The optical image stabilization system used on the S3 IS (and its predecessor) works, and it works pretty well for most everyday photography - though don't expect miracles. There are three modes: Continuous (IS on all the time), 'Shoot only' (IS is activated at the moment the exposure is made) and Panning (for horizontally panned shots).
The first option makes framing easier - the IS system steadies the preview image, and we found it more consistently effective at extremes (where the shutter speed is more than 3 stops slower than would normally be used for the focal length you're shooting at) than the shoot only mode. Shoot only - which doesn't steady the preview image, but is theoretically more efficient because it minimizes the amount of corrective movement required by waiting until the instant the picture is taken - is perfectly effective when you're nearer to the correct shutter speed (and once you're within 1 or 2 stops it is actually more reliable). I wouldn't read too much into the fact that our results for the relative effectiveness of the two modes are the opposite to what we saw with the S2 IS; we have no quantitative tests for stabilization, and every person's 'shake' is different; users often disagree on which mode works best, so you just have to find which best suits your style of shooting.
I certainly found it made handheld shots at 3, 4 or even 5 shutter speeds slower than normal perfectly possible. Impressive stuff (though the reliability is much better if you're within 3 stops), and if you take a couple of 'safety' shots when pushing the system hard (1/30th at 432mm for example), you'll usually get at least one 'keeper'.
Although we've no definitive test for IS systems in real-world use, I was very impressed with the S3's system, though I'm not convinced that it is as effective as that used on Panasonic's Lumix range. Of course the more megapixels in the image, the more you're likely to see any blur (viewed at 100% on-screen); each new generation of 'super zoom' camera is pushing the system's capabilities that little bit more. It's also worth noting that the weight and ergononics of the S5 IS mean that it is far less prone to camera shake than some of its featherweight competitors.
These tests are rather extreme - up to five stops slower than you could safely use without IS - and in 'real life' shots - where you are maybe using a shutter speed two stops slower than normal - the system is pretty much 100% effective.
The stabilization test
In this simplified version of our SLR IS test, four hand-held shots were taken of a static scene with the stabilization off and on. The shutter speed was decreased and repeated (from 1/1000 sec to 1/8 sec). The zoom was set to its maximum position (432mm equiv.), the test target was 6 m away from the camera. The test was repeated three times and an average taken.
The resulting images were then inspected and given a blur score - 'Sharp' (no visible blurring at 100%), 'Mild Blur' (the kind of camera shake that is tolerable at small print sizes) and 'Heavy Blur' (virtually unusable due to camera shake) and 'Very Heavy Blur' (little discernible detail).
The results below show that Canon's IS system gives a real two-stop advantage but it's not capable of working miracles, and that at anything two shutter speeds or more below the recommended minimum for a non-stabilized lens (around 1/450th at this focal length) you should take a couple of shots to be sure of getting one perfectly sharp one.
It's obvious from comparing these results with those we got from the S3 IS and the S2 IS that the perceived 'effectiveness' of the IS system falls with each increase in resolution, mainly because we're viewing at 100% (actual pixels), and as resolution increases so does the camera's ability to capture even the smallest amount of blur. For 'standard' print sizes (up to say 5x7 inches) you can ignore the difference between 'sharp' and 'mild blur', and many of the shots classified as 'heavy blur' produce acceptable prints.
Hand-held, no stabilization (432mm equiv.)
As you can see from the chart below only at 1/500th sec or above can we be confident of getting sharp results from the majority of shots, and once you get to 1/125th sec and below the majority of shots are blurred, and none are sharp.
Hand-held, stabilization on (432mm equiv.)
With stabilization on the results are better - we got no blurred shots at all 1/250th and above, and the more than half the shots taken at down to 1/60 sec have little or no blur (and are certainly printable at smaller sizes). Even at three or four stops below the recommended minimum for hand-holding a lens this long you will get a usable shot maybe 25% of the time - just don't expect miracles!
Specific image quality issues
The output from the S5 IS is very similar to its predecessor (so much so that at normal print sizes I doubt you'd be able to tell one from the other). Once again it's a slightly mixed bag, with lots to like but also enough to complain about to make the most demanding users think twice before handing over their credit card.
The good stuff is everything we've come to expect from Canon's consumer level cameras; bright, vivid colors that have real 'pop' right out of the camera without straying into 'over the top' territory. The fact that the exposure and focus systems are generally very reliable helps to ensure that - unlike so many cameras in this sector - even the novice user can 'point and shoot' to their heart's content safe in the knowledge that 9/10 times they'll get the shot they want.
But anyone thinking that a 12x zoom and 8MP sensor in such a compact form doesn't involve any compromises is in for a rude awakening, especially if they're the kind of person who wants image quality that stands up to the huge enlargements involved when looking at the output at a pixel level. Whilst better than many competitors, the effects of noise and noise reduction make anything over ISO 100 disappointing if you look too closely. But no one in their right minds judges a camera of this type based entirely on 100% crops unless they intend to produce giant poster prints - and in that case I'd say 'walk away now... you need an SLR'.
So what about the kind of issues that affect 'real world' use; the kind of things you'll see in a standard print? As I mentioned above the color, focus and exposure systems are remarkably reliable, so there's little to worry about there. There are, inevitably however, a couple of issues we need to talk about; fringing (the SX IS series' achilles heel since day one) and dynamic range (the bane of all small sensor cameras).
As with every camera in this range the S5 IS exhibits some fairly extreme red (and occasionally blue) fringing at the long end of the zoom in certain situations - around the edges of very bright (usually over exposed) areas. Thankfully it's not that common, but when it happens it can be extreme, to say the least (UK readers will know what I mean if I describe the effects as looking like the Ready Brek glow!). The size of the fringes means they are - just - visible in small prints (they average around 6-12 pixels wide), but in most cases not so badly that they completely ruin the image. That said, Canon: Sort it out!
|100% crop||432mm equiv., F3.5|
|100% crop||432mm equiv., F3.5|
All small sensor cameras suffer from highlight clipping in scenes with a very wide dynamic range; it's one of the things you have to learn to live with. The problem usually occurs when the S5 IS metering gets things a little wrong (though both the fringing examples above show how easy it is to get clipping if there's too much contrast in the scene). A little negative AE compensation (and watching the histogram) helps, and to put it into perspective, of just over 1000 shots we took for the gallery (nearly all without any exposure compensation), clipping was an issue on maybe 20 or 30 shots... so no worse than most cameras we test.
|100% crop||41mm equiv., F4.0|