Studio Tests - 35mm full frame
The 70-200mm F2.8 OS gives somewhat mixed results on full frame. It's very good indeed when stopped down, but relatively weak wide open, giving rather soft, low contrast results towards the edges of the frame. Distortion and chromatic aberration are both low, but wide-open vignetting is slightly higher than average.
Sharpness is generally high in the center at all focal lengths, but falls off markedly towards the edge of the frame at F2.8 (indeed wide open at 200mm, most of the frame is somewhat soft). However things improve dramatically on stopping down, and the lens gives excellent results right across the frame at its best apertures (F5.6 - F11). At smaller apertures sharpness falls off progressively due to diffraction, but even F22 is quite usable when maximum depth of field is paramount.
Chromatic aberration remains very low on full frame. There's a tiny bit of fringing measurable, but really nothing much you're ever likely to see.
We consider falloff to become perceptible when the corner illumination falls to more than 1 stop less than the center. Maximum vignetting wide open is about 2 stops at the extremes of the zoom range (a little lower towards the middle), which is higher than normal for this class of lens. However but it practically disappears on stopping down to F4. (The pattern measured in this test is noticeably asymmetric, due to slight occlusion by the 5D Mark II's mirror box.)
Distortion is still low on full frame, at 1.1% barrel at 70mm, through neutral at 100mm, to -1.5% pincushion at telephoto: a slight improvement over Sigma's older unstabilized design. This may occasionally be visible in real-world photos, but it's unlikely to be particularly disturbing.
Specific image quality issues
As always, our studio tests are backed up by taking hundreds of photographs with the lens across a range of subjects, and examining them in detail. This allows us to confirm our studio observations, and identify any other issues which don't show up in the tests. We were impressed by the 70-200mm F2.8 OS HSM in real-world shooting - it gave reliably good results under a wide range of situations, with the focusing and stabilization systems working extremely well.
The 70-200mm F2.8 OS has a complex optical design, and typical of its class this means it's susceptible to problems with flare under certain circumstances. In general, as long as you make a habit of using the excellent hood and avoid pointing the lens directly at bright light sources it has few problems, but if you push the lens too hard it doesn't fare so well.
The lens deals with being pointed into the sun about as well as can be expected - at larger apertures there's a general loss of contrast, and on stopping down, flare patterns become ever more defined. The bigger problem though appears with a strong light source outside the frame, but not shaded by the hood; large areas of the image can be destroyed by orange-colored flare. This is visible in the viewfinder, though, so easy enough to spot and attempt to combat when necessary.
|70mm F22, Canon EOS 5D Mark II||200mm F2.8, Canon EOS 5D Mark II|
To be fair this isn't a particularly unexpected performance for this type of lens, and whilst these issues are easy to demonstrate when explicitly looking for them, they're unlikely to reflect common shooting situations for many users.
Background Blur ('bokeh')
One genuinely desirable, but difficult to measure aspect of a lens's performance is the ability to deliver smoothly blurred out-of-focus regions when trying to isolate a subject from the background, generally when using a long focal length and large aperture.
The Sigma 70-200mm OS generally gives quite attractive bokeh, that's not hugely better or worse than any other lens in its class. Distant backgrounds tend to be nicely blurred, although with a hint of bright edging to out-of-focus highlights. For closeup shots the transition to out-of-focus regions is generally smooth, and notably there's practically no color fringing from longitudinal chromatic aberration.
|200mm F2.8, Canon EOS 5D Mark II||200mm F2.8, Canon EOS 5D Mark II|
|50% crop, upper left||50% crop, upper right|
Under certain circumstances, though, the Sigma can be prone to giving quite 'busy' bokeh, especially at longer focus distances, and we noticed this most clearly when shooting sports. In the sample below we can see both bright rings around point highlights and doubling of lines in slightly out-of-focus regions (so-called 'nissen' bokeh). Note though that, in the latter case, we're looking at things very closely indeed - a 100% crop from an 18Mp APS-C Canon EOS 7D - so it's questionable whether you'd ever see this in a print.
|50% crop, upper right|
|157mm F2.8, Canon EOS 7D||100% crop, lower right|
The 70-200mm F2.8 OS shows exceptionally low levels of chromatic aberration in normal use (clearly the FLD glass used in its construction is doing its job). The examples below illustrate this well - even on high contrast transitions at the edges of the frame, there's very little color fringing to be seen, indeed we struggled to find any real examples at all.
The importance of this can't be overstated - the lack of fringing results in a very clean-looking image right across the frame, and contributes substantially to the overall perceived image quality.
70mm, full frame
|F8, Canon EOS 5D Mark II||F10, Canon EOS 7D|
|100% crop, lower left||100% crop, lower right|
Edge / corner softness wide open
Our studio tests suggest the Sigma to be somewhat soft wide open towards the edge of the frame, and require stopping down to F5.6 to sharpen up well. This is apparent at all focal lengths on full frame, but is most marked in the 85-100mm region.
Having looked at plenty of real-world images shot wide open on full frame, these low MTF50 values appear to represent mainly a loss in contrast, and may also reflect a degree of curvature of field. The crops below show that fine detail is still being resolved into the corner of the frame, albeit at low contrast. In fact, thanks to the lens's extremely low chromatic aberration, the overall image quality in these regions is rather better than the MTF data alone might suggest.
|F2.8, Canon EOS 5D Mark II||F2.8, Canon EOS 5D Mark II|
|100% crop, point of focus (upper left)||100% crop, center of frame|
|100% crop, left edge||100% crop, top right|
The are plenty of shots at F2.8 in the samples gallery, and with this lens perhaps even more than usual, we'd advise taking a close look at them to see if the image quality will meet your needs.
Optical Image Stabilization
The 70-200mm features Sigma's latest image stabilization system, which claims to allow hand-holding at shutter speeds up to four stops lower than usual before blur from camera shake becomes apparent. The mechanism is practically silent in use, but with noticeable clicking noises when it activates and deactivates due to the OS group moving in and out of the 'at rest' position. The viewfinder image takes a moment or two to stabilize, but then 'locks' impressively when the OS is active.
Stabilization is available to users of all mounts, and offers an alternative to the in-body systems of Sony and Pentax SLRs, with advantages including a stabilized viewfinder image for precise composition, and stabilization of the image provided to the autofocus sensor (which in principle may improve results). It only works, though, on bodies which support ultrasonic-type focusing (SSM or SDM) and can therefore supply power to the OS module.
We've generally found the stabilization units in SLR lenses to be pretty effective in real-world use, and to quantify this, we subjected the 70-200mm OS to our studio image stabilization test at the wide and long ends of the range, using the EOS 5D Mark II as the test camera. The subject distance for these tests was approximately 2m at 70mm, and 6m at 200mm.
We take 10 shots at each shutter speed and visually rate them for sharpness. Shots considered 'sharp' have no visible blur at the pixel level, and are therefore suitable for viewing or printing at the largest sizes, whereas files with 'mild blur' are only slightly soft, and entirely usable for less critical applications.
|70mm OS OFF||200mm OS OFF|
|70mm OS ON||200mm OS ON|
Looking at these results, it's clear Sigma's OS system is performing well, and while in our tests it doesn't quite match the very latest Canon and Nikon incarnations, the difference is minimal. We found it to be a little less able to stabilize the image to absolute pixel-level sharpness on the 5D Mark II's highly demanding 21Mp sensor than the Canon EF 70-200mm F2.8 L IS II USM, giving a higher proportion of just-slightly-blurred shots, but this would only be noticeable at the very largest display sizes. Overall we'd estimate the OS system to be providing at least three stops of stabilization under these (controlled) conditions.
Setting the OS switch to the position marked '2' engages panning mode. The manual is a little vague about this option, saying it 'detects the vertical camera shake, and overcomes blurring. It is effective on subjects moving horizontal to the camera'. Which sounds pretty well what you might expect.
A little further investigation, however, reveals that the lens appears to be doing something subtly different from Canon and Nikon's systems. These stabilize in both dimensions until a panning motion is detected, at which point shake compensation is turned off in the corresponding direction. In contrast, the moment the lens is set to mode 2 the Sigma disables stabilization completely in the horizontal direction - i.e. there's no attempt to work in both directions when the camera isn't being panned. Also, if you attempt to pan the camera up or down rather than side to side, you'll suddenly be fighting against the OS system.
Now this all assumes the camera is being held in landscape format (which to be fair we suspect is generally preferred for panning). But if you turn the camera to the portrait format, the lens doesn't sense the change in orientation, so it now stabilizes movement in the horizontal direction, allowing vertical panning only.
This behavior is certainly something you need to understand, so as to select your OS mode correctly for the situation at hand. In practice, we didn't find it especially limiting - you just to need to remember that in mode 2 you should only pan in the direction of the long side of the frame. However it's undeniably less capable and intuitive than the Canon and Nikon systems.
Real world examples
To give some idea of how well the OS system works in the field, the samples below show the kind of slow-speed hand-held shots we were able to get in everyday shooting. The usual caveats apply, and any specific results will depend upon many factors - you're likely to shake more if it's wet and cold, or have just drunk lots of coffee - but it's clear that Sigma's technology works pretty well in practice.
|1/25 sec F22, Canon EOS 5D Mark II||1/40 sec F8, Canon EOS 7D|
|100% crop, center||100% crop, center|