I am trying to understand the practical relevance of CIPA stabilization ratings. I have read the white paper on measurement. In a nutshell, it seems that they measure something they call “bokeh”, which is blur on a black and white boundary, and check how much slower they can expose with the camera on a vibratory apparatus to get the same amount of this blur, compared to using no image stabilization.
I have performed a much simpler practical test using my own cameras and two bodies with IBIS I have borrowed. The test is simple: take 10 images at each shutter speed of something high-contrast with fine lines (eg an LCD, where I can discern showing pixels and fine black lines in between), and grade them as “sharp” (1):
“blurred” (0):
and “in-between” (0.5):
I calculate the average of 10 images at each exposure setting. I start at 1/focal length, and make slower and slower expositions (finer gradations than steps). I then plot the average sharpness vs log2(shutter speed), with and without image stabilization. The difference is quantified in stops.
I am confused because with the five cameras I tried, I get 2–3 stops lower values for stabilization than the CIPA rating. I am not as steady as a sniper, but I can handhold a 28mm (eq) lens for 1/40 without stabilization (tucking my hands in). A 2s shutter delay is equivalent to about 1 stop of stabilization.
Eg here is a graph for the GX9 with the 14mm lens (unstabilized, purely IBIS):
From this I read 1.5 stops. The official CIPA rating is 4 stops. (Just to clarify, this camera does not have the issue some people find with the GX9 IBIS). I tried a Sony 6700 borrowed from a friend, which is rated for 5 stops, and get around 2-2.5 stops.
It would be great to hear from others who want to perform this experiment. I may be misunderstanding what CIPA rates, or if not, I have to conclude that it does not translate to what I imagined in practice.
I have been reading up about this and various people have similar experience; effectively, with top-of-the line image stabilization, steady hands, and practice one can handhold 1/4–1/2 shots, but that is about the practical limit.
So, if you get handheld sharp images without stabilization at, say, 1/100s for a particular focal length, and your manufacturer claims a stabilization factor of 7–8 for some body + lens combo (not uncommon these days), this does not mean that you will be able to take unblurred 1.2s–2.4s shots under general conditions.
While I still see the value in stabilized bodies and lenses, now I believe that the window where IBIS+OIS matters is much more narrow than I originally thought, the practical details depend on the user, the focal length, and the kind of movement the lens experiences. Unfortunately, CIPA does not make the details of the testing standard available for research purposes or individuals, so it is hard to know what exactly they are testing.
So the bottom line is that I better test each body + lens combo myself if I want to rely on extreme IS.
If anyone is interested, a great video on the nitty gritty details of IS:
I read your posts quickly, so maybe these thoughts have already been covered in your links. My opinion is that ratings are always better than real life results. Like EV kilometrage / mileage. There are many reasonable explanations to this, one being that the tests were conducted under idealized laboratory conditions and two being that there is pressure to claim larger values.
The standard is also dated and likely needs comparative studies with newer makes and models. In addition, very likely, increases in stop recovery is not linear, and as identified above, peaks only occur in very specific scenarios and then drop off precipitously in performance in other situations. Lastly, it depends on the type of stabilization the camera decides to do. An algorithm or the moving parts can sometimes make the wrong choice or perform the adjustment poorly (edit: “suboptimally” may be the better word - stabilization should be magical these days - I would not know because I do not have such a camera).
