Crystal Vodka: the Canon 135/2.8 SF
Crystal Vodka: the Canon 135/2.8 SF
I do most of my shooting between 35 and 50 mm, with an occasional foray into wide-angle and an even more occasional foray into telephoto. For about a year and a half, the telephoto end of my operation has been handled by the bargain-bin Canon 55-200/4.5-5.6 USM. While it really is a damn good lens for its very low price, it's still clearly several cuts below the optical quality of my main-use "pocket primes." I have formerly owned a Canon 200/2.8L, but sold it because I found it a bit too long for my needs, a bit too bulky to take along where the pictures are, and a bit too expensive to leave sitting on the shelf. Therefore, especially for portrait and event type shooting, I'd been on the lookout for one of Canon's excellent short telephoto lenses. I finally broke down and bought a slightly unusual lens: Canon's venerable 135/2.8 Soft Focus. I chose it partly out of curiosity, since it was an unusual lens, partly out of an irrational fondness for the 1987-1989 early EF build and nostalgia for the no-nonsense type of optical designs that Canon appears to have abandoned lately, and partly due to cost considerations: what with the weak dollar and Canon's strange pricing policies, it was available new for about half the price of any of its immediate competitors. I hereby dub it Crystal Vodka, because it's clear and smooth and can make the world go blurry if you abuse it.
The 135/2.8 Soft Focus is one of Canon's oldest EF lenses. It was introduced with the EOS system in 1987, along with the 15/2.8 fisheye, 28/2.8, 50/1.8 Mk 1, 50/2.5 Compact Macro, a couple of standard zooms, and a few telezooms, of which the 100-300/5.6L is the most fondly remembered. A lens with serious pedigree, in other words. However, it's also one that's been seriously eclipsed by its younger siblings. The 135/2.0L, lauded by many as Canon's sharpest lens, is nearly legendary in optical quality, while the more affordable 85/1.8 USM and 100/2.0 USM sport brighter maximum apertures and more modern builds, in particular ring USM and full-time manual focusing. In this company, it does seem like the odd one out: its one truly differentiating feature is that it allows control over spherical aberration, which makes it possible to dial in any desired amount of a soft-focus effect that cannot be effectively duplicated in post-processing.
Canon's choice to include a 135 mm prime in its EOS starter line-up is indicative of the importance given to this focal length. It sits at the long end of the "portrait lens" spectrum, clearly separated from the 50 mm normal and long enough to make for pleasing tightly framed portraits, but yet not long enough to be deep inside telephoto territory. Shooting with a 135 still puts the photographer inside the scene and forces him to interact with his subjects in some way; go to 200 mm and beyond, and you become something of a passive observer. A 135 is the longest lens that can be comfortably hand-held even in less than ideal lighting, and a 135/2.8 is compact enough to be almost pocketable... if you're wearing a coat, at least. A versatile focal length, therefore -- suitable for anything from portraits to event photography to ringside sports photography, and even certain types of travel and street work, and of course "compressed" landscapes.
From an optical design point of view, a 135 mm lens for the 35 mm format is just about ideally specified -- long enough that most of the corner problems with shorter lenses go away, yet not so long that you have to jump through several hoops to keep aberrations under control. There are several standard optical designs that work well for this focal length, and the designers have a quite a bit of leeway to balance different characteristics against each other. What choices, then, has Canon made on their first EF telephoto prime?
Moonrise with Anthracite. Helsinki, January, 2007. Canon 135/2.8 SF at f/4.0.
Build and handling
The 135/2.8 SF has the classic 1987 EF build: metal mount, hard plastic barrel with a narrow focusing ring, and arc form drive (AFD) auto-focus. The barrel is sprayed with a fine-grained black "nubbly" paint and molded with ribbed textures here and there. The focus ring is narrow, ridged, and of a slightly grippier type of plastic, as is the ring controlling the amount of spherical aberration. Focus distance is displayed behind a clear plastic window with depth of field markings for f/22 and f/32, and a dot for the infra-red focus point. The lens has no full-time manual focus; switching between AF and MF disengages the focus ring. Like all AFD lenses, the focus ring has some play. The lens takes the ET-65 hood which clips onto a little groove on the front bezel; this system only permits circular hoods but is handy enough to clip on and off or reverse onto the lens.
The Canon 135/2.8 SF on the EOS-5D, with and without the ET-65 III lens hood.
A slightly unusual build feature is the front bezel of the lens: it is covered with a rubber ring. Also somewhat unusually, the filter threads continue pretty deep into the barrel; the front element is recessed about a centimer. It's almost as if the optics don't quite manage to fill the barrel. The lens is internally focusing; polarizers will not turn as the camera is focused, and the lack of external moving parts gives it a solid one-of-a-piece kind of feel. The lens feels quite heavy for its size, adding to the impression of workmanlike solidity. The AF switch and the spring-loaded switch locking the soft-focus ring in place are positioned to be easily accessible without taking the eye off the viewfinder. All in all a nice enough, solid lens to handle, albeit without the advantages of ring USM -- and definitely old-fashioned in its utilitarian, angular look. Next to a newer style lens like the 28-105/3.5-4.5 USM I'm looking at, it looks pretty Spartan: clearly designed by engineers rather than industrial designers. Which, I think, is one reason for my fondness for lenses from this period.
Canon's Arc Form Drive (AFD) lenses have a somewhat dodgy reputation for focusing performance. This is not entirely deserved: it is largely due to the fact that when USM was introduced, AFD was only used on the bottom-drawer lenses, which were not great performers for a number of reasons. AFD is simply a way to move the focusing elements: there's nothing inherent in it that makes it slower than USM. The main advantage of (ring) USM is full-time manual focusing: the possibility to tweak focus by adjusting the focusing ring without changing from AF to MF mode. No AFD lens has this feature. Another advantage of USM is that it is virtually silent at least to the human ear; AFD lenses buzz when focusing, due to the focusing motor and gear train driving the focusing elements. However, not all AFD lenses are created equal in this respect either. My 35/2.0 sounds like a wasp in a matchbox when it is "hunting," while my 50/1.8 Mk 1 sounds a good deal more pleasant -- and my 24/2.8 is actually rather quiet. As is the 135/2.8: its focusing buzz is easily audible in a quiet room, but gets lost in the ambient noise, even to the photographer, in a normal outdoors environment. The lens is quite quick to focus and tracks moving subjects quickly and precisely. I had no trouble shooting quite fast-moving skaters with it, and I'm not even very experienced at that sort of photography. Subjectively, the 135/2.8 SF doesn't feel any slower than the 200/2.8L USM I used to own (although I cannot do a side-by-side comparison, having sold the lens) and feels clearly faster than the ring-USM-equipped 28-105/3.5-4.5 USM (or 55-200/4.5-5.6 USM, no big surprise there). In other words, the downsides of the AFD on this lens are more in the features than in the performance.
Skater, Helsinki, January, 2007. Canon 135/2.8 SF at f/2.8. I had no trouble locking on or tracking this fast-moving kid on a skating rink, using the EOS-5D.
The lens's closest focusing distance is around 1.3 meters, although dialing in soft-focus reduces the distance somewhat (and stopping down eliminates the softness it causes). This isn't a macro lens by any stretch of the imagination.
Minimum focus distance at f/8.0 and two notches of soft-focus. This is as close as it gets.
Canon notes in its product literature that the 135/2.8 SF is physically incompatible with their teleconverters. I have no idea why, since there's plenty of air towards the back of the lens. I tried it with the Sigma 1.4x converter that I have knocking around in a drawer (if someone wants to buy it, drop me a line), and it worked well enough -- a bit muddy wide-open, but quite decent from one stop down.
An actual-pixels crop from a test frame shot with the Sigma 1.4x APO teleconverter at f/5.6 (f/4.0 displayed). Acceptable in a pinch if you need the reach. It's somewhat sharper than the 55-200/4.5-5.6 USM at a corresponding focal length and aperture, but I would expect the 55-200 to be nicer in other respects, such as bokeh. I haven't tried it, though, and probably won't bother.
Dialing In Soft-Focus
The 135/2.8 Soft Focus's main distinguishing characteristic appears in the name: the possibility to dial in variable amounts of spherical aberration, resulting in a "Hollywood style" (or soft-porn style, depending on how you look at it) gauziness to the picture. Despite what you may think, the effect cannot be easily duplicated in post-processing: it's a very complex phenomenon that depends on focus distance, distance to background, relative brightness, and so on. The strength of the effect depends strongly on the aperture: wide-open and at the maximum setting, the pictures turn out pretty over-the-top; as you stop down, the effect becomes much subtler, and is barely noticeable at f/5.6. Similarly, as you dial down the effect, it becomes less apparent; wide-open, only the barest nudge to the soft-focus control will produce visible effects. Changing the setting also shifts the focus point, which means that it's necessary to re-focus after each change. The ring controlling the aberration has two click-stops for better control over it, but it can be set to any position between them. At high levels, the viewfinder becomes very "dreamy," which makes it tricky to judge focus on it, even with the Ee-S screen I have installed; auto-focus seems to cope with it just fine, though.
Hollywood Missy, January, 2007. Cranked up to the max, the soft-focus effect is... hard not to notice. This is Missy experiencing a full blast, at the number 2 notch and f/2.8.
Soft Granite, Helsinki, January, 2007. This is how things look at f/2.8 and one notch of soft-focus -- still pretty in-your-face; it's unlikely you'll be wanting to use the lens with these settings most of the time. Unless you shoot soft-core porn, of course.
The soft-focus effect's impact on out-of-focus areas is at least as strong, if not stronger, than on in-focus areas. Bokeh becomes much smoother; highlights spread out and become "limb-darkened" rather than "bright-lined." The interesting thing is that even as the effect becomes less apparent on in-focus areas, the effect on bokeh persists. Dialing in one notch of soft-focus at f/4 will make a significant difference to out-of-focus backgrounds while leaving the subject looking pretty sharp. This setting will work nicely for flattering portraits.
I dialed in just a touch of soft-focus for the right-hand frame. This leaves the subject looking pretty sharp, but significantly improves the bokeh in this highly problematic background. The frames have been post-processed to normalize contrast and brightness.
Above, Joanna against a chaotic background, shot at f/4.0 and the soft-focus dial at zero; below, a frame shot from the same position with the same aperture, but at the dial set to 1. The bokeh is a good deal smoother, but Joanna is still looking quite sharp. Close-up, the effect has smoothed down things a bit, but the effect is rather subtle and not easy to spot without a comparison frame. For portraits, I think the second setting would work rather well.
The proof of a pudding is in the eating, they say, and the proof of a lens is in the optics. I was counting that Canon couldn't possibly release a lemon in their starting line-up for the EF system, and they didn't. The 135/2.8 SF is a very respectable lens optically. It has very much the same optical "feel" as the other lenses in the same series that I own (the 24/2.8, 35/2.0, and 50/1.8 Mk 1). It feels like a very balanced design, with no glaring weaknesses but no truly outstanding features either. It's exactly what you expect from a high-quality prime of its specification and price point. If I had to rank it, I'd say that it'll hold its own against most L zooms, and beat just about any consumer zoom in its focal length range in some not entirely trivial way. Compared to its heavier-duty L siblings -- the 135/2.0L and 200/2.8L, the latter of which I owned -- it visibly loses out at wide-open apertures. The L's are engineered to bite like a shark even wide-open; the 135/2.8 SF loses just the keenest cut of its edge there, although you have to look pretty hard, preferably with comparison shots, to tell.
The instruction sheet that comes with the lens claims that it's designed to work best at "close distances," meaning the typical 3-5 meters used to shoot portraits. I'll take Canon's word for it, but they could've fooled me; I haven't been able to tease out any performance differences between close-up, middle distance, and near-infinity.
The 135/2.8 SF is very contrasty indeed. It produces lovely deep blacks with subtle low-key tones; certainly not worse and perhaps even a hair better than the 50/1.8 Mk 1. This gives their images a "pop" that gives them immediate appeal, is visible at any enlargement, and is very hard to recreate in post-processing if it's not there to start with, wide-radius unsharp mask notwithstanding. I would have been disappointed had I seen anything else; in fact, I had a momentary sinking feeling when looking at my test shots, until I realized I had gotten the 135/2.8 SF and the 28-105/3.5-4.5 shots mixed up.
Crops of low-key detail rendered in rather tricky conditions. You probably need a decently calibrated monitor to be able to see this.
The 135/2.8 at top, the 28-105 at bottom. Both shot at f/8.0, and actual-pixels crops taken close to the center of the frame. The latter has a somehow murky feel, and is overlaid by a slight haze. This denotes poorer contrast.
Wide-open, resolution in the center of the frame is very good, and in the corners, good. Resolution rises to excellent in the center by f/4.0, and peaks at f/5.6. In the corners, it rises to excellent by f/5.6 and peaks at f/8. Diffraction becomes visible at f/16, and at f/22 the lens's resolving power drops below what it was wide-open; however, it is still certainly usable if the depth of field is needed. The resolution drop towards the corners is very graceful; there is no accompanying loss of contrast (other than that attributable to vignetting) nor appearance of chromatic aberration. This means that even in-focus subjects near the corners wide-open look good at moderate enlargements.
Center crop, wide-open at top, f/5.6 at bottom.
Corner crop, wide-open at top, f/8.0 at bottom.
In summary, the lens performs very well wide-open, although it does lack the shark-bite of a telephoto L, and stopped-down it is sharp enough to satisfy even the pickiest of critics.
As far as I can tell, the 135/2.8 SF is pretty neutral in color, with perhaps a touch to the warmer side of it. It's very similar in color rendition to my 50/1.8 and 24/2.8; my 35/2.0 is noticeably cooler and my 55-200/4.5-5.6 is much warmer. I certainly cannot see any color casts in my real-life shots.
Not a proper test for color, but it'll give a rough idea of what's going on. The 135/2.8 SF at top, the 50/1.8 Mk 1 at bottom. Manually white-balanced for the 135/2.8, with the same WB used on both frames. There are differences, but they're very subtle, and could well be due to something I did -- for example, I had to change position when shooting the frames, which may have affected the mix of lighting in the room.
I bought the ET-65 III clip-on lens hood to go with the lens, and used it on all of my test shots, including the ones for flare. I have not seen any flare effects in any of my real-life shots. Despite my best efforts with streetlights and the moon, I have not managed to produce any visible flare spots at all with the lens.
The flare spot torture test: a two-second exposure of the full moon in the corner of the frame. This makes it about twelve stops overexposed. No spots to be seen.
Torture-testing for veiling flare revealed that stopping down increases the lens's sensitivity to veiling; even so, I had to go to pretty extreme measures to bring it out. The lens is clearly excellent at resisting flare.
The veiling flare test, f/2.8 at top, f/5.6 at bottom. This is my computer monitor in a darkened room. The scene is starting to look washed-out from veiling, and it's worse stopped-down than wide-open. However, this situation is truly extreme; the monitor is overexposed a whopping ten stops.
The lens shows moderate vignetting at f/2.8. It is gone completely by f/8, and unnoticeable in almost all real-life photos at f/4.0. It is due to optical vignetting (occlusion of the pupil by the lens barrel).
Vignetting: a white wall lit by bounce flash. Left to right and top to bottom: f/2.8, f/4.0, f/5.6, f/8.0.
The amount of vignetting wide-open can be gauged by the "cat's eye effect" -- out-of-focus highlights near the edges of the frame take the shape of a cat's eye, as the lens barrel obstructs a part of the pupil. This happens with just about all lenses, although of course the edges get cropped out on APS-C digital SLR's. Some people very much appreciate the effect this has on bokeh.
The 135/2.8 SF is specifically designed as a portrait lens. This would lead one to expect that it has been designed for pleasing bokeh. Bokeh is a notoriously subjective area to examine, and tricky to pin down with any consistent tests. This is because it varies a great deal depending on the circumstances (focus distance, distance to background, foreground versus background bokeh, type of background) and there's a great deal of subjective preference involved as well: where one person is immediately drawn to the shape of out-of-focus highlights, another one barely notices them but is greatly bothered by "nisen bokeh" -- line-doubling. Personally, I hate color artifacts in bokeh, dislike nisen bokeh, and don't care about the shape of the highlights; in fact, I consider that to be a part of the lens signature rather than something that's objectively "better" one way rather than another. I also like just a bit of lens character to show in the bokeh: for example, the legendary Minolta 135/2.8 STF that has been specially engineered for "perfect" bokeh (at the cost of losing over a stop of brightness, too; the lens's T-stop is 4.5 at f/2.8) produces pictures that look unnaturally perfect -- almost like they're computer renders rather than photographs.
A fellow named Rick Denney with a larger lens collection than mine has done a pretty interesting "lab test" of bokeh, comparing a bunch of them against each other; his line-up includes the 135/2.8 SF. Worth a read and a close look at the photos, if you want to get an idea of where the 135/2.8 SF sits relative to the classics, like the Zeiss Sonnar 135/3.5. I'd also suggest doing a "blind test" with his test shots: hide the information about the lens used for each shot, and rank them by personal preference. My ranking was slightly different than the author's, which just goes to show how subjective the matter is.
My personal take of the 135/2.8's bokeh is that it's... rather good, really. Not jaw-droppingly wonderful, but on the whole definitely on the pleasing side, and judging by the samples I've seen, a good deal better than the Canon 85/1.8 USM for example. There's very little color pollution (caused by axial chromatic aberration) to be seen in most real-life shots, which gives it an overall clean and fresh look, nisen bokeh, while not completely absent, is kept pretty well under control, and the transition from sharp to soft happens quite gracefully, giving the pictures a nice feeling of three-dimensionality and subject separation. The iris is hexagonal, and the aperture blades aren't as curved as is common on more modern lenses, so the highlights do become polygonal as soon as you start stopping down. There's little to no bright-line effect on the highlights. Both front and back bokeh are rendered nicely enough. A magenta/green shift is visible in high-contrast black against white subjects, such as branches against the sky, but this effect is not strong enough to be visible in most real-life shooting.
Wooden Explosions, Helsinki, January, 2007. Canon 135/2.8 SF at f/2.8. The bokeh torture test: branches against the sky. If you look very carefully, you can see some magenta/green shifting between the front and back bokeh, although it's slight enough that the overall color looks pretty neutral. A certain amount of nisen bokeh (line doubling) is more readily apparent.
Bark and Blur, Helsinki, January, 2007. Canon 135/2.8 SF at f/2.8. Bokeh always gets better when you get closer to the subject and the background drops further back. This subject has a similar background, and under these conditions the bokeh looks pretty damn good to my eye.
Skaters, Helsinki, January, 2007. Canon 135/2.8 SF at f/2.8. This middle distance photo demonstrates the transition between in-focus and out-of-focus areas. It's quite smooth and gives the photo a nice feel of depth. However, the background rendering could be nicer -- there's fairly obvious line doubling apparent in parts of the background wall.
Hard Granite, Helsinki, January, 2007. Canon 135/2.8 SF at f/2.8. This approximates a somewhat typical portrait situation. The bokeh looks pretty pleasing here as well.
What's It Good For?
The 135/2.8 SF's main strengths are a highly useful focal length, compact size, reasonable cost, and very respectable optics -- and, of course, the unique soft-focus effect that gives the lens its name. On the other hand, it doesn't have the cleaner looks and ring USM of its younger siblings, nor does it pack the optical shark-bite of its three times more expensive and twice as heavy L sibling. In other words, if you're looking for an inexpensive, compact medium telephoto lens with high-quality optics that's especially suitable for portraits, and don't consider USM a deal-maker, I say go for it. Otherwise, you, probably like most people, are better served with one of its other three siblings. Compared to:
- the 85/1.8 USM, the 135/2.8 SF has (arguably) better bokeh, is longer, a bit lighter, a good deal cheaper (in the USA), and over a stop darker.
- the 100/2.0 USM, the 135/2.8 SF has (arguably) better bokeh, is a bit longer, a bit lighter, a good deal cheaper (in the USA), and a stop darker.
- the 135/2.0 L USM, the 135/2.8 SF is a lot lighter, a great deal cheaper, a bit softer, and a stop darker.
On its own merits, the 135/2.8 SF makes a top-notch portrait lens, a great travel lens, and a highly useful focal length to have around for compressed landscapes and such. It's light and compact, focuses fast and precisely, works great wide-open, and at the magic f/8 it's sharp enough to satisfy even the pickiest of critics. The soft-focus effect leaves a few little cards up your sleeve even if you're not a David Hamilton wannabie. However, to be perfectly honest, other than the price I can't think of that many really compelling reasons to go with the 135/2.8 SF rather than one of its newer siblings. But if you need or want the particular mix of characteristics that it provides, it's unlikely to disappoint either. Its strengths are all in the light weight, optical qualities, and shootability, and its weaknesses are written down right on the spec sheet. For me, Crystal Vodka is a definite keeper -- and I'll try not to abuse it.
Light In The Window. Helsinki, January 2007. Canon 135/2.8 Soft Focus at f/4.0.