INSPECTION & TEST: Practical Issues in Image Sensor Module Testing
03/01/2008
BY JOSE MENDEZ, Tessera
Image sensor module testing does not need to be a confusing collection of acronyms, test equipment, and charts. With a clear set of goals and a little bit of know-how, it can be simplified to a few key steps, including material selection, electrical set-up, and module quality measurement.
Begin by identifing the key parameters to be measured. These will generally be consequences of the specific module components being tested. For example, what is the device under test: the package, lens stack, image sensor itself, or a combination of all three? Once these components are identified, a relevant subset from the universe of possible tests tion be selected.
Material Selection Considerations
There is a basic set of equipment required for image module testing. First and foremost is a properly designed dark room. In addition to the space requirement, the room must be isolated from uncontrollable light sources such that light intensity levels well below 0.1Lux are easily obtainable. This will require a Lux meter with a resolution of no less than 0.01Lux and with a sufficiently wide measurement range — 0 to 20K Lux or more. Additionally, the room should be painted with a non-reflective surface, such as an 18% matte gray paint.
Next, obtain the proper light sources. These can vary greatly in color spectrum. Fluorescent lights have more blue and green content than their incandescent counterparts. The term “color temperature” is used to describe the temperature of a nearly spectrum-equivalent black-body light source. A pair of approximately 250W light sources with a color temperature below 5000K will be sufficient. Incandescent tungsten bulbs, rated at 2800K, generally fit within this range. These should be attached to fully adjustable stands, such that the height, vertical, and horizontal angles are customizable. To facilitate even lighting, always attach a semi-transparent diffractive medium to the light source, such as a studio umbrella. To control light intensity, procure two floodlight dimmer switches rated at a higher power than each individual light source. A frame grabber is another essential piece of equipment. Choose one that offers flexibility in terms of data bit width, data type (it should support RAW, Bayer, RGB, YCbCr), and includes a register interface communication protocol, such as I2C. One with at least four independently adjustable integrated low-noise power supplies provides additional convenience. Peripheral components, include two+ meters long testing bench, miscellaneous mechanical clamps with a sturdy base, two-sided tape, rulers/measuring tape, matte black cardboard, lens quality wipes, lens cleaner, an opal diffuser, extension cables, test charts, and, optionally, a colorimeter.
Electrical Set-up
The electrical set-up is the next step. Non-test-dependent issues, such as power supplies, data/power cables, socket/board design and basic register configuration, must also be taken into consideration Most modern image sensors require, or allow the use of, different supply voltages for different functional sections. For example, there might be one pin that powers the analog section, one that powers the digital section and another that powers the communication I/O ports. A flexible electrical set-up requires a minimum of four independently adjustable power supplies. Some sensors also require a specific, timed power-up sequence. Although it is possible to achieve this using PC-controlled bench power supplies, it is easier to use a frame grabber with integrated power supplies that already include this functionality. Finally, low-noise power supplies are required if image noise measurements are part of the project scope, as power supply noise can modulate directly to output data noise.
Because it is convenient, if not always necessary, to be able to move the image sensor during testing, cables should be used to connect the mobile sensor unit with the relatively anchored frame grabber/PC unit. A two-board approach (Figure 1) works well in most cases, allowing one board to connect directly to the frame grabber interface while another board interfaces with the image sensor/socket.
 Figure 1. Dual-board configuration with flex cable interface. |
The rule-of-thumb when designing test print circuit boards (PCBs) is to keep the design spacious so that all traces are always easily accessible. Provide jumper positions for power supplies and static digital inputs. Keep traces on the surface or bottom layers to facilitate any possible rework.
Image sensor socket design is not trivial. There are many issues to consider, such as focal plane lens alignment, light-blocking properties, reflection suppression and force distribution. For instance, image sensors tend to be fragile and will crack readily under high forces. A clam-shell type design is recommended to facilitate access to the sensor. Use as little force as necessary to achieve good electrical contact. However, for optimal results, it is best to use the services of an expert custom design company.
Register configuration can be the most time-consuming step in the testing set-up. Most modern image sensors include an ISP core, which is finely tunable through register access. Unfortunately, the default register settings are neither optimal, nor correct for testing or general use. To complicate things further, some sensor manufacturers do not fully disclose the ISP register space. Functions that are generally controlled by these register values include, gain control, automatic exposure control (AE), automatic white balance (AWB), color correction, edge enhancement, and data output configuration. For most testing configurations, AE and AWB should be turned off. Most sensors use a two-wire (clock and data) I2C communication protocol. Almost any I2C master device can be used to write and read register settings as long as the clock frequency and voltage levels are compatible. Also, refer to the image sensor datasheet for details on the device address (needed for all transfers) and any register page settings to ensure correct register access. Although it is sometimes possible to obtain appropriate register settings by contacting the sensor manufacturer, some iteration is generally needed on the color correction registers, depending on particular testing conditions.
Module Quality Measures
Module quality measures are another important step in the process. The attributes that may be extracted from a given image module system map directly to a myriad of measurement techniques and figures of merit. Modulation transfer function (MTF), color accuracy, and distortion are three basic measures of performance.
 Figure 2. Simplified illustration of MTF measurement and calculation. |
MTF is to an image sensor module as frequency response is to an electrical filter. It is best described as the spatial frequency response of the module, which inherently combines contributions from sensor resolution, focus accuracy, lens quality, and contrast. Thus, it is generally used as a universal measure of image reproduction quality. Figure 2 shows a simplified pictorial representation of MTF. The ability of the sensor to resolve the lines decreases as the spacing between them drops.
Because MTF is not constant across the image plane, most image analysis software packages use the slanted-edge MTF test, which enables localized measurements on different image areas.
Distortion is a higher order lens effect, which results in straight lines on the target plane being mapped as curved lines on the image sensor plane. This aberration generally becomes more pronounced as the distance from the image center increases. Pincushion and barrel distortion are two possible distortion types (Figure 3).
 Figure 3. The top row shows two examples of lens distortion. Below, measurable quantities used to calculate SMIA TV distortion. |
Color accuracy is a measure of how well the module reproduces an image with known color content at a given lighting temperature. The Gretag/Macbeth checkered color chart is normally used for this purpose. Each color square has an ideal RGB color space value, which can be found on the chart itself or online. This chart helps to optimize the sensor ISP color correction/balance settings for the given lighting conditions. To obtain higher accuracy and some immunity to light source variations or aging, a colorimeter can be used, which measures the spectral content of the light source. The response of the module can then be isolated from any light source spectrum effects.
The abundance of measurable parameters resulted in the ISO 12233 chart, an almost all-encompassing test case that attempts to provide much of the information needed to extract all non-color-related module characteristics, including MTF and distortion among many others. ISO 12233 charts can be easily obtained on the web. The most basic image sensor testing lab should include at least one Gretag/Macbeth chart, one ISO 12233 chart, and one distortion chart, all properly sized for the application.
Testing Example
Figure 4 offers a top view of a possible image sensor testing lab layout. An important set-up consideration is that the target must be evenly lit. To ensure this condition, measure the light intensity using the Lux meter at the points shown in Figure 5. Adjust the position and angle of the light sources until the intensity at the target edges is within 5% of the center intensity, although a 10% deviation is also acceptable. Use the floodlight dimmers to obtain the desired Lux level at the center of the target. 1000Lux is a good starting point. Because some tests charts require the full image plane to be occupied, the distance between the module and the target must be adjusted for this purpose. When proceeding with image capture, keep track of which tests were performed and at what light levels by either keeping a detailed log or using a smart file naming convention.
 Figure 4. Top view of a basic imager module test bench set-up. |
null
 Figure 5. The measured light intensity at the four corners should be within 5% of the center measurement. |
null
Conclusion
The first step when developing an image sensor module testing plan is to identify the devices to be characterized and decide on a set of relevant basic tests. In most cases, there is a fundamental set of materials and procedures that will be needed for any functional characterization. This basic collection of items and know-how enables a surprisingly broad testing capability, which can easily be expanded to include further improvements. AP
JOSE MENDEZ, product development engineer RF, may be contacted at Tessera, 3099 Orchard Drive, San Jose, CA 95134; 408/894-0700; E-mail:[email protected].
the short story
With a clear set of goals and a little bit of know-how, sensor module testing can be simplified to a few key steps. Parameters are generally consequences of the specific module components being tested. A fundamental set of materials and procedures are needed for functional characterization.