[richandfleur]

It’s certainly pretty interesting

Feels like composition adjust during video

Bug wise, the crop applied during video (for Movie SR to have some wiggle room) doesn’t go away when you turn Movie SR off.

I wonder if some other aspects of Movie SR don’t turn off here?

Cool example to discuss though 👍🏻

[Andrea K]

Hi,

I retried and I was wrong. The "Composition aid" there is only with Movie SR enabled then, as many of you already wrote, it is a side-effect of this.

I think that, according to the subject, it can be useful or harmful, in my case it was useful.

Cheers!

[richandfleur]

Yes because there are no physical components in use, Movie SR detects their movement as camera shake, and tries to correct it by pulling the subjects back into the middle of the frame.

Illustrates how much of a crop is present in video mode, even when Movie SR is turned off.

All Pentax stabilisation should be switched off when using a tripod, unless you want this effect. At other times it may not behave as usefully as this though.

[mcgregni]

Yes Richard, and just for the sake of clarity and consistency, the terminology "Move SR" is the name for the newer, software based, stabilisation feature during video recordings on Pentax DSLRs (from the K3 and newer). This is often referred to as the 'Jello' effect !!


The older type of system, used on the K5II and older models, is known as "mechanical shake reduction", meaning the use of the sensor shake system to move the sensor during video recording for stabilisation.


As you said, both types should be switched off when using a tripod. The "movie SR" is prone to try and stop actual subject movements, leading to poor effects. Mechanical SR does not respond to subject movements, only camera movements, but just as like for still photos, it is optimal for handheld situations, and has a negative effect, reducing sharpness, on a tripod.


BTW, I knew the K5 & K5II & continued the "mechanical" system, but what about the K30? Did that also use sensor based SR for video?

[house]

That's quite funny. It must be fooled by the synchronized movement of the group! Your camera gave the performers the best possible review by interpreting them as camera shake as this suggests they were very much in sync indeed ;-)

To trigger this must be very rare.

[richandfleur]

K-5 series was the last to have real sensor stabilisation.

K-30 and K-01 both feature the Movie SR software approach. So that’s 6 years now with no real IBIS on Pentax during video.

[monochrome]

Don’t use AF at all. Professional videographers manually focus, and pull focus while recording as necessary. That’s what QuickShift is for.

.:

[peterh337]

Surely the K1 shake stabilisation (by moving the sensor around) is done solely using accelerometers in the camera, not by looking at the image content?

[mcgregni]

Only for still photos. ..... In video mode the sensor SR system is deactivated and the software based "movie SR" comes into play.

We were given hope some time ago, from a Ricoh representative, that sensor SR in video mode would be reinstated on the K1 with a firmware update, but sadly this has not happened.

[peterh337]

I suspect the reason is that video stabilisation requires a large amount of sensor movement, whereas still stabilisation needs only a little bit to deal with the time the shutter is open.

Accordingly, video cams which do software stabilisation (usually they use accelerometers, not by looking at the image) have a substantially oversize sensor, because they use only the centre area of it. They can use the accelerometer signals to vary which part of the sensor is used for the video data. This is also why, for some years, none of the "action cams" offered stabilisation at 4K: they all use the same Sony 4K sensor which (when used at 4K) has no spare area around it. Recently this was solved, in e.g. the latest Go-Pro and Sony (X3000) models by adding optical stabilisation whereby a portion of the lens assembly is moved around electromagnetically. Optical has other problems though; it has a slow response and can produce "jello" if the camera is subjected to vibration. I had to modify the X3000 for use when mounted on an aircraft, due to the jello issue. And I suspect moving the sensor has exactly the same problem, of very limited capability for dealing with movement frequency spectrum.

The K1 sensor has very little movement available to it. For video stabilisation, to be like the action cams, you would need a movement of some 1/3 of the sensor size i.e. some 15mm.

Another thing is that moving the sensor needs a lot of power, and on a video this power is needed continuously. On a still, it is needed only for the shutter curtain movement duration. This affects the design of the camera's power circuitry, battery life, etc.

[mcgregni]

Some very interesting points there, thanks. I learned early on with my K7 to take great care with SR activated in both Live View and Video modes (when the SR system is working continuously), because any sudden or large movements of the camera cause a rather alarming sound from the sensor. It sounds very unhappy and I certainly got the feeling that some damage or strain could happen, especially at long focal lengths when the effect is more extreme.

[peterh337]

Compare this video, done with a stabilised camera

Vimeo

with the same or similar scene, done with a handheld camera

Vimeo

I would say the sensor displacement required to stabilise the background (which is all that a stabilised camera can do, or should do) is 10-20% of frame size. That is a lot of movement, which needs to be done very fast. Making wire connections to the sensor which withstand that flexing long-term would be a challenge, too.

If you do it optically, with a movable lens assembly which is electromagnetically manipulated, you can do it with much less displacement.