K7 composition adjustment = slight shift effect?

ricardobeat · 05-21-2009, 02:20 AM

Seeing that the sensor can move 1mm to 3mm in any direction, is it possible that it could have a shift effect? (I really don't know how to do the math)

*RuiC* · 05-21-2009, 04:05 AM

Originally Posted by ricardobeat

Seeing that the sensor can move 1mm to 3mm in any direction, is it possible that it could have a shift effect? (I really don't know how to do the math)

No math to do. Just drop the camera on a tripod, coarse adjust the position and fine tune the composition using the arrow buttons. No perspective correction.

Turo · 05-21-2009, 06:01 AM

If you move the sensor , you are making the same relative movement of a shifted shift lens. I think that you would get a similar shift effect.
The problem is that a shift lens normally can shift up to 11mm and the shift movement of the sensor only 1 to 3 mm.
So if the portion if the image you can capture with the lens and the sensor is proportional I think that you would get the same effect as if you shift a shift lens 1 to 3mm. A little shift effect I think.

Anyway we should wait and see how is designed this movement as I think this is a quite special feature, but they don't want to talk a lot about it... maybe because it is something new they don't have experienced and they want to know how this works.

Ricardo are you sure the shift movement is up to 3mm? I heard this was only 1mm. 3 would be much better...

Marcoz · 05-21-2009, 06:57 AM

my k10d had tilted and non-centered sensor ( what i saw through the viewfinder did not corrispond precisely to the actual image taken).
sent it twice to pentax (for other issues , too) but i got it fixed only for the tilt issue.
I still have to think by framing, that i will have more free space on the left and on the top sides of the image. It is so annoying.
I hope that this feature can guarantie a perfect correspondance viewfinder- sensor.

By the way, this is a common issue of k10d, whereas the k100d super of a friend of mine is perfrect in this regard! WTF

*muustuus* · 05-21-2009, 07:19 AM

Where is the 3mm mentioned? I only saw 1mm.
It would be nice if there would be a burst mode where the sensor takes a shot in each corner. :P

plop.

ricardobeat · 05-21-2009, 01:46 PM

Originally Posted by muustuus

Where is the 3mm mentioned? I only saw 1mm.
It would be nice if there would be a burst mode where the sensor takes a shot in each corner. :P

plop.

You can do that by hand

Or maybe you can automate that with Remote Assistant (if there is one).

asdf · 05-21-2009, 01:57 PM

Originally Posted by ricardobeat

Seeing that the sensor can move 1mm to 3mm in any direction, is it possible that it could have a shift effect? (I really don't know how to do the math)

Sounds like an almost larger frame camera.

You can shift the image, while the camera is on a tripod and stitch the pics together later.

ricardobeat · 05-21-2009, 02:02 PM

Originally Posted by muustuus

Where is the 3mm mentioned? I only saw 1mm.
It would be nice if there would be a burst mode where the sensor takes a shot in each corner. :P

plop.

Hm, thought I read that on DPR. Did a search here and can't find any info. Maybe it's 1 to 2mm or maybe that's the total travel distance (1.5mm each side). Anyone with a better memory than me?

What I was curious about is if this small shift is going to have any visible effect on perspective. I've never used a tilt&shift lens so I don't know how much change in perspective 1mm or 2mm warrants, couldn't find any examples.

update: found the post, it's 2mm in both axis. Some more calculations on PC: http://forums.dpreview.com/forums/re...ssage=31920552

*falconeye* · 05-21-2009, 02:13 PM

Originally Posted by asdf

Sounds like an almost larger frame camera.
You can shift the image, while the camera is on a tripod and stitch the pics together later.

PTGui and APP are so awesome that in practise, you always could do this already.

asdf · 05-21-2009, 02:19 PM

Originally Posted by falconeye

PTGui and APP are so awesome that in practise, you always could do this already.

I know, but it's just an extra way of doing it and the result could be more pixel-perfect.

*cousinsane* · 05-21-2009, 02:28 PM

Maybe be the next gen body will be able to shift more, say 5mm.That would make a pretty significant difference. As long as the lens has a large enough image circle.

*muustuus* · 05-22-2009, 06:15 AM

I think manually it takes some time. With quick burst you could use it easily without a tripod. Even on some moving targets. :P
Even 2mm could be useful on some situations.

.muus

*muustuus* · 05-24-2009, 01:48 AM

Here's a clip showing composition adjust on tv-screen.
I think that was a quite large movement. :P

YouTube - PENTAX K-7 ????????? in ?????

.muus

Olivier · 05-24-2009, 04:27 PM

Originally Posted by muustuus

Here's a clip showing composition adjust on tv-screen.
I think that was a quite large movement. :P

Hard to tell for sure which is which: the movement of the camera that took the video, the (physical) re-positioning of the K and the use of sensor shift. However, it looks like the sensor is moving +- 1mm (total 2mm). Also, it looks like the movement is pretty smooth (i.e. the sensor can be moved in small increments), which is great for accurate positioning when on a tripod.

Btw, you may also want to look at the K-7 preview on Imaging-Resource.com. They have pictures that nicely demonstrate the extremes of sensor shift.

If you merge 4 pictures taken at the extremes of the shift range you end up with a "crop factor" of about 1.4.

*falconeye* · 05-24-2009, 06:36 PM

Originally Posted by muustuus

Here's a clip showing composition adjust on tv-screen.
I think that was a quite large movement. :P

It probably is as large as with the K20D.

You can make the following experiment (with a K20D):
On a tripod, enable LV (with grid lines). Suddenly turn the camera by 45° or so: You will see LV pan "and crawl back" a bit. This back crawling is in order to bring the sensor back to zero position for further shake. So, its distance tells us by how much the sensor has been displaced.

From my observation, it is about 1/3 grid quarter horizontally and 2/3 vertically.

This translates to 1/12 or 2mm horizontally and 1/6 or 2.5mm vertically.

Or ~4-5mm from edge to edge. The video looks like the same amount of shift.

Maybe, the early information that shift is +/-2mm in every direction was correct and the newer, corrected info (1mm) is wrong?

BTW, the video confirms that K-7 has LV out over HDMI and/or AV out ... This hasn't been confirmed so far.

EDIT:
In the experiment above and with LV at 8x, you can see that SR ON makes the LV image just a little bit less steady than SR OFF. Must be in the subpixel range. And the source for the recommendation to switch SR off when on a tripod.