TMS320C27x Technology Overview

TI Semiconductors: The World Leader in DSP Solutions

TMS320C27x Technology Overview

Why a Mass Storage Processor? C27x Architectural Overview

For the first time in the history of DSPs a processor has been designed with features completely optimized for mass storage applications. The Texas Instruments TMS320C27x combines high speed multiply and accumulate (MAC) operations of Digital Signal Processors with the input/output (I/O) characteristics of microcontrollers. The result is a high performance processor with a world of capabilities ideal for current and future real-time embedded applications. The 'C27x is the latest addition to the TI family of fixed-point processor cores.

Contemporary data storage applications such as hard disk drives, tape drives, and digital video disks (DVD) are examples of real-time embedded systems. These storage devices are used in several end equipments with powerful central processing units. To keep up with the pace of these CPUs, the storage devices demand high performance from their local processor. The performance required from these local processors will increase as applications tend to become more sophisticated.

The 'C27x is a member of the TI line of customizable digital signal processors (cDSPs). The cDSP provides a simple and effective method for reducing a system by merging TI DSPs with TI application-specific integrated circuit (ASIC) gate-array or standard cell products. The TI ASIC design environment enables logic designers to combine custom logic with cDSP cores, other digital hardware macros, and linear hardware macros. This allows the designer to reduce part or all of a system design to a single device to improve performance and reliability while reducing system cost, hardware size, and power consumption.

Why a Mass Storage Processor?

Three distinct types of processors are employed in today’s storage devices -- MCUs with VonNeumann architectures, load/store RISC based architectures and computationally efficient Multiply/Accumulate (MAC) based architecture like DSPs.

MCUs using VonNeumann architecture have one large memory space (for both data and instruction) and a single bi-directional bus. A large register set is present for easy interrupt handling and context switching. While they may have the advantage of a simple architecture, they perform poorly in instruction cycle time due to a unified memory space and a single bus.

Conventional DSPs are based on a modified Harvard architecture with discrete buses for read and write operations for both data and instructions. This increases operational parallelism and enables pipelining of instructions and/or data during program execution. DSPs also include hardware Multiply/Accumulate (MAC) capability to handle intensive math operations.

Current microcontroller uniprocessor solutions for storage devices cater only to low cost, low performance disk drives or CDROM drives which require only 15-20 MIPS of processing power. Virtually all other storage devices like high end disk drives and DVD (Digital Video Disk) products require up to 30 MIPS. These products currently engage a dual processor solution to meet this need - an MCU for interface (I/F) control and a DSP for servo control.

But with increasing Tracks Per Inch (TPI), soaring spindle speeds and crunching cost constraints, the storage industry is desperately in need of a high performance, cost-effective uniprocessor.

Today’s semiconductor process technology will certainly allow two processors to be integrated on a single chip. But integrating two dissimilar cores will not prove to be cost effective.

Bolting a MAC unit into an MCU is a possible solution, however this approach falls short of delivering the required quality of servo/spindle MIPS.

From its ability to perform true single-cycle multiply and accumulate operations, a DSP can deliver the required quality MIPS. A DSP also includes special modes and instructions to simplify complex math operations for control systems.

The best solution for a storage uniprocessor is to combine the modified Harvard architecture of a DSP and it's inherent computational power, add the best features of a MCU to provide ease of programming and include new innovations to slash code size and time to market.

The TMS320C27x is the result.