This is a story about a lens family whose soft focus backgrounds were consciously designed, and the lens mount that made them possible.
For the last six months I’ve been working part time as a trainer for the ARRI Academy’s Camera Systems class. Every month or so I travel to either Burbank or Brooklyn and work with a c0-trainer on a 2.5 day intensive workshop covering all of ARRI’s current cameras. Recently we’ve added the Alexa LF to the lineup, and it’s been a treat to work with. To be honest, it’s not terribly different to the SXT when it comes to day-to-day operation. The real difference is in the optics.
This article covers two different, but related, subjects: why the Signature Prime design is so different to common film lens designs, and what effect this has on the out-of-focus portions of our images.
WHY A LARGER SENSOR MAKES SENSE
The SXT, Amira and Mini boast 3.4K sensors, and if you aren’t counting pixels they do the job very well. There are a huge number of TV shows and features shooting in 3.4K ARRIRAW and 3.2K ProRes, and I suspect that no one has noticed that they fall .6 to .8K short on resolution. ARRIRAW compares very well to other raw formats for the simple reason that, other than log encoding—which only optimizes the tonal scale—most other raw formats incorporate some sort of detail compression. This means the codec is discarding detail that it hopes you won’t notice: very fine detail in general, and coarse detail in shadows.
Compared to highly compressed 4K footage, uncompressed 3.4K compares very well. For those who insist on full UHD (3.8K) resolution, though, an super 35-style ARRI sensor wasn’t going to make the grade.
ARRI won’t compromise on color and noise. The ALEV sensor incorporates 8.25 micron photosites, which are fairly large by modern standards. Some cameras utilize five micron or smaller photosites to fit as many onto an S35 sensor as possible. The smaller the photosite, the noisier it tends to be.
IT’S ALL ABOUT THE NOISE
Dynamic range is determined by two things: the point at which a photosite can’t collect any more photons (where it saturates, or “clips”), and how few photons it can collect without the signal devolving into noticeable noise. Small photosites require a lot more cleanup when it comes to noise. Larger photosites are naturally subject to fewer noise issues because the ratio of signal to noise is higher. And this makes a big difference when it comes to color.
Imagine a color image as being comprised of red, green and blue layers. We can “print” a beautiful, three-color image as long as each color layer is fairly distinct. If one layer is fuzzy, the image will suffer: hues will drift, color becomes less saturated, and the image may appear muddy. Color fidelity and noise are inextricably linked.
Given that ARRI’s goal is to make each camera consistent with every other ARRI camera, they realized that changing the photosite size wasn’t going to work out. That led to the necessity of confronting new problems, because 35mm film lenses are generally not great for digital capture.
Film lenses are compromises. The farther a lens sits from the film plane, the more it suffers optically—and for quite a long time lenses have had to sit at least 52mm from the film plane due to the fact that they had to dodge a spinning mirror.
The exit pupil is the image of the iris when viewed from the perspective of the sensor. It’s the point from which light rays appear to exit the lens. Because film lenses sit so far from the film plane it has never been simple or cost effective to design a telecentric, or partially telecentric, lens, because the greater this distance is the harder it is to manufacture such a lens. The advantage to a telecentric lens is that the exit pupil appears to be at the front of the lens, or even in front of it, with the result that the light exiting the lens strikes the film at less of an angle, especially at the edges of the illumination circle.
When the exit pupil is at the rear of the lens, the light rays at the edges of the frame strike the film at an extreme angle. This doesn’t affect film emulsion, but sensors are a different matter.
A photosite has depth, so photons approaching at a steep angle may not find their way inside. This results vignetting. Photosites are often covered by micro lenses that attempt to direct photons into photosites to increase their light-gathering efficiency, and striking these lenses at an angle can cause certain wavelengths of light to scatter. This causes color fringing. Telecentricity solves all these problems nicely: photons strike the sensor either dead on or at much reduced angles, naturally improving efficiency. Color fringing largely disappears. Surprisingly, focus breathing also disappears.
It’s easy to tell if a lens is telecentric or not. If you look into the back of it and the iris appears to be well forward of the aperture ring, and possibly at the front or beyond the front of the lens, it is at least partially telecentric. (A fully telecentric lens would place the exit pupil at infinity from the sensor’s perspective, and this plays strange games with depth perspective. This works well for machine vision applications, but doesn’t lend itself to visual storytelling.)
The only caveat is that telecentric lenses work best with shallow flange depths. The standard 52mm flange depth isn’t optimal at all.
REINVENTING THE LENS MOUNT
ARRI engineers decided that adhering to a flange depth standard meant for lenses mounted in front of a spinning mirror shutter didn’t make much sense without the shutter, and since ARRI invented the PL mount more than 30 years ago they opted to replace it. They settled on a new flange depth standard of 44mm, which made partially telecentric lenses easier to build and design. This also left room for an S35-to-LPL adapter that placed film lenses 52mm from the sensor plane. Rather than making older lenses obsolete, they left room for the ones that could cover a large format sensor.
Now, though, a bigger problem arose: light emanating from an exit pupil at the front of a typical lens can’t reach the edges of a large format sensor because the PL mount itself is too small. The solution: build a wider PL mount.
PL stands for “Positive Lock.” Rotating the locking ring pulls the base of the lens tight against the lens mount, preventing any motion whatsoever. LPL stands for “Large (format) Positive Lock.”
The rear of a Signature Prime is huge, and this prevents the lens mount itself from shadowing a large format sensor. The natural vignetting of a film-style lens can be attractive but it’s also, physically, an image grading choice, which is always destructive. A more evenly illuminated sensor allows for greater flexibility in post. (Whether this is desirable or not is an artistic question, not a technical one.)
Enough talk. Let’s look at some images.
ULTRA PRIME VS. SIGNATURE PRIME
The reason I brought up the ARRI Academy is because it has offered me the only opportunity so far to shoot footage with an Alexa LF. We’ve started offering a new class specifically for the LF, and as the key differentiator between this camera and the others is sensor size I thought it would be informative to shoot a clip using an LPL-based Signature Prime and a PL-based Ultra Prime. The Signature Primes come in odd focal lengths as they are meant to pair with common S35 focal lengths, and in this case we employed a 47mm Signature Prime which is meant to be the equivalent of a 32mm S35 prime. (The “crop factor” is approximately the S35 focal length times 1.4.) For our purposes, though, we found that backing the camera up a bit and using a 50mm Ultra Prime gave us a better perspective match.
When ARRI set out to devise its own lenses (and create their own “signature” look) they did two things: they designed the lenses by looking at faces instead of charts, and they paid as much attention to the out-of-focus elements of the image (the bokeh) as they did to the focus plane.
The Ultra Prime has an interesting, almost random quality to it. In this shot, containing a curious little character we found on a shelf in the ARRI Burbank creative space set, the difference in bokeh characteristics is extremely noticeable (thanks to some holiday lights draped across the background). The Ultra Prime has an interesting look that’s a bit smeary, dreamlike and random: the chess piece highlights take on a “swirl” effect while highlights at the very edge of the frame become blobs.
Also, the highlight circles themselves show color fringing that appears to be due to light striking the sensor, or microlenses, at off angles. Notice that the inside edge of those circles appear to be one color while the outside edge appears to be another, and while this effect is most noticeable toward the edges of the frame we still see this kind of chroma distortion directionality toward the center of the frame.
The Signature prime, on the other hand, shows no chromatic distortions, likely due to its telecentric design. And while both lenses have a “swirly” feel to the bokeh, this is likely an accidental feature on the Ultra Prime vs. an intentionally designed feature on the Signature Prime. ARRI was asked to give the bokeh a compelling quality, and they obliged. (This effect dissipates as the lens is stopped down. Both lenses were used wide open—T1.9 for the Ultra Prime and T1.8 for the Signature Prime—and closing the Signature Prime to T2.8 or smaller offers the option of traditional round bokeh.)
Viewing highlights one at a time is interesting, but what’s more important is what happens when they are clustered together:
Both lenses show fairly well-defined bokeh with definitive edges, but the chromatic distortion around the Ultra Prime highlights makes it feel sharper to me. There’s a distinctive smearing effect that I don’t see in the Signature Prime, but the Signature Prime’s bokeh blends better to my eye. I also find the symmetry of the Signature Prime’s bokeh to be less distracting, as the consistent highlight shape brings a unified quality to the background. There’s a design principle, the Principle of Harmony, that states: “A collection of objects looks better to the eye if all the objects appear exactly the same.” That certainly seems to apply here.
Here’s something I don’t say all the time: “Let’s closely examine the cactus.”
The Ultra Prime bokeh is sharper than that of the Signature Prime. This could be due to not perfectly matching the framing between lenses because this was meant to be a quick look, not a scientific examination. Still, the sizes look fairly close in the wide shot. Once again, though, the chromatic fringing makes the cactus leaves look a bit sharper through the Ultra Prime. They look softer, and blend better, through the Signature Prime.
We rarely stare at dolls, cacti or chessboards in feature films or television, so let’s take a look at an actual human. One of the LF class attendees, Austin Sabado, volunteered to be our test subject. The class put him in the hot sun, placed a bounce card to one side, and photographed him in open gate mode using both a 75mm Signature Prime and an 85mm Ultra Prime. Both lenses were set to near wide open apertures and aimed through an ARRI FSND 2.1 neutral density filter. The camera was set to 5500K preset and I’ve done nothing more than apply the stock ARRI LogC-to-Rec 709 LUT in Resolve. The stop was slightly close down from wide open on both lenses, probably around T2 1/2.
The three things I notice right off are the quality of the bokeh, the contrast, and the saturation. The quality of the highlights in the leaves over Austin’s camera-right shoulder are very soft and delicate through the Ultra Prime, and are a little brighter and more defined through the Signature Prime. Still, I like the Signature Prime’s bokeh a little more because the brighter disks blend more easily, once again due to the lack of chromatic distortions around the edges of the disks.
This is most obvious in the car highlights at the bottom left of the frame:
The Ultra Prime disks have harder edges, and they blend poorly due to chromatic distortion around their edges and their differing shapes. The Signature Primes are much prettier.
The other thing to look at here is how much more dramatic chromatic aberration is around areas of contrast in the older lens vs. the newer lens. This is likely just a difference of lens coatings and optical improvements rather than an illustration of the different lens designs, but it’s still interesting:
It probably isn’t terribly fair to put a lens as old as an Ultra Prime against something as modern as a Signature Prime, but the comparison does illustrate the kinds of things one can look for when judging one lens family’s characteristics against another.
WHAT IT ALL MEANS
We pay a lot of attention to the quality of the in-focus image, but we don’t pay as much attention to the quality of the out-of-focus image. For me, one is as important as the other, because the sharp image we intend to capture almost always plays against a softer image that sets the background for the story being told. In an era where low light levels and large formats mean that bokeh literally sets the stage for every setup, it’s more important than ever to pay attention to what our lenses do to everything within the frame.
Film lenses create some interesting bokeh effects, but they are largely the result of a mismatch between their intended medium and the new medium that dominates nearly all of modern production. Filmmaking is about intentionality and choices, and both the new LPL mount and the Signature Prime family will certainly add to, and influence, those choices.
Disclosure: I have been paid by ARRI to conduct classes for the ARRI Academy. I was not paid to write this article. Any errors and omissions are my own.
Director of Photography