1

I was wondering what is the true reason for the rolling shutter effect that can be observed on many CMOS camera.

I've done some research on the internet and found an excellent article about it: link.

In this article (page 1, paragraph 3), it mentions that modern CMOS sensor uses ADC for each column of pixels and these ADCs work in parallel for higher frame rate.

However, the entire sensor array can only be converted one row at a time. This introduces a slight delay between each row's readout, and leads to a delay btw each row's next exposure, making them no longer simultaneous.

I have some questions about this explanation:

  1. Why the entire sensor array needs to be converted one row at a time? what is the limiting factor?

  2. What kind of conversion does it need?

  3. As the rolling shutter results from the quest for higher frame rate, when taking a still photo, it seems no requirement on frame rate. Why does the roll shutter effect still happen?

Jason Yang
  • 113
  • 4
  • 3
    Ummm. Consider how many pixels you have. To do it all at once, you'd need that many ADCs. – JRE Oct 19 '16 at 08:44
  • 3
    Take a VGA camera. 640x480. Row-wise you need 480 ADCs to sample an entire row at once. To sample all pixels at once, that would be 307200. That's a lot of ADCs. – JRE Oct 19 '16 at 08:53
  • @JRE For VGA camera, it only requires 480 column-wise ADCs for 640*480 pixels, doesn't it? I think the trick part is that rolling shutter effect happens when the exposure time of each row does not start synchronously. – Jason Yang Oct 20 '16 at 02:54
  • Global shutter does not need an ADC per pixel! Just a sample-and-hold cap and transistor per pixel, much cheaper. – pericynthion Oct 20 '16 at 05:32
  • @JasonYang: Rows and columns swapped. – JRE Oct 20 '16 at 05:42

2 Answers2

2

The reason rolling shutter mode exists is to allow a greater amount of overlap between making the exposure and reading it out. With global shutter, there can be no overlap at all.

For a given technology, the rolling shutter mode gives higher exposure times (i.e., higher sensitivity and SNR) and/or higher frame rates as compared to global shutter mode.

The number of sense amplifiers and ADCs is the same in either case. The cheapest sensors use a single row of sense amplifiers and a single ADC, while higher-end sensors split the sensor array into two halves and have two rows of sense amplifiers and two ADCs.

Dave Tweed
  • 168,369
  • 17
  • 228
  • 393
1

Most of the time it's the same reason as for virtually all compromises in performance. Cost.

There is no technical reason why you have to do it one row at a time, you can get global shutter sensors that don't. For the same resolution they are larger and more expensive.
As @JRE has indicated in the comments, you need a lot more ADCs to do this which means more space taken up on the silicon and more places things an go wrong leading to lower yields. A larger chip with a lower yield is going to cost more.

Given that most people are happy with a rolling shutter you then get economies of scale kick in, more people want rolling shutter because it's smaller and cheaper, that increases volume which in turn decreases unit price even more.

And for some of the applications where cost is less of an issue the increased size of a global shutter sensor and the physical requirements that would place on the sensor and lens arrangement can be a strong reason for using a smaller rolling shutter sensor.

Andrew
  • 6,872
  • 19
  • 25