Texas Instruments said a new series of AEC Q-100-qualified floating-point digital signal controllers (DSCs) introduced June 4 are among the first such devices on the market. Samples of the three devices in the TMS320F2833x series are expected to be available in September.
Andrew Soukup, worldwide marketing manager for TMS320C2000 DSCs, said the new devices, which are priced from $13.30 in 1,000-piece quantities, offer an average performance increase of 50% over current-generation DSCs from TI and others while operating at a 150 MHz clock rate. He added that some algorithms, such as a Fast Fourier Transform (FFT), will see a 200% improvement over an equivalent 32-bit fixed-point implementation due to tight integration of the F2833x floating-point unit with the C28x central processing unit (CPU).
In addition to a performance gain, Soukup said the devices will simplify and speed-up application development. He noted that software developers typically begin creating algorithms in a floating-point environment for validation and then must convert the code to run on fixed-point devices. He estimated that contending with scaling, saturation and adjustment of numerical resolution, as required in fixed-point implementations, can take weeks or months, and he added that developers can now eliminate that effort and bring products to market faster.
Automotive applications for the new DSCs include lane-departure warning, blind-spot detection and adaptive cruise control. Floating-point precision enables more sophisticated radar detection algorithms, thus enhancing object-detection capabilities. The devices will also permit longer-range detection for adaptive cruise control applications.
The TMS320F2833x series includes the 150 MHz F28335 ($19.95 in 1,000-piece quantities), 150 MHz F28334 ($17.65), and 100 MHz F28332 ($13.30). At 150 MHz, the DSCs are capable of 300 MFLOPS (million floating-point instructions per-second) performance.
The devices offer 512 KB, 256 KB and 128 KB of flash memory respectively. The F28335 and F28334 offer 68 KB RAM while the F28332 has 52 KB. The F28335 includes 18 pulse width modulation (PWM) and six high-resolution (150 picosecond) PWM (HRPWM) channels, the F28334 has 16 PWM and six HRPWM channels and the F28332 includes 16 PWM and four HRPWM channels. The F28335 features six event capture (CAP) interfaces and two quadrature encoder pulse (QEP) interfaces while the F28334 and F28332 each include four CAP and two QEP interfaces. The F28335 and F28334 offer three serial communications interfaces (SCI), while the F28332 has two.
Soukup said a six-channel direct memory access (DMA) controller increases overall system bandwidth by offloading from the CPU the need to service the on-chip 12-bit, 16-channel, 12.5 megasamples per-second analog-to-digital converter – a device he said that is faster than on-chip A/D converters in earlier DSCs. TI’s F2833x DSCs also feature a user-configurable 16- or 32-bit external memory interface. Communication interfaces include controller area network (CAN), inter-integrated circuit (I2C), universal asynchronous receiver transmitter (UART), serial peripheral interface (SPI) and TI’s multichannel buffered serial ports (McBSP).
TI’s F2833x controllers are software compatible with previous TMS320C28x controllers. Soukup said developers can program the new DSCs using any available F28x-based eZdsp development tool and IQ Math, TI’s floating point software library. The controllers will also be supported by TI’s C2000 digital motor control and digital power supply software libraries, which can be downloaded at www.ti.com/c2000appsw, and by the code composer studio integrated development environment.