DSP combines fixed- and floating-point operation
By Robert Cravotta, Technical Editor -- EDN, April 23, 2009
Texas Instruments’ TMS320C6743 low-power DSP features both high-precision, wide-dynamic-range, floating-point operations and higher-performance, lower-power, fixed-point operations in the same device. The C674x core uses a superset of the fixed-point C64x+ and the floating-point C67x+ instruction sets. The single-precision/double-precision floating-point operations support a wider dynamic range than do fixed-point operations. The floating-point unit also allows developers to more naturally transfer their work from analysis tools to the floating-point device and get their design to market faster because they need not perform an immediate conversion to fixed-point operations. When the design must squeeze more performance from the system, developers can convert the floating-point operations that do not require the full dynamic range to fixed-point operations.
The C6743 includes 192 kbytes of on-chip memory. The device consumes 490 mW in operating mode and 60 mW in standby mode. Peripheral support includes a 10/100-Mbps Ethernet MAC (media-access controller) with a management-data I/O module; a multimedia-card/secure-digital peripheral; enhanced, high-resolution PWM (pulse-width modulation); 32-bit enhanced quadrature-pulse peripherals; and two external memory interfaces—one for slower, asynchronous external memories and the other for higher-speed SDRAM.
The C6743 is object-code-compatible with all TMS320C6000 devices; it is scalable and pin-to-pin compatible with C674x and OMAP (Open Multimedia Applications Platform)-L137 devices. The TMDX320C6743 is available for sampling now and costs $7.85 (1000). The OMAP-L1x/C674x starter kit, TMDXOSKL137BET, which is available for $395, includes a board-specific Code Composer Studio IDE (integrated development environment) and supports DSP/BIOS kernels.
