Mode 1 or 2 on IS of Canon 70-300mm f4-5.6 IS USM

The manuals of this lens state Mode 2 is for:

- If you are taking continuous shots of a moving subject

- If you are tracking an erratically moving subject

otherwise to use Mode 1.

 

Now if I take single picture of a flying plane, a flying bird, or a moving car

setting the camera to AI-servo, should I choose Mode 1 or 2 of the IS on the

lens? Obviously the manuals choose mode 2 for a moving car in demo.

 

Thanks in advance.

For a moving car and the flying plane, use mode 2. For birds flying, turn IS off. Mode 2 disables IS from working on the X axis (horizontal axis, so if you are panning (like tracking a car), you'll still get stabilization for up and down movement, but not any for side to side because you are panning anyways. Flying birds move pretty much every which way, which is why you'll want IS off for that.

"Mode 2 disables IS from working on the X axis (horizontal axis, so if you are panning (like tracking a car)"

 

I don't think that is correct. My understanding is in mode 2 the IS unit determines the axis of movement and turns off stabilization on the axis with the largest/most linear movement.

 

So it should work OK for birds in flight as long as you pan on one axis or the other and not diagonally.

 

To quote Canon

"Figure-57 EF 70-200mm f/2.8L IS USM Image Stabilizer Effect Graph

(%)

100

50

Image stabilizer ON

Approximately 3 steps? worth

When at 200m

To resolve this problem, Canon developed Image Stabilizer

Mode 2. In this mode, if large movement such as panning

continues for a preset time, image stabilization in the

direction of the motion is shut off. As this stabilizes the finder

image during movement, accurate framing is possible. In

Image Stabilizer Mode 2, if you are panning, image

stabilization continues vertically relative to the movement of

the camera, making it possible to control vertical shaking

during panning. "

 

http://software.canon-europe.com/files/documents/EF_Lens_Work_Book_9_EN.pdf

<p>To expand a bit on the above answer, the IS system has two motion sensors (X and Y axes), and has actuators to move the IS elements in two axes to compensate.</p>

 

<p>In mode 1, IS assumes that all motion it detects is unwanted, and compensates accordingly. This mode should be used if you're trying to hold the lens still. If you pan in mode 1, IS will try to compensate for your panning. The IS elements have a limited range of motion and will bump up against the stops at the end of that range. I think most IS lens users have accidentally hit those stops at some point (I have and I've seen similar admissions posted here before by others) and it appears that the lens is designed to be able to withstand at least a small amount of such abuse, but I suspect it isn't wise to push one's luck by doing it frequently. And of course having the lens elements bouncing off the stops is not conducive to sharp pictures.</p>

 

<p>In mode 2, IS looks for one axis in which motion is basically linear. For instance, if you pan from left to right, the lens' motion in the X axis is always positive, while the motion in the Y axis will bounce around between positive and negative, more or less as it would if you were trying to hold the lens steady. In this example, IS tries to stabilize for vertical motion only. Mode 2 can cope with vertical or horizontal panning; it simply looks for one axis in which the motion is all on the same side of 0 and deduces that this is the axis in which you're panning. A corollary of this is that mode 2 works whether the camera is in portrait or landscape orientation.</p>

 

<p>But note that it expects one axis to be moving (reasonably) steadily in one direction while the other axis is unsteady around zero. So you have to pan along one axis. If you pan diagonally, both axes will show (reasonably) steady motion in one direction and IS cannot correct for this. Theoretically, it is possible to do so; if the IS system were to apply some sort of fitting algorithm to the motion it detects, it could figure out what the expected motion in each axis should be, and then correct for any variations from that. But every IS lens manual or other document from Canon that I've seen says that IS mode 2 only works if you pan along one axis, so it's apparent that this is not how it's implemented.</p>

 

<p>And note that it expects the panning to be more or less regular. If your subject is prone to sudden changes in motion which would cause you to change your panning direction, then as the previous posted pointed out, IS isn't going to be happy. It can cope with changes in panning speed; going back to the example I gave, you can pan to the right faster or slower if you want, but the motion IS detects will always be a positive number on the X axis, and since IS is not trying to correct for motion on the X axis in this example, there's no problem).</p>

 

<p>The manual's suggestion about taking continuous shots is misleading you. Your choice of IS mode does not depend on what drive mode or AF mode you select, only on what motion, if any, you are intentionally imparting to the lens. If you're trying to hold it steady, use mode 1; if you're panning horizontally or vertically, use mode 2; if you're panning erratically or diagonally, turn IS off.</p>

"If you pan diagonally, both axes will show (reasonably) steady motion in one direction and IS cannot correct for this"

 

Well it could, actually - just as easily as correcting along one axis or another. All it would need to do is work out the average direction of motion for the time period *before* exposure, and correct for deviations at right angles to this direction *during* the exposure.

 

It should be no harder in fact than correcting for x-axis or y-axis deviation.

<p>Yup, like I said, it's theoretically possible to correct for unsteadiness during diagonal panning. But Canon has had IS lenses on the market for over a decade now and has implemented new or improved IS features/capabilities a number of times since the first IS lens was released, and they have not done this. I take it that this means their research into what people want/need/will pay for has shown that this is not a priority.</p>

"Well it could, actually - just as easily as correcting along one axis or another. All it would need to do is work out the average direction of motion for the time period *before* exposure, and correct for deviations at right angles to this direction *during* the exposure. "

 

I think the problem with this is the IS correction element runs into its end-stops when there is a continuous motion.

 

Although it should be possible for the IS to shut both off when that happens. It may be late version IS systems can do this as they detect a tripod which causes a similar issue due to no movement, I am guessing this is due to inherent DC offsets in the control loops.

 

Just to clarify, turning IS off for erratic bird flight may be good advice. I just wanted to point out that the reasoning behind that was incorrect.

"it's theoretically possible to correct for unsteadiness during diagonal panning. But Canon has had IS lenses on the market for over a decade now and has implemented new or improved IS features/capabilities a number of times since the first IS lens was released, and they have not done this."

 

Steve, are you quite sure this isn't how it has always worked? it's totally brain-dead to design a pannable IS and *not* do this - and would have ben, even 10 years ago.

<p>Alec, Canon has been consistent in stating that mode 2 works if you pan horizontally or vertically, in explaining this by saying that mode 2 detects in which of these two planes you're panning and shuts off IS in that plane, and even in providing diagrams showing this. The EOS 20^th anniversary brochure I picked up at a photo show a week or so ago, for instance, has text and diagrams explaining it exactly this way.</p>

 

<p>Now, it's possible that they're lying to us and designed it to be able to pan diagonally as well. Frankly, I've never tested it to see what happens. I don't have time right now but perhaps some day I will play with this. But in the meantime, I suspect Canon has a reason for describing mode 2's operation the way they do, so unless proven otherwise, I'll take their word for it.</p>

 

<p>What you and I are suggesting is possible requires a different algorithm than what Canon says is the case. As Canon explains it, it doesn't need to figure out the magnitude of a (reasonably) steady motion; it merely needs to detect the <em>existence</em> of a reasonably steady motion. It then turns off the IS actuator in that plane. You and I are suggesting that they instead determine the magnitude of the motion using a fitting algorithm; they would then subtract this magnitude from the detected motion at any moment in order to determine what the correction should be. Not difficult at all, at least conceptually, but different.</p>