Dartmouth College

Thayer School of Engineering

Signal Processing

Problem Set #6

This assignment is worth 50 points. Please start each new problem on a new page and clearly
label the problem you are working on at the top of each page. Show all your work when
solving problems and answer any conceptual questions clearly and concisely, only addressing
the specific question. You may discuss these problems with others, but copying work from
others is prohibited. You may not use any solutions that address the specific question asked,
including from previous offerings of this course. List the names of any collaborators and
acknowledge all sources and tools you used outside the regular course readings (such as
other books, websites, computing tools, artificial intelligence, etc.).

Page 1 of 6
Signal Processing Problem Set #6

1. Discrete-Time Processing of Continuous-Time Signal with linear-phase FIR filter. In

the system shown below, the discrete-time system is a linear-phase FIR lowpass filter
designed by the Parks-McLellan algorithm with passband tolerance δ1 = 0.01, stopband
tolerance δ2 = 0.001, end of passband at ωc = 0.4π, and start of stopband at ωs = 0.6π.
The length of the resulting impulse response is 28 samples. The sampling rate for the
ideal C/D and D/C converters is 1/T = 10, 000 samples/sec.

(a) What is the constraint on the input signal so that the overall system behaves like
an LTI system with effective frequency response Hef f (jΩ):

Yc (jΩ) = Hef f (jΩ)Xc (jΩ)

(b) If the input signal satisfies the constraints such that this overall system behaves like
an LTI system, provide the passband and stopband specifications for the magnitude
of Hef f (jΩ).
(c) What is the overall delay from the continuous-time input to the continuous-time
output in seconds?

Page 2 of 6
Signal Processing Problem Set #6

2. Linear Phase FIR Filters. The linear phase constraint on FIR filters places constraints
on the impulse response and the location of zeros of the system function:

• Type I: no constraints on locations of zeros

• Type II: always has a zero at ω = π
• Type III: always has a zero at ω = 0 and ω = π
• Type IV: always has a zero at ω = 0

For each type of FIR filter above, list which of the following filters could be successfully
approximated: lowpass, highpass, bandpass, bandstop.

Page 3 of 6
Signal Processing Problem Set #6

3. FIR Filter Design Using Windows. Use the window design method to find the impulse
response h[n] of a linear phase FIR filter of order M = 24 to approximate the following
ideal frequency response magnitude:
(
1, |ω| ≤ 0.2π
|Hd (ejω )| =
0, 0.2π < |ω| ≤ π

(a) Start with specifying the ideal impulse response for this ideal filter and take into
account the appropriate delay α = M/2.
(b) What is the impulse response of the linear-phase FIR filter if a rectangular window
is used to truncate the ideal impulse response? Use the approximate width of the
main lobe (available in table 7.2) to estimate an upper bound on the width of the
resulting transition band of the FIR filter. The same table says that the stopband
attenuation will only be 21 dB.
(c) What is the impulse response of the linear-phase FIR filter if a Hamming window
is used to truncate the ideal impulse response? Formulas for common windows are
in section 5.7.1 of the Oppenheim and Schafer textbook. What is the resulting
stopband attenuation and the width of the transition band and how does this
compare to the results from the rectangular window?

Page 4 of 6
Signal Processing Problem Set #6

4. Optimum Equiripple / Parks-McClellan FIR Filter. The figures below depict four
frequency-response magnitude plots of linear-phase FIR filters, labeled |Aie (ejω )|, i =
1, 2, 3, 4. One or more of these plots may belong to equiripple linear-phase FIR filters
designed by the Parks-McClellan algorithm. The maximum approximation errors in the
passband and the stopband, as well as the desired cutoff frequencies of those bands,
are also shown in the plots. Please note that the approximation error and filter length
specifications may have been chosen differently to ensure the cutoff frequencies are the
same in each design.

(a) What type(s) (I, II, III, IV) of linear-phase filters can |Aie (ejω )| correspond to, for
i = 1, 2, 3, 4? Please note that there may be more than one linear-phase FIR filter
type corresponding to each |Aie (ejω )|, so if that is the case, list all possible choices.
(b) How many alternations does each |Aie (ejω )| exhibit, for i = 1, 2, 3, 4?
(c) For each i, can |Aie (ejω )| belong to an output of the Parks-McClellan algorithm?
(d) If you claimed that a given |Aie (ejω )| could correspond to an output of the Parks-
McClellan algorithm and that it could be Type I, what is the length of the impulse
response?

Page 5 of 6
Signal Processing Problem Set #6

5. Deriving Highpass FIlter from Optimal Lowpass FIR Filter. Consider a type I linear-
phase FIR lowpass filter with impulse response hLP [n] of length M + 1 and frequency
response
HLP (ejω ) = Ae (ejω )e−jωM/2 (3)
The system has the amplitude function Ae (ejω ) shown in the figure below.

This amplitude function is the optimal (in the Parks-McClellan sense) approximation to
unity in the band 0 ≤ ω ≤ ωp . where ωp = 0.27π, and the optimal approximation to
zero in the band ωs ≤ ω ≤ π where ωs = 0.4π.

(a) What is the value of M (the order of the filter)?

(b) Suppose now that highpass filter is derived from this lowpass filter by defining

hHP [n] = (−1)n+1 hLP [n] = −ejπn hLP [n]

Show that the resulting frequency response is of the form

HHP (ejω ) = Be (ejω )e−jωM/2

where Be (ejω ) is a shift of Ae (ejω ). You can use the frequency-shifting property of
the DTFT.
(c) Sketch Be (ejω ) for 0 ≤ ω ≤ π
(d) It is asserted that for the given value of M (as found in part a), the resulting
highpass filter is the optimum approximation to zero in the band 0 ≤ ω ≤ 0.6π and
to unity in the band 0.73π ≤ ω ≤ π. Is this assertion correct? Justify your answer.

Page 6 of 6