The manufacture of single-chip camera systems requires the use of advanced optical alignment systems; however, the method and approach to building multisensor cameras is far more challenging. Lenses and other optical elements must be aligned to each individual sensor; at the same time, accurate sensor-to-sensor alignment must be performed to ensure that images from each sensor are spatially coincident. To achieve this, systems with five and six degrees of freedom (DOF) are now being deployed.
In a typical camera module, a threaded lens barrel is used to attach the lens to the rest of the camera module (see Fig 1a). This type of design does not allow a lateral adjustment of the lens to be performed during alignment or any adjustment in pitch or roll to compensate for tolerance variations in the optical axis of the lens, resulting in less-than-optimal image quality for the camera module.
|FIGURE 1. a) In typical camera modules, a threaded lens barrel is used to attach the lens to the rest of the camera module. Such designs do not allow a lateral or pitch and roll adjustment to be performed during optical alignment. b) Changing the design by eliminating threads and employing adhesive to secure the lens enables adjustments using a five- or six-DOF active alignment system.|
By changing the design, lateral, pitch, and roll adjustment can be performed using a five or six DOF active alignment system (see Fig. 1b). In this design, there are no threads on the lens barrel. Instead, the lens barrel is attached to the camera housing using an adhesive bead located at the top of the camera housing (shown in red). The radial gap between the lens barrel and the camera housing allows active alignment to be achieved with five DOF. A sixth DOF can be also added to accommodate adjustments that may be needed in the yaw axis.
Adhesive selection is significant: It must exhibit low shrinkage during curing, a fast cure rate, high strength after cure, and the ability to tolerate wide temperature extremes. Companies building camera modules typically use dual-cure adhesives to achieve a high-strength bond with a short initial UV cure after active alignment. This is followed by a thermal batch cure or moisture cure.
Five- or six-DOF active optical alignment systems are superior to traditional one-DOF systems because they employ focus score evaluation at both the centers and corners of images acquired from the camera’s sensor during the alignment and assembly process. Using this analysis, the system can adjust for variances in manufacturing tolerances along the lens optical axis, compensate for variations in tip and tilt along the optical axis and lateral variations or centration. This produces a superior focus and image quality at the camera’s image plane.
Cameras inspect cameras
To achieve optical alignment, AEi’s CAMAT system uses a downward-looking camera to view the adhesive after it is dispensed. A side-looking camera is then used to verify the adhesive bond gap after active alignment to characterize and provide data compensating for adhesive shrinkage (see Fig. 2 and frontis on page 12). A pallet fixture is used to hold the camera housing and provide an electrical interface both to this camera and a machine-vision camera mounted below a five- or six-DOF motion stage.
|FIGURE 2. AEi’s CAMAT system uses a downward-looking camera (see frontis) to view adhesive after it is dispensed. A side-looking camera then verifies the adhesive bond gap after active alignment to characterize and provide data to compensate for adhesive shrinkage.|
This motion stage moves the preassembled camera lens through a locus of positions to determine the best overall focus position for optimal image quality. Focus quality is evaluated during these alignment motion scans by using five different regions of interest (ROI) from images acquired via the alignment target located above the stage.
Off-the-shelf hardware and custom AEi-developed software are used to implement multiaxis active optical alignment of either single-sensor or multisensor camera assembly applications (see Fig. 3). The Accelera DMC-18x6 PCI series motion-control card from Galil Motion Control is used to control the motion axes of both the alignment motion stage and the adhesive dispense and UV cure stages.
|FIGURE 3. To implement multiaxis active optical alignment of single-sensor or multisensor camera assemblies, off-the-shelf hardware and custom AEi-developed software are used.|
To image the camera and lens under test, image-acquisition boards such as the 1422/1433 parallel LVDS Camera Link Full board from National Instruments (NI) or the PXC200AF analog NTSC/PAL/SECAM/S-video card from Imagenation are used to support different video signal formats. A camera-control bus interface device such as the NI USB to CAN or LIN or Aardvark I2C bus controller from Total Phase is used to control the machine-vision cameras during image acquisition. These communication cards are used to set image format, exposure, and integration time of the machine-vision cameras, depending on the type of camera that is used.
The Camera Interface Board (CIB), PCB Interface Board (PIB), and the custom pallet PCB developed by AEi support the unique electrical interfaces of the to-be-assembled camera’s electrical connector and image output format. The CIB also supplies the extracted camera frame-sync signal to the motion controller card so images from the camera can be acquired with lens locations during the alignment procedure.
Custom software developed by AEi is used to control and test the assembly procedure (see Fig. 4). This software includes focus-analysis algorithms, machine-control and user-interface functions, and an alignment-process sequencing logic editor. The logic editor allows password-authorized users to optimize parameters and thresholds during the alignment process such as scan step size, starting position, and adhesive dispense-pattern definitions.
|FIGURE 4. Custom software developed by AEi is used to control and test the assembly procedure. This software includes focus-analysis algorithms, machine-control and user-interface functions, and an alignment-process sequencing logic editor.|
Since the traditional modular transfer function (MTF) method of evaluating focus can be adversely impacted by image noise, AEi developed a gradient-based focus score method for implementing active alignment. This algorithm evaluates the rate of change or gradient across edge transitions from light to dark or dark to light in the ROI captured by the camera under test.
After the active alignment process is complete, MTF is also measured and recorded for comparative purposes. This is implemented according to the ISO 12233 standard spatial frequency response method of measuring MTF in cycles per pixel. Using features in the standard CMAT align and test target, this algorithm examines the sides of a rotated rectangle feature within the ROI. The CMAT station records the MTF at the spatial frequency where MTF is 50% or “MTF50” as specified in the ISO standard.
For high-end consumer and professional video cameras, a three-sensor configuration is typically employed with a separating prism. Active optical alignment for these three-chip cameras requires electrical engagement and optical alignment of the three camera sensors. Thus, the sensors need to be interfaced at different positions requiring adhesive to be applied and UV cured at these different angles.
As with all multisensor cameras, centration alignment must ensure that the sensors are coincident laterally and at a tight tolerance so that each of the three sensors will image the same field of view. The range of motion of the five- or six-DOF motion stages of optical alignment systems needs to be greater in order to overcome these challenges. To achieve alignment, the prism is held in the same position during alignment rather than moved.
Aligning and attaching lenses for stereoscopic borescope cameras presents similar challenges. Here, the alignment of two optical lenses to two sensors must be with best focus and with pixel-to-pixel alignment in the image plane. However, this demands more restrictive tolerances since the acquired images are used to generate 3-D image results. The active alignment approach requires the electrical engagement of both sensors simultaneously.
A parallel electrical and mechanical connection to the camera is required. This is implemented in a more complex nest/fixture that accommodates the interface to both sensors. The alignment approach is similar to that of single-sensor cameras except that aligning optics for the second sensor and optics must be evaluated and adjusted for any shift in adhesive attachment, resulting in the first optical alignment.
Andre By is founder and chief technology officer at Automation Engineering Inc. (AEi), Wilmington, MA, USA.
Automation Engineering Inc. Wilmington, MA, USA
Galil Motion Control Rocklin, CA, USA
Imagenation Beaverton, OR, USA
National Instruments Austin, TX, USA
Total PhaseSunnyvale, CA, USA