New 'C2000 DSPs provide best price/performance ratio in industry Revolutionary 'C27x architecture combines the best of DSP and MCU technologies Celebrating 10 years of floating-point DSP excellence (1988-1998) New TMS320 applications make DSP development easier TI's C54x family extends reach from low cost to high performance Experience power-efficient performance with 100-MIPS 'C549 ICASSP show salutes 50th anniversary of IEEE Registration opes for DSPS Fest '98 Quad codec-filter combo brings multi-channel capabilities to CO line cards New Third-Party Hardware & Software TI leads the digital age of electronics with strategic acquisitions and alliances |
New TMS320 applications make DSP development easierTwo new applications for the TMS320 floating-point family of processors are available today. The Logarithmic Differential Compression and Sliding Fast Fourier Transform evolve two new DSP concepts as tools for enabling tomorrow’s power-intensive innovations. Logarithmic Differential Compression (LDC)A simple method of compressing audio signals is now available for the TMS320 floating-point processors which exhibits very little processor overhead for impressive audio quality. LDC dramatically enhances the quality per bit of audio, video, and other natural signals by differentially encoding signals using a short floating-point format. The bit trade-off for precision and quality exhibits a surprisingly high quality for the number of bits used. This effect occurs since floating-point representations are logarithmic by nature, which happens to match closely to how humans perceive sight and sound. Playback of high-quality audio is possible using only 8 bits, but additional bits will greatly improve the quality. In this case, the US $9.95 ’C32, which is capable of external 16-bit short float formats, becomes very cost and performance effective. In addition to being a simple compressor and decompressor, LDC is also a linear process. This allows an LDC data stream to be filtered, processed, or further compressed by more traditional methods. A patent application has been filed covering LDC as a new technology. Sliding Fast Fourier Transform (SFFT)The SFFT is a continuous time Fourier Transform which differentiates itself from block transform operations like the Discrete and Fast Fourier Transform (DFT) and (FFT). The SFFT requires only the frequency bins of interest to be calculated, making it particularly useful for narrow-band filtering, analysis, modulation, and demodulation, or when phase information is important. SFFT applications include:
DSP beginners and veterans will also find that SFFT derivation leads to the DFT, without using a rigorous DSP math approach. This may help in understanding where a Fourier Trans-form can be applied in an application. The SFFT computes the frequency response of a signal, which is the summation of a series of impulses, from the response of individual impulses. Since the response of each impulse within a signal is simple, with each impulse having a flat response across all frequencies, it is also a simple process to sum the responses in the frequency domain. Long standing problems concerning the stability of each frequency bin, non-rectangular windowing functions, and signal reconstruction have also been solved. The forward and reverse transform including a non-rectangular window, performed as a convolution in frequency, takes only 6 cycles per frequency bin of interest to calculate. Also, the SFFT can easily be distributed among multiple processors without requiring any external memory. |