數(shù)字信號(hào)處理教學(xué)課件：Chapter11 Supplement Analysis of Finite Wordlength Effect

1、1Supplement: Analysis of Finite Wordlength Effect Types of quantizationNumber representationsLimit cyclesAnalysis of Finite Word-length Effects There are three types of finite word-length effect in DSP.(1) A/D Conversion noise(2) Coefficient Quantization of Filter(3) The Quantization of Arithmetic O

2、perations 1.The Quantization Process and Errors Fig. A general (b+1)-bit fixed-point fraction Fig. The quantization process model is called as quantization step.x: the original data : (b+1)-bit (employed truncation or rounding) Quantization of Fixed-Point Numbers Fig illustrate quantization process.

3、 Range of quantization error.Rounding :Truncation or 2. A/D Conversion NoiseQuantization Noise Model For example 3-bit bipolar A/D conversion with twos-complement representation: full-scale range. and The analysis of error is based on assumptions:(b) is uncorrelated with (a) is a wide-sense stationa

4、ry random process, white noise, with uniformly distribution(c) is a sample sequence of a stationary random process. Fig 9.17 In the case of rounding error, the mean and variance are given by The corresponding parameters for the twos-complement truncation are as follows: Signal-to-Quantization Noise

5、Ratio where is the input signal variance representing the signal power. In the case of a bipolar (b+1)-bit A/D converter, for roundingHence So, : the rms value of input signal This expression can be used to determine the minimum word-length of an A/D converter needed to meet a specified SNRA/DNote:

6、SNRA/D increases by 6 dB for each added bit to the word-length Computed values of the SNR for various values of K are as given below:Effect of Input Scaling on SNRFor a given wordlength, the actual SNR depends on x , the rms value of the input signal amplitude and the full-scale range RFS of the A/D

7、 converterExample - Determine the SNR in the digital equivalent of an analog sample xn with a zero-mean Gaussian distribution using a (b+1)-bit A/D converter having RFS = Kx Coefficient Quantization of FilterNotch filter designNo coefficient quantization errorNo pole Coefficient Quantization of Filt

8、erNotch filter designNo coefficient quantization errorWith poleCoefficient Quantization of FilterNotch filter designWith coefficient quantization error, qfilt( )With poleLimit Cycles in IIR Digital FiltersThis type of instability usually results in an oscillatory periodic output called a limit cycle

9、The system remains in this condition until an input of sufficiently large amplitude is applied to move the system into a more conventional operationLimit cycles occur in IIR filters due to the presence of feedbackSuch oscillations are absent in FIR filters as they do not have any feedback pathLimit

10、Cycles in IIR Digital FiltersAssume the quantization operation to be rounding and the filter to be implemented with a signed 6-bit fractional arithmetic The nonlinear difference equation characterizing the filter is given by+Qaxn Consider the first-order IIR filter as shown rightLimit Cycles in IIR Digital Filters The limit cycle generated has a period of 1 For xn = 0.03dn, , and a = 0.9, the output of the filte