Real-time processing. “Time-critical events call for efficiently managing time synchronisation, latency and throughput,” says Bharat.
“To overcome this issue, we need high-speed processing and communication infrastructure. We also need to have very good audio-visual file-compression algorithms to reduce file size,” he adds.
Their storage space, the memory. “The desired features in a media processor are fast cores that can handle math-intensive graphical processing, along with plenty of random access memory (RAM),” explains Giovino.
Explaining the concept of memory, Bharat says, “There are two types of memories used in various processor based solutions. The current trend is permanent memory like embedded-multimedia-card, serial advanced technology attachment and solid-state memory, which are small in physical size but offer large storage capacities. Multi-chip package and package on package memories are also popular in mobile and consumer electronics applications.”
Adding to this, Giovino goes into more detail, “Larger amounts of on-chip memory, or fast external memory interfaces, are necessary to efficiently handle memory-intensive tasks like buffering images, streaming high-definition (HD) video, or dealing with compression.
Compression also helps with reducing the bandwidth required for wired or wireless data transmission; media processors must rapidly decompress or even decrypt multimedia data, especially if there is any digital rights management (DRM) encryption present.”
Smaller the node….
The technology node is at 14nm, even 10nm now. Rei feels that, as we go further down, challenge would be to design transistors of such small dimensions. “More transistors within a given area lead to more capabilities, and it is a matter of choosing the right design process,” he adds.
Newer materials used in manufacturing are also influencing changes for more suitable semiconductor designs, feels Srinath.
A media processor processes the inputs it gets, and hence sensors become vital. A few qualities expected in such sensors, from Bharat’s experience, are support for night vision, optical/digital zoom, area of interest, ambient light sensing, auto-detection of objects, filter option for video based noise and real-time data collection.
“Built-in analogue-to-digital converters, digital signal processors, digital interfaces instead of analogue interfaces, and outputs are as important,” chips in Srinath.