[GUB]

No it is a massive crop.

In that case I consider the droplet trail profile to be a record of the sensor movement.
And it is an example of sr working correctly.
We can not expect it to stabilise diverse motion.

---

Last edited by GUB; 06-28-2022 at 02:49 AM.

[beholder3]

In that case I consider the droplet trail profile to be a record of the sensor movement.
And it is an example of sr working correctly.
We can not expect it to stabilise diverse motion.

Oh, if it sounded like I said that it is not working properly, my bad, sorry.

Yes, I do believe it is working properly and we are seeing just the effects of it working properly. I just wanted to show that there can be effects in the image under very specific circumstances.

There have to be. During the exposure user shake moves the sensor out of the place where it should be and SR brings it back into position. All while light hits the sensor.
In my "lightpainting" scenario we can (I assume) see how there is movement and then compensation. Even then best sensors and magnets need some finite time due to inertia to bring the sensor back to where it needs to be.

[Digitalis]

The deviation can be human shaking, it can be shutter/mirror shock. The returning can be SR or it can be just the sensor or camera swinging back.

For any test to be valid these variables need to be quantified, mitigated or isolated. I'm not suggesting IS doesn't have deleterious effects upon image quality, I'm saying that in order to identify any such effects: there is a need be more rigor and control in testing parameters than a few photos of sunlit water droplets from a common garden sprinkler.

I posted a thread on the subject of optical testing - this equally applies to when one seeks to test cameras

If the droplet was oscillating it would be a sine wave and even then the wave would have a lower amplitude than the width of the droplet.

Variables such as droplet mass, air resistance and prevailing winds add a chaotic element to the droplets movement: noise if you will. Also droplets are capable of moving in three dimensions which may give false positives when translated into a two dimensional image.

[GUB]

Oh, if it sounded like I said that it is not working properly, my bad, sorry.

Yes, I do believe it is working properly and we are seeing just the effects of it working properly. I just wanted to show that there can be effects in the image under very specific circumstances.

There have to be. During the exposure user shake moves the sensor out of the place where it should be and SR brings it back into position. All while light hits the sensor.
In my "lightpainting" scenario we can (I assume) see how there is movement and then compensation. Even then best sensors and magnets need some finite time due to inertia to bring the sensor back to where it needs to be.

Yeah but something is still not right here. I get that the horizontal movement of the droplet won't be stabilised because the sr is concentrating on the background. But the droplet in a vertical sense is stationary against the background and so the camera shake / shutter shock would effect it identically as the background and thus the sr would stabilise it. So those notches then don't make sense. (They are after all vertical correction)

[beholder3]

Yeah but something is still not right here. I get that the horizontal movement of the droplet won't be stabilised because the sr is concentrating on the background. But the droplet in a vertical sense is stationary against the background and so the camera shake / shutter shock would effect it identically as the background and thus the sr would stabilise it. So those notches then don't make sense. (They are after all vertical correction)

With regards to "affecting this but not that" I believe that my reference to lightpainting is the answer.
The bright droplets leave a big mark on the sensor, while during the relatively long exposure the dark background is not impacted much by the sensor being briefly "somewhere else".
If the background was focused on and in bright sunlight versus the foreground was dark I think the situation would be different.
That is pure speculation from me of course.

[GUB]

The bright droplets leave a big mark on the sensor, while during the relatively long exposure the dark background is not impacted much by the sensor being briefly "somewhere else".

Do you mean you feel the background detail did the same jump?

[UncleVanya]

Yeah but something is still not right here. I get that the horizontal movement of the droplet won't be stabilised because the sr is concentrating on the background. But the droplet in a vertical sense is stationary against the background and so the camera shake / shutter shock would effect it identically as the background and thus the sr would stabilise it. So those notches then don't make sense. (They are after all vertical correction)

SR isn’t “concentrating on the background” it is sensing camera movements and compensating to keep the light falling on the sensor aligned to reduce camera movement from creating motion induced blur.

I think a pendulum against a static background is perhaps a more reliable test to try. I’m not in a position to perform the test as I’m away from home without my gear. I’d also be interested in seeing other systems subjected to the same test (Sony, Olympus, Panasonic etc. )

It is interesting and compelling as a subject of investigation and discussion.

[beholder3]

Do you mean you feel the background detail did the same jump?

I assume that yes. It looks like the bulk of the deviation only took about 1/8th or so of the total exposure time and with it being pretty darkish I could imagine that a brief shake did not create a very deep impression on the background image.

Lots of things one could test in different setups. :-)

[biz-engineer]

What should we do about it?

[GUB]

What should we do about it?

Nothing but observe at the mo. It is probably an example of sr working correctly.

[GUB]

I assume that yes. It looks like the bulk of the deviation only took about 1/8th or so of the total exposure time and with it being pretty darkish I could imagine that a brief shake did not create a very deep impression on the background image.

Lots of things one could test in different setups. :-)

Yes a bright reflective point as part of the background.

[leekil]

Yeah but something is still not right here. I get that the horizontal movement of the droplet won't be stabilised because the sr is concentrating on the background. But the droplet in a vertical sense is stationary against the background and so the camera shake / shutter shock would effect it identically as the background and thus the sr would stabilise it. So those notches then don't make sense. (They are after all vertical correction)

But it is also a 2-dimensional image capture of a 3-dimensional scene. If the camera isn't held perfectly perpendicular to the scene - for some definition of what elements you consider the "scene" plane - you can get different effects if things (like the droplets) are moving three-dimensionally and not staying in any particular plane, as the camera tries to compensate for the movement.

[GUB]

But it is also a 2-dimensional image capture of a 3-dimensional scene. If the camera isn't held perfectly perpendicular to the scene - for some definition of what elements you consider the "scene" plane - you can get different effects if things (like the droplets) are moving three-dimensionally and not staying in any particular plane, as the camera tries to compensate for the movement.

It would take quite a large movement of the camera to create that perspective shift. As in inches of movement.
The sensor is on the image plane so any movement of the sensor could not create a perspective shift.

[Digitalis]

SR isn’t “concentrating on the background” it is sensing camera movements and compensating to keep the light falling on the sensor aligned to reduce camera movement from creating motion induced blur.

Correct, the effects of IS systems are by design affect the entire image, and not select portions of it. The camera is compensating for its own movement, not the subject*. This and the aforementioned reasons I highlighted above are reasons alone to consider this "test" to be flawed.


* one small exception to this case is when panning while using IS, it is recommended to spin up IS during a horizontal panning arc so the camera can register the panning motion so the camera only corrects for vertical motion inconsistencies.

[house]

Correct, the effects of IS systems are by design affect the entire image, and not select portions of it. The camera is compensating for its own movement, not the subject*. This and the aforementioned reasons I highlighted above are reasons alone to consider this "test" to be flawed

You need to explain why the droplet test is flawed because IS affects the entire image. Can't follow the logic when considering light path, distances, oof etc.

I'll be setting this up with a uwa lens and checking corners. There has been mumblings about ibis and uwa's and I was for a long time considering if it made my da15 worse (now I know its the optics) Still interested in seeing the effect on various focal lenghts.