Measure dimensions faster than the brain can comprehend, record details faster than the hand can move and find flaws that the eye cannot see—these are routine feats performed by today’s machine-vision systems. Advancements in image-sensor technology, video-camera architecture and frame grabbers, combined with the high data-processing power of modern PCs, make these actions possible. Today’s systems can provide fine resolution (25 x 106 pixels), high light sensitivity (1 x 10-7 foot candles) and dozens of quantified results in seconds.
For some time, vision systems have automatically determined the correct application of solder paste and proper placement of SMT components. But more PCB inspection and evaluation tasks can be performed by today’s systems.
For example, one new application checks plating purity on device leads or connector fingers, which requires ultra-high color resolution. The mechanical anomalies encountered during vibration testing can be assessed using high-speed imaging.
While PCs can perform equally well in almost any of these applications, the same is not true for cameras or, more specifically, sensors. The most commonly used image sensor today—the charge coupled device (CCD) array—comes in a variety of configurations, each matching a specific range of operational conditions. Application-specific requirements also affect the format and read-out rate of the video signal the camera produces.
As a result, selecting the right camera is of utmost importance when you assemble a new video-based inspection or analysis system. The following information on image-sensor CCD operating principles, camera types and output formats can help you make wise choices.
CCD Principles
CCD image sensors provide the primary means of acquiring video images today, having surpassed video tubes a decade ago. Their operation is based on the principle that electric charges can be generated via photon-to-electron holes/pairs conversion.
A photosensitive CCD cell consists of a MOS capacitor that stores charges accumulating by photon-to-electron conversion during a controlled time period. Since a stored charge produces a voltage in accordance with the relationship V = Q/C, an output voltage develops that is proportional to the light intensity and exposure time. The charge could also produce a current of i = dq/dt.
Alternatively, the charge may be transferred to an adjacent cell. A series of interconnected cells—coupled, hence the name charge coupled device—may form an analog shift register. Sequential shifting of the charges to an output sense amplifier will result in an analog signal which is a pixel-by-pixel representation of the image projected on the CCD array.
Individual CCD cells may be as small as 7.0 m m x 7.0 m m and be interconnected to form a single line or a 2-D area array. While many area arrays feature a rectangular aspect ratio of 4 x 3 to match the common television wider-horizontal-than-vertical screen format, square arrays with square pixels are preferred for many measurement applications.
Line Scan and TDI Arrays
Line-scan arrays contain a single line of pixels which are repeatedly exposed to light and read-out to a sense amplifier. The image of an object is sequentially assembled as the object is moved across the array perpendicular to the line of pixels. Resolution may range from 128 to 6,000 pixels and line transfer rates may be as high as 50 kHz with data rates of up to 60 MHz.
High-definition images are best obtained if the object is still during exposure and moved during read-out, or if the exposure time is short compared to object motion. Line- scan arrays are typically used for scanning documents, as in fax machines, and continuous-process inspection applications.
If exposure time is very short or lighting is inadequate, the number of photons gathered and converted during a single exposure may not be sufficient to generate a usable signal. The time delay and integration (TDI) array solves this problem: It takes multiple exposures of the same object at successive locations.
The TDI array consists of an interconnected set of CCD rows, referred to as stages. After an initial exposure is taken at the first stage, and while the object is moving toward the second stage, all accumulated charges are transferred to corresponding cells in the second stage. Another exposure is taken, but this time using the second row of CCDs to accumulate more charges. This process is repeated as many times as there are stages, which may be as many as 96.
The output signal is in a format equivalent to that produced by a line-scan CCD. Tight synchronization between object motion and charge transfer is essential for proper TDI operation.
Area Arrays
An area CCD array consists of M columns of N pixels, forming an M x N rectangular or square matrix. Three architectures are in common use—full frame, frame transfer and interline transfer. Each has its own way of transferring, storing and accessing charges.
Full-Frame Architecture
In the full-frame architecture, subsequent to projecting an image on the CCD, the charges accumulated in each pixel are sequentially shifted downward into a read-out CCD shift register. The scene is transferred one line at a time until the entire image has been read out. All light should be blocked during the transfer, requiring exposure control via strobe light or mechanical shutter. Resolutions range from 256 x 256 to 4k x 4k pixels.
Frame-Transfer Architecture
A frame transfer sensor has an active pixel region onto which the image is projected, plus another M x N storage region that is covered with opaque material. Subsequent to exposure, the entire frame is transferred at high speed from the active region to the storage cells. From the storage cells it is shifted into the read-out CCD.
Since there are two arrays, image gathering time and image transfer time are independent of each other. This separation makes it possible to introduce electronic control of charge accumulation on the active array, enabling programming of shutter speed and synchronizing exposure with external events.
Interline Transfer Architecture
The interline transfer architecture also enables electronic shutter-speed programming but does not employ a separate M x N array. Instead, the sensor’s active pixel area and storage register are contained within the active imaging area.
There is an obvious disadvantage here. The photosensitive area of each pixel is smaller in size relative to frame transfer implementations, and consequently, less light is gathered. This can result in a pixel fill factor of only 35%. However, some interline transfer CCDs now feature microlenses over each cell, bringing the fill factor back up to 70% or more.1
Advantages include faster transfer time and greater read-out implementation flexibility. Transfer of charge packets to the storage register can be as fast as 5 ns, a fraction of the time required by frame transfer implementations. Individual pixels also have a lower dark current, providing a wider dynamic range.1
Area Camera Output Formats
Most general-purpose monochrome video cameras provide an output that conforms to the RS-170 (U.S. television) or CCIR (European television) format. An RS-170-compliant image contains 525 horizontal lines. These lines are not consecutively transmitted, but are comprised of two separate fields, each containing 262½ lines. Fields are transmitted every 1/60 of a second.
To obtain the full 525-line resolution, two images and the resultant two fields are combined into one frame. This is not objectionable for visual observation since the eye integrates the two fields, but may cause problems for machine-vision applications.
If the object moves in the time interval between the two fields, blurring may occur, which could lead to inaccurate quantitization. If a succession of short exposures must be followed by immediate full-frame processing, only data from one field may be taken each time, providing just 262½ lines of resolution.
There are also speed limitations. Fields are available only every 16.6 ms, full frames only every 33.3 ms. Even if a higher scan-out rate is used, the single output of RS-170-compliant frame transfer CCDs cannot emit data any faster than the slowest individual charge-transfer occurrence in the entire array.
To circumvent these problems, application-specific cameras use variable scan for demanding applications. Some also use multiple output CCDs. Variable scan imposes no restrictions on the organization of the pixels in the image. Although general by definition, there is a set of required and agreed-to features defined in a User Bus Standard.2
Scan output consists of valid pixels, valid line and valid frame data and synchronization signals. In contrast to RS-170, there is neither a standard time between lines or frames nor a fixed blanking time. Compatibility between the frame grabber and camera model you select must be verified to assure interoperability.3
Cameras
Features, performance, capabilities and implementations of cameras used for inspection vary widely today. After establishing the performance requirements of your system, the first step in a selection process is to determine whether a nonstandard or an RS-170/CCIR camera should be used.
“Nonstandard CCD cameras are separated by their ability to realize improved speed and acquisition accuracy beyond that achievable with their standard RS-170 video camera cousins,” remarked Dave Litwiller, Manager of Development Engineering at Dalsa. “And quality and yield requirements of the electronic industry make extra demands on machine vision.”
Most cameras of either type are operated in conjunction with frame grabbers. Some include internal pre-processing electronics and interface with a PC via a plug-in card, and a few have a built-in computer. Output is most often analog, although some cameras include internal A/D converters providing an 8-b quantized gray-scale signal. Control and synchronization signals are generated internally or furnished by the frame grabber and PC.
Some products are based on specific techniques or processes perfected by the manufacturer. Pulnix’s product line, for instance, includes not only a variety of interline transfer CCDs but also progressive scan models they developed.
“Progressive scan simply means non-interlace and leaves the RS-170 world behind in speed and resolution,” explained Richard Hofmeister, Manager, Industrial Products Division at Pulnix. “These cameras offer 1,024 x 1,024 pixels resolution, partial scanning and electronically controlled shuttering to 1/16,000 s.”
Analyzing objects in motion requires taking images at higher speed. “There is a misconception that using a faster shutter increases the camera’s ability to capture motion,” said Gordon Kent of Vision Research. “It is true that the shutter will reduce or eliminate blur, but a lot of motion may take place between the time images are recorded. Imaging your subject at 500 pictures/s shows details you cannot capture at an RS-170 rate.”
Since high-speed shuttering results in low light levels, some companies, such as Xybion, have integrated image intensifier (night-vision) technology with CCDs. “The electron gain of these devices can attain levels in the tens and hundreds of thousands,” said Richard Sturz, Director of Sales at Xybion. “At these sensitivities, shuttering for periods as short as 5 ns is attainable.”
Trends
In the past, data transfer and processing rates presented the major speed limitations. With the advent of the PCI bus and fast Pentiums, the camera output data rate is the limitation. As time goes on, more cameras will make use of multiple-output CCDs. More cameras will also provide digital outputs, reducing noise levels and supplying more accuracy and a greater dynamic range.
Active pixel sensors will become more prominent, providing better data transfer capabilities. Charge Injection Devices (CID) may also be used increasingly. CIDs provide features not available with CCDs, such as addressable pixels and nondestructive read-out. By accessing only a selected number of pixels representing an area of special interest, much faster read-out, processing and update rates can be achieved.
Acknowledgment
Some of the background material for this article was provided by Dave Litwiller, Manager of Development Engineering at Dalsa. We thank him for his contribution.
References
1. Barret, J., “Selecting the Correct Video Camera for Test and Measurement Applications,” Cohu, Electronic Division.
2. CCD Image Sensors and Cameras, Dalsa.
3. Jacob, G., “Shedding Light on Image Analysis System Components,” EE-Evaluation Engineering, July 1995, pp. 115-121.
Products
Interline Transfer CCD Camera
Occupies Minimal Space
The 2100 Series Monochrome Camera features a 1/2″ interline transfer CCD with a resolution of 580 HTVL. It measures 1.5″ x 3.75″ x 2.12″ without the lens. The size of the CCTV camera makes it well suited for installations where space is at a premium. Cohu Electronics Division, (619) 277-6700.
Area-Scan Camera
Provides High Resolution
The CA-D4-1024A (dual output) and CA-D7-1024A (single output) Cameras provide a resolution of 1,024 x 1,024 with 12 m m x 12 m m pixels. The IA-D4-1024A sensor features a frame transfer architecture with an on-chip storage region, obviating the need for an external shutter. A 100% fill factor is realized since the entire image area is photo-sensitive. Providing 8-bit digital outputs at a rate of 25 MHz/output, the CA-D4 generates up to 40 frames/s and the CA-D7 20 frames/s. Dalsa, (519) 886-6000.
Camera Features Progressive
Scan, Multiple Output Formats
The TM-1040 Series Progressive Scan Camera enables full-frame asynchronous dynamic image capturing at 30 frames/s with high resolution (1,024 x 1,024 pixels). The 1″ interline transfer CCD features a cell size of 9.0 m m x 9.0 m m. Output is provided with 8- or 10-bit gray levels over the RS-422 digital bus or via the RS-343 interlace analog video. Remote communication over RS-232-C is standard. Pulnix America, (408) 747-0300.
Digital Motion Analysis System
Employs Antiblooming CCD
The Phantom V2.0 is a digital high-speed motion acquisition and analysis system. It features high sampling rates and rapid shuttering, and employs an antiblooming 512 x 512 CCD pixel array offering 500 full-resolution pictures/s. Image playback, enhancement and data-reduction capabilities are built into the camera. Image files can be transferred from camera memory or an internal hard drive to a PC serial/parallel port, Ethernet or the camera’s built-in PCMCIA card drive. Vision Research, (800) RESOLUTION.
Camera Interfaces Via PC Card,
Without Frame Grabber
The EDC-1000 Camera Series encompasses seven models offering asynchronous and subarray scanning, an 8-bit/pixel gray scale, minimum exposure time of 1 ms and operation in the TDI mode. The medium-resolution EDC-1000M features a frame-transfer CCD with square pixels and low noise (60 electrons rms). The EDC-1000L has lower noise (35 electrons rms) and higher resolution (753 x 484 pixels). The EDC-1000D combines the features of the EDC-1000L with 24-bit color. Camera power is derived from the computer bus. Electrim, (609) 683-5546.
Intensified CCD Video Camera
Features Electronic Gating
The ISG Series intensified CCD camera offers 18-, 25- and 40-mm intensifier versions. Images may be acquired synchronously at a selectable 30- or 60-images/s rate or may be obtained asynchronously. Event-triggered imaging is also possible. For high-speed imaging, gating to 5 ns may be internally or externally generated. The camera’s frame flag permits internal or external gate timing for frame-grabber compatibility. On-array time integration for increased sensitivity is also offered. Xybion Electronic Systems, (619) 566-7850.
Cooled Color Camera
Contains Three CCD Chips
The C5810 Cooled Color CCD Camera is built around three 1/2″ 768 x 495 pixel interline CCDs with microlenses. It features on-chip integration and exposure times from 1/10,000 to 30 s. The control unit provides two image memories to enable real-time background subtraction. Direct digital output via SCSI bus connections facilitates interfacing with PCs. Multiple video outputs include Y/C, RGB and composite video signals. The C5810 is suited for demanding low-light level imaging applications. Hamamatsu Photonic Systems, (908) 231-1116.
Camera Offers Electronic
Shuttering, Enhanced Image
The Kodak ES 1.0 Megaplusä Camera offers electronic shuttering, real-time 30- frame/s operation and 8 bits/pixel. Two channels operate at a 20-MHz rate each. The camera’s CCD interline transfer sensor provides a spatial resolution of >106 pixels (1,008 x 1,018 array). Each pixel measures 9 m m2 and has a 60% effective fill factor. The ES 1.0 comes in a compact housing, weighs less than two pounds and interfaces with PC-compatible, Macintosh, Sun and other platforms. Eastman Kodak, (619) 535- 2909.
Board-Level Camera
Provides SVHS Output
The MB-1060C is a high-resolution SVHS digital modular board-level camera. It features electronic shutter-speed control, a backlight compensation function, an auto gain control, an adjustable white balance mode, and separate Y, C, (SVHS) and composite video outputs. The two-piece PCB assembly is interconnected by a flex cable. The design allows maximum mounting flexibility and the freedom to set the lens at any desired angle. Polaris Industries, (800) 752-3571.
Sidebar
Glossary of Terms
Area Array—solid-state video detector consisting of rows and columns of light-sensitive semiconductors; sometimes referred to as a matrix array or focal plane array; used in most television cameras.
Charge-Coupled Device (CCD)— semiconductor device capable of transporting finite isolated charge-packets from one position to adjacent position by sequential clocking of an array of gates; charge-packets considered minority carriers with respect to semiconductor substrate.
Charge-Injection Device (CID)—specific fabrication scheme for solid-state image sensors; photo-generated charge sensed by injecting it from sensor into substrate.
Dark Current—current flowing or signal amplitude generated in a photosensor placed in total darkness; dark noise expressed as current flow.
Fill Factor—indication of image-to-electron conversion efficiency; influenced by ratio of light-sensitive pixel area to total pixel area.
Frame Rate—number of times per second that frame is scanned; RS-170 standard—30 frames per second.
Gray Scale—variations of values from white through shades of gray to black in digitized image; black usually assigned digital value 0, white designated as some non-zero value.
HTVL—horizontal television lines.
Interlaced Scanning—for standard video cameras; two consecutive fields per frame; 262½ horizontal lines per field; fields offset to provide frame resolution of 525 lines.
Linear Array—solid-state video detector consisting of single row of light-sensitive elements; used in linear array cameras.
Progressive Scanning—method of scanning image information out of CCD in a sequential, line-by-line basis.
RS-170 Standard—Electronic Industries Association standard governing monochrome television studio electrical signals; specifies maximum amplitude of 1.4 V, peak-peak including synchronization pulses.
Shuttering—does not necessarily refer to a mechanical device; images electronically “shuttered” by collecting or scanning out portion of charge accumulated over one field time.
Copyright 1996 Nelson Publishing Inc.
May 1996