13.07.2015 Views

Passive, active, and digital filters (3ed., CRC, 2009) - tiera.ru

Passive, active, and digital filters (3ed., CRC, 2009) - tiera.ru

Passive, active, and digital filters (3ed., CRC, 2009) - tiera.ru

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

VLSI Implementation of Digital Filters 21-3Program <strong>and</strong>data RAMProgramcacheROMProgram addressProgram data busData addressData data busMultiplyArithmetic logicunit (ALU)AccumulateShiftsDMAcontrollerTimerI/OportsPeripheralbusFIGURE 21.1Texas Inst<strong>ru</strong>ments TMS320C30 architecture.recent examples of the TI family are the TMS320C67x DSP chips (TI, 2006a), <strong>and</strong> more recent AnalogDevices parts are the ADSP-2116x SHARC chips (Analog Devices, 2006).The architecture of the TI TMS320C30 is illustrated in Figure 21.1. The floating-point word size usedby this processor was 32 bits. The most prominent feature of this chip was the floating-point arithmeticunit, which contains a floating-point multiplier <strong>and</strong> adder. This unit was highly pipelined to support highthroughput, at the cost of latency; when data is input to the multiplier, for example, the results will notappear on the output from that unit until several clock cycles later. Other features included a separateinteger unit for control calculations, <strong>and</strong> significant amounts (2k words) of SRAM for data <strong>and</strong> on-chipinst<strong>ru</strong>ction memory. On-chip ROM (4k words) was also optionally provided in order to eliminate theneed for an external boot ROM in some applications. This chip also included a 64-word inst<strong>ru</strong>ction cacheto allow its use with lower speed memories. The modified Harvard architecture, that is, the separate data<strong>and</strong> inst<strong>ru</strong>ction buses, provided for concurrent inst<strong>ru</strong>ction <strong>and</strong> data word transfers within one cycle time.The TMS320C30 offered inst<strong>ru</strong>ction cycle times as low as 60 ns. A code segment which implementsportions of an finite impulse response (FIR) filter on this device wasRPTS RCMPYF3 *AR0þþ(1),*AR1þþ(1)%,R0jj ADDF3 R0,R2,R2ADDF R0,R2,R0where the MPYF3 inst<strong>ru</strong>ction performs a pipelined multiply operation in parallel with data <strong>and</strong>coefficient pointer increments. The ADDF3 inst<strong>ru</strong>ction is performed in parallel with the MPYF3inst<strong>ru</strong>ction, as denoted by the ‘‘jj’’ symbol. Because these operations are in parallel, only one inst<strong>ru</strong>ctioncycle per tap is required. An FIR filter tap was benchmarked at 60 ns on this chip. Similarly, a typicalbiquad infinite impulse response (IIR) filter code segment wasMPYF3MPYF3MPYF3*AR0,*AR1,R0*þþAR0(1),*AR1––(1)%,R1*þþAR0(1),*AR1,R0

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!