This forum's wisdom is that SR autodetects when the camera is panning and only does SR in one direction then.
So, when SR fails to detect the pan it will certainly ruin the image.
We would need some experience reports here when the SR actually does correctly determine a horizontal pan
- what (minimal) horizontal panning speed?
- what (maximum) vertical panning speed?
- how long (is it panning time or turning angle?) before the SR switches into panning mode?
If it doesn't detect the pan, SR will cause the image to "jump" and must ruin it. Also, one should only pan strictly horizontally, i.e., only turn around the z-axis thru the camera body.
Update:
You can actually hear it.
With a K20D, set focus length to a high number like 800mm (with no lens or manual lens). The test works with any length, but at 800mm, it is most impressive. Now switch LiveView on. You will hear some bad noise from the SR at the slightest move of the body (much worse than at 50mm
).
No noise at a constant pan, though.
If you shake the camera during the pan: noise again.
Actually, I got the impression that it still does try to do SR in both directions, but with the panning rotation set as the new normal (it is other vendors known to switch to 1D SR -- maybe Pentax does even better?). This would mean that pans should have a stunning quality.
To test it out, one would have to go to a kid's carousel (merry-go-round) and shot the opposite side (the gyro sensors measure angular velocity, so the test would be valid). Any other testing propositions?