MIT OpenCourseWare http://ocw.mit.edu 2.161 Signal Processing: Continuous and Discrete Fall 2008 For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.MASSACHUSETTS INSTITUTE OF TECHNOLOGY DEPARTMENT OF MECHANICAL ENGINEERING 2.161 Signal Processing - Continuous and Discrete Fall Term, 2008 Quiz #1 — October 16, 2008 Notes: • The quiz is closed-book, two sheets of notes are allowed. • The time allowed is ninety minutes. • There are 4 problems, answer them all. • Partial credit will be given. • Some possibly useful information is given at the back of this quiz. 1Problem 1: (20 points) A low-pass filter is to be designed from the power frequency response function 12 |H(jΩ)| = 1 + Ω4 (a) Determine a stable system transfer function based on this specification. (b) Make a pole-zero plot for the system and determine the filter class (low-pass, etc). (c) What type of design is this (Butterworth, Chebyshev Type I, etc). (d) Transform your filter design to a new filter H1(s), using the substitution 2s s2 + 4 −→ s in H(s). (e) How many p oles and zeros does H1(s) have, and what class of filter (low-pass etc) is it?. Problem 2: (20 points) Your advisor (who should remain nameless!) went out and bought you a new lab computer complete with A/D converters that will sample up to 10,000 samples/second. Without consulting you he also went and bought a pressure sensor that produces a sinusoidal output with a frequency that is proportional to pressure, in the range 0 to 7 kHz. He expects you to use the computer with the sensor. Your heart skips a beat. “Oh, no – it cannot be!”, you cry in despair, “The output of this sensor exceeds the maximum Nyquist rate allowed by my computer!” But, in fact, is everything lost? Assume that you have software that will allow you to take a record of sinusoidal data with a sampling interval ΔT , and then report back its frequency. Determine and describe an experimental procedure that will allow you to determine the frequency of the sensor output in the range 0 – 7 kHz, using a maximum sampling rate of 10 kHz. Hints: 1) Remember that you are not asked to reconstruct an arbitrary unknown waveform, therefore you know the effect of aliasing. 2) You may want to consider taking more than one data record with different sampling rates. Problem 3: (35 points) We are going to design a simple speech scrambler, that will encode an audio signal for “secure” transmission (that is make it unintelligible) over the internet, yet allow it to be decoded at the receiving end by a similar process. Let’s start by investigating a Hilbert transformer, which is a linear system with a frequency response function � −j, Ω > 0 H(jΩ) = j, Ω < 0, 2y ( t ) a u d s i g u d i o f ( 9 A 1 (a) Determine the impulse response h(t) of the Hilbert transformer. H i l b e r t t r a n s f o r m e r d ( t ) h ( t )H ( j W ) Is this a causal filter? Hint: F{sgn(t)} =2/(jΩ) where sgn() is the signum function. (b) If the input to the Hilbert transformer is a sinusoid A sin(Ω1t), what is the output y(t)? H ( j 9 ) What is the general effect on sinusoidal signals? (c) Now consider the block diagram of the complete speech scrambler below: y ( t )A s i n ( 9 t ) 1 H i l b e r t t r a n s f o r m e r X X S s i n ( 9 t )c - + H i l b e r t t r a n s f o r m e r H i l b e r t t r a n s f o r m e r i o n a l O s c i l l a t o r s a S p e e c h s c r a m b l e r t ) m u l t i p l i e r c r a m b l e d Internally an oscillator generates a constant frequency sinusoid sin(Ωct). Assume for now that the audio signal is a sinusoidal waveform A sin(Ω1t). Determine the output of the system y(t). (d) If the audio signal has a (one-sided) spectrum shown below | F ( j 9 ) | 0 9 3 o r i g i n a l s p e e c hf ( t ) t 1 sketch the magnitude spectrum |Y (jΩ)| of the output signal y(t). What will be the amplitude and frequency of the component shown with frequency Ω1? Explain how the scrambling operation works. (e) For extra credit, suggest what the “descrambler” needs to do to restore the original signal. Problem 4: (25 points) A linear interpolator is a hardware, non-ideal, data reconstructor in which successive samples are connected by straight line segments (“connect-the-dots”). If the sample set is {f(nΔT )}, the reconstructed continuous waveform fˆ(t) can be written fˆ(t)= f ((n − 1) ΔT )+ f(nΔT ) − f ((n − 1) ΔT )(t − nΔT), nΔT ≤ t< (n + 1)ΔT ΔT t f ( t )^ 0 D T 2 D T 3 D T 4 D T 5 D T 6 D T 7 D T Notice that when t = nΔT , the interpolator output fˆ(t)= f((n − 1)ΔT), that is it is one step ΔT behind. If we treat the linear interpolator as a filter, (a) Consider the response of the interpolator to the digital pulse train below: - 3 D T 2 D TD T 3 D T- 2 D T - D T 0 Show that the linear interpolator (with a single-step delay as above ) has an pulse response ⎧ ⎨ t/ΔT, 0 ≤ t< ΔT hˆ(t)= 2 − t/ΔT, ΔT ≤ t< 2ΔT, ⎩ 0, otherwise. Sketch the pulse response (to help you in the next part). 41 0 T t  1 0 T 2 T t 1   (b) Find the frequency response function Hˆ(jΩ) of the linear interpolator. You can use any method but, consider the two following possible methods: Method 1: The waveform x(t) to the left x ( t ) t/T, 0 ≤ t<T x(t)= 0, otherwise. has a Fourier transform e−jΩT (jΩT +1) − 1 X(jΩ) = Ω2T (i) Use X(jΩ) to find the Fourier transform of the waveform x1(t) below x ( t ) (ii) Combine these results to find Hˆ(jΩ). Method 2: • Consider h(t) as the convolution of two rect() functions... (c) Make a sketch of  Hˆ(jΩ), and compare it to the frequency response of the ideal (cardi-nal) reconstructor. 5Some useful/useless information: θ sin θ cos θ tan θ 0 π/6 π/4 π/3 π/2 0 1 0 1/2 √3/2 √3/3 √2/2 √2/2 1 √3/2 1/2 √3 1 0 ∞ x loge(x) log10(x) 2 0.6931 0.3010 3 1.0986 0.4771 5 1.6094 0.6990 10 2.3026 1.0000 sin(a + b) …