15
\$\begingroup\$

Correct me if I'm wrong:

When I take a picture, the sensor gathers the light information "as a whole" during the exposure time and then saves it in the card. When I make a movie, the sensor scans line by line and then saves the frame in the card.

So, if this is correct, why does the camera switch from global shutter (as in photo mode) to rolling shutter (movie mode) when filming?

Of course there must be a technical reason, but why bother using a rolling shutter instead of a global shutter, which is much more convenient?

\$\endgroup\$
4
  • 3
    \$\begingroup\$ You're wrong. When you take a picture a CMOS sensor scans line-by-line exactly the same way as it does when shooting video. \$\endgroup\$ Feb 24, 2011 at 16:31
  • \$\begingroup\$ The explanations aren't clear...the old photograph explains why film registers action later due to physical shutter movement, but not why electronic shutters should, as the problem is the sequence of the line acquisition. Without knowing the capacity of processor in the present camera models vs the data quantity, I would presume the reason why the sensor is registering one line at a time is a processor-capacity dictated technology? That is, when processors become fast enough, the rolling shutter effect will go away, because the camera can register all lines simultaneously in every "frame"?? \$\endgroup\$
    – user5969
    Jul 15, 2011 at 22:43
  • 1
    \$\begingroup\$ @Terra: See Jerry Coffin's answer. The problem isn't processor power, it's wiring. \$\endgroup\$
    – mattdm
    Jul 16, 2011 at 2:56
  • 1
    \$\begingroup\$ Aye, the problem is indeed a wiring/physical layout problem. I think this is an area that is being improved upon for CMOS sensors, however. Around this time last year, Canon demonstrated a 120mp APS-H sensor that had a significantly higher readout rate than the largest modern sensors that exist in cameras today. They did not release the technical details of their prototype, however the assumption is that they restructured the wiring to provide more parallel sensor readout than was possible before. \$\endgroup\$
    – jrista
    Jul 16, 2011 at 18:21

3 Answers 3

14
\$\begingroup\$

In both cases, the actual reading from the sensor is done line by line. It's done that way (largely) because doing otherwise would be excessively expensive -- to read all the pixels in parallel, you'd need a separate connection from the sensor to memory for each pixel. 12 million connections (for example) from sensor to memory would be horrendously expensive -- and almost never provide any real benefit.

As to why there appears to be a difference between movie mode and still mode, it's pretty simple: in still mode, you're using a physical shutter, and the read-out from the sensor to memory happens when the shutter is closed. As @Matt Grum pointed out, you still get some of the same effect above the X-sync speed, due to physical limitations on the shutter.

The reason you don't use the physical shutter in movie mode is more of those same physical limitations. While the shutter can have a very short exposure time, there's a recovery time between activations, so it becomes difficult to achieve more than about 10 frames per second or so. Getting to the 24 frames per second or so needed for video would, again, add a lot of expense with little benefit. Therefore, in movie mode the physical shutter remains open, and the camera uses an electronic shutter instead -- and once it does that, artifacts of the line-by-line readout from the sensor to memory can become visible.

\$\endgroup\$
6
  • \$\begingroup\$ So, it's correct to say that in stills mode, the CMOS acts PARTIALLY as a global shutter (As all the light exposes the sensor as a whole, and then, when there isn't any more light hitting it, each line is read)? But why, if the camera uses an electronic shutter (which I guess commands the sensor when has to be light sensitive), still there is skew artifacts? I mean, if the sensor receives the light as a whole, then the electronic shutter hits, and then the lines are read, why there are rolling shutter artifacts? \$\endgroup\$
    – Andres
    Feb 24, 2011 at 17:10
  • 2
    \$\begingroup\$ Because there's isn't a command to "turn off" light sensitivity. There's just reading the data from a cell, which drains the charge in that cell. \$\endgroup\$ Feb 24, 2011 at 17:20
  • 13
    \$\begingroup\$ It might help a bit (or confuse) if you think that light = rain and a pixel = bucket. And in a CMOS sensor, the only way to check how much water you have in each bucket is to empty the bucket. You can only empty the buckets one by one, and it is raining all the time. The "movie mode" simply means that you are running around in the rain, emptying the buckets. The "photo mode" is different: first set up a cover on top of your field of buckets; then you go around and empty each bucket; then you remove the cover for a while and put it back ("exposure"); then you empty each bucket and measure them. \$\endgroup\$ Feb 24, 2011 at 20:43
  • \$\begingroup\$ @Jukka Suomela: Nice analogy, except for one minor point: it can actually empty an entire "row of buckets" at once, not just one. \$\endgroup\$ Feb 24, 2011 at 20:45
  • 1
    \$\begingroup\$ Makes it sound tiring, running around in the rain emptying buckets. \$\endgroup\$ Jul 15, 2011 at 22:48
15
\$\begingroup\$

When taking a still you have a physical shutter which controls light hitting the sensor. The image is still read line by line when taking stills but because of the physical shutter closing no extra light is recorded during readout.

The line by line readout is a consequence of the CMOS design found in large sensor cameras, and thus is unavoidable (until they find a way to make cheaper/bigger CCDs).

It's worth noting that at shutter speeds faster than the cameras x-sync speed (usually 1/250s) the shutter starts closing at the bottom before it's fully open at the top. The result of this is that for really fast speeds like 1/4000s you get a slit that traversed the frame and gives you a similar rolling shutter effect for stills. However because the time to traverse the frame by the shutter (1/250s) is ten times faster than the time to read the sensor during video (1/25s) you need a really fast moving object to notice it.

Here's a very old photograph which demonstrates the effect well:

The effect is also more noticable in videos if you pan back and forth, which doesn't happen with stills.

\$\endgroup\$
6
  • \$\begingroup\$ Uhmmmm, what about 1 second long (or more) exposures? Does the sensor takes more time to read each line or it re-reads the lines when hits the bottom? \$\endgroup\$
    – Andres
    Feb 24, 2011 at 15:35
  • \$\begingroup\$ @Nick You have just pointed out why there is a limit on shutter speeds with (non-HSS) flash: the shutter can be entirely open for exposures of about 1/250 s or longer (depending on the camera), but for shorter exposures the entire sensor is never exposed at any given time. \$\endgroup\$
    – whuber
    Feb 24, 2011 at 15:49
  • \$\begingroup\$ @Nick it's only at speeds faster than 1/250s that the shutter scans the frame, at 1/200s there's a moment when the whole sensor is exposed, thus enabling flash at this speed \$\endgroup\$
    – Matt Grum
    Feb 24, 2011 at 15:51
  • 1
    \$\begingroup\$ @Andres, when shooting stills the frame is scanned after the shutter is closed, so with a 1 second exposure, the shutter is open for 1 second, then it closes and then the whole frame is scanned in approx 1/25s \$\endgroup\$
    – Matt Grum
    Feb 24, 2011 at 15:52
  • \$\begingroup\$ @matt I know, I realised my mistake and deleted the comment! But this doesn't explain why the shutter has to roll in video mode - why can't the camera capture the whole frame then? \$\endgroup\$
    – user456
    Feb 24, 2011 at 15:56
3
\$\begingroup\$

You may find this link useful:

http://www.dvxuser.com/jason/CMOS-CCD/ (by Barry Green /via@SFGPhoto)

It's a CMOS compromise: heat, power, slower refresh rate, cheaper than CCD. CCDs are faster, but need more energy and cost more...

From my basic understanding of chip manufacturing, image sensors and processors are created using similar tech. Processor fabrication favours shrinking: smaller = cheaper. But photographers want BIG sensors. Can't have it both ways without it costing a lot more. Plus packing in lots of pixels into a sensor can be too much of a 'good' thing (less light etc.)

@mogwailun

\$\endgroup\$
2
  • \$\begingroup\$ But why in photo mode the light is readed as a whole and in video mode line by line? The article explains the diferences between rolling and global, but I can't understand the switch between photo and movie mode. \$\endgroup\$
    – Andres
    Feb 24, 2011 at 15:19
  • 2
    \$\begingroup\$ @Andres, the image is still read line by line in photo mode, you just don't notice it! \$\endgroup\$
    – Matt Grum
    Feb 24, 2011 at 15:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.