1 / 34

Digital Signal Processing Introduction Course R.WEBER SP2 ESI module Traitement Numérique

Digital Signal Processing Introduction Course R.WEBER SP2 ESI module Traitement Numérique. A] Time and Frequency description of a digital signal B] Digital Filtering : Analysis and Design Tools. PART A Time and Frequency description of a digital signal. Temporal Description. ?.

rane
Download Presentation

Digital Signal Processing Introduction Course R.WEBER SP2 ESI module Traitement Numérique

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Digital Signal Processing Introduction Course R.WEBERSP2 ESI module Traitement Numérique A] Time and Frequency description of a digital signal B] Digital Filtering : Analysis and Design Tools

  2. PART ATime and Frequency description of a digital signal

  3. Temporal Description ? What really happens here ? x=[ 2.1 -0.5 1.3 0.5 0 1.7 -0.4 0.3 -1.9 -0.8 ]; -1 0 1 2 3 4 5 6 7 8 9 10 n (time index or relative time) Dr. R.Weber / Digital Signal Processing

  4. Theoretical spectral description !  Remarks : ! X(f+1)=X(f) • periodicity of the spectrum • f can be limited to [0, 0.5] for real signal (spectrum symmetry) • f can be limited to [- 0.5 , 0.5] or [0 , 1] for complex signal ?  The problem : Yes, with the Fourier Transform  The answer : Dr. R.Weber / Digital Signal Processing

  5. Theoretical spectrum of basic signals  impulse x = [ 1 zeros(1,1023)]; f 0 1.5 -1.5 1 -0.5 -1 0.5  Complex exponential x = A.*exp(2*pi*j*(0:1023)*fo); f fO 0 1.5 -1.5 1 -0.5 -1 0.5 x = A.*cos(2*pi*(0:1023)*fo);  sine wave f fO 0 1.5 -1.5 1 -0.5 -1 0.5 Dr. R.Weber / Digital Signal Processing

  6. Theoretical spectrum of basic signals  White Gaussian Noise x = .*randn(1, 1024); f 0 1.5 -1.5 1 -0.5 -1 0.5  General case : time and frequency are related f ? • How will appear • a pure sine wave, • a pure impulse ? n specgramdemo(x,Fs); Dr. R.Weber / Digital Signal Processing

  7. Basic spectral properties (1)  Pure delay :  Modulation or frequency shift: Dr. R.Weber / Digital Signal Processing

  8. Basic spectral properties (2)  Product and Convolution:  Parceval relation : Dr. R.Weber / Digital Signal Processing

  9. Practical spectral description (1) ! No digital implementation possible X=fft(x); FFT(x) = DFT(x) with M a power of 2  Theoretical Formulation :  Approximations : The Discrete Fourier Transform • n[0, M-1] • Only f / f=k/M Dr. R.Weber / Digital Signal Processing

  10. Practical spectral description (2) Theoretical spectrum (n]-,+ [) |X(f)| M Windowed spectrum (n]0,M]) h(n) H(f) n f 0 0.5 Practical spectrum (n]0,M], f k/M) |X(k)| n n k/M 0 0.5 1/M 2/M Possibility to reduce distortion by using different kinds of windows, h(n) Dr. R.Weber / Digital Signal Processing

  11. Windowing Hamming Rectangular Blackmann-Harris Hann WINtool sous Maltab Dr. R.Weber / Digital Signal Processing

  12. DFT properties  Inverse DFT : x=ifft(X);  Parceval relation : Power spectral density cov(x) Dr. R.Weber / Digital Signal Processing

  13. Power Spectral density(1) with ? What is the measured power spectral density of a sinus wave when using a classical DFT ?  Spectral line case :a complex exponential or s(n) is sine wave DFT so then When s=sin(2*pi*(0:1023)*20/1024); plot((abs(fft(s)).^2)/1024) Dr. R.Weber / Digital Signal Processing

  14. Power Spectral density (2) ! Problem : each slice of M samples will give a different power spectral density estimation  Solution : averaging of all the local power spectral density L samples M samples 50% overlap  Impact of the windowing psd(s,M,Fs,[h(0) h(1) … h(M-1)]) =  Random noise case : f=k/M Dr. R.Weber / Digital Signal Processing

  15. Signal Power ? What is the mean power ? psd psd f 0 f 1 0.5 - 0.5 0.5 0 ? What is the mean power of the sum of signals ? Dr. R.Weber / Digital Signal Processing

  16. Signal Processing goals • s’(n)  s(n) • Power(b’(n)) 0  Goal : y(n)=s’(n)+b’(n) x(n)=s(n)+b(n) processing Signal to noise ratio :  Example : Spectral point of view Temporal point of view s(n) x(n)=s(n)+b(n) Dr. R.Weber / Digital Signal Processing

  17. PART BLinear Digital Filtering :Analysis and Design Tools

  18. Linear Digital filtering y = filter(b,a,x); b = [ b0 b1 … bN]; a = [ 1 a1 … aD]; 1. Filter if x1(n) y1(n) Filter thenx1(n)+x2(n)  y1(n)+y2(n) Filter if x2(n) y2(n) Filter y(n-T) then x(n-T) Filter if x(n) y(n)  Formulations : temporal view x(n) Filter y(n) non recursive part recursive part  Basic properties : Linearity Temporal reproducibility Dr. R.Weber / Digital Signal Processing

  19. Impulse response The impulse response (n) h(n) =(n) =h(n) x = [ 1 zeros(1,29)]; h = filter(b,a,x);  Formulation : y(n) x(n) Filter  Consequences : Non recursive Finite impulse response (FIR) Recursive Infinite impulse response (IIR) Dr. R.Weber / Digital Signal Processing

  20. Application to stability Always stable h(n) c=0.985 c=0.95 c=0.995 Example : c=0.999 c=1 c=1.001  Formulation :  Consequences : Finite impulse response (FIR) Infinite impulse response (IIR) Risk of instability Example : x = [ 1 zeros(1,400)]; c=0.985; b= 1; a=[1 -2*c*cos(2*pi*0.01) c.^2]; h = filter(b,a,x); plot(h) Dr. R.Weber / Digital Signal Processing

  21. Convolution Filter with impulse response h(n) x(n) x(0)(n) + x(1)(n-1) + x(2)(n-2)  Remarks : + which is coherent ! x(3)(n-3) +  Formulation : = -- A recursive filter can be approximated by a non recursive one h(n) Dr. R.Weber / Digital Signal Processing

  22. The Z-transform  Application to the filtering: ZT ZT  Formulation :  useful (and simple) properties: • Given X(z)=ZT(x) and Y(z)=ZT(y) • linearity : ZT(x(n) + y(n)) = X(z) + Y(z) • convolution : Zt[(x@y)(n)] = X(z) Y(z) • ZT(x(n+1)) = z X(z) ; ZT(x(n+k)) = zk X(z) • ZT(x(n-1)) = z-1 X(z) ; ZT(x(n-k)) = z-k X(z) These make the ZT very interesting ! Dr. R.Weber / Digital Signal Processing

  23. Stability rule o zeros x poles Im(z) Instable filter Stable filter Re(z)  Formulation : o 1 x Filter stability if |zp,i|<1 o x ? x Verify this rule with the previous example 1 x o x o zplane(b,a); Dr. R.Weber / Digital Signal Processing

  24. Frequency response (1) Filter with impulse response h(n)  Formulation : x = [ 1 zeros(1,1023)]; c=0.985; b= 1; a=[1 -2*c*cos(2*pi*0.2) c.^2]; h = filter(b,a,x); H= fft(h); f=(0:1023)/1024; plot(f,10*log10(abs(H).^2)); Dr. R.Weber / Digital Signal Processing

  25. Simple filter design method h(n) happrox(n) n n L values • Draw the frequency response, H(f), you are looking for. • Compute the impulse response, h(n), by applying an inverse Fourier Transform • By truncating the length of this response, you obtain a sequence of values which are also the non-recursive coefficients of an approximate FIR version of your filter. H(f) FT-1 rectangular windowing f Checking : Ripple can not be removed even with L large Solution : other type of windowing (hamming, hanning, blackmannharris…) Dr. R.Weber / Digital Signal Processing

  26. Frequency response (2)  Another formulation : ? c=0.985; b= 1; a=[1 -2*c*cos(2*pi*0.2) c.^2]; %zero padding to increase frequency resolution b=[b zeros(1,1023)]; a=[a zeros(1,1020)]; H= fft(b)./fft(a); f=(0:1023)/1024; plot(f,10*log10(abs(H).^2)); [H,f]=freqz(b,a,1024,1); Dr. R.Weber / Digital Signal Processing

  27. Frequency response (3) zplane(b,a);  Another formulation : f=0.25 X O f=0.5 f=0 O X Dr. R.Weber / Digital Signal Processing

  28. Frequency response (4) Dr. R.Weber / Digital Signal Processing

  29. Phase response (1) x(n) Filter  pure delay : ? ? ? ?  general case : Time responses for the given frequencies grpdelay(b,a,1024,1); freqz(b,a,1024,1); Dr. R.Weber / Digital Signal Processing

  30. Phase response (2)  linear phase filter : Yes, if FIR filters with symmetric or antisymmetric coefficients No, if IIR Dr. R.Weber / Digital Signal Processing

  31. Filtered Power Spectral density ? Problem : What is the psd after filtering ? x(n) Filter y(n) |H(f)| 2 1 0 f 0.5 f0 Dr. R.Weber / Digital Signal Processing

  32. Filter Specifications(1) Passband Stopband Ripple Attenuation Transition band 2 1 3 4 5 Dr. R.Weber / Digital Signal Processing

  33. Filter Specifications (2)   Order or complexity Attenuation or • If  Ripple or  Transition Band  General rules : Ripple Attenuation Transition band then • For a given specification, IIR is always less complex than FIR But be careful to the stability (and the phase linearity) ! Dr. R.Weber / Digital Signal Processing

  34. Basic Design Methods Bandpass :  Basic FIR methods : ordre 63  64 coefficients, linear phase Least-square Equiripple order 12  26 coefficients  Basic IIR methods : order 24  50 coefficients Cauer or Elliptic Butterworth Dr. R.Weber / Digital Signal Processing

More Related