Imaging Electronics 101: Camera Types and Interfaces for Machine Vision Applications
As imaging technology advances, the types of cameras and their interfaces continually evolve to meet the needs of a host of applications. For machine vision applications in the semiconductor, electronics, biotechnology, assembly, and manufacturing industries where inspection and analysis are key, using the best camera system for the task at hand is crucial to achieving the best image quality. From analog and digital cameras, to progressive scan and interlaced scan formats, to FireWire and GigE interfaces, understanding parameters such as camera types, digital interfaces, power, and software provides a great opportunity to move from imaging novice to imaging expert.
Camera Types and their Advantages
Analog vs. Digital Cameras
On the most general level, cameras can be divided into two types: analog and digital. Analog cameras transmit a continuously variable electronic signal in real-time. The frequency and amplitude of this signal is then interpreted by an analog output device as video information. Both the quality of the analog video signal and the way in which it is interpreted affect the resulting video images. Also, this method of data transmission has both pros and cons. Typically, analog cameras are less expensive and less complicated than their digital counterparts, making them cost-effective and simple solutions for common video applications. However, analog cameras have upper limits on both resolution (number of TV lines) and frame rate. For example, one of the most common video signal formats in the United States, called NTSC, is limited to about 800 TV lines (typically 525) and 30 frames per second. The PAL standard uses 625 TV lines and a frame rate of 25 frames per second. Analog cameras are also very susceptible to electronic noise, which depends on commonly-overlooked factors such as cable length and connector type.
Digital cameras, the newest introduction and steadily becoming the most popular, transmit binary data (a stream of ones and zeroes) in the form of an electronic signal. Although the voltage corresponding to the light intensity for a given pixel is continuous, the analog-to-digital conversion process discretizes this and assigns a grayscale value between 0 (black) and 2N-1, where N is the number of bits of the encoding. An output device then converts the binary data into video information. Of importance are two key differences unique to digital and not analog cameras types:
- The digital video signal is exactly the same when it leaves the camera as when it reaches an output device.
- The video signal can only be interpreted in one way.
These differences eliminate errors in both transmission of the signal and interpretation by an output device due to the display. Compared to analog counterparts, digital cameras typically offer higher resolution, higher frame rates, less noise, and more features. Unfortunately these advantages come with costs - digital cameras are generally more expensive than analog ones. Furthermore, feature-packed cameras may involve more complicated setup, even for video systems that require only basic capabilities. Digital cameras are also limited to shorter cable lengths in most cases. Table 1 provides a brief comparison of analog and digital cameras types.
|Table 1: Comparison of Analog Camera and Digital Camera Types|
|Analog Cameras||Digital Cameras|
|Vertical resolution is limited by the bandwidth of the analog signal||Vertical resolution is not limited; offer high resolution in both horizontal and vertical directions|
|Standard-sized sensors||With no bandwidth limit, offer large numbers of pixels and sensors, resulting in high resolution|
|Computers and capture boards can be used for digitizing, but are not necessary for display||Computer and capture board (in some cases) required to display signal|
|Analog printing and recording easily incorporated into system||Signal can be compressed so user can transmit in low bandwidth|
|Signal is susceptible to noise and interference which causes loss in quality||Output signal is digital; little signal loss occurs during signal processing|
|Limited frame rates||High frame rates and fast shutters|
Interlaced vs. Progressive Scan Cameras
Camera formats can be divided into interlaced, progressive, area, and line scan. To easily compare, it is best to group them into interlaced vs. progressive and area vs. line. Conventional CCD cameras use interlaced scanning across the sensor. The sensor is divided into two fields: the odd field (rows 1, 3, 5..., etc.) and the even field (rows 2, 4, 6..., etc.). These fields are then integrated to produce a full frame. For example, with a frame rate of 30 frames per second (fps), each field takes 1/60 of a second to read. For most applications, interlaced scanning does not cause a problem. However, some trouble can develop in high-speed applications because by the time the second field is scanned, the object has already moved. This causes ghosting or blurring effects in the resulting image (Figures 1a – 1b). In Figure 1a, notice how TECHSPEC® Man appears skewed when taking his picture with an interlaced scanning sensor.
In contrast, progressive scanning solves the high-speed issue by scanning the lines sequentially (rows 1, 2, 3, 4..., etc.). Unfortunately, the output for progressive scanning has not been standardized so care should be taken when choosing hardware. Some progressive scan cameras offer an analog output signal, but few monitors are able to display the image. For this reason, capture boards are recommended to digitize the analog image for display.
Figure 1a: Ghosting and Blurring of TECHSPEC® Man's High-Speed Movement Using an Interlaced Scanning Sensor
Figure 1b: TECHSPEC® Man's High-Speed Movement Using a Progressive Scanning Sensor
Area Scan vs. Line Scan Cameras
In area scan cameras, an imaging lens focuses the object to be imaged onto the sensor array, and the image is sampled at the pixel level for reconstruction (Figure 2). This is convenient if the image is not moving quickly or if the object is not extremely large. Familiar digital point-and-shoot cameras are examples of area scan devices. With line scan cameras, the pixels are arranged in a linear fashion, which allows for very long arrays (Figure 2). Long arrays are ideal because the amount of information to be read-out per exposure decreases substantially and the speed of the readout increases by the absence of column shift registers or multiplexers; in other words, as the object moves past the camera, the image is taken line by line and reconstructed with software.
Figure 2: Illustration of Area Scanning Technique (left) Illustration of Line Scanning Technique (right)
|Table 2: Comparison of Area Scan Cameras and Line Scan Cameras|
|Area Scan Cameras||Line Scan Cameras|
|4:3 (H:V) Ratio (Typical)||Linear Sensor|
|Large Sensors||Larger Sensors|
|High-Speed Applications||High-Speed Applications|
|Fast Shutter Times||Constructs Image One Line at a Time|
|Lower Cost than Line Scan||Object Passes in Motion Under Sensor|
|Wider Range of Applications than Line Scan||Ideal for Capturing Wide Objects|
|Easy Setup||Special Alignment and Timing Required; Complex Integration but Simple Illumination|
Time Delay and Integration (TDI) vs. Traditional Line Scan Cameras
In traditional line scan cameras, the object moves past the sensor and an image is made line by line. Since each line of the reconstructed image is from a single, short exposure of the linear array, very little light is collected. As a result, this requires substantial illumination (think of a copy machine or document scanner). The alternative is Time Delay and Integration (TDI) line scan cameras. In these arrangements, multiple linear arrays are placed side by side. After the first array is exposed, the charge is transferred to the neighboring line. When the object moves the distance of the separation between lines, a second exposure is taken on top of the first, and so on. Thus, each line of the object is imaged repeatedly, and the exposures are added to each other (Figures 3a - 3b). This reduces noise, thereby increasing signal. Also, it demonstrates the concept of triggering, wherein the exposure of a pixel array is synchronized with the motion of the object and the flash of the lighting.
Digital Camera Interfaces
Digital cameras have gained in popularity over the past decade because transmission noise, distortion, or other signal degradations do not affect the information being transmitted. Since the output signal is digital, there is little information lost in the transmission process. As more and more users turn to digital cameras, imaging technology has also advanced to include a multitude of digital interfaces. The imaging landscape will be very different in another decade, but the most common interfaces available today are capture boards, FireWire, Camera Link®, GigE, and USB (Table 3).
As with many of the criteria for camera selection, there is no single best option interfaces, but rather one must select the most appropriate devices for the application at hand. Asynchronous or deterministic transmission allows for data transfer receipts, guaranteeing signal integrity, placing delivery over timing due to the two-way communication. In isochronous transmission, scheduled packet transfers occur (e.g. every 125μs), guaranteeing timing but allowing for the possibility of dropping packets at high transfer rates.
Image processing typically involves the use of computers. Capture boards allow users to output analog camera signals into a computer for analysis; or an analog signal (NTSC, YC, PAL, CCIR), the capture board contains an analog-to-digital converter (ADC) to digitize the signal for image processing. Others enable real time viewing of the signal. Users can then capture images and save them for future manipulation and printing. Basic capturing software is included with capture boards, allowing users to save, open, and view images. The term capture board also refers to PCI cards that are necessary to acquire and interpret the data from digital camera interfaces, but are not based on standard computer connectors.
FireWire (IEEE 1394/IIDC DCAM Standard)
FireWire, aka IEEE 1394, is a popular serial, isochronous camera interface due to the widespread availability of FireWire ports on computers. Although Firewire.a is one of the slower transfer rate interfaces, both FireWire.a and FireWire.b allow for the connection of multiple cameras, and provide power through the FireWire cable. Hot-swap/hot-plugging is not recommended, as the connector’s design may cause power pin shorting to signal pins, potentially damaging the port or the device.
Camera Link® is a high speed serial interface standard developed explicitly for machine vision applications, most notably those that involve automated inspection and process control. A Camera Link® capture card is required for usage, and power must be supplied separately to the camera. Special cabling is required because, in addition to low-voltage differential pair LVDP signal lines, separate asynchronous serial communication channels are provided to retain full bandwidth for data transmission. The single-cable base configuration allows 255 MB/s transfer dedicated for video. Dual outputs (full configuration) allow for separate camera parameter send/receive lines to free up more data transfer space (680 MB/s) in extreme high-speed applications.
Camera Link® HS (High Speed) is an extension to the Camera Link® interface that allows for much higher speed (up to 2100MB/s at 15m) by using more cables. Additionally, Camera Link® HS incorporates support for fiber optic cables with lengths of up to approximately 300m.
GigE (GigE Vision Standard)
GigE is based on the gigabit ethernet internet protocol and uses standard Cat-5 and Cat-6 cables for a high-speed camera interface. Standard ethernet hardware such as switches, hubs, and repeaters can be used for multiple cameras, although overall bandwidth must be taken into consideration whenever non peer-to-peer (direct camera to card) connections are used. In GigE Vision, camera control registers are based on the EMVA GenICam standard. Optional on some cameras, Link Aggregation (LAG, IEEE 802.3ad) uses multiple ethernet ports in parallel to increase data transfer rates, and multicasting to distribute processor load. Supported by some cameras, the network Precision Time Protocol (PTP) can be used to synchronize the clocks of multiple cameras connected on the same network, allowing for a fixed delay relationship between their associated exposures. Devices are hot-swappable.
USB (Universal Serial Bus)
USB 2.0 is a popular interface due to its ubiquity among computers. It is not high speed, but it is convenient; maximum attainable speed depends upon the number of USB peripheral components, as the transfer rate of the bus is fixed at 480Mb/s total. Cables are readily available in any computer store. In some cases, as with laptop computers, it may be necessary to apply power to the camera separately.
USB 3.0 features the plug-and-play benefits of USB 2.0, while allowing for much higher data transmission rates.
CoaXPress is a single cable high bandwidth serial interface that allows for up to 6.25Gb/s transmission rates with cable lengths up to 100m. Multiple cables can be used for speeds of up to 25Gb/s. Much like PoE, Power-over-Coax is an available option, as well. A CoaXPress frame grabber is required.
|Table 3: Comparison of Popular Digital Camera Interfaces|
|Digital Signal Options||FireWire 1394.b||Camera Link®||USB 2.0||USB 3.0||GigE|
|Data Transfer Rate:||800 Mb/s||3.6 Gb/s (full configuration)||480 Mb/s||5Gb/s||1000 Mb/s|
|Max Cable Length:||100m (with GOF cable)||10m||5m||3m (recommended)||100m|
|# Devices:||up to 63||1||up to 127||up to 127||Unlimited|
|Connector:||9pin-9pin||26pin||USB||USB||RJ45/Cat5e or 6|
|Capture Board:||Optional||Required||Optional||Optional||Not Required|
|Power:||Optional||Required||Optional||Optional||Required (Optional with PoE)|
Powering the Camera
Many camera interfaces allow for power to be supplied to the camera remotely over the signal cable. When this is not the case, power is commonly supplied either through a Hirose connector (which also allows for trigger wiring and I/O), or a standard AC/DC adapter type connection. Even in cases where the camera can be powered by card or port, using the optional power connection may be advantageous. For example, daisy chaining FireWire cameras or running a system from a laptop are ideal cases for additional power. Also, cameras that have large, high-speed sensors and on board FPGAs require more power than can be sourced through the signal cable.
Power over Ethernet (PoE)
Currently, power injectors are available that allow, with particular cameras, the ability to deliver power to the camera over the GigE cable. This can be important when space restrictions do not allow for the camera to have its own power supply, as in factory floor installations or outdoor applications. In this case, the injector is added somewhere along the cable line with standard cables running to the camera and computer. However, not all GigE cameras are PoE compatible. As with other interfaces, if peak performance is necessary, the power should be supplied separately from the signal cable. In PoE, the supply voltage is based on a standard that uses a higher voltage than standard cameras can supply; this necessitates more electronics and causes more power dissipation which requires sophisticated thermal design to avoid an increase in thermal noise and thus loss of image quality.
Analog CCD Output Signal
There are a few different formats for analog video signals. The format defines the frame rate, the number of display lines, time dedicated to display and blanking, synchronization, the bandwidth, and the signal specifics. In the United States, the Electronic Industries Association (EIA) defines the monochrome signal as RS-170. The color version is defined as RS-170A, more commonly known as National Television Standards Committee (NTSC). Both RS-170 and NTSC are composite signals. This means that all of the color and intensity information is combined into one signal. There are some component signals (Y-C and RGB) which separate chrominance (color) from luminance (color intensity). CCIR is the European monochrome standard while PAL and SECAM are the European color standards. Note: The camera and display formats must be the same to get a proper image.
Laptops and Cameras
Although many digital camera interfaces are accessible to laptop computers, it is highly recommended to avoid standard laptops for high-quality and/or high-speed imaging applications. Often, the data busses on the laptop will not support full transfer speeds and the resources are not available to take full advantage of high performance cameras and software. In particular, the ethernet cards standard in most laptops perform at a much lower level than the PCIe cards available for desktop computers.
In general, there are two choices when it comes to imaging software: camera-specific software development kits (SDKs) or third-party software. SDKs include application programming interfaces with code libraries for development of user defined programs, as well as simple image viewing and acquisition programs that do not require any coding and offer simple functionality. With third-party software, camera standards (GenICam, DCAM, GigE Vision) are important to ensure functionality. Third party software includes NI LabVIEW™, MATLAB®, OpenCV, and the like. Often, third-party software is able to run multiple cameras and support multiple interfaces, but it is ultimately up to the user to ensure functionality.
Though a host of camera types, interfaces, power requirements, and software exist for imaging applications, understanding the pros and cons of each allows the user to pick the best combination for any application. Whether an application requires high data transfer, long cable lengths, and/or daisy chaining, a camera combination exists to achieve the best results. To learn more about imaging electronics, view our additional imaging electronics 101 series pertaining to camera sensors, camera resolution, and camera settings.