Originally posted by sledger Yes, sorry. What I meant was that the K-30 amplifies the effect, the K-5 reduces it. Rolling shutter is indeed created by reading the sensor line by line as opposed to one frame at a time. You will only see the effect when there is movement in the frame, when the frame has moved between the time you start reading it until the time you stop reading it. The K-30 and K-5 should have pretty much the same sensor (thus a similar rolling shutter effect), but:
In the K-5 the sensor is moved around when you are shaking, in an attempt to get a stable image. What then arrives on the sensor itself is a stable image, with no movement (within the limits of the sensor shift system of course). As if the camera was mounted on a tripod. Thus there is no wobbling around (coming from small random movements left, right, up and down). The only way to get a visible rolling shutter effect is to either have a fast moving object (like a train) move through the image or by panning fast. If you look for it you'll see it, but it is not very irritating. Movements need to be pretty fast, and the amount at which lines tend to lean is pretty stable.
In the K-30, and probably K-01, the sensor stays where it is. That means all the time when you shake the camera (i.e. handheld usage) the image area has moved before a single frame has been read. This happens randomly, so where it wobbles to also changes, sometimes it is to the left, to the right, or the video is stretched or squeezed, and everything in between. You can see it on most of the Canon and Nikon DSLR videos, unless they use things like a steadicam or a stabilized lens. Now this is pretty irritating, but since the whole frame shakes anyway that disguises the effect a bit. But when you activate SR in the K-30 it does make use of the sensor that is used for the sensor shift SR system (making it superior to those pure software ones), however it only stabilizes according to the top most line, it takes the final frame/crop based on that line. However due to the rolling shutter effect the lines below that have moved on. So you do get a half way stabilized video, where the top part of the frame stays perfectly stable. But the rest of the frame... nope. Moves around. And THAT is seriously irritating.
And that's not the only problem with this sort of digital image stabilization that simply crops a part out of the sensor and moves around the part of the sensor. You might be able to eliminate this effect by selecting a different part of the sensor for each line (this could also work for the rolling shutter effect in general), at least when it comes to panning, though it would introduce a rolling shutter effect if you actually follow a driving car for example. However when your shutter speed isn't very short you do get motion blur when you move the camera when shaking. Again, since the sensor is recording an unstabilized frame, that's it. You can't avoid it. And again, if you don't stabilize at all it is perfectly acceptable, and even concealing the camera movement to a certain degree, making it more acceptable. But the K-30 picks a part of the sensor for the frame, thus stabilizing it more or less. Then you have a stable video that a) has the bottom of the frame move around randomly and b) gets blurry sometimes, with motion blur going in different directions, randomly (especially weird when there are bright spots of light).
The K-5 as I said before doesn't get any of these negative effects because what sensor sees is perfectly stable, or at least only moving in one direction, smoothly.