Image stabilization and tall buildings

[Busted Knuckles]

Pat, I think you explained quite nicely. Sort of how does the IS system know what is an intentional pan vs. an unwanted shake - assuming you haven't set the IS mode or the lens doesn't have multi modes.

Higher frequency, large amplitude vibrations are going to tend to create higher accelerations. And within reason, higher frequencies are going to be a bigger contributor to higher accelerations as opposed to increased amplitudes, but IS is going to respond better to larger amplitude changes only if the frequency is high enough to produce a detectable acceleration. That was a very confusing, and loaded sentence... better said, it is the accelerations that create the forces needed to measure vibrations... and that goes for IS systems as well. With a building there a couple of different vibration modes - 1) those induced by the buildings equipment (motors, elevators, HVAC, roof mounted equipment, etc.) which tend to be higher frequency and lower amplitude; and 2) the sway of the building due to environmental inputs (wind, thermal expansion of sunny side vs the other side, etc.) which tend to be higher amplitude and very low frequency. The sway of the building can have some rather large amplitudes, but the frequency is very low and is not likely to be seen by IS (unless there is some higher frequency torsional component). However, I would say that the man made/induced vibrations from mechanical systems would be something that the IS would handle.

Back when I was in college (pre-invention of the wheel, but dirt was invented that year) we had a 12 story Mechanical Engineering building. Being the nerds we were, we put some accelerometers on the upper floor and a few of us would run back and forth to get it swaying... then measure the acceleration and calculated the displacement. I was really surprised how much it moved. I do not remember the number, but it was in the inches. And others had measured in high winds and found the displacement to be quite high. If you were on the 24th floor in any kind of wind, I would think you would have problems with blur, but only if you were shooting perpendicular to the direction of the wind or if the wind set up some torsional motion (which does not seem good for a building).

There... my nerd fix for the day.

Pat

[conropl]

Pat, I think you explained quite nicely. Sort of how does the IS system know what is an intentional pan vs. an unwanted shake - assuming you haven't set the IS mode or the lens doesn't have multi modes.

I am not an IS expert, but how I would think it works:

1. For Canon the IS system senses either 2 axis of rotation (pitch-yaw) or the newer IS systems have 3 axis of rotation (pitch-yaw-roll). Hybrid IS for macro will also sense accelerations in x and y axis (I am not sure about z axis... I would doubt it since the resulting IS lens shift is only in the x-y axis).
2. Those movements are then translated to an x-y plane shifts to compensate.
3. If you go to mode 2 for panning, I would think you would ignore the yaw or reduce the sensitivity of the yaw sensor when reporting back yaw position differences (I suspect the later). You may still want some yaw effect for rapid positional changes over time (less amplitude based) and cyclic acceleration (vibration), but ignore the slower rate of change in the yaw rotational position cased by panning. Panning induced movement is going to be relatively lower accelerations. Ideally panning would consist of, starting and stopping accelerations with constant speed - no acceleration - during the pan. So the input for the x-axis motion of the lens would be filtered, or the yaw sensitivity to rate of change would be reduced such that you would get an output to the shift axis only if a given threshold rate of change in the yaw is reached or a cyclic acceleration in the yaw is detected. I would guess it is strictly a reduced sensitivity in the Yaw rate of change because anything else would take to much computing power that I doubt the lens has.
4. If it doesn't have multi mode IS, then I do not think it does know you are panning, and it will happily mess you up. Try panning a large lens with out mode 2 active, and it does seem to jump more because it is try to compensate for your intentional movement.

Some conjecture on my part because researching the inner thoughts of the Canon engineer that programmed the IS system is not readily available, but that is how I would think it works.

Pat