Introduction to Digital Signal Processing Paolo Prandoni LCAV - - PowerPoint PPT Presentation

Introduction to Digital Signal Processing Paolo Prandoni LCAV - EPFL Introduction to Digital Signal Processing – p. 1/2

Inside DSP . . . Digital Brings experimental data & abstract models together Makes math very simple i.e. implementable Signal Measurement of a varying quantity Experimental data (physics, electronics, astronomy, etc.) Processing Manipulation of the information content Abstract model (math, computer science, etc.) Introduction to Digital Signal Processing – p. 2/2

A Bit of History and Philosophy Egypt, 2500 BC: Introduction to Digital Signal Processing – p. 3/2

A Bit of History and Philosophy Egypt, 2500 BC: the Palermo stone. Introduction to Digital Signal Processing – p. 4/2

A Bit of History and Philosophy USA, 2005 AD: the Dow-Jones Industrial Average 10000 9000 8000 7000 6000 5000 4000 3000 2000 1000 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 Introduction to Digital Signal Processing – p. 5/2

A Bit of History and Philosophy What do these measurements have in common? Life-changing phenomena Unpredictable patterns Discrete set of observations = Digital Signal Processing Is a discrete set of measurement a sufficient representation? Can we formalize this concept? Introduction to Digital Signal Processing – p. 6/2

A Bit of History and Philosophy The Platonic schizophrenia of Western thought. Dichotomy between the ideal and the real Zeno’s paradoxes An odd synergy: calculus and ballistics Introduction to Digital Signal Processing – p. 7/2

A Bit of History and Philosophy Calculus: a lofty ideal at the service of war. b b b b b b b b b b g t 2 � x ( t ) = � v 0 t + (1 / 2) � Galileo, 1638 Introduction to Digital Signal Processing – p. 8/2

Ideal Signals vs. Real Signals How does an ideal signal look like? Tuning fork: It’s a function of a real variable! f ( t ) = A sin(2 πωt + φ ) As such, 3 parameters completely describe the signal. Introduction to Digital Signal Processing – p. 9/2

Ideal Signals vs. Real Signals Tuning forks are boring; Bach is not: Unfortunately (or fortunately): f ( t ) =? How do we deal with real-world signals? Introduction to Digital Signal Processing – p. 10/2

Ideal Signals vs. Real Signals Sampling: we measure the signal value at regular intervals b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b x [ n ] = f ( nT s ) Can we do this or are we in one of Zeno’s paradoxes? Yes, we can if the signal is “slow enough”. Introduction to Digital Signal Processing – p. 11/2

Ideal Signals vs. Real Signals The Sampling Theorem (Nyquist 1920). Under appropriate “slowness” conditions for f ( t ) we have: ∞ x [ n ]sin( π ( t − nT s ) /T s ) � f ( t ) = π ( t − nT s ) /T s n = −∞ In a way, the sampling theorem solves one of Zeno’s paradoxes: the infinite and the finite have been reconciled. The sampling theorem is the ”revolving door” into the digital world. We will therefore operate in the digital world only. Introduction to Digital Signal Processing – p. 12/2

The Digital Revolution Digital signals make our life simpler: Processing: Sequence of numbers: ideal for computations Development easy (general-purpose hardware) Storage: Storage is basically media-independent Perfect duplication Digital compression is miraculous Communications: Transmission schemes independent of data Error correction techniques make it noise-free Introduction to Digital Signal Processing – p. 13/2

The Digital Revolution: Processing Computing the average value of a signal. a b Introduction to Digital Signal Processing – p. 14/2

The Digital Revolution: Processing Computing the average value of a signal. a b � b 1 x = ¯ f ( t ) dt b − a a Introduction to Digital Signal Processing – p. 14/2

The Digital Revolution: Processing Computing the average value of a digital signal. b b b b b b b b b b b b b b b b b b b b 0 N − 1 Introduction to Digital Signal Processing – p. 15/2

The Digital Revolution: Processing Computing the average value of a digital signal. b b b b b b b b b b b b b b b b b b b b 0 N − 1 N − 1 x = 1 � ¯ x [ n ] N n =0 Introduction to Digital Signal Processing – p. 15/2

b b b b b b b b b b The Digital Revolution: Processing Computing (vertical) speed the “Platonic” way. x ( t ) t x ( t ) = v 0 t − (1 / 2) gt 2 v ( t ) = ˙ x ( t ) = v 0 − gt Introduction to Digital Signal Processing – p. 16/2

The Digital Revolution: Processing Computing speed the DSP way. x [ n ] b b b b b b b b b b n Introduction to Digital Signal Processing – p. 17/2

The Digital Revolution: Processing Computing speed the DSP way. x [ n ] b b b b ∆ x b b b b b b n ∆ T v [ n ] = ( x [ n ] − x [ n − 1]) /T s Introduction to Digital Signal Processing – p. 17/2

The Digital Revolution: Processing The ”Speed Filter”: Processing Speed Position Introduction to Digital Signal Processing – p. 18/2

The Digital Revolution: Processing Inside the ”Speed Filter”: 1 /T s x [ n ] + v [ n ] x [ n − 1] z − 1 − 1 This is a general results: filters’ building blocks are just delays, multiplications and additions. Introduction to Digital Signal Processing – p. 19/2

The Digital Revolution: Storage How do you store a signal? In the (not so) old days: Build a physical system (wax cylinders, magnetic tapes, vynil...) Fragile, data dependent Nowadays: Quantize the signal values into binary digits Store in any digital memory support Perfect copies Signal to noise ratio for digital signals: SNR ≈ 6 dB / bit Introduction to Digital Signal Processing – p. 20/2

The Digital Revolution: Storage How do you deal with large amounts of data? Compression! Signal Type Default Rate Compressed Rate 4.32 Mbps 128 Kbps Music CD audio MP3 64 Kbps 4.8 Kbps Voice AM radio CELP 20 Mb 600 Kb Image this image JPEG 170 Mbs 600-800 Kbs Video PAL video DiVx Introduction to Digital Signal Processing – p. 21/2

The Digital Revolution: Transmission The Agamemnon, 1858 Introduction to Digital Signal Processing – p. 22/2

The Digital Revolution: Transmission Digital data allows for large throughputs: Transoceanic cable: 1866: 8 words per minute ( ≈ 5 bps) 1956: AT&T, coax, 48 voice channels ( ≈ 3Mbps) 2005: Alcatel Tera10, fiber, 8.4 Tbps ( 10 12 bps) Introduction to Digital Signal Processing – p. 23/2

The Digital Revolution: Transmission Digital data allows for large throughputs: Transoceanic cable: 1866: 8 words per minute ( ≈ 5 bps) 1956: AT&T, coax, 48 voice channels ( ≈ 3Mbps) 2005: Alcatel Tera10, fiber, 8.4 Tbps ( 10 12 bps) Voiceband modems: 1950s: Bell 202, 1200 bps 1990s: V90, 56000bps Introduction to Digital Signal Processing – p. 23/2

DSP Friends and Partners Electronics Computer science Physiology Music Medicine Photography And many more... Introduction to Digital Signal Processing – p. 24/2

Conclusions Digital signal processing is FUN! It’s a fresh new take on what you already studied in theory. Just turn on a computer and you have a “mad scientist lab” where you can bring everything you know, and nothing ever blows up. Introduction to Digital Signal Processing – p. 25/2