White Paper

the cooling capability of the modules. Since this application was limited to conduction cooling, this would require the use of heavy copper heat exchangers. Since this approach would fail to comply with the weight limit, the developer would need to request a waiver from their customer, reducing the capability of the customer's system. Using the Mercury modeling tool, our first step was to work with our customer to understand the algorithm that would be installed on the FPGA modules. By applying some specific characteristics of the customer's algorithm to the EchoCore Power Load IP, we determined the inputs to the modeling tool. After running our automated script, which populates the manufacturer's power estimation spreadsheet, we calculated the predicted power dissipation of each FPGA device. With this information our mechanical engineering team applied standard thermal modeling techniques for a known heat source to determine that in order to sufficiently cool the devices, for a given rail temperature, one module could use a light-weight aluminum heat exchanger while the other higher-powered module required a copper heat exchanger. By using the heavier, more thermally conductive material only where it was needed, we were able to optimize for both weight and cooling, meeting the program requirements and positioning our customer for success. Additionally, through the scalability of our FPGA power predicting tool, we found the range of FPGA device sizes that could be sufficiently cooled through our planned cooling approach. This knowledge allowed us to quickly provide technical guidance to our customer when they determined that their application would require a larger FPGA device. When compared to the traditional method of making conservative assumptions, it is obvious that accurate modeling is critical to define the architecture and achieve first-pass success. In our current environment of rapidly emerging threats our prime customers cannot afford to settle for sub-optimal designs with extended development lead time. Using this tool, we can quickly understand the power dissipation trade-offs and work collaboratively with our customers to find the best solutions. 5 This tool enables us to compare and correlate the power modeling predictions with the empirical measurements. This step refines the assumptions used as inputs to the model, thereby improving the accuracy of the power model for future FPGA designs. Additionally, this technology allows us to validate production hardware prior to shipping to the customer. This testing step provides the customer with the confidence that the device will operate reliably with their algorithm and reduces the probability of integration issues and failures in the field. Case Study: Maximizing FPGA Performance in a SWaP- Constrained Application As new applications require advanced processing capabilities in smaller footprints, we are seeing our customers increasingly focus on power management. Through the adoption and utilization of the previously described power modeling method, we are able to partner with our customers to provide critical technical guidance on the preliminary architecture definition. As an example of this type of collaboration, we recently partnered with a key customer to support the development of an advanced, SWaP-constrained EW system for an airborne application. This system consisted of two FPGA-based transmit/receive modules, was limited to conduction cooling and included a very strict weight limit. In order for our customer to be successful we had to determine the optimal balance between FPGA processing capability and the weight of the cooling hardware. Before describing how we apply our modeling technique to address this design, we start with the traditional industry process. Without the ability to accurately understand the power dissipation as a function of the customer algorithm, the prudent approach would be to make a variety of conservative assumptions. This would lead to a design that maximized Standardized Universal 6U VPX PCBA Air Cooled Packaging [VITA 48.1] Conduction Cooled Packaging [VITA 48.2] Conduction Cooled Packaging w/LFT [VITA 48.4] Air Flow-by Packaging [VITA 48.7] Air Flow-by Packaging w/LFB [VITA 48.X]

• Cover