Você está na página 1de 11

Handout 6

ELEN133/233E/COEN201E: Spring ‘17


4/15/2017

ELEN133/233E/COEN201E: HW #1 (87 points) SOLUTIONS

Problems (graded for correctness):

1) (12 points) MATLAB:


a. Download and run the MATLAB program hw2.m. Understand the steps.
b. Once you are comfortable moving forward with MATLAB - use MATLAB to create a
short script that creates a sequence and plots it.
c. Plot 𝒙𝒙[𝒏𝒏] vs. n where 𝒙𝒙[𝒏𝒏] = 𝐜𝐜𝐜𝐜𝐜𝐜(𝟐𝟐𝟐𝟐(𝟓𝟓)𝑻𝑻𝑻𝑻) and where −𝟐𝟐𝟐𝟐 ≤ 𝒏𝒏 ≤ 𝟐𝟐𝟐𝟐 and T=0.01.
d. After plotting this with the “plot” command, type “hold on”, then run the stem function
with the third parameter as ‘green’ to highlight what these samples look like. Make sure
you plot these vs. n – meaning, the x-axis should range from -25 to 25. The plot
command connects the dots, while the stem command shows you where the samples
occur.
e. In a new figure (type “figure()” to create a new figure), repeat steps c and d with T =
0.02. Has the frequency of your sampled signal changed? Describe why this looks
different and what “n” represents in each case.

SOLUTION:
a. Honor system - I hope that MatLab is more familiar after reviewing this sample code
b, c, and d. Solution appears in the .m file and the figures below.

Code:
%% Section 1 of HW2: Problem 1
n=-25:1:25;
T=0.01
X=cos(2*pi*5*T*n);
figure(1)
plot(n,X);
hold on;
stem(n,X,'green');

%% Section 2 of HW2: Problem 1


n=-25:1:25;
figure(2);
T=0.02
X=cos(2*pi*5*T*n);
plot(n,X);
hold on;
stem(n,X,'green');

Figure 1 (parts c and d):


Figure 2 (part e):

Solution to question e:
The fundamental period and frequency of the underlying continuous time function has not
changed. We are sampling at a different frequency, which means we are simply grabbing
samples at a different rate. The result of this means different samples… and, if we plot those
samples vs. n, it will appear that our signal has a different frequency or period. In the first
case, an increment of n represents a time change of 0.01 seconds. In the second case, an
increment in n represents a time change of 0.02 seconds. So, for example, n=10 from the
first sequence should map n=5 from the second sequence.

It should be noted that the discrete time frequency (that we have not talked about yet) IS
different. In the first case, the discrete-time period is 20 samples – thus, the discrete
frequency is 1/20. In the second example, the discrete-time period is only 10 samples – thus,
the discrete frequency is 1/10. Since the question could be interpreted in different ways, both
answers will be accepted.

2) (8 points) Discrete-time processing schematics: Draw the discrete time processing schematics
(block diagrams) for the following difference equations.
a. 𝑦𝑦1 [𝑛𝑛] = 𝑥𝑥[𝑛𝑛] + 2𝑥𝑥[𝑛𝑛 − 1] + 𝑥𝑥[𝑛𝑛 − 2]

SOLUTION:

One possible solution is here:

b. 𝑦𝑦1 [𝑛𝑛] = 𝑥𝑥[𝑛𝑛] − 𝑥𝑥[𝑛𝑛 − 3]

SOLUTION:

Here are two possible solutions:

OR

3) (6 points) Output of a discrete time system


Compute the first eight outputs for y[n] for the 8-sample input sequence x[n]={6,1,-4,1,6,1,-4,1} for
𝒏𝒏 = 𝟎𝟎: 𝟕𝟕.

SOLUTION:

First, let’s determine the expression for the output sequence.

𝒚𝒚[𝒏𝒏] = 𝟐𝟐𝟐𝟐[𝒏𝒏] + 𝟐𝟐𝟐𝟐[𝒏𝒏 − 𝟐𝟐] This is developed by looking at the two branches. The top-most
branch multiplies 𝒙𝒙[𝒏𝒏] by 2 – this yields the first term of 𝒚𝒚[𝒏𝒏]. The downward branch delays
𝒙𝒙[𝒏𝒏] by two units and then multiplies it by 2, this yields 𝟐𝟐𝟐𝟐[𝒏𝒏 − 𝟐𝟐]. The summer adds these two
results.

Our input is {6,1,-4,1,6,1,-4,1}

Similar to above, the first two terms are going to be created from twice the first two terms of
x[n], respectively. So, y[0]=12 and y[1]=2.

𝒚𝒚[𝟐𝟐] = 𝟐𝟐𝒙𝒙[𝟐𝟐] + 𝟐𝟐𝟐𝟐[𝟎𝟎] = 𝟐𝟐 ∗ −𝟒𝟒 + 𝟐𝟐 ∗ 𝟔𝟔 = −𝟖𝟖 + 𝟏𝟏𝟏𝟏 = 𝟒𝟒


𝒚𝒚[𝟑𝟑] = 𝟐𝟐𝟐𝟐[𝟑𝟑] + 𝟐𝟐𝟐𝟐[𝟏𝟏] = 𝟐𝟐 ∗ 𝟏𝟏 + 𝟐𝟐 ∗ 𝟏𝟏 = 𝟐𝟐 + 𝟐𝟐 = 𝟒𝟒

The pattern continues to create the rest of the terms of y[n].

FINAL ANSWER: 𝒚𝒚[𝒏𝒏] = {𝟏𝟏𝟏𝟏, 𝟐𝟐, 𝟒𝟒, 𝟒𝟒, 𝟒𝟒, 𝟒𝟒, 𝟒𝟒, 𝟒𝟒} for 𝒏𝒏 = 𝟎𝟎: 𝟕𝟕

4) (6 points) Convolution: Assume that h[n] ranges from –M to N and w[n] ranges from –L to –R.
M, N, L and R are all positive integers with L>R. What is the length and range of ℎ[𝑛𝑛] ⊗ 𝑤𝑤[𝑛𝑛]?

SOLUTION:
Here we are asked to find the length and the range of the index n for a function (we will call it
h[n]) that ranges from –M to N and we’ll say w[n] that ranges from –L to –R. M, N, L and R are
all positive integers with L>R.

First, we need to know the lengths:


𝒉𝒉[𝒏𝒏] 𝒉𝒉𝒉𝒉𝒉𝒉 𝒂𝒂 𝒍𝒍𝒍𝒍𝒍𝒍𝒍𝒍𝒍𝒍𝒍𝒍 𝒐𝒐𝒐𝒐 𝑵𝑵 − (−𝑴𝑴) + 𝟏𝟏 = 𝑵𝑵 + 𝑴𝑴 + 𝟏𝟏
𝒘𝒘[𝒏𝒏]𝒉𝒉𝒉𝒉𝒉𝒉 𝒂𝒂 𝒍𝒍𝒍𝒍𝒍𝒍𝒍𝒍𝒍𝒍𝒍𝒍 𝒐𝒐𝒐𝒐 − 𝑹𝑹 − (−𝑳𝑳) + 𝟏𝟏 = 𝑳𝑳 − 𝑹𝑹 + 𝟏𝟏

So, the length of the convolution, as we’ve already learned, will be the sum of these lengths
minus 1. Thus, the length is 𝑵𝑵 + 𝑴𝑴 + 𝟏𝟏 + 𝑳𝑳 − 𝑹𝑹 + 𝟏𝟏 − 𝟏𝟏 = 𝑵𝑵 + 𝑴𝑴 + 𝑳𝑳 − 𝑹𝑹 + 𝟏𝟏

The range can be found by conceptualizing the left-most and right-most points of overlap.

Flipping w[n] places a signal that ranges from positive R to positive L.

Sliding w[-k] will reveal the overlap. Note that the initial flip is the n=0 term of the convolution.
So, sliding left, we would continue sliding until the term that flipped to +L has reached the
furthest left term of h[n] – which is at –M. So, this would take L+M shifts to the left – thus, the
furthest left term of the convolution is at –(L+M).

If R<N, it will take (N-R) units of right shift of w[-k] until the term that flipped to positive R
reaches N. If R>N, the sliding actually goes to the left until R reaches N and the right-most edge
of the convolution sum will be negative. This is still captured by (N-R). So, regardless of the
magnitudes, the right-most edge of the convolution will be at N-R.

So, the edges are (-(L+M), (N-R)) and allows us to confirm the length answer above. The length
is right-edge, minus left edge plus 1. So, 𝑵𝑵 − 𝑹𝑹 – (−(𝑳𝑳 + 𝑴𝑴)) + 𝟏𝟏 = 𝑵𝑵 + 𝑳𝑳 + 𝑴𝑴 − 𝑹𝑹 + 𝟏𝟏.
Same answer confirmed.

NEAT TRICK: It can actually be shown that the indices of the non-zero values go from the sum
of the left most points of the two signals (-L + -M) to the sum of the right most points (N + -R).
5) (6 points) Convolution – part 2: Find the convolution of {2, 4, 6} with {−0.5, 1, −0.5}.
Assume both sequences are defined for n=0:2.

SOLUTION: There are many short-cut methods that can be found to perform the convolution. I use a
methodical flip, then shift, multiply and add. It is not efficient, but it is clear.

Flipping 0.5 1 0.5 about the n=0 access results in:

n -2 -1 0 1 2
1st sequence, x[n] (0) (0) 2 4 6
2nd sequence, h[n] (0) (0) -0.5 1 -0.5

For the first step of convolution, we move to the k axis.

k -2 -1 0 1 2
1st sequence, x[k] (0) (0) 2 4 6
2nd sequence, h[k] (0) (0) -0.5 1 -0.5

Next, we flip h[k] around the k=0 axis which yields:

k -2 -1 0 1 2
1st sequence, x[k] (0) (0) 2 4 6
2nd sequence, h[-k] -0.5 1 -0.5 (0) (0)

Now we shift, multiply ones that overlap, and add.


For n=0 there is no shift and the only overlap is (2)(-0.5) which equals -1.

We also note that for all left shifts of h[-k] (that is, h[n-k] for all negative n) the result is 0 because no
terms overlap.

So, we continue to determine the rest of the convolution values by shifting right and multiplying and
adding.

For n=1, we have:


k -2 -1 0 1 2
1st sequence, x[k] (0) (0) 2 4 6
2nd sequence, h[1-k] (0) -0.5 1 -0.5 (0)

This yields (2)(1) + (4)(−0.5) = 2 − 2 = 0. Again, this is y[1].

For n=2, we have:


k -2 -1 0 1 2
1st sequence, x[k] (0) (0) 2 4 6
2nd sequence, h[2-k] (0) (0) -0.5 1 -0.5

The above sequence results in three products at are summed: (2)(−0.5) + (4)(1) + (6)(−0.5) = −1 +
4 − 3 = 0. This is the value at 𝑛𝑛 = 2, i.e. shifting to the right by 2. So, 𝑦𝑦[2] = 0.

For n=3, the pattern continues and we have:


k -2 -1 0 1 2
1st sequence, x[k] (0) (0) 2 4 6
2nd sequence, h[3-k] (0) (0) (0) -0.5 1
So, 𝑦𝑦[3] = (4)(−0.5) + (6)(1) = −2 + 6 = 4.

Finally, for n=4, we have:


K -2 -1 0 1 2
1st sequence, x[k] (0) (0) 2 4 6
2nd sequence, h[4-k] (0) (0) (0) (0) -0.5

𝑦𝑦[4] = (6)(−0.5) = −3.

Thus, the final convolution is: 𝑦𝑦[𝑛𝑛] = {−1,0,0,4, −3} 𝑓𝑓𝑓𝑓𝑓𝑓 𝑛𝑛 = 0: 4.

Another way to do this is with a shortcut method that can be found online. Put the coefficients of the
original sequences on the top and side of a matrix. Then create a multiplication table as follows:

2 4 6
-0.5 -1 -2 -3
1 2 4 6
-0.5 -1 -2 -3

Now, add along the diagonals from upper right to lower left. The first diagonal only hits the upper left -1.
The next hits the -2 and the 2 which get added to 0. The next hits all three values and we add them to get
0. The next diagonal hits 6 and -2 (which sum to 4) and the final diagonal hits the last value in the lower
right: -3. So, again, the final sequence is {−1,0,0,4, −3}. The indices need to be taken care of externally
and we can reason our way through determining that these align with indices n=0:4.
6) (6 points) Circular Shift: Consider {𝑥𝑥[𝑛𝑛]} = {2, −5, 6, −3, 4, −4, 0, −7, 8} for −5 ≤ 𝑛𝑛 ≤ 3.
Let y[n] denote the sequence obtained by a left circular shift of 12 sample periods. Preserve the
range (i.e. y[n] should have the same range as x[n]). Determine the sample value of 𝑦𝑦[−3].

SOLUTION:
A left shift by 12, for a sequence of length 9, is the same as < 12 >9 = 3 So, a left shift of three yields:
{-3, 4, -4, 0, -7, 8, 2, -5, 6} And, thus, since the range is preserved as −5 ≤ 𝑛𝑛 ≤ 3, 𝑦𝑦[−3] = −4.

Another way to directly access this value is to use the equations straight from the definition:
𝑦𝑦[𝑛𝑛] = {𝑥𝑥[< 𝑛𝑛 + 12 >9 ]} for n=-3 is: 𝑦𝑦[−3] = 𝑥𝑥[< −3 + 12 >9 ] = 𝑥𝑥[< 9 >9 ] = 𝑥𝑥[0] = −4

Note that here, unlike in class when the range was from 0 to N-1, our goal is to stay inside the original
range of -5 to 3.

Remember, to solve the modulos, think “add or subtract 9 until the result is in the range -5 to 3 (original
range).

7) (5 points) Odd/Even Parts: Find the odd and even parts of {2, −3,1,4, −2,4, 5} for −2 ≤ 𝑛𝑛 ≤ 4

SOLUTION:
This problem is looking for the odd and even components of the signal. To do this, we apply the
conjugate symmetric and conjugate antisymmetric equations. For real numbers, this results in
simply even and odd functions.
𝟏𝟏
𝒙𝒙𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆 [𝒏𝒏] = (𝒙𝒙[𝒏𝒏] + 𝒙𝒙[−𝒏𝒏])
𝟐𝟐
And
𝟏𝟏
𝒙𝒙𝒐𝒐𝒐𝒐𝒐𝒐 [𝒏𝒏] = (𝒙𝒙[𝒏𝒏] − 𝒙𝒙[−𝒏𝒏])
𝟐𝟐

So, first, let’s make our sequence long enough so the even and odd components will look
symmetric by putting the two zeros on the right hand side and saying the range of listed
parameters is -4 to 4.

n -4 -3 -2 -1 0 1 2 3 4
x[n] (0) (0) 2 -3 1 4 -2 4 5
x[-n] 5 4 -2 4 1 -3 2 (0) (0)
𝒙𝒙𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆[𝒏𝒏] 2.5 2 0 0.5 1 0.5 0 2 2.5
𝟏𝟏
= (𝒙𝒙[𝒏𝒏]
𝟐𝟐
+ 𝒙𝒙[−𝒏𝒏])

𝒙𝒙𝒐𝒐𝒐𝒐𝒐𝒐 [𝒏𝒏] -2.5 -2 2 -3.5 0 3.5 -2 2 2.5


𝟏𝟏
= (𝒙𝒙[𝒏𝒏]
𝟐𝟐
− 𝒙𝒙[−𝒏𝒏])

x[n] 0 0 2 -3 1 4 -2 4 5
reconstructed
by adding
even and odd
8) (5 points) Fundamental Period:
Find the fundamental period of: 4𝑠𝑠𝑠𝑠𝑠𝑠(0.12𝜋𝜋𝜋𝜋) − 𝑐𝑐𝑐𝑐𝑐𝑐(0.15𝜋𝜋𝜋𝜋 + 0.4𝜋𝜋)

SOLUTION:
First, we should find the period of each, then combine by using the rule for combining periodic
sequences.

The period of 𝟒𝟒 𝐬𝐬𝐬𝐬𝐬𝐬(𝟎𝟎. 𝟏𝟏𝟏𝟏𝟏𝟏𝟏𝟏) is found by solving the equation 𝟎𝟎. 𝟏𝟏𝟏𝟏𝟏𝟏𝑵𝑵𝟏𝟏 = 𝟐𝟐𝟐𝟐𝟐𝟐 such that 𝑵𝑵𝟏𝟏
and r are the smallest integers possible.

𝟎𝟎. 𝟏𝟏𝟏𝟏𝟏𝟏𝑵𝑵𝟏𝟏 = 𝟐𝟐𝟐𝟐𝟐𝟐


𝟎𝟎. 𝟏𝟏𝟏𝟏𝑵𝑵𝟏𝟏 = 𝟐𝟐𝟐𝟐
𝟎𝟎. 𝟎𝟎𝟎𝟎𝑵𝑵𝟏𝟏 = 𝒓𝒓
The minimum integer 𝑵𝑵𝟏𝟏 that results in an integer result is 𝑵𝑵𝟏𝟏 = 𝟓𝟓𝟓𝟓 (and, for that 𝒓𝒓 = 𝟑𝟑).

Remember that and let’s find the period of the second term:
−𝐜𝐜𝐜𝐜𝐜𝐜(𝟎𝟎. 𝟏𝟏𝟏𝟏𝟏𝟏𝟏𝟏 + 𝟎𝟎. 𝟒𝟒𝟒𝟒)

The 𝟎𝟎. 𝟒𝟒𝟒𝟒 term only introduces phase and does not impact the period or frequency of the cosine.
So, we can use the same logic above and only focus on the 𝟎𝟎. 𝟏𝟏𝟏𝟏𝝅𝝅𝝅𝝅 term. This results in:

𝟎𝟎. 𝟏𝟏𝟏𝟏𝟏𝟏𝑵𝑵𝟐𝟐 = 𝟐𝟐𝟐𝟐𝟐𝟐


𝟎𝟎. 𝟏𝟏𝟏𝟏𝑵𝑵𝟐𝟐 = 𝟐𝟐𝟐𝟐
𝟎𝟎. 𝟎𝟎𝟎𝟎𝟎𝟎𝑵𝑵𝟐𝟐 = 𝒓𝒓

And, the first integer 𝑵𝑵𝟐𝟐 that results in an integer is 𝑵𝑵𝟐𝟐 = 𝟒𝟒𝟒𝟒 (and this results in n=3).

Now, 𝑵𝑵𝟏𝟏 = 𝟓𝟓𝟓𝟓 and 𝑵𝑵𝟐𝟐 = 𝟒𝟒𝟒𝟒.

When you combine two periodic signals, the key is to think about the first time when these two
signals are going to match with where they started AT THE SAME TIME. So, a multiple of
each period that aligns with the other. This occurs after the first signal has gone through five
periods (then, n=200) and after the 2nd signal has gone through four periods (also n=200). Thus,
the answer is 200.

Two other ways to view this.

One is to think about the lowest common multiple of 40 and 50 – that is LCM(40,50) and that is
200.

Another is to use the equation given in the lecture notes:


𝑵𝑵𝟏𝟏 𝑵𝑵𝟐𝟐 (𝟒𝟒𝟒𝟒)(𝟓𝟓𝟓𝟓) 𝟐𝟐𝟐𝟐𝟐𝟐𝟐𝟐
= = = 𝟐𝟐𝟐𝟐𝟐𝟐
𝑮𝑮𝑮𝑮𝑮𝑮(𝑵𝑵𝟏𝟏 , 𝑵𝑵𝟐𝟐 ) 𝑮𝑮𝑮𝑮𝑮𝑮(𝟒𝟒𝟒𝟒, 𝟓𝟓𝟓𝟓) 𝟏𝟏𝟏𝟏

All three ways yield the same answer. The period of the combined sequence is 200
samples/cycle.
9) (6 points) Energy/Power: 𝒙𝒙𝟐𝟐 [𝒏𝒏] = 𝝁𝝁[𝒏𝒏] that is, the unit step sequence. Calculate the energy
and the power. Is this an energy signal or a power signal?

SOLUTION:
The expression for energy is: 𝑬𝑬𝒙𝒙 = ∑∞
𝒏𝒏=−∞|𝒙𝒙[𝒏𝒏]|
𝟐𝟐

For our signal, this turns into: 𝑬𝑬𝒙𝒙 = ∑∞ 𝟐𝟐


𝒏𝒏=𝟎𝟎|𝟏𝟏| = ∞

𝟏𝟏
The expression for power of an aperiodic sequence is: 𝑷𝑷𝒙𝒙 = 𝐥𝐥𝐥𝐥𝐥𝐥 ∑𝑲𝑲
𝒏𝒏=−𝑲𝑲|𝒙𝒙[𝒏𝒏]|
𝟐𝟐
𝑲𝑲→∞ 𝟐𝟐𝟐𝟐+𝟏𝟏

𝟏𝟏 𝑲𝑲+𝟏𝟏 𝑲𝑲 𝟏𝟏
For our sequence, this is: 𝑷𝑷𝒙𝒙 = 𝐥𝐥𝐥𝐥𝐥𝐥 𝟐𝟐𝟐𝟐+𝟏𝟏 ∑𝑲𝑲
𝒏𝒏=𝟎𝟎 𝟏𝟏 = 𝐥𝐥𝐥𝐥𝐥𝐥 𝟐𝟐𝟐𝟐+𝟏𝟏 = 𝐥𝐥𝐥𝐥𝐥𝐥 𝟐𝟐𝟐𝟐 = 𝟐𝟐
𝑲𝑲→∞ 𝑲𝑲→∞ 𝑲𝑲→∞

Another way to think about this is that, starting at the -1th and 0-th term, you have a 1 and a 0,
which average to ½. As you move outward, you continue to add a 1 for every 0 and the average
power remains 1/2 forever… including in the limit to infinity which is the definition.

A signal of all ones, would have an average power of 1. This has ½ as many ones – thus, an
average power of ½.

This is a power signal because the energy is infinite and the power is finite.

10) (8 points) Unit sample and Unit Step Sequences: Plot the following signals
a. x[n] = 2δ [n] − 3δ [n − 2] + 5δ [n − 3]
b. x[n] = µ [n] + µ [n − 2] − 2 µ [n − 5]

SOLUTION:
a) This yields samples of height 2, -3 and 5 at n=0, 2 and 3, respectively.

-1

-2

-3

-4
-1 0 1 2 3 4 5
b) The first unit step sequence has ones from n=0 to ∞. The second term has ones from n=2
to ∞. Finally, third term has -2 from n=5 to ∞. This leaves: {1,1,2,2,2} for n=0:4.

11) (10 points) For the system 𝑦𝑦[𝑛𝑛] = 𝑥𝑥[𝑛𝑛] + 0.5𝑦𝑦[𝑛𝑛 − 1], find the values of the output 𝑦𝑦[𝑛𝑛] for
0 ≤ 𝑛𝑛 ≤ 8 for each of the following inputs. Assume that 𝑦𝑦[−1] = 0.
a. x[n] = µ [n]
b. x[n] = 2δ [n] − 3δ [n − 1] + δ [n − 2]

SOLUTION:
n -1 0 1 2 3 4 5 6 7 8
𝑥𝑥[𝑛𝑛] = 𝜇𝜇[𝑛𝑛] 0 1 1 1 1 1 1 1 1 1
y[n]=x[n]+0.5y[n-1] 0 1 1.5 1.75 1.875 1.9375 1.9688 1.9844 1.9922 1.9961

𝑥𝑥 [𝑛𝑛] 0 2 -3 1 0 0 0 0 0 0
= 2𝛿𝛿 [𝑛𝑛]
− 3𝛿𝛿 [𝑛𝑛 − 1]
+ 𝛿𝛿 [𝑛𝑛 − 2]
y[n]=x[n]+0.5y[n-1] 0 2 -2 0 0 0 0 0 0 0

12) (10 points) Sampling/Aliasing: For the following cases, determine the normalized frequency in
cycles per sample and determine whether or not aliasing occurs. If aliasing occurs, identify the
unaliased frequency in cycles/sec that produces the same sample sequence.

signal frequency, sampling normalized freq. normalized radian aliasing? unaliased


f0 frequency, FT ν0 frequency ω0 yes/no frequency

units: cycle/sec sample/sec cycle/sample rad/sample

(a) 6000 Hz 20,000 Hz 0.3 0.6𝜋𝜋 No 6000 Hz

-2000 Hz (or
1.5𝜋𝜋 (or
(b) 6000 Hz 8,000 Hz 0.75 (or -0.25) Yes 2000 Hz for a
−0.5𝜋𝜋)
cosine)
SOLUTION: The normalized frequency, 𝜈𝜈𝑜𝑜 , is in cycles per sample, is found by dividing the
signal frequency by the sampling frequency. Because of aliasing and the periodicity of a
sinusoid, the range of non-duplicated values of normalized frequency is -0.5 to 0.5… meaning –
the sampled signal, relative to the normalized frequency value, is periodic with a period of 1 (just
like 𝜔𝜔𝑜𝑜 is periodic with a period of 2𝜋𝜋 and non-duplicated values fall between – 𝜋𝜋 and 𝜋𝜋. So,
here, the first value is 6000/20000 = 0.3. The second value is 6000/8000 = 0.75, but, this
can be shifted by 1 down to -0.25.

The normalized angular (radian) frequency, 𝜔𝜔𝑜𝑜 , converts the normalized frequency to
radians/sample by multiplying by 2𝜋𝜋. If this original value of 𝜔𝜔𝑜𝑜 (before shifting 𝜈𝜈𝑜𝑜 to the -0.5
to 0.5 range) is between –𝜋𝜋 and 𝜋𝜋 then there is no aliasing. If, however, the value is outside of
that range, there is aliasing and we need to perform a mod 2𝜋𝜋 until we get in the range of –𝜋𝜋 and
𝜋𝜋 as described in the lecture notes.

Doing this for the second case, where 𝜔𝜔𝑜𝑜 = 1.5𝜋𝜋 brings it down to −0.5𝜋𝜋. This matches a
normalized frequency (𝜈𝜈𝑜𝑜 ) of -0.25 which yields a signal frequency of -2000 Hz. With no other
information, assume this is a cosine with no phase and you would get a frequency of 2000 Hz. If
there was phase or if this wasn’t an even function, the negative frequency would make a
difference.

Você também pode gostar