Quantization Noise Level Research Articles

On detection of quantization noise in low level reproduction of digital audio system

AbstractOn recent audio equipment, a wide dynamic range of nearly 100 dB is achieved by using 16 bit linear quantization. But quantization noise is still detectable, because relatively few bits are used to represent low‐level signals. When the signal is quantized directly, the quantization noise consists of harmonics of the source frequency. This quantization noise is detectable at 55 dB for a pure tone and at just about 40 dB for complex and modulated tones, if we assume that the maximum reproduction level is 120 dB. This harmonic distortion can usually be reduced by adding dither at quantization. When dither is added, the quantization noise becomes white noise and spreads over the entire frequency range. Experimental results indicate that at frequencies lower than 2 kHz, the detection level of quantization noise is raised by several dB when dither is introduced. But at about 4 kHz, the detection level falls because of the dither; in this case, the white noise induced by the dither seems to function as a detection cue.

Chirp subbottom profiler for quantitative sediment analysis

A wide‐band, frequency‐modulated, subbottom profiling system (the chirp sonar) can remotely determine the acoustic attenuation of ocean sediments and produce artifact‐free sediment profiles in real time. The chirp sonar is controlled by a minicomputer which performs analog‐to‐digital and digital‐to‐analog conversion, correlation processing, and attenuation estimation in real time. The minicomputer generates an FM pulse that is phase‐ and amplitude‐compensated to correct for the sonar system response. Such precise waveform control helps suppress correlation noise and source ringing. The chirp sonar, which has an effective bandwidth of 5 kHz, can generate chirp (Klauder) wavelets with a tuning thickness (Rayleigh’s criterion for resolution) of approximately 0.1 ms. After each return is correlated, a computationally fast algorithm estimates the attenuation of subbottom reflections by waveform matching with a theoretically attenuated waveform. This algorithm obtains an attenuation estimate by minimizing the mean‐square error between the autocorrelation function of the theoretically attenuated wavelet and the autocorrelation function of the subbottom reflection. The chirp sonar was tested in Narragansett Bay, R.I. along a line that had been previously cored. Experimental results show that correlation noise from the seafloor reflection was below −60 dB, the quantization noise level, thereby allowing detection of small subbottom impedance contrasts and accurate estimation of attenuation. Attenuation coefficient estimates from this sandy region agree with in‐situ measurements made by other investigators.

Simultaneous and nonsimultaneous masking within natural speech

Auditory masking functions are well known only for steady‐state tones and noises, not for complex sounds such as speech. To investigate the contribution of simultaneous and nonsimultaneous masking in natural speech, thresholds were obtained for 15‐ms probe tones placed in the closure portion of naturally spoken VCV utterances. The utterances contained a long closure (stop/d/) or a short closure (flap/P/). Two additional conditions were tested: flap closure was replaced by a short portion of stop closure, and the surrounding vowels were digitally attenuated. Critical‐band filtering and spread of masking characterize the simultaneous‐masking results. Some conditions reveal significant amounts of nonsimultaneous masking. These effects are less straightforward than the effects of simultaneous masking, primarily due to the time‐varying nature of the natural speech maskers. Results of this line of research may help refine perceptually weighted filters used in speech coders to reduce the perceived level of quantization noise [cf. M. R. Schroeder, B. S. Atal, and J. L. Hall, J. Acoust. Soc. Am. 66, 1647–1652 (1979)].

Enhancement of ADPCM Speech by Adaptive Postfiltering

Adaptive Differential Pulse Code Modulation (ADPCM) systems can provide high-quality digitizations of telephone-bandwidth speech at a bit rate of 32 kb/s. At a lower bit rate such as 24 kb/s, the quality of the speech is limited by an easily perceptible level of quantization noise. This paper proposes an adaptive postfiltering procedure that can significantly enhance the quality of lower bit rate ADPCM. The coefficients of the postfilter are easily derivable from the predictor coefficients in the ADPCM decoder. In a subjective test involving 18 listeners and two sentence-length test inputs, the enhanced 24-kb/s speech with an optimized postfilter design ranks very close to conventional 32-kb/s speech. A suggested application of the postfiltering procedure is in packet voice or mobile radio systems where substandard bit rates such as 24 kb/s or 16 kb/s are sometimes necessary. The postfiltering algorithm has also been successfully tested in non-DPCM situations, such as in the enhancement of speech degraded by additive white Gaussian noise.

The validity of the geological interpretations of marine magnetic anomalies

Summary. In order to urlderstand the relative importance of the various phenomena which contribute to marine magnetic anomalies, an analysis was performed on magnetic profiles measured simultaneously at the sea surface and near the sea floor. The data come from three long geophysical surveys made over spreading centres in the eastern Pacific with the Deep Tow instrument package. At low spatial frequencies the two sets of observations agree very closely when related by the mathematical model of upward continuation from magnetic sources below, which are assumed to be perfectly heated, i.e. ‘two-dimensional’. However, spectral analysis reveals the presence of a discrepancy at high spatial frequencies. In all three surveys at wavelengths shorter than about 4km the observed surface field power exceeds the upward continued power by several orders of magnitude. Several causes for this discrepancy are examined, including deviations from twodimensionality , numerical problems, and the noise background of the ocean environment, and the most likely candidate is shown to be ionospheric activity. Spectra of magnetograms from the nearest land observatories intersect the upward continued spectra at about 4-km wavelength, and exhibit power at high spatial frequency which agrees with the extra power in the surface data in each survey. For the observed surface data, the regime of valid geological interpretation is limited by ionospheric noise to wavelengths longer than about 4km. The near-bottom data do not suffer from such a limitation primarily because they were obtained near the geological sources, not just because of the attenuation of the ionospheric noise by the conducting sea water above. The interpretation regime of the near-bottom data is much wider, extending to wavelengths of a few hundred metres where the observed power nears the quantization noise level and where the validity of the assumption of perfectly heated sources probably breaks down. The behaviour of the observed spectra suggests an analytical model which portrays how the bandwidth of the interpretation regime of the surface field depends on the towing speed, source depth and ionospheric activity level. It is shown that a faster towing speed reduces the effect of the ionospheric noise.