White Paper

w w w. t m s . m r c y. c o m WHITE PAPER Figure 3. The modern integrated on-screen display A critical attribute of next generation video server performance is their level of programmability. The better they perform defines how well the system can cope with evolving video formats and the types of operations they perform. GPU based video servers inherently have a lot of programmability and versatility built-in to them, making them a good choice to offset evolving requirements. Characteristics of a modern on-platform video server • Modular, open systems architecture for tech refreshes • Contemporary GPU and CPU processors for performance • Software applications defined for upgradability • Hardware accelerated for performance • Low-risk flight safety certifiable for air worthiness On-platform video systems should use a modular and scalable hardware architecture for upgradability of the computing elements enabling them to scale across more input/output video channels, support higher resolutions and offer evolving processing options. Efficient video over IP As cameras proliferate air-platforms and their resolutions increase, the volume of raw video data is becoming very large and difficult to manage effectively over the airframe's network. A tactical approach to resolving this is to compress the video stream within the camera itself, drastically reducing the quantity of data transmitted. The reduced video stream bandwidth (ideally a few 10s Mbps) can then be transferred to the video server over an Ethernet link. Enabled by video server decoding engines already present within current GPUs, video servers can readily capture and distribute additional video channels. A helicopter's tail rotor as an example, may be monitored using this approach. An Ethernet enabled camera placed near the rotor streams its video over an Ethernet link to the video server for displaying in the cockpit. Achieving low-latency Video latency, the delay between the capture of the video input to the actual video display, must be minimized within the video server architecture and made as close to real-time (RT) as possible. Video latency is driven by the type of conversion to be performed and on the video format itself. Interleave versus progressive or packetization of video stream will incur a few frames of latency. Whereas in an analog to digital conversion channel the latency is 2-3 frames or below 100ms. In situations where the input and output formats are identical, the video channel latency can be reduced to almost instantaneous. Defining the on-platform video architecture for perfor- mance and upgradability Good video server architectures are all about efficient data flow. Video signals comprise a continuous stream of high bandwidth data, measured in gigabits per second of pixels flowing across hardware elements that process and compose them on-the-fly. Ultimately, to be recorded and/ or synced to a video port. To achieve server performance, the system architect has to limit the video data movement while performing the various video processing functions. Basically, the architect has to capture the video with hardware elements and copy the data to the video memory and frame buffer with minimal CPU intervention. The frame buffer should hold the video content in a common format to allow video composition from multiple sources and perform compression and conversion in place without creating extra data copies. Video memory must be tightly coupled with the processing elements via dedicated, high-speed/bandwidth busses. 2 Tag/Mark Video Stream Record and Play Seek to Tag

• Cover