4

Regular DSLRs make use of mechanical shutters and progressive electrical readouts, and hence images are captured in a rolling shutter.

But, when used to capture pictures with a long exposure, where the entire sensor is open for an extended duration, are all pixels of a sensor exposed simultaneously? The readout will definitely be sequential rows of pixels, but will the data acquired in every pixel correspond to a simultaneous event?

The reason to ask this question is to understand whether a DSLR camera may be suitable for an academic research purpose.

11

I think you've slightly misunderstood how the shutter works. Initially, the first shutter covers the sensor and the second does not. The first shutter then slides away to expose the sensor, and the second shutter follows it to cover the sensor again. The shutters always move at the same speed, regardless of the exposure time.

For a short exposure, the second shutter will start to close off the sensor before the first shutter has finished opening it. The effect is that an open slit moves across the sensor, with the width depending on the desired exposure time. For longer exposures, though, there are three phases: the first shutter opens all the way, the sensor spends some time completely uncovered, and then the second shutter closes. In this case, all pixels are indeed exposed simultaneously, except briefly at the start and end of the exposure.

  • Also, at very high speeds, (1/8000 for example), the two shutters, which are independent, can be moving at slightly different speeds so you may see a subtle difference in light levels from the bottom of the image to the top as the effective shutter aperture changes. Usually not noticeable but can be mostly compensated by a multiplying gradient in Photoshop. The larger effect is the timing differences which can elongate or shorten moving objects along the vertical. – doug Sep 10 '18 at 18:47
  • 2
  • 1
    @MichaelClark Thanks for the info on the film SLRs. I wasn't aware they had horizontal shutters. It always seemed to me that horizontal shutters would be a better choice than vertical since images typically have horizontally moving objects like people or cars and vertical shutters cause a slight tilt which is more noticeable than the elongation from horizontal shutters which only tilts things moving up/down. Always assumed that physics/engineering provided compelling reasons for the vertical placement overriding horizontal placement. – doug Sep 11 '18 at 19:22
  • 1
    Horizontal worked better when the lever used to wind film to the next frame in a horizontal direction also recocked the spring driven shutter curtains. For the same curtain speed, transit times are longer for horizontal travel (in a 3:2 'landscape' format such as 135 film) because the curtains have 36mm to travel instead of 24mm. That's the main advantage of vertical travel shutters, which were first introduced to the 135 format in significant numbers when shutters became electronically, rather than spring, driven. Shorter transit times enable shorter X-sync. – Michael C Sep 11 '18 at 19:28
  • 1
    The other issue with film SLRs was that they usually had cloth or rubberized fabric curtains that spooled or remained flat on either side of the opening it the back of the light box. There wasn't room for a vertical travel cloth curtain in SLRs (because the viewfinder prism needed that space directly above the light box). Only when metal shutter curtains with segmented 'leaves' were introduced was there room in SLRs for a vertical travel shutter. Medium format cameras, on the other hand, normally wound film and the focal plane shutter (if the camera had one) in the horizontal direction. – Michael C Sep 11 '18 at 19:34
7

Most DSLRs have shutter curtains that transit the face of the sensor in approximately 2-4 milliseconds. For any particular camera model, barring a mechanical malfunction, the transit time for each shutter curtain will always be the same, regardless of the exposure time. This is the case with shutter times anywhere from 1/8000 second, where the second curtain is 'chasing' the first curtain across the sensor a small fraction of a millisecond behind it, to Bulb mode, where the offset between when exposure begins and ends on one side of the sensor to the other is still that same 2-4 milliseconds.

For a 1/8000 second exposure, a 4 millisecond offset is roughly 32X the length of the exposure.

For an 8 second exposure, the same 4 millisecond offset is a miniscule 0.0005X of the total exposure time.

How readout is done is inconsequential for still images, since it is performed when no light is reaching the sensor after the shutter curtains are closed.

Reason to ask this question is to understand whether a DSLR camera be suitable for an academic research purpose.

Even top level professional cameras are not what would be considered 'lab grade'. The precision required for creative/historical/documentary photography is relatively low. Anything within one-sixth stop is considered acceptable for a camera with minimum adjustments of 1/3 stop for aperture (Av), exposure time (Tv), and sensitivity (ISO). In terms of exposure time measured in seconds or fractions thereof, one-sixth stop is roughly 12%. So if a camera is set to take a photo using a shutter time of 1/250 second (4ms), anything between 1/285 second (3.52ms) to 1/225 second (4.48ms) is considered acceptable.

The other thing to remember is that the numbers we use in the "standard" shutter speed scale are not the actual target numbers. They're rounded versions of the powers of two that were selected at a time when cameras were even much less precise than they are today. For example, what we call 1/250 second is actually 1/256 second (which means we give or take that allowable 12% time variation from 1/256, rather than from 1/250)! So the allowable window for our named 1/250 shutter time is actually +/- 12% of 1/256 seconds, which figures out to between 1/229 seconds (rounded to the nearest whole denominator) and 1/291 seconds (rounded to the nearest whole denominator). But since we like to round numbers to the nearest "5" in photography, we'll call them 1/230 and 1/290 seconds, respectively.

For more about marked numbers versus target numbers versus actual numbers used and why it doesn't really matter for creative/documentary/historical photography, please see: Is there a sane reason why ¹⁄₁₂₅ is not, instead, exactly half of ¹⁄₆₀?

My application involves using an extremely short pulse laser to freeze motion(as short a duration as 9ns). So I was worried whether rolling shutter will kick in even at long exposures and make my regular camera unsuitable for the experiment.

Yes, as long as the first curtain is already fully open when the light pulse begins, and the second curtain has not yet begun to close when the light pulse ends, the entire sensor will be illuminated by the entire duration of the light pulse. For still images the sensor is only read out once following the end of the exposure. There is no 'scanning' going on while the shutter is open.

As long as your laser pulse is significantly shorter in duration than the camera's flash sync speed (typically around 1/200-1/250 seconds or 5-4 milliseconds), you might consider using the camera's flash sync capability to trigger the laser. (How to do that would probably need to be a separate question.)

We also have some existing questions about high speed photography using very short pulses of light, rather than short shutter times, to freeze the motion of high speed objects. Among them:

What problems might there be with using a laser as a flash source for high speed photography?
How can I capture the movement of a bullet?
How do I trigger the flash in a way that takes luck out of high-speed photography?
Why doesn't a shutter speed of ¹⁄₂₅₀th freeze motion when a flash of that duration does?

  • 2
    Wow! Thanks for the extended answer about clarifying about the camera grades. My application involves using an extremely short pulse laser to freeze motion(as short a duration as 9ns). So I was worried whether rolling shutter will kick in even at long exposures and make my regular camera unsuitable for the experiment. Actually I think the dslr is superior in terms of number of pixels it can pack. I have to give it a try to see if it will be suitable. – karthikeyan Sep 10 '18 at 20:48
3

It is still the same shutter, even in long exposure and pixels are not exposed simultaneously.

However, the time difference (in relation to the total exposure time) becomes smaller as the exposure time increases. It might become insignificant depending on your requirements.

If you cannot tolerate rolling shutter at all, you still have 2 options:

  • use external global shutter
  • use flash (the shutter opens and closes in darkness, flash illuminates the whole scene simultaneously)
  • I am looking at trying the 2nd option. Hope rolling shutter would not create any issues – karthikeyan Sep 10 '18 at 8:47
  • Would rolling shutter even be a problem with long exposure? I mean, what does 1/16000 second matter when the shutter is open a total 2 or 3 seconds, or 10 seconds? – user69960 Sep 10 '18 at 14:05
  • 1
    realistic shutter travel time is probably closer to 1/500 sec but yes, i doubt it really makes a difference for long exposures where most of the exposure is captured simultaneously for all pixels. – szulat Sep 10 '18 at 14:11
2

CMOS pixels are reset and then gather the cumulative charge from the light they receive until they are read & reset again. The CMOS is usually reset, the shutter process takes place (front curtain, delay, then rear curtain) and then values are read.

Because of the mechanical process, there will always be a short period during the shutter opening/closing events where the sensor is partially obscured.

If this affects you will depend on what type of event you are attempting to capture with a long exposure. If you're hoping to capture a bright but unpredictable event like lightning then it's possible that a shutter curtain could be half-closed during the event. If you're looking for something under more constant lighting conditions then the 1/125 or 1/250th sec that are needed would be inconsequential in an exposure of 1 sec or more.

2

For the purpose you describe: Yes, all pixels are exposed simultaneously; they work independently, in parallel. Even for the short periods of time at start and finish, they're still all working together - but the shutter will need some time to get out of the way completely.

So if a short event were to happen during the long exposure, what you will get is a simultaneous read across the surface of the sensor.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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