0

I want to ask about the starting time of a new frame in a sequence of frames captured by a CMOS sensor. Assume that the readout time of the entire sensor is 10ms, and the exposure time is less than 10ms. Will the next exposure of the first row of the sensor begin at the starting time of the next frame which is described in the first figure below? Or the exposure begin right after the readout of the previous frame complete (I think it's not possible)? Sequence of frames

Or the starting time of the exposure is aligned to the right to ensure that beginning of the readout of the first row of the second frame starts right after the readout of the last row of the first frame has finished, which is described in the second figure below.

enter image description here

@mattdm: I have never asked "When a CMOS sensor is read line-by-line, which line is considered to mark the starting time?". It is the first line which marks the starting time, of course. But my question is when the exposure of the first row begin.

  • 3
    Context? Still or video camera? Mechanical or electronic shutter? – Michael C Jul 14 '15 at 22:29
  • 2
    Video is off topic here - this site is dedicated to still photography. – Michael C Jul 15 '15 at 5:08
  • 3
    There is no reputation gained from comments, which is all I have done regarding this question. There is a more appropriate place for video questions: video.stackexchange.com – Michael C Jul 15 '15 at 6:36
  • 2
    possible duplicate of Readout time vs Frame rate of CMOS sensor – mattdm Jul 15 '15 at 9:36
  • 2
    If you're talking about still photography, then there's only one frame to consider -- there is no next exposure. – Caleb Jul 15 '15 at 17:09
1

The answer is electronic shutters on CMOS simply cannot do shutter speeds faster than their readout time, which in your example sensor is 10ms (1/100s). The readout time is fixed, and it's not possible to start readout of the next frame before the readout of the last frame has finished or to somehow speed up readout.

The only exception is if the CMOS sensor supports a feature called global shutter, which is still relatively rare, in which case it does not read according to your diagram but instead operates completely differently (and I don't think it's relevant to shooting video).

Edit


To answer your updated question, the dotted lines on the graph showing where the frame "starts" is just convention. If you consider the timestamp of the start of a frame to be the head end of the exposure, ie when the first line starts receiving light through the lens, then the first diagram is correct. If you consider the timestamp of the start of a frame to be at the tail end of the exposure, ie when the first line's exposure ends and is read out from the sensor, then the second is correct.

I don't know if there is a standard for what point in time is considered the timestamp for the start of a frame. There may well be one, but that is very motion picture (film/video) related and getting off topic here. I'd assume that the former chart is more likely to be the logical way you'd define the start of a frame.

  • What do you mean by "it's not possible to start readout of the next frame before the readout of the last frame has finished"? – James Do Jul 15 '15 at 4:33
  • I may have misunderstood your question. It looked like you were asking what happens if your frame rate (ie exposure time) is faster/shorter than one readout period. The answer to that question is that it's not possible. I'm beginning to think that's not what you were asking. Perhaps you could add more information to the question. – thomasrutter Jul 15 '15 at 4:39
  • @JamesDo What he is saying is that the readout of the last line of the first frame must be complete before the readout of the first line of the second frame can begin. – Michael C Jul 15 '15 at 5:10
  • @thomasrutter Thank you for your answer. But I'm asking about the starting of the exposure of the first row of the next frame. Assume that the readout of the last row must complete before the readout of the first row of the second frame as Michael Clark said. Is it that the exposure of the first row of the second frame is aligned to the right as I describe in the (edited) question? – James Do Jul 15 '15 at 5:19
  • 1
    @mattdm not really but I can see how you might get that from the diagram. The blue bars actually represent the duration the line is "exposing" whereas the readout of the line happens at the end of that blue bar - he has added red rectangles in the second diagram to represent when readout occurs for each line. Those red areas do not overlap. A frame does not begin its readout until the previous frame's readout is complete. However due to the rolling nature it does begin exposing its first line while lines in the previous frame are still exposing. Hope that helps. – thomasrutter Jul 15 '15 at 11:29
1

The basic answer here is: usually, no one cares. For the art of photography, cameras are meant to record images for the purpose of viewing, not metadata for the purpose of scientific instrumentation. The prevailing EXIF standard for metadata doesn't even record timestamps more precisely than one second, and there's usually no provision for network time synchronization.

So, if you're using an instrument built around a camera, consult its documentation or the manufacturer. If you're building such an instrument, pick a point a document it. If you're using a general-purpose camera and need extra precision for some reason, one with a GPS built-in may provide better temporal resolution with GPSTimeStamp — but still probably not enough to distinguish between very fast frames in succession.

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.