TI boosts Piccolo MCUs with control-loop-centric accelerators

October 23, 2014 // By Graham Prophet
You can, says Texas Instruments, step up control system performance and increase integration while decreasing system cost with it latest C2000 Piccolo F2807x miccrcontrollers: they are designed to add flexibility and platform scalability for industrial control applications such as digital power, renewable energy and automotive control

TI positions this series of MCUs as providing (up to) a doubling of system performance without doubling cost or changing a hardware or software design. They use the established C2000 Piccolo architecture and maintain compatibility within that family, but add dedicated accelerator blocks, and uprate the core DSP engine from 90 MHz to 120 MHz. This, TI says, boosts execution speeds of control tasks in industrial applications such as telecom rectifiers, server power, solar micro inverters, frequency inverters and automotive HEV/EV. The F2807x MCUs also provide many analogue and control peripherals to enable more integrated control applications. They are scalable with the previously announced C2000 Delfino F2837xS and F2837xD MCU generations.

One of the added function blocks is the CLA, or real-time control accelerator. This unit relieves the main core of real-time-intensive and -critical functions, avoiding the need to rigorously partition cycles from the main core's execution. It can double the throughput of the F2807x MCU, providing an additional 120 MHz of floating-point processing capability. This additional bandwidth allows designers to run parallel math-intensive or time-sensitive signal processing tasks. Offloading these tasks to the CLA enables the CPU to focus on general system tasks such as diagnostics, communications, and motion profiling. The CLA has its own dedicated instruction set, and is supported in the TI development environment with its own function libraries.

A further accelerator block, the trigonometric math unit (TMU) places in hardware many of the trigonometric or geometrical functions – for example, sin, cos, atan, divide, square root – needed for transforms and motion profiling; it typically cuts the clock cycles needed to carry out, for example, angle calculations, by 80-90%. Instructions that are best executed by the TMU are automatically trapped and directed by the code compiler.

next; analogue integration...