The Impact of VR on IQ at Faster Shutter Speeds

[J-see]

It amazes me that you post a video that shows exactly what I've been describing and you still don't get it.

The difference between 1/100 and 1/1000 is the gap between the opening of the 1st curtain and closing of the 2nd curtain. The curtains don't move faster and slower; the 2nd curtain just starts closing sooner. If they did a 1/2000 video, the gap between the curtains would be 1/2 that of the 1/1000; at 1/4000, the gap would be 1/4 that of the 1/1000.

If you look at athe gap between the curtains in the 1/1000 video, it appears to be fairly close to 1/4 the travel which is precisely what would be required if the travel time is 1/250 and you need to make a 1/1000 second exposure. Get it?

Something doesn't make sense and maybe it's me. If, I'd like to find out.

Let's make this as simple as possible. I set an exposure of 1/4000s. Which implies each pixel can only be exposed for that duration regardless how long it takes for the whole sensor. Let's ignore that part for now.

Now we make the slit as small as possible (one sensor pixel = 5.9micron) which implies the second curtain will starts directly after the first. Faster is not possible. At which speed needs that slit to pass over my pixel to correctly expose it at 1/4000s?

If my slit moves at that speed, at which speed do my curtains move?

 

[WayneF]

I think it may be hopeless.

but for a focal plane shutter and a 1/4000 second shutter speed, there is a narrow open slit (between the opening front curtain, and the closing rear curtain - see the previous links, Wikipedia is good), that exposes a narrow moving slit on the sensor for 1/4000 second duration (under the narrow slit). 1/4000 second probably can stop any reasonable motion. It is 1/4000 second.

However, the total curtain travel time is more like 1/200 (or some are 1/300) second... as the slit moves down the frame.

That means as the slit moves down the frame (exposing), that the early 1/4000 second exposure at the top of the frame (inverted to be seen at the bottom) occurs a few milliseconds before the 1/4000 second at the bottom (inverted to top). That can distort the moving subject - for example, the classic oval racing car wheels. The wheel moves sideways while the curtain slit moves down. Top and bottom of wheel are both exposed for 1/4000 second, but at different points in time (different points in curtain travel - which is also different points in time of wheel travel).

The curtains always move at the same speed, but the timing when the second closing curtain starts varies. 1/4000 second is a very narrow slit, and 1/1000 second is a 4x wider slit (allowing 4x longer exposure). At about 1/200 second (much wider slit - but the duration corresponding to shutter travel time - the front fully opens, and THEN the rear starts shutting), the total frame height is open for an instant, allowing instantaneous flash to expose the total frame (called maximum flash sync speed). This is still a slit moving across the frame, a slit width equal to the frame width, shutter speed equal to the curtain travel time. For longer exposures, the full frame remains fully open longer, maybe seconds (a very wide slit)

shutter speed (duration) = slit width / curtain velocity

1/200 second is 5 milliseconds curtain travel time. 3 milliseconds is possible in top end models. FWIW, the D300 and D800 curtains travel different directions, some models move down, some move up.

This is the fundamental basics of focal plane shutters.

 

[Scott Murray]

Interesting (Not) and if I was not working I would be drunk by now and its 7am lol

 

[J-see]

I think it may be hopeless.

but for a focal plane shutter and a 1/4000 second shutter speed, there is a narrow open slit (between the opening front curtain, and the closing rear curtain), that exposes a narrow moving slit on the sensor for 1/4000 second duration (under the narrow slit). 1/4000 second probably can stop any reasonable motion. It is 1/4000 second.

However, the total curtain travel time is more like 1/200 (or some are 1/300) second... down the frame.

That means that the early 1/4000 second exposure at the top of the frame (inverted to be seen at the bottom) occurs a couple of milliseconds before the 1/4000 second at the bottom (inverted to top). That can distort the moving subject - for example, the classic oval racing car wheels. The wheel moves sideways while the curtain slit moves down. Top and bottom of wheel are both exposed for 1/4000 second, but at different points in time (different points in curtain travel).

The curtain always moves the same speed, but the timing of the second closing curtain varies. 1/4000 second is a very narrow slit, and 1/1000 second is a 4x wider slit (allowing 4x longer exposure). At about 1/200 second (slower and wider slit), the total frame height is open, allowing instantaneous flash to expose the total frame (called maximum flash sync speed).

1/200 second is 5 milliseconds curtain travel time. 3 milliseconds is possible in top end models. FWIW, the D300 and D800 curtains travel different directions, some models move down, some move up.

This is fundamental basics of focal plane shutters.

There's still something that doesn't make sense but I'm out of time for now.

 

[WayneF]

I don't understand this. VR kicks in before you take a shot, not during. This is why one has to wait a second for VR to settle down.
Your second sentence has yet to be proven. Especially the contra productive stuff.

VR has to continue correcting during the exposure. This is why we might be able to handhold way too slow shots (within reason at least). We still shake the camera during the exposure, but VR compensates with equal opposite motion of a lens element to make the image appear steady on the sensor. That is what VR is.

But VR speed is finite, it has a limit, and Nikon advises turning it off for fast exposures (when we don't need it anyway). VR is a feature for slow shutter speeds.

My notion is that like most things (auto ISO, auto WB, auto FP, Vivid, etc), the default ought to be OFF. Turn it On only when you know you need it, when you have a reason, and it might do something for you, instead of hurting.

 

[J-see]

I couldn't sleep. I need to get my head around this and find the logic. I assume (hope) I'm almost there.

My flash sync speed is 1/250s. That means at a shutter of 1/250s, the first curtain has fully opened since my sensor is fully exposed and the second curtain still has to start closing.

So the speed my curtain travels is 4000 pixels (ignoring the 16) during 1/250s since it moves vertically. (if we assume 1/250s to be the correct speed)
If I then pick a shutter of 1/500s, I need half the exposure time which means the slit has to become half the sensor; 2000 pixels.

If I keep upping the shutter, my slit becomes smaller and at 1/4000s that should be some 250 pixels. That's 250*5.9 micron = about 1.5mm

Is that about right this far? In this case the movement of the curtains remains identical but because the slit gets smaller, the exposure decreases.

 

[Eyelight]

I couldn't sleep. I need to get my head around this and find the logic. I assume (hope) I'm almost there.

My flash sync speed is 1/250s. That means at a shutter of 1/250s, the first curtain has fully opened since my sensor is fully exposed and the second curtain still has to start closing.

So the speed my curtain travels is 4000 pixels (ignoring the 16) during 1/250s since it moves vertically. (if we assume 1/250s to be the correct speed)
If I then pick a shutter of 1/500s, I need half the exposure time which means the slit has to become half the sensor; 2000 pixels.

If I keep upping the shutter, my slit becomes smaller and at 1/4000s that should be some 250 pixels. That's 250*5.9 micron = about 1.5mm

Is that about right this far? In this case the movement of the curtains remains identical but because the slit gets smaller, the exposure decreases.

I think you have the basic idea.

 

[J-see]

I think you have the basic idea.

I could not get my head around it until I realized I kept using the full sensor size as the travel path for the curtains and that simply made no sense at such a slow speed. But once I realized the actual distance halves each time you double the shutter, it started to make sense.

That does imply that regardless my shutter speed VR should indeed be able to kick in at least once during all fast shutters.

But it doesn't change the fact that shooting high speed with VR enabled seriously compromises my focus accuracy during tracking and bursts.

 

[Paganman2]

Here is a pic taken with my old D90/55-300Vr its a slight crop and its with VR on at 1/640th.

P.

 

[Paganman2]

Here is another same cam/lens VR on and 1/500th.

P.

 

[Eyelight]

I could not get my head around it until I realized I kept using the full sensor size as the travel path for the curtains and that simply made no sense at such a slow speed. But once I realized the actual distance halves each time you double the shutter, it started to make sense.

That does imply that regardless my shutter speed VR should indeed be able to kick in at least once during all fast shutters.

But it doesn't change the fact that shooting high speed with VR enabled seriously compromises my focus accuracy during tracking and bursts.

There's a lot going on during a burst. Does the power going to one system mess with another?? Does VR keep sampling and correcting? If it does then during a 6 shot per second burst, it's sampling a lot more than once per shot.

Not sure if it is the same with all cameras, but back button focus does not engage the VR on some. So, if using BBF and VR, the VR is not active until the shutter release is pressed, which may start VR without it really having a chance to do it's best. I can see this being a good reason to have VR off if you prefer using back button focus.

 

[Paganman2]

Old nikon D70s with a Siggy 70-300APO No Stabilization no exif on this one but from memory i think it was a slight crop and about 1/500th.

P.

 

[Woodyg3]

Good article on VR

This may have been posted here before, but I didn't see it. Considering all the speculation about VR, this clears things up quite a bit.

Nikon VR explained

 

[WayneF]

I couldn't sleep. I need to get my head around this and find the logic. I assume (hope) I'm almost there.

My flash sync speed is 1/250s. That means at a shutter of 1/250s, the first curtain has fully opened since my sensor is fully exposed and the second curtain still has to start closing.

So the speed my curtain travels is 4000 pixels (ignoring the 16) during 1/250s since it moves vertically. (if we assume 1/250s to be the correct speed)
If I then pick a shutter of 1/500s, I need half the exposure time which means the slit has to become half the sensor; 2000 pixels.

If I keep upping the shutter, my slit becomes smaller and at 1/4000s that should be some 250 pixels. That's 250*5.9 micron = about 1.5mm

Is that about right this far? In this case the movement of the curtains remains identical but because the slit gets smaller, the exposure decreases.

Well, pretty close, better idea than before. Yes, as slit gets smaller, exposure decreases. Except it is not about pixels. It is about mm of sensor height, and velocity over mm of travel. I suppose 4016 pixels is a D750, which is 24mm tall. The D750 and D8x0 actually sync at 1/320 second, making curtain travel time be (at least) 24mm / 1/320 second = 7.68 m per second. You can compute pixel microns, but it is unorthodox. Yes, the pixels are exposed when they are under the slit, for the duration as the slit passes over them.
But sensors vary arbitrarily in number of pixels, where FX is always 24 mm tall (more universal meaning). D750 and D8x0 have very different pixel counts, but have the same FX 24mm and same 1/320 sec sync speed. The way below describes both.


EDIT: Jsee pointed out that the D750 does not sync at 1/320 sec.
So the following numbers instead apply to D800 and D700, which do.


For 1/500 second, second curtain starts closing 1/500 second after first curtain starts opening. Technically, nominal 1/500 second has to be 1/512 second, but that's only about 2%, I'm ignoring that.
Crystal controlled, very accurate. One curtain speed for every case, very consistent.
A leaf shutter is no good past 1/500 second, if even that.
So focal plane is best possible shutter accuracy, even at very fast speeds, but flash sync can be an issue. But sync is still very adequate in most situations except in bright sun.

Slit width = shutter duration / curtain velocity = 1/500 sec / 7.68m/sec = 7.68/500 = 15.36 mm, or 64% frame height (slit).

Twice as fast starts sooner, half that time, for half this slit width.

For 1/4000 second:

Slit width = shutter duration / curtain velocity = 1/4000 / 7.68m/sec = 7.68/4000 = 1.92 mm, or 8% frame height.

Which is 500/4000 = 12.5% of 1/500 slit (12.5% of 15.36mm is 1.92mm slit)

You can confirm the 7.68 m/sec number by measuring the pixels in the non-dark band of flash at 1/500 or 1/4000 second. It could be a bit faster? All we know is that it does sync at 1/320 second.

My notion is this: You're obviously pretty sharp, and clearly from a technical background, but which definitely was not photography. New at forums, and you tend to explain in detail to others about how things work, for things that you are seeing for the first time. Because you're Jsee, you know? You're used to explaining, but that has downsides when you make things up. It is common, we all tend to do that (to imagine how it might work), but we can learn to really pay close attention to if we actually know or not. We do know the difference.
Tip: Try asking: How did I learn what I am about to say? How sure am I? Did I know it yesterday?

Hope that helps.

​

 

[mikew_RIP]

Re: Good article on VR

This may have been posted here before, but I didn't see it. Considering all the speculation about VR, this clears things up quite a bit.

Nikon VR explained

Thats the one that explains if Vr works during high shutter speeds its down to poor tripod or hand holding technique,as i have said i do have poor hand holding,the other thing i dont think he considers is that one reason pro sports shooters turn it of at high shutter speeds could be to avoid waiting for VR to settle or taking while it is settling

 

[J-see]

My notion is this: You're obviously pretty sharp, and clearly from a technical background, but which definitely was not photography. New at forums, and you tend to explain in detail to others about how things work, for things that you are seeing for the first time. Because you're Jsee, you know? You're used to explaining, but that has downsides when you make things up. It is common, we all tend to do that (to imagine how it might work), but we can learn to really pay close attention to if we actually know or not. We do know the difference.
Tip: Try asking: How did I learn what I am about to say? How sure am I? Did I know it yesterday?

Hope that helps.

​

Well Wayne sometimes I get it wrong, sometimes I get it right. I need to get it wrong to get it right.

It doesn't help that I can't rely on any information out there and have to double-check everything myself. Which is why I know my manual says 1/250s for max flash sync speed.

Maybe there's some truth in your words that goes for the both of us.

 

[J-see]

There's a lot going on during a burst. Does the power going to one system mess with another?? Does VR keep sampling and correcting? If it does then during a 6 shot per second burst, it's sampling a lot more than once per shot.

Not sure if it is the same with all cameras, but back button focus does not engage the VR on some. So, if using BBF and VR, the VR is not active until the shutter release is pressed, which may start VR without it really having a chance to do it's best. I can see this being a good reason to have VR off if you prefer using back button focus.

During this period of the year the Tam has a harder time grabbing and holding focus. f/6.3 at the long end isn't helping either. I always notice that during burst which presents itself as a reasonable fail-rate. But the moment I forget to switch VR off, the number increases quite a lot. I tested it and each time I forget it, it's immediately obvious VR does do no good there.

I don't know if that is the result of too many things happening at once, VR trying to correct for shake while the lens has to correct for subject movement and the one compromising the movement of the other or maybe the one simply causes a delay which slows tracking down and as a result, I shoot more out of focus shots. At some birding site I read VR slows down AF when using AF-C but there's no explanation to it.

At 600mm it doesn't need much to not be in focus.

That's why as a general rule, I don't use VR when shooting above 1/400s. At lower speeds, my subject is normally static.

To add: I use BBF but the moment I press focus, VR jumps into action since I see the viewfinder stabilize. If Tam follows the same logic as Nikon, lens correction starts the moment I release. I read on other cams it doesn't activate VR but that's clearly not the case for me. At least not in regards to the first algorithm.

 

[Vixen]

I soon get bored with all this technical stuff I leave my VR on except if using a tripod. My hands are not steady so even at 1/50000000000000000 I reckon any help is good

I'm sure there is a technical answer but my brain couldn't handle wanting to find out

 

[sonicbuffalo_RIP]

I can get too confused for all this nano tech talk. Oz

 

[WayneF]

It doesn't help that I can't rely on any information out there and have to double-check everything myself. Which is why I know my manual says 1/250s for max flash sync speed.

Sorry, that is my mistake then, assuming the D750 also has the D800 shutter. The D700 did sync at 1/320.

Assuming always gets us in trouble.