The-Digital-Picture has completed an exhaustive comparison between the Canon EOS R5 and Sony α1 and IBIS performance between the two cameras.
The comparison has 680 highly analytical test subject results, which should quash any risk out outliers. Both cameras perform extremely well in these tests, but it looks like the Canon EOS R5 performs just a little bit better.
From The-Digital-Picture
I experience a 1/3 – 2/3 stop advantage with the Canon camera. That difference is not dramatic, and the bottom line is that IBIS is valuable in both camera brands. This feature adds substantially to the versatility of non-stabilized lenses, such as the 50mm f/1.2 models tested here. IBIS is one more reason to love the latest mirrorless camera models. Read the full comparison
Canon EOS R5 $3899 || Sony α1 $6498
I still guess that they did not use OIS lenses with the R5 to keep it fair.
My reading of the data is around 4 stops on this test, but it depends where you choose your start and end points from.
I do wonder if there is a difference between focal lengths. Maybe so, maybe not. That would be interesting to test.
As stated in the article:
So you really can't generalize this result. For one, based on testing with just one focal length you can't extrapolate into how each system fares with for example a 24 and a 100 mm lens. It also can't be applied to consumer grade lenses like the Canon RF 50 mm 1.8 STM, as those constrain the image sensor motion further.
I would also expect Canon to further their advantage when it comes to combining lens IS with IBIS, since the larger mount gives them more freedom to move the sensor around. So these results also don't apply to that.
But for the specific case of the 50 mm 1.2 lens, this looks like a nice test and as that is a lens that does benefit from stabilization, it is good to see that Canon demonstrates a striking absence of Doom once more.
Well, the speed, resolution and dynamic range of the EOS 20D is good enough for 90% of the shooters out there. Actually, at least 95% of all shooters out there don't actually need anything more capable than the 1D mkII when it comes to stills. Whenever we think we do, it's just because we're generally not skilled enough.
If we specifically want the best image stabilisation, Sony just isn't it. "good enough" isn't even close to being the best and so "good enough" just isn't good enough.
So may guess is, that the electronics of IS-Systems are adapted stepwise by trial and error depending on long series of experiments: make thousands of photos with a primitive initial status, under ramdom conditions. Change parameters and compare whether something happens, change again until you see some improvement and repeat it so much times until you have good results. If this is in some way true, then IS-Systems can only have optimal results under certain conditions (those similar to the test-setting) and improvements cannot easily be done, because it would require another long series of testing.
Only guessing. :-)
They should just be based on motion sensors (Inertial Measurement Units and Gyroscopes) in the body and lens to detect how the camera moved in a given time. The sensor is suspended on a movable platform that is then offset by the measured amount of movement to effectively put it back into the place where it had been originally. Do this at an sufficiently high rate and you'll limit the relative motion between the subject and sensor.
To eliminate motion (blur) more effectively (and this is definitely done in Canon's system) methods from control theory can be used to better regulate the relationship between measurements of movement and movement of the sensor.
That does not negate your conclusion that the conditions under which IBIS is most effective are limited. But your reasoning does not sound related to the reality of engineering.