Oddities of Image Stabilization

By Ctein

Two weeks ago I bought a Panasonic 45–200mm lens for my Olympus Pen EP-1. Along with my Panasonic 20mm and the Olympus 14–45mm zoom, that gives me a pretty complete kit (save for wide-angle and macro work). I have found the image stabilization in the Olympus worked very well with my other lenses; I've had utterly spectacular results when I've used shorter focal lengths handheld in the 1/8–1/15 second range. The first illustration here was photographed with the Panasonic 20 mm lens at ƒ/1.7 and 1/7th of a second, and the companion frames are similarly sharp.

Image stabilization works wonders...(go here for a larger, higher-quality JPEG)...

Consequently, I was hoping that I'd be able to handhold the new zoom down to some fairly low shutter speeds, especially with the choice of on-lens and in-body image stabilization. That would go some ways to compensate for its modest aperture (otherwise, it's a very nice lens). Well, that's not what I found. At low shutter speeds there was some benefit to image stabilization, but the higher the shutter speed went the less this was true.

At shutter speeds above 1/250 of a second, I am half-convinced that image stabilization produced less sharp results than no stabilization at all. At 1/125 of a second and above, image stabilization is not obviously better than no stabilization. It is most definitely not foolproof; I would say only half of the frames were pixel-level sharp, even at 1/250 of a second.

At 1/60 of a second and lower, image stabilization is clearly advantageous, although not by as much as one would think. It looks like the lower the shutter speed, the better image stabilization does relative to ordinary handholding; the problem is that with a 200mm lens, low shutter speeds are going to be problematical (that's an understatement).

...Except when it doesn't! 100% sections from some test photos. All photos made at ƒ/6.3, 1/250th sec., 200mm focal length. Photos 1–3 used Olympus in-body stabilization, 4–6 Panasonic in-lens stabilization, 7–9 both stabilization systems and 10–12 no stabilization. No series of three photos is entirely sharp.

Note: The poor showing by in-camera stabilization in this series was not replicated in other test series. Sometimes, in fact, it was better than in-lens stabilization. All test series, though, showed that having both systems on produced no additional problems, but that having no stabilization at all worked as well as or better at high shutter speeds than any combination of stabilizing systems.

Several hundred test photographs indicated that there was no clear-cut advantage to one stabilization system over the other. Interestingly, having both systems on didn't perform any worse (didn't perform any better, either). The manual says not to do that. Sometimes manuals are wrong. No doubt you are all shocked to hear that. What I have to conclude is that image stabilization is of modest benefit with this lens used at its longest focal length, but it's not really a game-changer. That's somewhat disappointing. Most notably, getting consistently tack-sharp photographs at 200mm is going to require me to use exposure times even shorter than 1/250 of a second, unless I can improve my handholding skills. The methods and stances I've used in the past worked extremely well, but they may not be optimal for this camera and lens combination.

I've never had a chance to do this kind of careful comparison before, and I'm a little surprised at these results. But, after discussing this with other folks it seems like my results aren't abnormal. I posted a query about this on the mu-4/3 forum and quickly got some very helpful and informative comments, as well as valuable links to relevant tests other people had done. For those of you with Micro 4/3 systems, I really have to recommend that user group. Most of the time, when I post a question about a technical problem on a user forum, all I hear is the faint susurration of chirping crickets. Not with that bunch!

Anyway, if there's a reader out there who's got some real expertise on the technical innards of these systems, I'd love to hear your comments. My hypothesis: the frequency response of these systems is limited, so when the exposure gets into the few millisecond range the system can't compensate for the dominant motion frequencies. With long exposures, those high frequency movements are less dominant and they tend to average out into a symmetric slight blur. With short exposures, there isn't enough time for random motion to average out; you get a smear. The problem's exaggerated by long focal length lenses, where a little bit of camera movement translates into a lot of image movement.

That's a wild-ass guess. I'd love to hear more educated opinions.

A cautionary note: if you want to chime in with your opinions about what's going on, it would be good to read the thread on the mu-4/3 forum first and follow the links there. Your thoughts may very well have been anticipated.

Whatever, it's looking like image stabilization is not a "set it and forget it" kind of thing. Dang.

Ctein

Ctein's weekly column on TOP is published every Thursday morning.

Featured Comment by Thom Hogan: "All stabilization systems have a 'sampling and correction frequency.' On the Nikon VR systems, for instance, that frequency is a bit above 1/500, which is why I've long recommended that you turn VR off at 1/500 and above.

"As you get shutter speeds closer to the sampling frequency, you start to run into what I call 'stutter' problems. Basically, the system is still correcting for a -XY movement when a +XY movement is now needed, but the sampling/movement frequency isn't fast enough to compensate in time for the shutter speed (or only partially compensates).

"What the frequency is on the Micro 4/3 bodies is I don't know, but it seems lower than on the Canon/Nikon in-lens systems.

"Makers promote stabilization as a marketing checklist item. Users interpret that as meaning 'always use it.' Just the opposite. Stabilization should never be on unless you know you need it and know it will be effective."