Anamorphic Lens Tests

Anamorphic cinematography, first dabbled with in the 1920s, was popularised by Twentieth Century Fox in the fifties as CinemaScope. Television was growing in popularity and the studios were inventing gimmicks left, right and centre to encourage audiences back into cinemas. Fox’s idea was to immerse viewers in an image far wider than they were used to, but with minimal modifications to existing 4-perf 35mm projectors. They developed a system of anamorphic lenses containing elements which compressed the image horizontally by a factor of two. By placing a corresponding anamorphosing lens onto existing projectors, the image was unsqueezed into an aspect ratio of 2.55:1, or later 2.39:1.

Since those early days of CinemaScope, anamorphic cinematography has become associated with the biggest Hollywood blockbusters. Its optical features – streak flares, oval bokeh and curved horizontal lines – have been seared into our collective consciousness, indelibly associated with high production values.

I’ve not yet been fortunate enough to shoot anamorphic, but I was able to test a few lenses at Arri Rental recently, with the help of Rupert Peddle and Bex Clives. Last week I wrote about the spherical lenses which we tested; our anamorphic tests followed the same methodology.

Again we were shooting on an Alexa XT Plus in log C ProRes 4444 XQ, this time in 4:3 mode, a resolution of 2048×1536. Since all of the lenses had a standard 2:1 anamorphosing ratio, the images unsqueezed to a super-wide 2.66:1 ratio. (This is because the lenses were designed to be used on 35mm film with space left to one side for the optical soundtrack.) You can see the full width of this ratio in the first split-screen image in the video, at 2:08, and in the second image below, but otherwise I have horizontally cropped the footage to the standard 2.39:1 ratio.

We tested the following glass:

Series Length Speed CF* Weight
Hawk V 35mm T2.2 30″ 5.6kg
Cooke Xtal 30mm T2.8 ? 3kg
Kowa Mirrorscope 40mm T2.2 36″ 1.15kg
Kowa Mirrorscope 30mm T2.3 ? ?

* CF = close focus

For consistency with the spherical lenses, we used lengths around 32mm, but in the anamorphic format this is a pretty wide lens, not a mid-range lens. We shot at T2.8, again for consistency, but I hear that many anamorphics don’t perform well wider than T4.

We were only able to test what Arri Rental happened to have on the shelves that day. The biggest and presumably most expensive was the Hawk V-series. Next  in size and weight was the Cooke Xtal – pronounced “crystal” – a 1970s lens based on the much-loved Speed Panchros. The smallest and lightest, was the Kowa Mirrorscope, with a list price of £1,200 per week for a set of four. (Sorry, I couldn’t find any pricing info for the others online.) Note that there isn’t really a 30mm Mirrorscope; to get this length you put a wide angle adapter on the 40mm. As this extra element decreases the optical performance, we tested it with and without, hence the two lengths.

Here’s the video…


Skin tones

Click on the image to see it at full quality.

To my eye, the Hawk has a fairly rich, warm skin tone, while the Cooke – as with the spherical S4 tested last week – seems a little grey and flat. The Kowa is inexplicably brighter than the other two lenses, which makes it hard to compare, but perhaps it’s a little cooler in tone?



Focus is more critical with anamorphic lenses than spherical ones. From a forum posting by Max Jacoby:

Anamorphic lenses have what is known as a “curved field of focus” that works similarly to the curved movie screens in some large Cinerama theatres. This is one reason that one needs to expose these lenses at a deeper stop. If one doesn’t, the curved field will not be covered by depth of field and either the edges or centre of the frame will be soft.

One day I’d like to re-test these lenses at a lower stop, T4 or T5.6, where they will all undoubtedly perform much better. But in this T2.8 test, on Bex’s face in the centre of frame, the Hawk V and the Kowa Mirrorscope 40mm – both almost a full stop from their maximum apertures – are clearly the sharpest of the bunch. The Cooke Xtal, which is wide open, is unsurprisingly softer. The 30mm adapter on the Mirrorscope completely destroys the image, not only making it very soft but also introducing colour aberration.

Now let’s look at the checkerboard at the side of frame and see if we can spot any differences in sharpness there…

It seems to me that the Kowa, both with and without the adapter, has a greater difference in sharpness between the centre and edges of frame than the the Hawk and Cooke. With the latter two lenses, the checkerboard is reasonably sharp, at least on the lefthand side, with some ghosting/blur visible towards the righthand side. The same thing can be observed on the chart in the flare tests at the end of the video.


Breathing & Bokeh

All of these lenses have a noticeable degree of breathe, which I suppose is to be expected from anamorphics. The Hawk V has roughly oval bokeh, the Cooke’s is more circular, while the Mirrorscope has interesting D-shaped bokeh.



The Hawk V doesn’t flare much at all, which is apparently due to the anamorphic element being in the middle of the lens, rather than at the front. The Kowa has a nice streak and glow around the light source, with a funky purple artefact on the opposite side of frame. But it’s the Cooke Xtal which provides the most classic lens flare, with a horizontal line across most of the frame and a partial star pattern around the source, despite the lens being wide open.

At the end of the video you can see how the flares develop on each lens as the light source moves horizontally across frame.



A bulging effect is very obvious on all of these lenses, due to the focal lengths being quite wide for anamorphic. Notice how at 40mm on the Kowa Mirrorscope this curvature of the image is significantly reduced.

It’s hard to compare the levels of distortion because none of the focal lengths are exactly the same, except for the Cooke Xtal and the Kowa Mirrorscope with the 30mm adapter on. The Cooke’s top right and bottom left corners appear to be stretched away from the centre relative to the other two corners. I suppose that strange and funky stuff like this is exactly why you choose vintage glass.

Interestingly, the Cooke’s image appears a little tighter than the Kowa’s, which combined with my inability to find any evidence online of the existence of a 30mm Xtal, leads me to suspect we may have been given a mislabelled 32mm.



When we got to the end of our spherical tests and started putting the anamorphics on, I was shocked by the drop in sharpness. But as noted earlier, this is because anamorphics really need to be used with a smaller aperture than the T2.8 I often shoot at. If I learnt nothing else from this test, I learnt that anamorphic needs more light!

I would love to put the Cooke Xtal’s lovely flares and general vintage look to good use on a period movie one day. The Hawk V would be a good choice if I wanted the anamorphic look with warm, dynamic skin tones. The Kowa system seemed a little cheap and cobbled-together, but could well be a good solution for anamorphic on a budget, as long as I stayed away from the 30mm adapter!

I hope you’ve found these tests useful. Thanks again to 1st AC Rupert Peddle, 2nd AC Bex Clives and Arri Rental UK for making them possible.






Anamorphic Lens Tests

Spherical Lens Tests

The other week I spent a day at Arri Rental in Uxbridge, in the Bafta Room no less, conducting various camera and lens tests. I’ve done a number a productions now where I wanted to test but there wasn’t the time or money, so for a while I’ve been meaning to go into Arri on my own time and do some general tests for my education and edification. An upcoming short provided the catalyst for me to get around to it at last.

Aided by 1st AC Rupert Peddle and 2nd AC Bex Clives, I tested a dozen lenses, some spherical, some anamorphic. Today I will cover the spherical lenses; next time I’ll look at the anamorphics.



We shot on an Alexa XT Plus in log C ProRes 4444 XQ at 3.2K. In the video the image has been downscaled to 1080P and a standard Rec.709 LUT has been added.

I set the Alexa to ISO 800 and lit Bex to a T2.8 using a 650W tungsten fresnel bounced off poly. For fill I caught a little of the spill from the fresnel with a matte silver bounce board on the opposite side of camera. I placed fairy lights in the background to observe the bokeh (out of focus areas) and turned on a 100W globe during each take to see what the flare did.

We shot all the lenses at 2.8 – the stop I most commonly use – and also wide open (compensating with the shutter angle), but the direct 2.8 comparison proved most useful, so that’s mainly what you’ll see in the video. We tested a single length: 35mm or the closest available to it.

What we didn’t do was shoot grey-scale or colour charts, or do any testing of vignettes or distortion. (The day after doing these tests, Shane Hurlbut, ASC published an Inner Circle post about how to tests lenses, so I immediately learnt what my omissions were!)

We tested the following lenses:

Series Length Speed CF* Weight Price
Leica Summilux-C 29mm T1.4 18″ 1.7kg £27K
Arri/Zeiss Master Prime 35mm T1.3 14″ 2.2kg £16K
Cooke S4 32mm T2 6″ 1.85kg £14K
Leica Summicron-C 35mm T2 14″ 1.3kg £13K
Zeiss High Speed
(a.k.a. Superspeed Mk III)
35mm T1.3 14″ 0.79kg £12K
Arri/Zeiss Ultra Prime 32mm T1.9 15″ 1.1kg £10K
Zeiss T2.1 32mm T2.1 24″ 0.45kg £4K
Canon 35mm T1.5 12″ 1.1kg £3K

* CF = close focus

Here’s the video…


Skin tones

Click the image to see it at best quality.

The Arri/Zeiss Master Prime and the two Leicas seem to have the most vibrant skin tones. To my eye, the Leicas have a slight creaminess that’s very pleasing. The Canon looks just a little cooler and less dynamic. I was surprised to find that the Cooke S4, the lens I’ve used most, appears to have a grey, flat skin tone compared with the Master Prime, Leicas and Canon. I would rank the Ultra Prime and Superspeed next, on a par except that the Ultra Prime has a noticeable magenta cast. My least favourite skin tones are on the Zeiss T2.1, which comparatively makes poor Bex look a little bit ill!

Some of the nuances will be lost in the YouTube and Jpeg compression, but this is a very subjective assessment anyway, so feel free to completely disagree with all of the above. Any of the differences noted above could be corrected by grading, to some extent . But remember that the lens is at the very start of the light’s journey from set to screen, and any wavelengths that don’t get through it are lost forever. It’s like fluorescent lamps with colours missing from the spectrum; you can’t put those back in in post.



I have to say, I’m unable to detect any difference in sharpness between the Master Prime, Cooke S4, Canon and Leicas. The Ultra Prime and Superspeed both look a hair softer, while the T2.1 is very soft.



Breathing is the slight zooming effect that you get with some lenses when you pull focus. Looking at 4:44 in the video you can clearly see the differences in breathing between the eight lenses. Because this part of the video is showing a crop of the bottom left corner of the image, the breathing manifests as a shift to the left (zoom in) as the lens is racked closer (goes soft) and a shift to the right (zoom out) as it’s racked deeper (goes sharp).

All the Zeiss lenses except the Master Prime have a significant amount of breath when seen in isolation like this, but not enough to be noticeable to an audience in most real-world situations. The Cooke S4 has a little bit of breathe, and the Canon a hair less. The Master Prime and the Leicas are rock solid.



Small points of light, when thrown out of focus, most clearly demonstrate the bokeh pattern of a lens. The shape of the bokeh is determined by the number of iris blades and the shape of those blades. Generally a circle is preferred, because it’s a natural shape, but for certain stories a more unusual shape might be appropriate. The shape of the iris changes with the T-stop, hence the T2.8 and wide open images above.

Immediately noticeable is the difference in the Cooke S4’s bokeh between wide open (circular) and T2.8 (octagonal). All of the other lenses have round bokeh at T2.8, apart from the Superspeed, which has heptagonal (seven-sided) bokeh.

It’s entirely subjective which bokeh you prefer. The only other thing I’ll point out is that the Canon’s bokeh wide open is very fuzzy, with noticeable colour aberration, though this may be due to the bright highlight rather than the defocusing.



Flare patterns also vary with aperture. The smaller the aperture, the more of a star effect you will get, as the light interacts with the corners in the iris blades. The Summilux shows this most clearly, with a pronounced star at T2.8 (two stops down from its maximum aperture) and almost none when wide open. The Cooke S4 also has a nice star pattern at T2.8. With the other lenses it’s much more subtle, and the Canon has almost none.



The real revelations in these tests, for me, were the Leicas. The Summilux in particular is a beautiful lens, with rich, dynamic skin tones, nice bokeh, no breathing, plus the bonus of nice star flares. I will definitely be looking to work with this glass in the future, although given the price tag that may be optimistic!

The Summicron also performed incredibly well, matching the more expensive Summilux and Master Prime in every respect except speed. I can see this becoming my new go-to lens.

The Master Prime of course produced a beautiful, sharp, clean image, but it lacks character. It might work nicely for science fiction, a drama requiring a neutral look, or something where filtration was being used to give the image character.

The Canon impressed me too – no mean feat given that it’s the cheapest lens we tested. With nice skin tones and attractive flares, I could see this working well for a romantic movie.

The Zeiss T2.1 did not appeal to me, with poor sharpness and cold, washed-out skin tones, so I would avoid it.

The Superspeed is a decent lens, but in most cases I’d plump for an Ultra Prime instead. Ultra Primes are certainly easier to work with for the 1st AC, and have proven to be a good workhorse lens for drama. (I shot Above the Clouds on them.)

The Cooke S4 has been my go-to glass up to now, and while it will probably remain my first choice for period pieces, due to its gentle focus fall-off, I’m excited to try some of the other glass in this test on other productions.

I’ll say it one last time: this is all subjective. Our visual preferences are what make every director of photography unique.

Tune in next week when I’ll look at the anamorphic lenses: Hawk-V, Cooke Xtal and Kowa Mirrorscope.




Spherical Lens Tests

Alexa ProRes ISO Tests

My Cousin Rachel

I’ve shot three features on Arri Alexas, but I’ve never moved the ISO away from its native setting of 800 for fear of noise and general image degradation. Recently I read an article about the cinematography of My Cousin Rachel, in which DP Mike Eley mentioned shooting the night scenes at ISO 1600. I deliberately set off for the cinema in order to analyse the image quality of this ISO on the big screen. Undoubtedly I’ve unwittingly seen many things that were shot on an Alexa at ISO 1600 over the past few years, but this was the first time I’d given it any real thought.

To my eye, My Cousin Rachel looked great. So when I was at Arri Rental the other week testing some lenses, I decided to shoot a quick ISO test to see exactly what would happen when I moved away from the native 800.

But before we get to the test footage, for those of you unsure exactly what ISO is, here’s an introduction. The more experienced amongst you may wish to skip down to the video and analysis.


What is ISO?

ISO is a measure of a camera’s light sensitivity; the higher the ISO, the less light it requires to expose an image.

The acronym actually stands for International Organization for Standardization [sic], the body which in 1974 combined the old ASA (American Standards Association) units of film speed with the German DIN standard. That’s why you’ll often hear the terms ISO and ASA used interchangeably. On some cameras, like the Alexa, you’ll see it called EI (Exposure Index) in the menus.

A common ISO to shoot at today is 800. One way of defining ISO 800 is that it’s the sensitivity required to correctly expose a key-light of 3 foot-candles with a lens of T-stop 1.4 and a 180° shutter at 24fps, as we saw in my Barry Lyndon blog.

If we double the ISO we double the effective sensitivity of the camera, or halve the amount of light it requires. So at ISO 1600 we would only need 1.5 foot-candles of light (all the other settings being the same), and at ISO 3200 we would need just 0.75 foot-candles. Conversely, at ISO 400 we would need 6 foot-candles, or 12 at ISO 200. Check out this exposure chart if it’s still unclear.

ISO is one of the three corners of the Exposure Triangle, well-known to stills photographers the world over. You can read my posts on the other two corners: Understanding Shutter Angles and F-stops, T-stops and Optical Density.

Just as altering the shutter angle (exposure time) has the side effect of changing the amount of motion blur, and altering the aperture affects the depth of field, so ISO has its own side effect: noise. Increase the ISO and you increase the electronic noise in the picture.

This is because turning the ISO up causes the camera to electronically boost the signals it’s receiving from the sensor. It’s exactly the same as turning up the volume on an amplifier; you hear more hiss because the noise floor is being boosted along with the signal itself.

I remember the days of Mini-DV cameras, which instead of ISO had gain; my Canon XL1 had gain settings of -3dB, +6dB and +12dB. It was the exact same thing, just with a different name. What the XL1 called 0dB of gain was what today we call the native ISO. It’s the ISO at which the camera is designed to give the best images.


ISO and Dynamic Range

The Alexa has a dynamic range of 14 stops. That means it can simultaneously record detail in an area of brightness x and an area of brightness times 2 to the power 14. At its native ISO of 800, those 14 stops of dynamic range are equally distributed above and below “correct” exposure (known as middle grey), so you can overexpose by up to 7 stops, and underexpose by up to 7 stops, without losing detail.

If you increase the ISO, those limits of under- and overexposure still apply, but they’re effectively shifted around middle grey, as the graphic to the left illustrates. (The Pro Video Coalition post this graphic comes from is a great read if you want more detail.) You will see the effects of this shifting of dynamic range very clearly in the test video and images below.

In principle, shooting at ISO 1600 is the same as shooting at ISO 800, underexposing by a stop (giving you more highlight detail) and then bringing it back up a stop in post. The boosting of the signal in that case would come right at the end of the image path instead of near the beginning, so the results would never be identical, but they’d be close. If you were on a bigger project with a DIT, you could create a LUT to bring the exposure up a stop which again would achieve much the same thing.

All of the above assumes you’re shooting log ProRes. If you’re shooting Raw then everything is simply recorded at the native ISO and any other ISO you select is merely metadata. But again, assuming you exposed for that other ISO (in terms of iris, shutter and ND filters), you will effectively get that same dynamic range shift, just further along the pipeline.

If this all got a bit too technical for you, don’t worry. Just remember:

Doubling the ISO

  • increases overall exposure by one stop,
  • gives you one more stop of detail in the highlights,
  • gives you one less stop of detail in the shadows, and
  • increases picture noise.

Halving the ISO

  • decreases overall exposure by one stop,
  • gives you one less stop of detail in the highlights,
  • gives you one more stop of detail in the shadows, and
  • decreases picture noise.


The Test

I lit the subject, Rupert “Are You Ready?” Peddle, with a 650W tungsten fresnel bounced off poly, and placed a 40W candle globe and some LED fairy lights in the background to show highlight clipping. We shot the tests in ProRes 4444 XQ on an Alexa XT Plus with a 32mm Cooke S4, altering the shutter angle to compensate for the changing ISOs. At ISO 400 the shutter angle was maxed out, so we opened the lens a stop for ISO 200.

We tested five settings, the ones corresponding to a series of stops (i.e. doublings or halvings of sensitivity): 200, 400, 800, 1600 and 3200. I have presented the tests in the video both as recorded in the original log C (mislabelled as S-log, sorry!), and with a standard Rec.709 LUT.

You’ll need to watch the video at full-screen at 1080P to have any chance of seeing the differences, and even then you might see the compression artefacts caused by the noise more than the noise itself. Check out the stills below for a clearer picture of what’s going on. (Click on them for full resolution.)



To me, the most important thing with every test is how skin tones are rendered. Looking at the original ProRes of these comparisons I think I see a little more life and vibrance in the skin tones at lower ISOs, but it’s extremely subtle. More noticeable is a magenta shift at the lower ISOs versus a green shift at higher ones. The contrast also increases with the ISO, as you can see most clearly in the log images.

At the lower ISOs you are not really aware of any noise in the picture. It’s only at ISO 1600 that it becomes noticeable, but I have to say that I really liked this level of noise; it gives the image a texture reminiscent of film grain. At ISO 3200 the noise is quite significant, and would probably be unacceptable to many people.

The really interesting thing for me was the shifting of the dynamic range. In the above comparison image, look at the globe in log – see how it starts off as one big white blob at ISO 200 and becomes more detailed as the ISO rises? Now look at the dark wall around the globe, both here and in the previous image – see how it subtly and smoothly graduates into darkness at the lower ISOs, but becomes a grainy mess at the higher ones?

I can see an immediate benefit to shooting at ISO 1600 in scenes lit predominantly with practicals. Such scenes tend to have a low overall level of illumination, while the practicals themselves often blow out on camera. Going to ISO 1600 would give me extra exposure and extra detail in the practicals. I would be sacrificing shadow detail, so I would have to be a little more careful not to underexpose any faces or other important elements of the frame, but I can deal with that. In fact, I often find myself determining my exposure in these types of scenes by how blown out the practicals are, wishing I could open up a little more to see the faces better but not wanting to turn the lamps into big white blobs. Increasing the ISO would be the perfect solution, so I’m very glad I did this test to alleviate my ungrounded fears.

What about scenarios in which a lower-than-native ISO would be useful? Perhaps a scene outside a building with an open door, where the dark interior is visible in the background and more detail is required in it. Or maybe one of those night scenes which in reality would be pitch black but for movie purposes have a low level of ambient light with no highlights.

I hope you’ve found this test as useful and interesting as I have. Watch this space or subscribe to my YouTube Channel for the lens test.

Thanks to Rupert Peddle, awesome steadicam op and focus puller – check out his site at – for appearing in front of the lens. Thanks also to Bex Clives, who was busy wrangling data from the lens tests while we were shooting these ISO tests, and of course Arri Rental UK.



Alexa ProRes ISO Tests