Scribd
Upload
117 views3 pages

Rectangular Pulse Convolution-Update

The document summarizes the convolution of a rectangular pulse function with itself. It shows that: 1) The convolution results in an isosceles triangle function with a height of A2(b-a) and a base width of 2(b-a). 2) The triangle is centered at (a+b, A2(b-a)) with vertices at (2a,0) and (2b,0). 3) The convolution is broken into pieces depending on the value of t, with the triangle rising for 2a ≤ t ≤ a+b and falling for a+b ≤ t ≤ 2b.

Uploaded by

khaled lotees
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
117 views3 pages

Rectangular Pulse Convolution-Update

The document summarizes the convolution of a rectangular pulse function with itself. It shows that: 1) The convolution results in an isosceles triangle function with a height of A2(b-a) and a base width of 2(b-a). 2) The triangle is centered at (a+b, A2(b-a)) with vertices at (2a,0) and (2b,0). 3) The convolution is broken into pieces depending on the value of t, with the triangle rising for 2a ≤ t ≤ a+b and falling for a+b ≤ t ≤ 2b.

Uploaded by

khaled lotees
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Convolution of a Rectangular ”Pulse” With Itself

Mike Wilkes
10/3/2013

After failing in my attempts to locate online a derivation of the convolution of a general rectangular
pulse with itself, and not having available a textbook on communications or signal processing theory, I
decided to write up my attempt at computing it. I expect, however, that it is the first example one would
find in any textbook that discusses convolution.
Recall the general definition of the convolution f ∗ g of two real-valued functions:
! ∞ ! ∞
(f ∗ g)(t) = f (u) g(t − u) du = f (t − u) g(u) du. (1)
−∞ −∞

We apply this to the problem where f and g are both given by




0, t < a,
f (t) = g(t) = A, a ≤ t ≤ b, (2)


0, t > b,
where [a, b] is a time interval on the real line, with a < b. In signal processing this represents a rectangular
pulse of amplitude A and width or duration T = b − a. The convolution of this function with itself is the
time-dependent function
! ∞
(f ∗ f )(t) = f (u) f (t − u) du. (3)
−∞

How is f (t − u) related to f (u)? Define g(u) = f (u + t), which for t > 0 represents a horizontal
translation of f (u) to the left by t. Then h(u) = g(−u) = f (−u + t) = f (t − u) is a reflection of g(u)
across the vertical axis u = 0. Thus, as a function of u, f (t − u) is a replica of f (u) which, for t > 0, is
translated to the left a distance t, then reflected across the vertical axis u = 0. Thus, f (t − u) is a function
whose values depend on both u and t. The convolution is calculated for each value of t on the real line by
integrating over the real line (with respect to u) the product of the two functions f (u) and f (t − u) at that
value of t. For the rectangular function defined by (2), graphs of the functions f (u), f (u + t), and f (t − u)
for t > 0 are illustrated below:

f(u+t) f(u) f(t u)


A

u
a t b t a b t b t a

Referring to the Figure, observe that as the distance t from a increases, the translated and reflected pulse
f (t − u) moves to the right toward +∞. On the other hand, as the distance t to the left of a decreases,

1
the translated pulse f (u + t) moves toward the right, and the translated and reflected pulse f (t − u) moves
toward the left.
Focusing first on the left edge u = t − b of f (t − u) (represented in the next two Figures by the rectangle
with ”dashed line” sides), we see that for u = t − b > b, the original pulse has value f (u) = 0, so the
convolution will be zero for t > 2b, corresponding to u > b.
When a < t − b ≤ b, or a + b < t ≤ 2b, both f (u) = A and f (t − u) = A. But f (t − u) = 0 for u ≤ t − b,
so the integrand is nonzero only for t − b ≤ u ≤ b, as shown in this Figure:

f(u) f(t u)
A

u
a t b b t a

The two pulses coincide exactly when t − b = a, and t − a = b. that is, when t = a + b.
For t > a + b, we focus on the right edge u = t − a of f (t − u) as it moves thru f (u) to the left. For
a ≤ t − a < b, or 2a ≤ t < a + b, both f (u) and f (t − u) have amplitude A, but f (t − u) = 0 for u > t − a,
hence the integrand is nonzero only for a ≤ u ≤ t − a. The situation is illustrated in this Figure:

f(t u) f(u)
A

u
t b a t a b

Finally, for u = t − a < a, or t < 2a, the original pulse f (u) = 0, so the convolution is again zero for
u < a.

These results are summarized in the following calculations of the convolution for any value of t, listed
in the opposite order of the above discussion:


 (f ∗ f )(t) = 0, for t < 2a,





 ! t−a



 (f ∗ f )(t) = A · A du = A2 (t − 2a), for 2a ≤ t ≤ a + b,


 a
(4)

 ! b



 (f ∗ f )(t) = A · A du = A2 (2b − t), for a + b ≤ t ≤ 2b,



 t−b





(f ∗ f )(t) = 0, for t > 2b.
The graph of this piecewise-defined function is an isosceles triangle of height A2 (b − a) at the vertex point
( (a + b), A2 (b − a) ), and base of width 2(b − a) with vertices at the points (2a, 0) and (2b, 0) on the t–axis.
These details are illustrated in the next Figure.
A2 (b − a)

(f ∗ f )(t) = A2 (t − 2a) (f ∗ f )(t) = A2 (2b − t)

t
2a a+b 2b

You might also like