Loading... Please wait...

Categories

Our Newsletter


CMOS Sensor Shuttering

Since most CMOS sensors are 3T or some equivalent or, if they are 4T, have severe constraints on the use of the photodiode independently from the storage well, the availability of global shuttering is severely limited. What this means in practice is that the initiation of the exposure can be controlled electronically – it is just the end of reset – but the termination of the exposure cannot. The exposure for a particular row ends only when its contents are sampled and transferred to the row of readout capacitors. Therefore, the exposure time of each row as the scan moves down the sensor is one row longer than the previous row. If the exposure is set to match the duration of the scan, for instance, then the last row will get twice the exposure of the first row.   There are some ways to mitigate this problem.

Illumination Control

The simplest mitigation in instrumentation settings is to control the illumination so that the sensor is in the dark after exposure. A mechanical shutter can provide this darkness if it blocks the light from the entire sensor simultaneously. Such a condition is not necessarily easy to provide because any shutter near the sensor will obscure different parts of the sensor at different times in a circular or some sort of sweeping pattern. This lack of simultaneity may not be a problem if the time to close is short compared to the exposure time. Real simultaneous exposure can only be accomplished if the shutter is in the position of the iris in the lens or in some similar optical plane. In situations where the object to be viewed can be enclosed, the exposure can be controlled entirely by turning an illuminator off and on at the right times. Care must be taken to exclude stray light, especially if the exposures are short compared to the scan time. It is important to note that the onset of illumination, either by a shutter opening or a light turning on, does not have to be the start of the exposure because the sensor can be held in reset and used as the starting time. This can be very useful when an object is anticipated to arrive for capture but its exact time is not known. The illumination can be initiated slightly in advance while the exposure start can be held until the object is actually detected. Ending reset is generally much faster and more accurate than changing the illumination state.

Rolling Shutter

CMOS can provide a fully electronic exposure method if the sensor is provided with two independent row counters. If one of the counters is used to command reset of the selected row and the other counter controls reading, then simply by establishing a fixed delay (usually measured in multiples of the row period) between reset and read, all rows can have the same exposure time. For example, if row 25 is reset as row 1 is read, then, in the next cycle, row 26 would be reset and row 2 read. After 25 repetitions of this, row 25 would be read after having accumulated exposure for 25 row periods. This spaced arrangement is moved down the array so that all rows see an exposure of 25 rows. The start has to be handled to be sure the rows there get the right exposure but this requires only a simple logical control. The effect, though, is of providing a variable, uniform exposure simply be setting the number of rows between reset and read. A little additional logic can stop the scanning to provide exposure of any length, not limited by the number of actual rows in the sensor. This process is called rolling shutter scanning.

The difficulty with the rolling shutter is that exposure is not simultaneous. Each row has an exposure time (in the absolute sense) that is slightly delayed from the exposure of the previous row. The total difference in starting times will be exactly the time needed to scan the sensor once. While the delay is not important if static objects are to be imaged, moving objects will suffer geometric distortion. Objects moving perpendicular to the rows will be stretched or compressed depending on their direction relative to the scanning and objects moving along the lines will be tilted one way or the other. The significance of this depends on relative speeds and the requirements for geometric accuracy in the images. Oddly enough, the digital still camera with a mechanical shutter is often offered as the ideal way to use a sensor that does not have electronic exposure control. However, the shutters in such cameras are of the curtain type, consisting of two metal leaves that move across the image – one to start the exposure and one to end it. At speeds above about 1/125 sec, the terminating blade starts moving before the initiating blade has completed its travel. By the time the shutter gets to 1/1000 sec, the aperture the leaves form is a vertical slit that traverses the image, providing an exact mechanical equivalent to the electronic rolling shutter and producing the same geometric distortions.

Global Shuttering

To provide electronic exposure control at both initiation and termination, an extra storage node is needed in each pixel. This implies at least a 4T pixel and perhaps more. The extra transistors take space and will reduce fill factor at least until sensors can be built with transistors buried under the photodiodes. Because the storage node is in essentially the same plane as the photodiode, there will be paths for light arriving at the sensor surface to reach the storage node even though it is under an opaque shield.  This leads to generation of additional signal when light is present even if the exposure has already been ended.  The ratio of the signal generated by some exposure to the photodiode to the same exposure to the storage node is called the shutter efficiency. 

The shutter efficiency can be expressed as a ratio:

se-ratio.gif

Typical simple CMOS processes can provide ratios of a few thousand. 

Alternatively, the shutter efficiency can be expressed as a percent:

se-precent.gif

 

Generally, these are 99.9% or more.

If the storage node can be more effectively shielded, shutter efficiency ratios of one million or more (99.9999+%) are possible.

The actual test methods to determine shutter efficiency are under consideration by various standards organizations.