As you have read here before, Cadence has been working closely with Xilinx to create an extensible virtual prototype for the Zynq extensible platform. I have previously written about the need and value for extending virtual platforms at the transaction level. According to the Xilinx Zynq website the main application domains for Zynq are "Intelligent Video," "Comms," "Control" and "Bridging." A couple of these sub-application domains are directly targeted to the world of "Industrial Automation." Given that we in EDA are often fascinated by tablets and phones to motivate the value of our products, I had to learn about the industrial domain a little bit more.
From a pure numbers perspective, according to dataBeans numbers from July 2011, the industrial segment accounts in 2011 for about $530 billion in end products, which is about one quarter of the $2.1 trillion electronics pie (only wired and wireless communications together are bigger). The associated semiconductor content accounts for about $37 billion, which is less than 12% of the overall semiconductor pie. Within the segment dataBeans counts "Test & Measurement" (17%), "Process Control" (20%), "Mil/Aero" (10%), "Medical" (11%) and "Other" (41%).
So how does system-level design and software play in this domain? At the 11th annual edaForum held in September 2011 Berlin, I saw an interesting presentation on the "Strategic Importance of Microelectronics in Industrial Automation" given by Andreas von Schwerin, Director for Electronic System Simulation at Siemens Industry in Nuremberg. Mr. von Schwerin described the main reasons to use ASICs in this domain to be the protection of proprietary IP and a unique selling proposition -- high performance combined with lowest latency, which is crucial for motion control applications, lower power, space requirements and cost.
Success factors for the industrial products sold by Siemens are real time behavior with 100% determinism, quality of results, reliability and safety. Seamless HW/SW-interaction is required in spite of high system complexity, high system diversity with nearly unlimited configuration options, the mix and match of standard and proprietary components and short development cycles. As a result Mr. von Schwerin called out advanced design methodologies as key for their success, especially when it comes to verification and system test through the value chain of companies involved and with the use of failure injection to test the designs. He continued on to describe a design flow as shown in this post, which parallelizes hardware and software development after the concept is explored and the architecture is defined. Virtual prototyping is used not only for software development but also for the development of system tests.
Some other interesting aspects were mentioned in Mr. von Schwerin's presentation. For one, the interaction within the design chain has become more and more complicated. The Project SANITAS has been formed as a cooperation project of several players targeting to provide a seamless verification process along the development chain, capable of capturing the effects of micro-/nanoelectronic components and embedded software, identified as key drivers for building safe, reliable, and highly energy efficient products . The goal is to increase verification efficiency and effectiveness by moving to a higher level of model abstraction in conjunction with developing verification mechanisms acting on that level. Another project called "EffektiV" is in the definition phase and will work on a stress testing methodology for motion control systems using virtual prototypes. The outcome of both projects will be very interesting!
Going back to the title of this post, it turns out that there is probably no new revolution to be expected. Design flows for ASICs in industrial applications already seem to be heavily influenced by addressing software with the use of virtual prototypes. Success factors in this segment include system know-how and the ability to master complex value chains -- with strongly interdependent development processes for hardware and software. It will be interesting to see how well the mix of a compute sub-system and a programmable FPGA fabric will address the requirements for the industrial automation domain.
One thing is sure -- the ability to use advanced techniques like the extensible virtual prototype for the Zynq extensible platform will fit right into existing design flows and will not be in the way!