Global shutter vs. rolling shutter: which is best for your application?

JAI recently added six new camera models to its Go-X Series, bringing the total number of models in this series to 30. But there was something different about these last six models, namely, these new models featured CMOS imagers with "rolling shutters" as opposed to the "global shutters" found in all of JAI's other area scan cameras. In this blog we will briefly explain exactly what that means from both a technology and an applications perspective.

Roll with it
The first thing to point out about rolling shutter cameras is the attractive cost-per-megapixel value they provide. In fact, this is generally what attracts a vision system designer to them in the first place. Rolling shutter sensors have a simpler readout method than global shutter sensors (we'll talk about that in the next section). This means they can utilize a simpler transistor-level architecture, which costs less to design and produce. This makes the cameras they go into cost less than global shutter cameras on a megapixel-to-megapixel basis.

So, for engineers designing vision systems where the camera cost is a significant portion of the total system, and where competitive pricing is extremely critical either due to competition or to market adoption, a rolling shutter camera would seem to be an obvious choice, right? But it's a little more complicated than that.

A tale of two readouts
In order to know whether a rolling shutter camera or a global shutter camera is best for your application, you have to know a little about the differences in the exposure and readout methods of the two sensors.

In a global shutter sensor, all pixels/lines are exposed simultaneously. When the exposure is complete, each pixel transfers its charge into an on-pixel storage node to wait for readout. During readout, the pixels are assembled into lines and certain checks are performed (typically via correlated double sampling) to make sure that the right noise reference is maintained for each pixel. Because the pixel data has been moved out of the pixel wells, global shutter cameras typically allow the next exposure to start during the readout period so that reasonable frame rates can be maintained.

Global shutter method

The simpler design of a rolling shutter sensor has no storage nodes, so a different approach is required. Instead of all sensor lines being exposed simultaneously, the exposure periods are staggered slightly from line-to-line. In other words, after exposure of the first line has started, there is a small delay before the second line starts its exposure. Similarly, there is a delay between the start of the second line's exposure and the start of the third line's exposure, and so on, until all sensor lines have been exposed and individually read out. The length of the line-to-line delay is equal to the time it takes to readout a line, thus each line can be read out as soon as its exposure is complete. This method avoids the need for any pixel-level storage nodes, but it means different lines are capturing slightly different moments in time.

Rolling shutter method

Because a sensor can read out a single line quite rapidly, the individual line-to-line delay on a rolling shutter sensor might only be 10-20 µs, but on a sensor with 1000 or 2000 lines, that means the exposure of the last line in an image might not occur until 40 ms after the first line in the image.

This may not sound like much, but when the object being imaged, or the camera itself, is moving rapidly during the exposure period, the staggered, line-by-line exposures of a rolling shutter camera create various distortions of the image. The most common are "skew," where moving objects look slanted, and "spatial aliasing," where rapidly spinning objects like fans or propellers appear bent or even detached.

Rolling shutter effect (example)

Simplified animation of rolling shutter image capture shows the distortions that can be created when viewing moving or rotating objects. Source: Wikimedia. By Cmglee - Own work, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=45850383

It's a small (pixel) world
The simpler architecture of a rolling shutter sensor results in one other noteworthy phenomenon. With less circuitry using up space on the sensor, pixels can be smaller overall yet still have room for moderately-sized pixel wells with good dynamic range and signal-to-noise properties. But more pixels per square millimeter of sensor isn't always better. In order to take full advantage of the extra megapixels residing on the sensor, a camera with small pixels may require more expensive optics to deal with the Modulation Transfer Function (MTF) requirements of the smaller pixel pitch. Depending on the image quality requirements of a particular application, this could negate the cost-per-megapixel savings of one camera versus another.

The right shutter for the job
Now that we have a basic understanding of the differences between a global and rolling shutter, it should be easier to choose the right type of camera for a particular application.

As noted at the beginning, if you are looking for the absolute best cost-per-megapixel for your application, chances are you will start by looking at a rolling shutter camera. But to make your final decision, you'll need to answer a few questions:

• Does your application involve continuous motion of the target or the camera?  Whether it's capturing images of fast-moving items on a conveyor belt, or if your camera is mounted on a vehicle or a robot arm, movement can create problems for rolling shutter cameras (see the next question). But if your application is a "stop-and-go" application, where each item is paused briefly while the image is captured, or if the camera is moved to various "stop points" over a stationary target, a rolling shutter camera should be an acceptable option.

• Is the shape, size, or exact location of objects in your images important to your application? Even if your application does involve movement, a rolling shutter camera might still be acceptable, but only if some degree of spatial distortion can be tolerated. As noted, rolling shutter cameras may make moving objects look skewed or oddly disjointed within an image. This would be completely unacceptable for a metrology, barcode reading, or facial recognition application, for example, but might still work for some kinds of presence/absence checking or situational awareness applications.
• Is a high level of detail and contrast critical to your application? As noted, rolling shutter cameras typically utilize a smaller pixel size (and not just a simpler design) to maximize their cost-per-megapixel advantage. But a smaller pixel pitch (lp/mm) requires more expensive optics than with larger pixels to achieve the same contrast in the same optical circle. For applications where the use of low-cost optics and the resulting lower contrast images is not a problem, the use of rolling shutter cameras should not be a problem. But if higher MTF (contrast) is required, the slightly larger pixel sizes (and lower resolution) on comparably-sized global shutter sensors might make it easier to put together a combined camera-lens solution.

Rolling shutters with moving objects?
While continuous motion typically rules out the use of low-cost rolling shutters for most applications, the Go-X Series rolling shutter cameras include a feature called "global reset" that makes it possible to eliminate the spatial distortion problems in some cases – but only under special conditions and with the use of flash illumination. We'll have more details about that in a future blog. For now, if this sounds like it might be a possibility for your application, please contact JAI to discuss.

You can learn more about the Go-X Series rolling shutter cameras at: https://news.jai.com/jai-expands-its-go-x-series-with-new-rolling-shutter-cameras

For a complete listing of all Go-X Series cameras, including links to individual product pages where you can download datasheets and other information, please go to: https://www.jai.com/go-x-series