The Impact of VR on IQ at Faster Shutter Speeds

[J-see]

Another thing i was thinking - could the physical handling of the lens have any impact on Vr? E.G if you are holding the lens pointing upwards with the wight of the lens falling towards the back, could this have any affect when compared to holding the lens flat where the weight is uniform?
Just an idea probably a stupid one...

P.

It's hard to check. While shooting vertical might imply gravity has an effect on VR, it at the same time makes the cam more vulnerable to shake because now gravity "pulls" the cam down while it can much easier move in all other directions. We're no longer stabilizing it the same way.

I assume shooting vertical handheld induces more shake simply because of the nature of such sort of shooting.

The gravity will have the same effect on the VR system as it has on the AF. Which will be very minimal at best. For the full AF system itself it might be slightly worse since now the motor has to start lifting weight. VR does micro adjustments in comparison.

 

[mikew_RIP]

Another fast shutter speed 1/3200th but i let it run on continuous,images have had auto ACR because they where under exposed but i have done no other PP

Obviously Highish ISO

 

[J-see]

Another fast shutter speed 1/3200th but i let it run on continuous,images have had auto ACR because they where under exposed but i have done no other PP

If you processed them all identical, it is interesting that again there are exposure differences being applied by Adobe.

 

[Eyelight]

If you processed them all identical, it is interesting that again there are exposure differences being applied by Adobe.

The shots are all slightly framed different, which will cause and auto functions, in camera and post processing, to see a different image.

Another fast shutter speed 1/3200th but i let it run on continuous,images have had auto ACR because they where under exposed but i have done no other PP

Obviously Highish ISO

Couple interesting things with the series.

They all appear to acceptably sharp, at least at this viewing size.

The change in framing, I assume is the VR moving the image, which at the same time, means the camera was moving. We know this is how it's suppose to work, but interesting to see the evidence of how much compensating it can do.

Was VR activated for a moment before the shooting began?

 

[mikew_RIP]

The different framing will be down to my hand holding,i always try to let VR settle with static subjects, with moving ones its not possible to tell when it has, as normally everything happens too fast.

 

[J-see]

I noticed the same when shooting high speed series of a static yesterday. You always move slightly during when hand holding.

 

[Eyelight]

I suppose it's a combination of camera movement and VR compensating. I have read somewhere that, when using VR, the image recorded may not match what you saw in the viewfinder. I assume it would be a small difference, but makes sense the potential difference would be greater as the focal length increased.

 

[mikew_RIP]

Not really sure what else this thread needs,i was happy with VR and high shutter speeds and still am.I mentioned in another thread and i think J-see said something similar in this one,i wonder what problems would be induced with a combination of AF-C,high shutter speeds.VR and a fast moving subject,not sure how you could test that and quantify any results.

 

[J-see]

I suppose it's a combination of camera movement and VR compensating. I have read somewhere that, when using VR, the image recorded may not match what you saw in the viewfinder. I assume it would be a small difference, but makes sense the potential difference would be greater as the focal length increased.

I read that too but that seems to be about the image in the viewfinder when VR is using that algorithm contrary to the image that will be projected on the sensor. Which makes me wonder that if I point my focal points at an eye using VR and that does not match the actual image, am I still focusing on the eye?

 

[J-see]

I was doing a very basic calculation of VR corrections during my exposure. Now to keep things simply let's assume my curtain speed is 1/250s, VR detects every 1/1000s and corrects 1/1000s after detection.

What happens at a 1/1000s is roughly like this (BC - back curtain, FC - front curtain):

1/1000s (duration 1/222s)


0,000s : BC 0 - FC 0
0,000s : VR reset and active


0,001 : BC 0 – FC 25
0,001 : shake detected – correction01 signal


0,002 : BC 25 - FC 50
0,002 : shake detected – correction02 signal
0,002 : correction 01 applied


0,003 : BC 50 - FC 75
0,003 : shake detected – correction03 signal
0,003 : correction 02 applied


0,004 : BC 75 - FC 100
0,004 : shake detected – correction04 signal
0,004 : correction 03 applied


0,005 : BC 100
0,005 : shake detected – correction05 signal
0,005 : correction 04 applied
0,006 : correction 05 applied


The commands executed at 0.005s is unclear to me because there's no way I know what occurs first. Still, once the signal is sent, it will go through the routine since there's nothing that can override it.

Anyways; what becomes obvious is that any correction that occurs is only applied to a part of the shot since the back curtain is already on the move together with the front.

Up to 1/500s the full sensor is exposed at the time of the first correction which applies it to the full shot but the higher in shutter speed we go, the less that area will become. Each doubling of the speed halves the gap between the FC and BC.



The curtain speeds for cams can differ and it's a little different than this rough example but at higher shutters it'll result in the same behavior. The main issue I see with this is that at a high enough speed to freeze any movement, the actual correction could do more harm than good.

If anything is wrong in what I mention here, feel free to point out.

To visualize the difference in a somewhat exaggerated example:

Without VR shake would result into a distortion of the image during the exposure at higher shutters since it occurs gradually:



While a sudden correction would be very different, especially since the correction is done for a situation that already changed: