Machine vision synchronized with the control cycle
Increasingly, the historically established separation between automation and vision is proving to be a hindrance precisely where precision and speed are required. With EtherCAT as a standardized communication backbone, however, image processing becomes an equal participant in the network and a key component of the control architecture. For machine builders, application engineers, and vision integrators, this means: highly precise synchronization with all machine processes, reduced engineering and hardware costs, and simplified commissioning and diagnostics.
Anyone integrating an inspection task into a design today rarely struggles with the algorithm – they struggle with the interfaces. Separate vision controllers are connected to their own network, exchange status words with the PLC via a gateway, and operate on an uncoupled time base. A flash controller triggers the illumination via a third signal chain, while another engineering tool manages the camera parameters. Each of these limitations costs installation space in the control cabinet, terminal points, commissioning time – and, above all, determinism. Until now, communication between image processing and the control system has been complex and difficult to manage because latencies arising from network stacks and operating system factors cannot be reliably calculated. A twofold problem: The system runs more slowly than necessary, and it requires specialist knowledge in two disciplines at the same time.
“The separation of image processing and control was never a technical necessity, but rather the result of two engineering cultures that developed independently,” emphasizes Bernd Stöber, Senior Product Manager for Vision at Beckhoff. “Anyone who still maintains two separate worlds today pays the price in complexity – the most expensive ‘material’ of all in machine building.”
Reason 1: A standardized time base
The technical basis for integrating PLCs and vision systems is EtherCAT’s distributed clocks (DC) mechanism. Unlike purely synchronous communication, distributed clocks operate with a high tolerance to jitter. Synchronization is entirely hardware-based: The time of the first DC-capable slave is distributed cyclically to all other participants; the propagation delay in the network is measured and compensated for during start-up – and, if required, continuously during operation. The resulting jitter is well below 1 µs, typically in the double-digit nanosecond range. The absolute system time is stored with 64-bit resolution; time measurement is carried out with a resolution of 1 ns.
This is the key factor in image processing: Cameras and illumination are operated at precisely defined times in relation to all other machine processes – both relative to the devices on the network and absolutely relative to a higher-level reference time. This allows image capture to be triggered deterministically based on a conveyor belt position or an axis position, rather than being initiated via a photoelectric sensor and an undefined software handshake. Track and trace applications and precise product tracking in EtherCAT-synchronized applications thus become standard practice.
Reason 2: Triggers and lighting in real time
Consequently, the light also originates from the same time base. The EL2596 LED strobe control terminal contains a flexible power supply unit that supplies the illumination with the required current and voltage – ranging from continuous light to short light pulses in the kHz range. Each individual flash is triggered in a controlled manner by the control system via the distributed clocks/timestamp function. The terminal has a trigger output for triggering the camera and fast current and voltage control, so that even line scan cameras, for example, also benefit from constant illumination. Comprehensive real-time diagnostics of the input and output currents and voltages enable detailed monitoring of light intensity. If the load falls outside a specified range, the terminal switches off – a function that can be reset – to protect the LED. The timestamp-based EL2262 output terminal is also available to send a hardware trigger signal with microsecond accuracy to the camera.
Beckhoff’s multicolor LED lighting – in area, ring, and bar lighting designs – produces constant illumination conditions and creates the best possible contrast between the inspection feature and its surroundings in spectrally adjustable pulse mode. As synchronization via distributed clocks does not require an external trigger loop, pulse operation can be implemented without the need for additional hardware or software interfaces. In effect, very high light intensities freeze movement without causing thermal overload in the illumination system. Components no longer need to be decelerated and then accelerated again for testing, which saves space in the system. At the same time, power supply requirements, cooling demands, and energy consumption decrease, while the service life of the LEDs increases and the sensitivity to ambient light is reduced.
Reason 3: Shorter reaction times
The fact that vision results immediately translate into action is, first and foremost, a property of the bus. EtherCAT overcomes the drawbacks of standard Ethernet through its operating principle: As a rule, a single frame is sufficient to update the output data in all participants and to read in the input data using the same frame. Each participant reads the data addressed to it “on the fly” and places its input data into the forwarded frame; the telegram is delayed only by the hardware processing time. There is no store-and-forward latency caused by switches, no protocol stacks at the intermediate layer, and no unpredictable delays, because only the master is allowed to transmit actively.
Image processing therefore does not run alongside the control system, but within it. With no time lost due to interfaces, it is possible to react directly to the results – literally in the very next line of PLC code and with instant communication to the motion components.
Reason 4: Reduced effort and costs
This integration has a positive financial impact on the parts list. There is no longer any need for a separate vision controller, flash controller, or the associated interfaces. All vision hardware components are configured via EtherCAT in an engineering environment. Equally important is the freedom in system layout: EtherCAT allows for a free choice of topology and combines line, star, and tree structures, so that cameras and illumination systems can be positioned where the process requires them, rather than where a switch infrastructure permits.
EtherCAT P takes this concept all the way to the camera: communication and power are transmitted via a standard four-wire Ethernet cable. All the properties of EtherCAT are retained, such as the free topology, on-the-fly processing of telegrams, high-precision synchronization, and extensive diagnostics. Unlike with traditional Power over Ethernet, participants such as cameras and their associated illumination systems can be connected in a cascaded configuration to a single power supply unit. The result is reduced engineering and hardware costs, simplified commissioning, and a significantly smaller system footprint. This is particularly evident in Beckhoff’s vision units, which combine a camera, multicolor LED illumination, and focusable optics with liquid-lens technology in a sealed IP65/67 housing. The PLC uses EtherCAT to control all the settings of the functional components at runtime – focus position, light color, intensity, and pulse width, as well as the camera parameters.
Reason 5: Consistent diagnostics
Vision benefits directly from the bus’s inherent diagnostic capabilities. In each participant, the EtherCAT Slave Controller checks the passing telegram using a checksum; only telegrams received without errors are made available to the application, while those containing errors increment a counter and are flagged for subsequent participants. By reading these error counters, the master pinpoints the exact location of the error – a significant advantage over traditional fieldbuses, where malfunctions propagate along the shared line and the source remains unknown. Rare sources of interference are detected and pinpointed before they affect machine operation.
Even a loose plug can be quickly identified using the Link-Lost-Counter. Additional features include scanning and verification of the network topology during start-up, as well as diagnostic mechanisms across all ISO/OSI layers specified in the fieldbus standard. Because machine data and image data are aligned on the same timeline, and images are saved with reliably traceable timestamps, root cause analysis can then be conducted across axis profiles, I/O signals, and recordings. Unlike a conventional GigE Vision camera, in which fixed user settings are pre-programmed, EtherCAT-controlled hardware always remains open to the parameters specified by the PLC at runtime. This not only makes replacement easier in the case of service, but also eliminates the need to search for faulty parameters in the connected devices.
Conclusion: From an add-on to an integral part of the machine
It is not automation technology that lays the foundations for efficient vision integration, but the underlying communication technology. EtherCAT is standardized as an open standard – including the application layer and device profiles – and is being further developed by the EtherCAT Technology Group as the official standardization partner of the IEC. It is precisely this standardization that makes a camera an equal participant in the automation system: It is assigned the same 64-bit system time as the servo axis and digital terminal, it resides in the same cyclic process image, and it is monitored using the same diagnostic mechanisms. Where there used to be a gateway between the vision controller and the PLC, there is now a frame that is processed as it passes through. EtherCAT typically requires just a single telegram for the entire network, making it more streamlined and faster than other fieldbus or Industrial Ethernet technologies. Vision thus evolves from an add-on to an integral part of the machine, without the designer having to maintain a second system environment.