Control Intelligence Improves Renewable Energy Efficiency

Sep 1, 2007 12:00 PM
By Arefeen Mohammed, C2000 Applications Engineer, Texas Instruments, Dallas

Integrated peripherals such as analog-to-digital converters (ADCs) and pulse-width modulators (PWMs) make it possible to directly sense inputs and control power MOSFETs, saving system space and expense. On-chip flash memories aid in programming and data collection, and communication ports simplify design for networking with units such as meters and other inverters. Fixed-point DSCs in solar-power inverters have been demonstrated to achieve significant cost reductions while cutting power-efficiency losses in half; newly available floating-point controllers will push these results even further.

An example of a DSC with floating-point capabilities is Texas Instruments' (TI's) TMS320F2833x, a 32-bit device that operates at frequencies up to 150 MHz and provides up to 300 mega floating-point operations per second (MFLOPS). Integrated features include on-chip direct memory access (DMA), fast interrupt handling, a 32-bit enhanced memory interface (EMIF), fast 12-bit ADCs supporting up to 16 input channels, multiple timers, standard communication ports, and 12 individually controlled enhanced PWM (ePWM) channels, each with its own timer and phase register (Fig. 5).

The F2833x differs from its predecessors in the F28x generation because it is based on a floating-point architecture. With a single-sign bit, 8-bit exponent and 23-bit mantissa, the device handles a normalized value range of ±~1.7 × 10-38 to ±~3.4 × 1038. For control tasks in particular, the extensive range is valuable in that it deals with scaling and saturation more efficiently than fixed-point does.

TI's testing of F2833x floating-point assembly versus F28x fixed-point assembly indicates an average improvement of nearly two and a half times in cycle efficiency for basic operations such as division, square roots and trigonometric functions, all of which scale and saturate routinely. Some signal-processing algorithms that are commonly used in control systems, including infinite impulse response (IIR) and fast Fourier transform (FFT), show nearly as much improvement.

The controller's C compiler has been designed to take advantage of the floating-point architecture with pipelining that largely eliminates wait states. The DMA controller keeps the pipeline fed while handling multiple routine transfers without interrupting the DSP core. Some examples of algorithms that are frequently used in inverter control, and in control systems in general, are listed in the table, showing the improvement in cycle counts of floating-point C code over fixed-point C.

The F2833x compiler allows the same code to be recompiled as fixed-point code with a simple switch. Programmers can enjoy the faster development that comes with floating point, and then let the compiler make the adjustments necessary to change the code to fixed point.

Want to use this article? Click here for options!
© 2007 Penton Media, Inc.