Silverbased

Many photographers got a good chuckle this week, when the dating site OkCupid posted an analysis of which camera settings make you romantically attractive.

OkCupid harvested all this camera information from the EXIF data embedded in their members’ photos. This means, unfortunately, that we can’t get numbers proving that being photographed on film is the sexiest of all—but I assure you it’s true.

Since I often sputter about how harsh and unflattering on-camera flash is, I’m tickled to finally see hard numbers spelling out how bad the damage really is. Having your picture taken with flash is equivalent to adding 5 or 10 years to your age, as far as your attractiveness goes. (Hmm, plotting my age on that graph, I fear I’d better be photographed in total darkness.)

The analysis also showed an effect where viewers preferred photos taken at wide lens apertures—notice all the green boxes for f-numbers f/2.8 and below? Of course, taking shots at wider apertures means shallower focus, allowing a photographer to de-emphasize extraneous background clutter.

But I think OkCupid misunderstood something about the f/stop effect. Point-n-shoot cameras and DSLR kit zooms generally don’t offer apertures of f/2.8 or larger. Only a photographer who went out and bought a different, more advanced lens will ever have “f/1.4″ show up in their EXIF data. Such people are more engaged, and presumably more skilled, as photographers—thus, they probably shoot somewhat more flattering portraits.

Another problem is that without knowing the sensor size and the subject distance, the f/number alone will not tell you how blurred the background actually was.

Selective focus—will it get you more dates?

Nonetheless, when Christian Rudder comments, “because the photos with the low f numbers feel more intimate and personal, they get a better viewer response” I do agree. A portrait where twinkling eyes are sharp, but beyond is a soft halo of blur… it does look very stylish and appealing—even romantic.

So will running out and buying that f/0.95 Leica Noctilux convert you into an irresistible sex bomb? (Albeit one with an empty bank account?)

Anything relating to Depth of Field inevitably turns out to be a bit more complicated than you might imagine. As I mentioned, f/numbers alone are not the whole story.

Now, Dr. Hubert Nasse of Carl Zeiss has helped us out tremendously here, by writing a detailed analysis (1.7 Mb PDF) regarding focus depth, background blur, and bokeh. It’s published in the Zeiss newsletter Camera Lens News No. 35. This document is a humbling read, for anyone who innocently believed they already understood depth of field.

But perhaps I can spare you the 45 pages of graphs and diagrams. Regarding the topic at hand, “how can I blur the background the most,” let’s jump straight to Dr. Nasse’s answer (on page 30):

If by “bokeh’ you mean principally the ability to be able to represent the  background as very blurred, soft and lacking detail, it is necessary to have an entrance pupil which is sufficiently large.

[Misusing "bokeh" just to mean "blurry background" is regrettably widespread today, despite my blog's futile protests],

And:

The decisive parameter for the quantity of the blurriness is therefore the physical size of the entrance pupil.

Hmm, what is this “pupil” thing Dr. Nasse is talking about?

Well, we all know that the pupil of your eye is the black part, where the light gets in. Likewise, a lens’s entrance pupil is just the apparent diameter of its clear opening, looking from the outside.

Entrance pupils of two lenses

The lens on the left is the one I used to make the portrait shown above. As you can see, the clear opening is quite large.

Imagine all the rays of light originating from one point on the subject which fall into that opening: They form an imaginary cone. The “fatter” that cone is, the bigger the disk of blurred light becomes in the un-focused parts of the image.

On the right side we have a very standard “kit zoom,” ubiquitous on today’s DSLRs. I set it to give the widest possible entrance pupil; but even so, that’s nowhere close to the diameter of the front element. So we see at a glance this lens can’t give us much background blur.

And point-n-shoot cameras, whose entrance pupils are practically pinholes, will be utterly hopeless. (Perhaps another reason OkCupid users found those cameras’ snapshots less appealing?)

Now, is this just because the left-hand lens can open all the way to f/2.0? The zoom as shown above is a couple of stops dimmer.

Not so fast. Entrance pupil diameter also depends on the focal length. Or to be precise, an f-number is defined as the focal length divided by the entrance pupil width, at a given diaphragm setting.

So different lenses can have different entrance pupil diameters, even when they’re all set to the same f/number. See here:

Three lenses, all at f/2.8: Focal lengths 24, 50, 100 mm

Each doubling (roughly) of the focal length requires a doubling of the entrance pupil diameter. If you’d like to click on the photo above to open a larger version, you can measure and confirm that yourself.

So even though all these are set to/2.8, it’s clear that the 100 mm will blur the background the most.

Thinking in terms of entrance pupils, not f/numbers,  helps resolve one classic paradox in understanding depth of field:

As you switch focal lengths, you must move nearer or farther away to keep your subject the same height in the frame. When you take this into account, and play with a depth of field calculator, suprisingly you discover that changing lens focal lengths has practically no effect on the DOF (as long as you maintain the same f/number).

Yet that seems nutty! Our common-sense experience tells us that for blurring distracting backgrounds, you’d always reach for the telephoto first.

The answer to the paradox is simple. Yes, the depth of sharp focus depends simply on the f/number. But the amount of blurring of distant backgrounds (say, 25 feet or more behind the subject) depends largely on the lens’s entrance pupil size. This the distinction Dr. Nasse was trying to clarify.

We now understand why many lenses considered classics for portraiture are both bright in f/ratio and longer in focal length. For pleasing portraits, I tend to reach for the first two lenses shown below: First, the Olympus Zuiko 85/2.0; next, the Canon new-FD 100/2.0.

Quartet of blur kings

Both remain sought-after gems from their respective lens lineups; on eBay you would expect to spend a couple hundred dollars for either. So alongside those, I’ve shown a couple of alternatives you might find interesting.

While lens #3 does not quite match the entrance pupil diameter of the Canon, it’s still pretty respectable. This is a Mamiya 110 mm f/2.8 lens, for their 645 medium-format system. And it only cost me $60 from KEH. As I’ve written before, medium format systems can give you nice background blur at quite an affordable price.

Even more entertaining is the fourth lens. My apologies that its engravings are dingy and hard to read—I estimate it’s at least 70 years old, maybe more. That’s a Kodak Projection Anastigmat, which “only” opens to f/4.5. But its ten-inch focal length means it has the widest entrance pupil of all. (10″ = 254 mm)

That one came from an eBay auction, where the word “projection” scared off every other buyer but me. So my low starting bid won it: Just eleven bucks, shipped.

I admit, it’s not very convenient to use a 1.5 pound, shutterless large-format lens. But oh! Isn’t the background blur lovely?

4×5 sheet film; Kodak 10″ Projection Anastigmat

Lets compare that to the blur from the plasticky DSLR zoom I showed you before. To make this a fair fight, I actually stopped down the old Kodak to f/5.6, the best the kit lens can manage when zoomed in.

Background blur with an undersized entrance pupil

Now I promise you: These are both f/5.6 shots, taken from the same distance. It’s just the entrance pupil diameter that changes the background blur. You need a longer focal length lens on the bigger 4×5 film format; thus at any given f/number, the entrance pupil will be larger.

It’s unfortunate that “entrance pupil” is such geeky-sounding optical jargon. If we could spread awareness of its importance, photographers would automatically know how to get that lovely, soft, selective-focus look.

Maybe we could just talk about “wide eyed” lenses instead?

After all, everyone knows—wide open eyes are sexy!

Imagine taking two disks of glass and rubbing them together, along with a slurry of abrasive grit. As glass is ground away, what happens next? If the grinding motion is completely even, both surfaces remain flat. But with even a slight change in pressure, the grinding surfaces begin to take on a curve.

A moment of thought should convince you that the curve must be a section of a sphere. That’s the only shape where the two surfaces will always stay in contact as they move. Any high points that deviate from a sphere would eventually be ground away.

This is the reason why a spherical surface is the easiest (and least expensive) curve to manufacture glass lenses to.

This insight is well understood by many old-school amateur astronomers, the ones who go through the long process of grinding their own telescope mirrors. But there’s a problem: the figure actually required for a telescope mirror is not a section of a sphere—it’s actually a parabola.

Although the difference between the two is infinitesimal, the telescope-builder needs to hand-polish subsections of the mirror to reach the right final shape. And the polishing and testing to reach this special curve accounts for a large fraction of the total labor required. The effort turns the optical surface into an aspherical one—simply meaning, any curve that deviates from a simple sphere.

Camera lenses are made using multiple glass elements (at least three are needed for a reasonably aberration-free image). Those surfaces are typically all spherical. But an “aspheric” camera lens includes one of these specially-polished aspheric surfaces (or in rare cases, a couple of them).

Why Aspherical?

But since aspherical surfaces are harder to manufacture, they cost more. So why use them?

A recent aspherical lens

There’s a misconception, often repeated in internet discussions, that aspherical surfaces are needed to “correct spherical aberration.” This statement is very misleading.

We should back up and explain that spherical aberration is when light rays passing through the edge of a lens focus at a different distance than ones from the center. This is mostly undesirable, because then the fine details of your photo subject will lack contrast. (We should note, however, that a bit of uncorrected spherical aberration can improve a lens’s bokeh.) Spherical aberration is particularly hard to cure at fast f/ratios, since the ray paths must bend so steeply at the edges of the lens.

A lens designer needs to balance many factors when creating a new product. There are multiple aberrations to correct, across the whole image, including spherical aberration. But there are also issues of weight, size, and manufacturing cost to consider.

It’s easier to design a good lens when you have more “degrees of freedom.” The more different glass types you can add, and the more surface curvatures you can tweak, the more flexibility you have to cancel out aberrations.

So if a lens design has enough complexity, it can give perfect correction of spherical aberration—even using only spherical surfaces. But it may prove impractically large and expensive to manufacture.

What aspheric surfaces offer is simple: More degrees of freedom.

An aspherical surface can improve a lens design without adding extra glass. Computer ray-tracing can adjust the curvature across an element’s width, reducing all aberrations (spherical being just one). If an aspherically-polished surface can replace three or five spherical ones, it can justify its cost—yielding a smaller, lighter lens that may be be less prone to flare.

Long ago, aspheric lenses were pretty exotic: Polishing the special shapes required lots of extra labor. However technology has moved on, and now there are automated processes that can produce good-quality non-spherical surfaces. As a result, aspheric optics have gone mainstream.

A compact 35mm f/1.7 lens with one aspheric surface

So aspheric lenses aren’t made from some crazy unobtanium, and they don’t have magical powers. They’re just a way to build a better lens using fewer elements, and that’s all.