot of time, so I tried to choose things that seemed practical. Or cool. Or both.

Starting with cool, I mounted a Canon 14mm f/2.8 II lens, which with the adapter should give us the equivalent of a 10mm f/2.0 NEX lens. That sounds cool to me. Plus I thought we should look at things as extreme as people are likely to get.

We’ll start by comparing simple shots of an ISO1223 chart shot with the lens on a plain adapter and on the Speed Booster. To even things up I moved my position so that the chart filled the image each time, so that we could compare resolution directly. First an overall picture of the chart, followed by center and near corner crops with each adapter.

Canon 14mm f/2.8

Again, the images were from different distances so that the chart filled the image with both shots. It’s not an optically critical test, but I’d call it a complete success for the Speed Booster. Even spotting the original image 1 stop of light, there’s no significant difference in resolution to my eye.

The change in perspective is impressive. These are shot from the same location with the Canon 14mm f/2.8 mounted to a straightforward adapter first and the Speed Booster second.

Canon 50mm f/1.2

That was pretty impressive, now lets stress things a little bit more. The Canon 50mm f/1.2 lens brings a few aberrations to the table and with the adapter will be an f/0.9 equivalent. I can’t think of anything that would stress an adapter more than f/0.9. Again, I’ll reposition myself so both shots fill the frame with the chart. We’ll compare the 50mm f/1.2 on top to the Speedboosted 35.5mm f/0.9 below.

Again, these are not critical tests, but are carefully lined up, best focus of several shots. And again, the Speed Booster comes out very well.  It may be the illumination boost but the acutance in the center, at least seems a little better with the Speed Booster. I would draw your attention, though, to the difference between vertical and horizontal lines in the corner crop of the Speed Booster image. That’s not an artifact of the shot or alignment. With this lens and camera, at least, the astigmatism seems to be showing up a bit. I also note that the image looks oversharpened, but it’s an unsharpened JPEG and this appearance was consistent on multiple shots.

 Canon 135 f/2.0

I wanted to try the other extreme, and made the assumption that this would be about the longest focal length people would want to use the Speed Booster with. I may be wrong about that, but was running out of time today.

Again, if there’s any deterioration in image quality with the Speed Booster, even though it’s a stop of aperture wider, I’m having difficulty seeing it. I also don’t notice the astigmatism with this combination.

But let’s test the resolution a bit more critically.

Imatest results

We don’t have multiple copies of the Speed Booster yet, so this is what we did. We took a Canon 50mm f/1.2 lens and tested it on a Canon 5D Mk II camera. Then we tested that same copy on an NEX-7 using a standard adapter. Finally, we tested those same copies (camera and lens) with our Speed Booster.

Usually when talking about Imatest results I’m sampling dozens of copies and give you the average (mean) resolution in the center and a weighted average of all the test points on the lens. But usually we NEVER test lenses on adapters if we can avoid it. Why? Because even the very best adapter still introduces and extra variation in tilt and centering between the lens and the camera.

Let me word this more carefully because it’s important. When the imaging sensor is placed in the camera, it is carefully lined up to be completely parallel to, and centered with, the lens mount of the front of the camera. A tilt of 20 microns may be visible on a very wide angle lens. A tilt of 40 microns almost certainly will be visible. From repair manuals we know that the sensor can be made parallel  to the lens mount within a few microns so that’s taken care of.

But when the big, heavy lens mount rotates into the big, heavy camera mount, chances are it’s not accurate within a few microns. Let’s assume it’s getting close to the 20 micron limit, because we know with high-quality, wide angle lenses we can often see some side-to-side variation. Sometimes obvious with some pixel peeping, sometimes not at all, but frequently enough that I’ve assumed we’re getting close to tolerance.

This is one of those sources of lens-camera variation I talk about so often. Lens 12345 looks great on camera 54321 but not so great on camera 112233. It may tilt a bit more on that second camera mount.

When we add an adapter we’re adding another heavy duty mount and making it more likely there’s a bit of tilt. It’s rarely apparent (with a high quality adapter) at standard or telephoto ranges, but often can be detected with high resolution wide-angle lenses. It may cause no harm at all. It may create too much tilt. I mention all of this because it’s going to explain some of our test results.

Canon 50mm f/1.2 on a Canon 5D Mk II

On a Canon 5D Mk II and our lens shot at f/1.4, the MTF50 was 590 line pairs / image height in the center; 460 lp / ih averaged over the entire lens, and 265 lp / ih in its worst corner with a barrel distortion of 1.2%.

Canon 50mm f/1.2 on a Sony NEX-7

On the NEX-7 with a standard adapter shot at f/1.4, it resolved 625 lp / ih in the center, 485 averaged over the entire lens, and 210 in its worst corner with a barrel distortion of 0.825% (smaller sensor). The difference in the center and overall isn’t surprising – the NEX has a higher pixel density and is using the ‘sweet spot’ from the center of the lens. The fact that the lowest corner is worse is a bit surprising until you see the overall graph of results:

Notice the upper left corner is worse and the upper right better, and that the right side of the image resolves better than the left. The lens, which behaved very nicely on a Canon camera, is tilted when shot on this particular NEX-7 with this particular adapter. So, of course, we went and got another adapter. It tilted the other way. And we were out of time. From experience I can say the center resolution number is going to be accurate, the actual weighted average should be a few points higher and the worst corner about 275 or so rather than 210 (275 is the resolution on the less affected corner).

So it is with adapters. And before you scream that the adapter was bad, it wasn’t. The guys put those adapters on other cameras and lenses later and they were either dead even or tilted another way. It is what it is. Also remember we’re churning MTF 50 numbers. You need a big difference to be able to actually see the difference in a photograph, and an even bigger difference for it to affect video.

Canon 50mm f/1.2 on NEX-7 with Speed Booster

On the NEX-7 with Speed Booster adapter and shot at f/0.9, it resolved 720 lp / ih in the center, 410 averaged over the entire lens and 230 in the worst corner. Stopped down to f/1.3 the MTF 50 increased to 800 center, 510 weighted average, and 300 for the worst corner. As advertised, the MTF 50 increased compared to the same lens on no adapter.

However, barrel distortion increased to 1.9%. Remember, however, this is in effect now a 35mm lens, so that number isn’t as big a jump as you would think at first glance. Just to note, there was no sign of tilt with this adapter on this camera.

One thing to note – there was definitely a bit of astigmatism, with horizontal and vertical resolution quite different along the edges of the lens. One other interesting note – we measured primarily MTF 50 as this is the number we work with most frequently. We also checked the MTF 20 numbers and for these combinations the Speed Booster had a similar effect: slightly improved but with greater astigmatism.

A reminder for everyone again – we had one copy of the adapter to play with and limited time. But I’ll have to admit that it seems the folks at Metabones pulled off what they claimed: with the adapter a lens is wider, faster, and even a bit sharper.

A Few Images

There’s not a lot of photogenic material around the Lensrentals Lab, but how often do I get to shoot with a 35mm f/0.9 lens? I’ve included some 100% crop areas in the corners of the scaled down shots.

Now, to get completely subjective, there’s something about the images as far as photography that I don’t like, but it’s hard for me to put a finger on it. But if you look at the crops there’s a bit of a glow around highlight areas, both in-focus (in Sarah’s shirt) and out of focus (the rubber bands).

My first thought was perhaps shooting on an NEX -7 could be the issue, since we know that camera has had some problems with adapted retrofocus lenses. Shooting with a Canon 50mm f/1.2 may also be the culprit – that lens is, well, different. But we repeated the rubber band shot with an NEX-6 and then with both cameras and a Canon 85 f/1.2 and the Speed Booster. To my eye (and remember I’m a techie, so I don’t do subjective all that well), there’s a real tendency for highlights, in and out of focus, to bloom a bit at the widest apertures. It seems to go away by f/2.0 affective aperture.

I’d also add, for those who plan on using it, that the autofocus worked well as far as accuracy. Eventually. You won’t be catching any moving targets unless they are turtles, though. I doubt this is of great import to many people, though.

Conclusions

I think it was pretty obvious that I came armed for battle, ready to slam this product as some marketing overhype. I was wrong less correct than I might have been. The Speed Booster does what they claimed it would do, much to my shock and surprise. It creates a wider-angle, greater aperture lens while retaining resolution and acutance.

It does increase astigmatism a bit, although I doubt this will cause anyone problems unless someone is trying to shoot landscape photography with it. It also seems to create some highlight blooming at very wide apertures. Again, nothing that can’t be worked around and probably not something that will be noticeable with anything but the widest aperture lenses.

It is going to take a while and a lot of people experimenting before we find out what combinations of lenses and cameras are awesome with it, which are fairly good, and which fairly bad. They won’t all be the same. But I suspect most of them are going to be pretty good. And this is going to be a very useful tool. 

Most of the little foibles I’ve seen (including the part about adapter tilt) really only apply to photographers trying to tweek every drop of resolution out of their high-resolution sensor. Video, even 5k video, is more forgiving of a slightly weak corner or a bit of astigmatism.

Roger Cicala

Lensrentals.com

January, 2013