Image Sensor Basics - Operation and Testing

An image sensor is a device which converts light into an electronic image. There are two primary types of image sensors in use:
                  CCD – Charge-Coupled Devices
                  CMOS – Complementary metal oxide semiconductors.

Although the design complexity and performance may vary, the basic operation of both of these types of sensors is similar, and neither is categorically better than the other. In this article we will discuss the operation and testing of a highly integrated CMOS sensor which contains the digtizing circuitry on the same die as the pixel array.

The Image Pipeline

Here is a simplified view of the “Image Pipeline”.  We will briefly discuss each stage of operation and testing from the light source to the pass/fail binning selection.

Figure 1 - The "Image Pipeline" diagram showing how we get from Photons to test result.

Illuminator

A fully calibrated light source is used to stimulate the pixel array.  The light source offers several important features:

• Highly collimated uniform light source
• Selectable color wheel for red, blue, green, and white
• Adjustable wedge and ND filters
• Variable intensity from 0 to 6000 Lux
• Automated control from the ATE test program.

 

Filter

On its own, a pixel has no color. To achieve wavelength selectivity, a filter must be applied to the pixel.

Most of us interact with image sensors that have a Color Filter Array such as the Bayer filter shown on the left.  This type of filter is designed so that the sensor will produce an image close to what the human eye sees.

Visible light is just a portion of the radio spectrum.  Many other types of filters exist to capture the infrared, ultraviolet, and X-ray wavelengths.

 

Pixel

A pixel consists of three basic components

1. Photo Diode.  This is the device converting photons into electric charge.
2. Capacitor.  This will store the charge which correlates to the number of photons entering the pixel.
3. Sample and hold circuit.  This usually involves a type of Correlated Double Sampling.

To Sample the light coming into the pixel, the following procedure is used:

1. Reset. Close the reset switch and charge the capacitor to VDD_RST. Open the reset switch after the capacitor is fully charged.
2. Integration. Close the integration switch.  As the photo diode is stimulated it will cause the charge in the capacitor to drain through the photo diode.  Open the integration switch after the integration time has passed.
3. Sample. Close the sample switch.  This will cause the sample and hold circuit to take a snapshot of the charge stored on the capacitor.  Open the sample switch. The pixel value is now ready to be scanned out.

 

Digitizing Path

Depending on the type of image sensor, portions of the digitizing process may be performed by external integrated circuits. In our example, one digitizing has been fully integrated into the CMOS sensor.
Driven by a state machine, the column buffer and line driver scan through each pixel, sending its value to the PGA. The PGA will then amplify and level shift the sample to match the full scale input of the ADC.
After being digitized, there may be additional processing stages such as digital filtering or serialization before the data is outputted.

ATE

Most ATE testers are capable of capturing image data.  The Teradyne IP750 contains several Protocol-Aware instruments designed to capture image data from a variety of designs making it an ideal choice for a variety of image sensor designs.

As the image sensor streams out data the ATE instrument can trigger off of the H-sync and V-sync signals provided by the image sensor to capture the desired number of frames.

Once the data has been stored into memory as image planes, the image tests can be performed.

Fixed Pattern Noise
Using a median filter we can identify failing rows or columns.

Algorithm: Source->Median Filter->Subtract filtered plane from the source plane.

After performing these manipulations, most of the pixels are black and the defects are easy to locate.

Point Failures and Blotches
Using the same median filter technique, we can also identify point failures and blotches.

The defects are easy to identify in the final image.
 

Uniformity
We can easily compute the mean and standard deviation for the image.  This helps identify if the image plane is showing a gradient, tint, or broken regions.

Figure 12 - Uniformity failure showing a gradient

Image Response
We can sweep the intensity of the illuminator while capturing several frames.  This helps determine the linearity of the pixel’s response to the changing stimulus.

Figure 13 - Example image response data

Integration Linearity
For integration linearity testing we sweep the integration time while holding the light source intensity constant.  This is effectively the same as image response.  In both cases we are increasing the photon count.

Figure 14 - Example integration linearity data

Extinction Ratio
By sweeping the hold time between integration and sampling we can see how much leakage the pixel has. Ideally, the pixel would retain the charge indefinitely.

Figure 15 - Example extinction ratio data.  3 different integration times are shown.

Noise
By capturing several consecutive frames we can determine the inherit noise for each pixel.

Pass/Fail
The pass fail criteria for each device is different.  A combination of the image defects and other digital, DC, and functional tests will be used to determine if the part is good or bad.