"Reason: low shutter speeds can cause blurred images on individual frames when the camera moves."

A long known about problem, the motion blur can also create trails that move around static objects for no apparent reason. Also the blur can make it harder for the trackers used in the stabilizer to lock on properly.

On the other hand if your shot was an acceptable speed pan at 25 or 30fps the lack of motion blur would result in a lot perceived judder.

There is no one solution. If you're intending to stabilise a static shot in post and there's little movement in the frame then a fast shutter speed is definately the way to go although depending on the camera you might also need to consider the attendant impact on DOF. If you're doing a handheld pan then sticking to a less agressive shutter speed than 1/2000 would be the way to go.

Personally I've never found it that hard to find a solution to keeping a camera reasonably static even to pull off a pan. A picnic table can double as a tripod and a plastic bottletop as a pan head. For static shots there's a considerable number of cheap solutions that are no drama at all to carry around, at a pinch most of them can also help you get a reasonably stable and useable pan.

Bob.

Wonderful, brilliant, and very useful demonstration!!

As Bob said, when you know you will need to do stabilization in post, the need to use a faster shutter speed than the default 1/50 or 1/60 is something that is widely known. However, I don't think anyone has ever taken the time to do a test to show the difference.

This post will be a very useful reference.

Personally I've never found it that hard to find a solution to keeping a camera reasonably static even to pull off a pan.It always seems that whenever anyone has a post on stabilization, someone always says something like this, namely that you should never need stabilization if you are a pro (because a pro always uses a tripod, or a pro has good, steady camera technique). I can't argue with that, because it is a true statement. However, budget does rear its ugly head for most of us, and too much coffee (or advancing old age) can render the steadiest hand a little jumpy.

Also, many of us do things with cameras beyond just standing there and panning the camera or holding it steady while shooting a distant object. For instance, I take aerial video. How do I hold the camera steady when the small plane, at 300 feet above the ground, is bouncing around so badly that the pilot is handing out barf bags? Yes, if I had lots of money, I'd get one of those rigs the news helicopter crews use, but I don't have that kind of money.

Same goes for shooting from a car and walking with the camera (I can't afford a dolly, and usually don't have time to set up for run-and-gun video).

Finally, I mostly restore other people's video (and film) and have to deal with what I'm given.

I spent a lot of time writing all those guides and scripts for Deshaker because I realized a lot of people need this technology. I then purchased Mercalli when Jerry showed how rock steady it could make footage -- far better than Deshaker.

And, I am not upgrading to Vegas 11 because of the way they "broke" the stabilization workflow (hopefully it will be changed back to the way it was).

That is so interesting! Wow!

Heh, heh. Yeah, I stacked the deck on this demo. By turning off my camera's optical stabilization and zooming to the max, the footage was really shaky. I did try to hold the camera as steady as I could, however.

The reason I made this because of what Bob & John said... "this is a commonly known problem." However, to my knowledge, no one has ever put together a side-by-side to illustrate exactly what happens.

Therefore, next time someone posts some footage that doesn't stabilize properly, we now have a reference video to show the effect of low shutter speed.

...Jerry

PS: The corollary to the crappy job of stabilizing the 1/60 footage is what a fantastic job Mercalli does of stabilizing the 1/2000 footage (imho). The source was really, really shaky and stabilized footage is tripod-like.

Now, if ProDad (Mercalli) could only fixed their border building algorithms to match Deshaker's.

I stabilize handheld footage routinely. Add to this my DSLR's tendancy to use really short shutter speeds in bright light to make up for the lack of built it ND filters, and you have something that can be quite useful: the idea that handheld static shots without intentional movement that you are going to want to stabilize later can benefit from very fast shutter speeds.

This also might reduce cmos wobble which can be terrible on handheld footage.

This also might reduce cmos wobble which can be terrible on handheld footage.It might be interesting to start a new thread and see if anyone has useful information on whether this is true. I had this same thought a few months ago, but since I don't yet own a CMOS-sensor camera, I haven't been able to test it out. It sure seems like you could virtually eliminate the vertical line bending and jello effects if you used a fast shutter speed, but the only way to know for sure would be to do some tests.

Has anyone tested this idea?

I haven't tested this directly, but I do know that I have been using way faster shutter speeds on my DSLR than I would use on my HDV Camcorder because the DSLR doesn't have built in ND filters and I wanted to use wider apertures. On the Nikon D5100, the prevailing knowledge is that you don't have manual control over the settings of shutter speed, aperture and ISO in video modes, and for a while I agreed with this and let the camera adjust shutter speed and ISO on it's own (while I just set aperture in AP mode). The auto control of the Nikon tends to use very fast shutter speeds. More recently I have begun to realize that this isn't totally the case. For instance, you can set a minimum shutter speed in the master settings. If you set this value to 1/60th of a second, the shutter speed will be 1/60 in all but the darkest environments, at which point it will go up towards 1/30th of a second, something that I sometimes do on camcorders anyway when I need more exposure and don't mind locking down on a tripod. On the Nikon D5100, with the minimum shutter speed set to 1/60, in AP mode I have control over aperture and the exposure is varied by the ISO value which is automatic. Unless it's really dark, the shutter speed stays at 1/60 and I can see what's going on in the display. Not bad at all really.

Anyway, the point of this rambling is that I had grown accustomed to quick shutter speeds for a while and did notice more rolling shutter after I learned how to set the shutter speed to a standard half the frame rate of 1/60 for 30p. Not a formal test, but yes, I believe that that is the way it is.

With the faster shutter speeds everything looks fine as long as you really slow down the camera movement. Yes there is still a little judder on the moving people in the shots, but it really looks fine. Not at all like the horrible judder you get on a quick pan or zoom.

With the DSLR, I really try not to zoom at all. Among other things, the exposure does not change smoothly when you zoom. I do a little zooming in post on the ads I do because they are displayed in the theater at 720p but shot at 1080p, so I can get away with it.

Very interesting and useful comparison.
Makes me feel a bit more comfortable having advocated higher shutter speeds in another thread.

I don't mind using lower shutter speeds either if the camera is locked down. Better than adding gain in many cases.

Look, if you are doing a smooth pan, a shutter rate of twice the frame rate makes for a bit of motion blur between the points of action. This motion blur is not good though if you are stabilizing because it will be blurred in the wrong direction. It makes sense if you are outside and handholding a shot that you know you are going to stabilize later, that you use a much faster frame rate. It will give you the ability to make it almost tripod steady later on. Yeah, there are lots of times when I would use this now that I know.

"It sure seems like you could virtually eliminate the vertical line bending and jello effects if you used a fast shutter speed, but the only way to know for sure would be to do some tests.

This has been tested. The critical value that affects skew with a CMOS sensor is readout time i.e. the time from when the first row of pixels is read until the last row is read. That value has no connection to how long each row is exposed. For the EX1 I think readout time has beed tested and verified by Sony as 1/60th second.

To some extent fast shutter speeds can make skew look worse as the normal motion blur to some extent masks the skew. One solution to skew is to use a mechanical shutter which is expensive. The only camera I know of that does this is Sony's F65.

"It always seems that whenever anyone has a post on stabilization, someone always says something like this, namely that you should never need stabilization "

My point wasn't that you should never need it, rather that with a bit of creative thinking you can reduce how much of it needs doing. I've used pillows, backpacks and old rags. I've had lots of problems with skaky shots despite using expensive tripods. Not too much you can do when the floor moves.

Bob.

"Not too much you can do when the floor moves."

Steadycam?

"Steadycam?"

For two hours???

Too old, too fat for that :(

What can help is to put shot / sand bags under the legs of the tripod, the more the better. That adds a bit of shock absorbancy and it loads down the sprung floor.

Bob.

Can a Steadycam be mounted on a tripod?

Mercali has a preset for stabilizing a shaky tripod. It actually works quite well...assuming you used a fast shutter speed.... ;-)

Does anyone have any thoughts on whether or not the stabilization process adds any sort of pseudo-motion blur, so that the high shutter speed footage looks more natural stabilized (for judder, etc) than it would as unstabilized?

(and thus blends well with slower-shutter footage)

Can a Steadycam be mounted on a tripod?

Technically yes, but practically no. A "real" steadicam with the sprung arm mounted to a body brace and also be mounted to any other bracket you can rig/manufacture. So you could hard mount it in a car or a helicopter. or a heavy camera dolly. But because this setup isn't actually balanced it'd tip over a tripod.

I think it was pretty much agreed that if you need high frame rates you should use them, and that stabilization was one of those special applications where high frame rates help.

Rob

The critical value that affects skew with a CMOS sensor is readout time i.e. the time from when the first row of pixels is read until the last row is read. That value has no connection to how long each row is exposed.Wow, is that really the case? (I have absolutely no information to the contrary, so I am not questioning what you say.)

So, if I understand what you are saying, the top to bottom scan rate is fixed, no matter what the shutter speed. Given that the camera is nominally shooting either 25 or 30 frames per second (when shooting progressive), and that there are approximately 1,000 vertical lines (in HD), then each 1/25,000 or 1/30,000 (or maybe its 1/50,000 or 1/60,000) of a second the next scan line is exposed using whatever shutter speed has been set. Thus, all the artifacts caused by the scanning should be completely independent of shutter speed, and actually might be far worse at a higher shutter speed.

Hmmmm ....

I still don't think I fully understand what is going on with all these CMOS artifacts. In fact, the question I had from the very beginning was (and is):

"Why didn't we have this problem with original analog video, because that too scans each and every scan line at a different moment in time, and in fact every point along the scan line in NTSC or PAL analog video, taken using an old-fashioned scanning tube camera comes from a different point in time."

Therefore, we should have been seeing exactly the same skew, wobble, jello, flash, and other artifacts in all of our old videos. The skew of vertical lines during a fast pan should certainly have been evident in analog video taken with raster tube cameras because each line was taken at a different moment in time.

So, speaking strictly for myself, I still don't think I fully understand what is really going on with these CMOS sensors. Also, some CMOS cameras apparently have a much worse problem than others. Shouldn't they all exhibit exactly the same degree of problem when filming the same subject, in the same way, under the same conditions?

You've got me thinking Bob.

[edit] I'm trying to research to better understand this and came across the most bizzare rolling shutter artifact I've seen to date:

Kind of a motion version of moire.

"the most bizzare rolling shutter artifact I've seen to date"My God! That is bizarre. Looks like the aliens are deploying their stealth bombers!

...Jerry

Here is a video that explains the airplane propeller thing:

Amazing animation, Laurence. Thanks for sharing that.

I've done a little research this morning, while waiting for various things to render (how many of us post here during renders??), and found a few things.

On page 2 of this thread (I don't see a way to link directly to the specific post):

Just Say NO to CMOS Cameras!

the forum sysop says that he did a PhD dissertation on this problem in Vidicon cameras, so I guess perhaps the problem does exist in all of our old video. I'll have to go back and look critically at the background in my old sports videos (I had a Saticon tube in my first 1981 video camera) and see if I can find evidence of it when the camera is panning rapidly to follow the motion. The lines on a football field, or flag pole in the background should provide a good test.

As always, johnmeyer gives us an informative perspective on an issue.

Besides, assuming that you are shooting at 30p, below 1/30 shutter you just get frame accumulation. That creates a nice special effect with relatively motionless shots of flowing water, if you like that kind of thing. Above roughly 1/500 (some even say 1/250) you start to get staccato "Saving Private Ryan"-like motion. Think about it. At 1/000 sec for example, you have a gap of ~(1/30)*(999/1000) of a second between frames.

The shutter speeds in video cameras are pretty much limited by their frame rates and although they have settings for high and low shutter speeds, they are just for creating special effects (which personally, aren't that special).

If you want less motion blur (and hence more effective stabilization), the usual solution is to buy a camera that shoots at 50p/60p or grater. I'm a little surprised at the results in the OP, but that may just be a result of stabilization algorithm in VP11.