The introduction of 2Gpbs (S1600) FireWire products this past fall including cameras from Sony1 & Point Grey3, OHCI host adapter and Firespy Analyzer from DapTechnology, and others. 4Gbps (S3200) FireWire was demonstrated by DAP Technology2 at Vision 2011 in November. FireWire has reached the pinnacle of performance defined in the IEEE 1394-2008 standard; we no longer have to wait. FireWire is now running at S3200.
S1600 FireWire is in full production, all required tools are available including protocol analyzers, computer node supporting OHCI with a PCIe x4 add-in card, tested cables, connectors, evaluation boards and IP. Performance improvements, however, need not end at 4Gbps. As long as FPGA technology allows higher performance, FireWire can scale to higher throughputs as we will soon see.
The first application to exploit this performance is industrial vision. S3200 FireWire supports the high throughput and robustness required for next generation industrial vision systems, precision robotics, industrial control, studio, and military applications. Implementing S1600 and S3200 FireWire and an FPGA, allows broad flexibility for widely ranging applications and innovations, including a roadmap to even higher performance.
The pull from industrial vision came from several markets making three general demands; higher resolution, faster, and more flexible.
More detailed inspection for industrial assembly and inspection lines, for example, is driving the demand for higher resolution imaging. High-resolution cameras and associated imaging algorithms are increasing reliability, throughput, consistency and cost of product quality assurance. The security industry also is demanding higher resolution systems to allow for reading of vehicle license plates from an individual vehicle of interest and facial recognition of an individual in a crowded scene. Higher resolution also aids security systems since it allows greater digital zooming of a captured image, which can be done much faster and more quietly than mechanically zooming an optical lens or mechanically panning and tilting a security camera.5 The higher the pixel count, the further an image can be digitally zoomed with acceptable resolution.
The next demand, faster, is for higher resolution and frame rates, i.e. faster assembly lines, faster inspections, faster robotics. With computing speed increasing every few months, the computing power is becoming less and less of a bottleneck in the throughput of vision systems. Even if the computer is not fast enough today, it will be faster in the future, therefore the transport providing the image from the camera to the computer must scale to support the needs of tomorrow.
Higher resolution images place high demands on industrial control systems throughput. One approach to lessen the throughput is to compress the high-resolution images before transport and then decompress before processing at the computer, thus saving bandwidth on the transport. This approach introduces compression hardware in the camera and decompression hardware and/or software in the computing system, i.e. more cost and more SW to setup and control. Most importantly for feedback control systems compression and decompression introduce additional latency into the control loop. Many control systems cannot tolerate much latency and for these systems uncompressed images with their enormous bandwidth needs are required. The Industrial Imaging Digital Camera (IIDC) specification ("IIDC 1394-based Digital Camera Specification") from the 1394 Trade Association supports uncompressed industrial video cameras. Both the 1394 Trade Association document "BT.601 Transport Over IEEE-1394" and the International Electrotechnical Commission standard number 61883 (IEC 61883) support compressed and uncompressed video transport over FireWire. Both cameras used in the S1600 FireWire demo are based on the IIDC specification.
The third demand is greater configuration flexibility to make it easier to adapt to meet conditions of a particular industrial automation cell. For example, after installation, it is found that an additional camera is needed, or an entire second automation cell is needed for the next step in the process. Integrators want the flexibility to add additional machines as required without making additional long cable runs. IEEE 1394 inherently has the ability to daisy-chain cables, so with the addition of more bandwidth, control for an additional machine can be added just by daisy-chaining a FireWire cable from the previous FireWire cell.
The quadrupling of bandwidth from S800 to S3200 will enable numerous new FireWire-based applications to flourish.
Taking Advantage of FPGAs
S1600 and S3200 FireWire implementations are Beta only, they only support IEEE-1394-2008 Beta (1394b) physical connections. Electing to not support bilingual (1394a and 1394b) physical connections was done for several reasons: Most implementations requiring these high data rates won’t fully operate at lower speeds (S100 to S400), simplified digital logic increased reliability and eliminated analog circuitry not supported by most Field Programmable Gate Arrays (FPGA) technologies.
If the capability to interface to legacy 1394a devices is needed, the simple addition of one of the current bilingual physical layers or standalone bilingual repeaters will allow interfacing to legacy 1394a devices.
Integration of this IP actually simplifies current camera designs, which already include FPGAs by integrating the IEEE-1394 physical (PHY) layer into the same FPGA as the link layer and camera specific logic.
This provides all the typical FPGA advantages including easy, quick prototyping and cycles of learning fixes. There is no need to wait until you have "enough" problems identified to justify going through another ASIC FAB cycle. Best of all, there is no need for a multi-month trip through the cycle. The entire FPGA library of intellectual property is at your disposal.
The support ecosystem is in place for FPGAs, including power ICs, clock ICs, filter ICs, debugging tools, and others. You can configure the FPGA for your specific application with your specific IP, they can be encrypted for protection and FPGAs can be upgraded in the field.
In addition, implementing the PHY and Link layers in FPGA IP has enabled FireWire to move beyond S800 (1Gbps), which has been available for almost a decade now. The S1600 implementation falls below the 3.125 Gbps sweet spot for FPGAs allowing for low cost small implementations today and with FPGA advancements S3200 promises to be in the sweet spot next year. Enabling this high level of integration means S1600 and S3200 implementations promise to be smaller and more cost effective than current slower implementations. Also gone is the ASIC support limit for only PCI/PCIe and SATA digital interfaces for 1Gbps FireWire. Now designers have all the flexibility of the FPGA IP library of digital interfaces.
Unlike USB 3.0, IEEE-1394b’s architecture supported these higher data rates since the beginning. The same software (example: Windows 7 supports these higher data rates) can be used as with every other FireWire device. If not already done, updating custom software to support these higher data rates is straightforward because it scales exactly like existing 1394.
There are other advantages, FPGA implementation allows for more than three PHY ports in a single device. This allows for less hierarchy in networks making for fewer PHYs and faster networks since not as many cable hops must be traversed. ASIC implementations that used discrete 1394b PHYs had two single-ended 100MHz clocks that needed to be routed on the PCB. Unless care was taken this was an EMI issue. These clocks are now internal to the FPGA, eliminating the EMI concern.
There are also all the typical advantages of FireWire. The top ten are:
1. High Performance, now S3200 for data transfer, with no overhead at all
2. Isochronous Support, best QoS bus technology for in time, frame-by-frame image delivery
3. Peer-to-Peer communication that allows any node to initiate a transaction, not just the master
4. Flexible Cable Topology, the hardware supports redundant cable connections connected in any manner to support robust, failure tolerant systems
5. Packetized Data Delivery allows many different protocols to be supported at the same time
6. Plug and Play ease of use allows new instruments to be plugged into the system and self-identify to the rest of the system.
7. Memory Architecture supporting DMA, this was part of the original 1394 definition, FirewireTM asynchronous transactions are 64-bit memory reads and writes
8. Open Industry Standard, No companies restrict access to the standards - everyone has access and the opportunity to innovate. This has resulted in extensive support for many protocols, importantly for vision the Industrial Camera Digital Camera (DCAM) industrial vision standards.
9. FireWire is the basis for SAE AS5643 used in multiple military aerospace applications ensuring availability of tools and technology in the future.
10. Modified IBM 8B10B DC-balanced encoding allows easy Galvanic Isolation and many different physical cables including:
a) Shielded Twisted Pair up to 10m@4Gbps available
b) Co-ax cable with EqcoLogic4 Modulators up to 40m@1Gbps
c) UTP5 up to 70m@500Mbps with EqcoLogic4 Equalization
d) Glass Optical Fiber, as fast and as far as is possible
e) Plastic Optical Fiber/Hard Polymer Clad Fiber (POF/HPCF) up to 50m@250Mbps
Finally the path to performance faster than S3200 is clear. FPGA serial I/Os already support 8-Gbps PCI-Express and 6-Gbps SATA I/O cells. But since faster than S3200 is outside the current IEEE standard, the only nodes that can be guaranteed to work are the nodes of the company’s particular implementation. In other words, if you are creating a closed system, where all nodes that will be placed in the system are under your control, a custom implementation going faster than 4Gbps is possible. Migrating IEEE-1394 to support higher data rates and/or higher throughput is always an option. IEEE-1394 is inherently scalable therefore adding additional speeds are relatively straightforward.
About the Author
Burke Henehan is a consulting engineer with more than 25 years of industry experience. He has created project plans, device requirements, test plans and boards, and has worked on standards, strategy formation, silicon validation, debugging customer problems, conducted training seminars and written application notes, data sheets, and white papers. He holds a master’s degree in systems engineering from Texas Tech University and a bachelor’s degree in electrical engineering from South Dakota School of Mines and Technology, and can be reached at firstname.lastname@example.org .
1 Sony S1600 1394b Camera
3 Point Grey Research
5 "FireWire Makes a Move" – Advanced Imaging Pro, January 12, 2011
TM All trademarks are the property of their respective owners