Parasitic Light Sensitivity (aka Shutter Efficiency)

Image sensors in modern cameras usually have electronic shutter control. This means that the shutter is part of the sensor itself. Once the image has been integrated, the charge is moved to a storage node (floating diffusion readout node), which is shielded from light and can be read out. Ideally, no further integration takes place during this time, but in reality this is not the case. This effect is called Parasitic Light Sensitivity (PLS) or sometimes Shutter Efficiency.

To understand what PLS actually is, we need to look at how a global shutter image sensor works.

These image sensors have the following states (which may overlap, but to keep things simple, we will assume that they occur step by step):

1. Resetting the photodiode → Start of exposure time
2. Exposure: The photons of the incoming light generate electron-hole pairs in the photodiode of a pixel, and the electrons are accumulated.
3. End of exposure time → The accumulated electrons are transferred to the storage nodes shielded from light.
4. Line-by-line readout of the image

Ideally, the image is frozen as soon as the transfer to the storage node takes place. In reality, however, the storage node is not 100% shielded from light. Photons can still enter the storage node, generate electron-hole pairs, and accumulate in the storage node. Thus, the readout time represents the effective exposure time.

Since the first line of an image is read out immediately after charge transfer (i.e., exposure end), the first line is not affected, but the subsequent lines are increasingly affected. This is why a brightness gradient can be perceived in the image from top to bottom. The longer the wavelength, the easier it is for photons to enter the storage nodes, which results in a color dependence in the brightness gradient.

Figure 1: The brightness change caused by the PLS in the Y direction (here 20% for better visibility)

High-quality CMOS image sensors have good shielding of the memory node, which is typically expressed in dB using the suppression factor. This factor usually ranges from -70 dB to -95 dB. This means that the memory node is 3200 to 60000 times less sensitive than the photodiode itself.

For the IMX sensor families, for example, Sony specifies the following PLS suppression factor values:

• Gen 1, 2 & 3: ~ -94 dB
• Gen 4: ~ -74 dB (optimized pixel performance and back exposure lead to losses in terms of PLS)

With very short exposure times and sensors that read out slowly, the effect can still be seen; or as soon as the readout time and exposure time (in the range of parasitic light sensitivity suppression) are of the same order of magnitude. The additional brightness can be expressed as follows:

indicates how much brighter the last line of the image is compared to the first.

Figure 2: The signal caused by PLS for different PLS values at a readout speed of 60 fps (17 ms)

The PLS is a sensor parameter and unfortunately, the shielding of the memory node cannot be influenced. However, the readout time is a second important factor that leads to the PLS. Balluff cameras have an image memory to buffer the images during transmission. This means that the readout of the images can be decoupled from the transmission, which helps in this case. This is because the "PLS readout time" is not the same as the image transmission time, but is the same as the readout time in the image memory.

Figure 3: Min. exposure time at 1% brightness gradient (PLS factor) with Sony Gen4 sensor with different hardware interfaces GigE/USB3: max. sensor readout due to interface transmission limits

Figure 3 shows the minimum exposure time allowed to obtain a PLS effect of 1%. With longer exposure times, the PLS effect becomes smaller. The curves have the following meanings:

• GigE and USB3: Curve at sensor readout time = transfer time (without image memory).
• MV: Actual curve for BVS CA-SF / BVS CA-GX2 cameras with image memory, which enables a faster readout time. If the BVS CA-GX2 is operated with a network connection, the sensor readout time remains constant due to the image memory.
• MV EC 2.3: Actual curve for BVS CA-GT and BVS CA-BN4. The BVS CA-GT also has image memory and a 10GigE interface. In a 1GigE network, the sensor readout time remains constant thanks to the image memory of the BVS CA-GT. The BVS CA-BN4 does not have an image memory, but DMA transfer allows the sensor to be read out at maximum speed and prevents the interface from becoming a bottleneck.
• MV EC 4.7: Actual curve at an optimal readout time of the sensor, which is supported by the BVS CA-GT series if desired.

While it is possible to compensate for brightness changes from top to bottom, there is another serious effect: motion blur from top to bottom. Once the exposure time has been adjusted so that only a certain amount of blur occurs, this can be worsened by the PLS. Although motion blur is tolerable in the upper regions of the image, it worsens the further down you go. This can then be difficult to compensate for.

It is possible to compensate for the brightness change caused by the PLS using a linear improvement function. The best way to do this is to measure the actual change in brightness caused by the PLS and then use the resulting curve for compensation. It is also possible to calculate the curve using PLS values provided by the sensor manufacturer. However, since these are general values and may vary in detail depending on the sensor, this leads to inaccurate correction results.

As already mentioned, PLS also depends on wavelength (as does QE). This means that color sensors exhibit different brightness deviations depending on the color channel, resulting in color shading from top to bottom. If compensation is to be performed, a correction must be calculated separately for each color channel.

Every image sensor has to contend with PLS. Whether this is decisive depends on the respective application. The PLS value alone cannot be used to determine whether the image sensor is suitable for the application or not. This is because the readout time of the sensor, which is ultimately the integration time in the memory node, must also be taken into account (this is normally reciprocal to the maximum burst rate of the camera; the camera manufacturer can provide information on this).

For some applications, it may make sense to use 10GigE.

For most applications, however, PLS is irrelevant with today's sensors.

PLS Parasitic light sensitivity

fps Frames per second

QE Quantum efficiency