You are on page 1of 88

# 馬大電機不負責任資格考參考密笈

To 2007 Fall
Preface

## Probability，Circuits，Linear Systems 和 Digital Logic 為主，討論過後

力，把這份參考解答錯誤的地方或是新的考題部分在做更新與補充，

不是嗎？

2007/02/03
Basic Mathematics
Contents

1 math 2
1.1 sample (a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 sample (b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Fall 2007 (i) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Fall 2007 (ii) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5 Fall 2007 (iii) . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.6 Fall 2007 (iV) . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.7 spring 2007 (1) . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.8 spring 2007 (2) . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.9 fall 2006 (1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.10 fall 2006 (2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.11 spring 2006 (1) . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.12 spring 2006(2) . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.13 Fall 2005 (a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.14 Fall 2005 (b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.15 Fall 2005 (c) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.16 spring 2005 (1) . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.17 spring 2005 (2) . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1
Chapter 1

math

## 1.1 sample (a)

d2 f df
2
+ 5 + 4f = 2exp(2x) (1.1)
dx dx
the feature of linear differential equation is f duplicates itself, so f is of
the form exp(rx). plug into the equation, we find f is a linear combination
of ’two’ exponents. It is two because we are solving second order DE.

## f (x) = Aexp(−x) + Bexp(−4x) (1.2)

Above is the so call homogeneous solution. You can say it is the natural
response of the equation no matter if there is driven source. The other
solution is the particular/inhomogeneous solution. It is the response driven
by the source. In our example it is 2exp(2x). Since it is the driven source, the
’frequency’ of the particular solution is the same. so we model the particular
solution to be of the form kexp(2x) plug into the equation, k = 1/9 is
carried out directly. Coefficients A, B are determined by boundary condition
or initial condition. without the possible conditions, the general form is
2
f (x) = exp(2x) + Aexp(−x) + Bexp(−4x) (1.3)
9
since given f (0) and f 0 (0), directly differentiate the above equation and
evaluate both at x = 0. two unknowns with two equations A, B suggests
these are sufficient to find them. solution:
2 2 4
f (x) = exp(2x) + exp(−x) − exp(−4x) (1.4)
9 9 9

2
1.2 sample (b)
¯ ¯
¯ 1−λ 1 1 ¯
¯ ¯
|A − λI| = ¯¯ 1 1−λ 1 ¯
¯
¯ 1 1 1−λ ¯
the above matrix is set to be zero becuase of definition of eigenvalue
and eigenvector. the corresponding eigenvalues are λ = 0, 0, 3. same values
are called degenerate. substitute eigenvalue and use the technique of Gauss
elimination to get the corresponding eigenvector. If one eigenvalue is degen-
erate, one can find more than one eigenvectors which are orthogonal to each
other. orthogonal means dot product is zero. Remember to normalize those
eigenvectors as required.

 √ √ √ 
1/√3 1/ 2 1/ √2
A[v1 , v2 , v3 ] = A  1/√3 0√ −1/ 2  = [λ1 v1 , λ2 v2 , λ3 v3 ] = [3v1 , 0v2 , 0v3 ]
1/ 3 −1/ 2 0

## 1.3 Fall 2007 (i)

To evaluate limit, remember x → 4 means√x can approach to 4 arbitrarily.
However, x nevers equals 4. Since x 6= 4, x 6= 2 and denominator is never
zero. We can safely rewrite

x−2 1
lim = lim √ = 1/4 (1.5)
x→4 x − 4 x→4 x+2

By definition,

## f (x) ≡ f (−1)+f 0 (−1)/1!∗(x+1)+f 00 (−1)/2!∗(x+1)2 +f 000 (−1)/3!∗(x+1)3 +...

(1.6)
2 3
= 1 + 2(x + 1) + 3(x + 1) + 4(x + 1) + ... (1.7)
While Taylor expansion is a systematic way, sometimes it suffers from
inefficiency. This example serves as an example, using identity equation
−1 1
= = 1 + (x + 1) + (x + 1)2 + (x + 1)3 + ... (1.8)
x 1 − (x + 1)

3
differentiate it,
1
= 1 + 2(x + 1) + 3(x + 1)2 + 4(x + 1)3 + ... (1.9)
x2
we reach the same result.

## 1.5 Fall 2007 (iii)

Two ways possible. 1)determine disk area with radius x and then integrate
over y. 2) determine cylindrical shell with height y and then integrate over
x. method 2 is preferable because y has been given explicitly in terms of x.
Express x in term of y is not easy.
the cylindrical shell has surface area (y2 − y1) × 2πx. The thickness of
the shell is dx. so volume is
Z 1
V = dx × 2πx × (−1/2x3 − x2 + 3/2x) = 3/10π (1.10)
0

## 1.6 Fall 2007 (iV)

consider
d
(y × exp(−2x)) = (y 0 − 2y)exp(−2x) = −4xexp(−2x) (1.11)
dx
Z
∴ yexp(−2x) = −4xexp(−2x)dx = 2xexp(−2x) + exp(−2x) + C (1.12)
y = 2x + 1 + Cexp(2x) (1.13)
applying condition x = y = 0,
y = 2x + 1 − exp(2x) (1.14)
the above method is the most general technique to deal with first order
DE, no matter linear or not. However, it only applies to first order. In
this example, it is linear because coefficient of y is independent of x. When
DE is linear, one can use Laplace transform even the DE is of higher order.
When DE is linear and one is asked steady-state solution, one can use Fourier
transform. In this example, it is not steady, we should not expect the same

## L [y 0 − 2y] = (s − 2)Y (s) = L [−4x] = 4δ 0 (x) (1.15)

Inverse Laplace transform gives
4δ 0 (x)
y(x) = L−1 [ Y (s)] = 2x + 1 − exp(2x) (1.16)
s−2

4
1.7 spring 2007 (1)
similar to fall 2007 last question. Consider
d 2 2 2
(xe−t ) = (x0 − 2tx)e−t = te−t (1.17)
dt
Z
t2 2 2
x=e te−t dt = −1/2 + Cet (1.18)

## applying initial condition,

Z
t2 2 2
x=e te−t dt = −1/2 + 1/2et (1.19)

## another technique is separation of variable, observing that

dx
= t(1 + 2x) (1.20)
dt
Therefore,
dx
= tdt (1.21)
1 + 2x
integrate both side,
Z Z
2 dx
t = tdt = = 1/2 ln (x + 1/2) + C (1.22)
1 + 2x
Applying initial condition, we arrive the same result.

## 1.8 spring 2007 (2)

When a plane is tangent to a surface, both normal is same. Use gradient
to determine the normal of the surface so that we obtain the normal of the
plane.
F ≡ x2 + 2y 2 + 3z 2 − 6 OF = (2x, 4y, 6z) (1.23)
the normal at (1,1,1) is thus
n = OF k(1,1,1) = (2, 4, 6) (1.24)
since the plane pass through (1, 1, 1), the plane is
(2, 4, 6) • (x − 1, y − 1, z − 1) = 2(x − 1) + 4(y − 1) + 6(z − 1) = 0 (1.25)
after manipulating,
z = −1/3x − 2/3y + 2 (1.26)

5
1.9 fall 2006 (1)
because A is symmetric, modeling matrix A as
· ¸
a c
A=
c b

· ¸· ¸ · ¸ · ¸
a c 3 3 15
Av = =5 =
c b 4 4 20

## the final condition gives

ab − c2 = 50 (1.27)

## At a glance, three unknowns a, b, c with three conditions suggests sufficiency.

one systematic way is to express a, b in terms of c. from the second conditon,
we get
15 − 4c 20 − 3c
a= b= (1.28)
3 4
substitute a, b into the third condition, we get c = −12/5 as a result,
· ¸
41/5 −12/5
A=
−12/5 34/5

## 1.10 fall 2006 (2)

The curve is symmetric about x, y axes. As a result, we only need to deal
the first quadrant, If we define total length to be L, then
Z p Z a p
1/4L = 2
dx + dy =2 1 + y 02 dx (1.29)
x=0

## The given curve equation is in polar coordinate. if we wish to apply above

method, we should transform it back to x, y coordinate. (One can also do
it in polar coordinate, but I think difficulty is the same. Hope someone has
better method.) There is cos2 (θ), suggesting multiply both sides by r2 . So,
3
(x2 + y 2 ) 2 = r3 = ar2 cos2 (θ) = ax2 (1.30)
p
or y = a2/3 x4/3 − x2 (1.31)

6
p p
The logic is, we want 1 + y 02 and we have y. So we construct 1 + y 02
from y:
4 2/3 1/3
0 3
a x − 2x
y = √ (1.32)
2 a2/3 x4/3 − x2
... (1.33)
s
p 4 4/3 2/3
a x − 13 a2/3 x4/3
9
y 02 + 1 = (1.34)
a2/3 x4/3 − x2

the above form is tedious. One good way is to track each term’s dimension.
For example, y and x have the same dimension (length). So we expect y 0 is
dimensionless. That is why y 02 can be added with number one. Similarly, a
and x have the same dimension. In the last equation, we see both denomi-
nator and numerator have the same dimension as x2 . It is very instructive
to divide both denominator and numerator by x2 or a2 as well. As a result,
we are dealing with

s
Z 4
1 a
9
(a/x)4/3− 13 (a/x)2/3
L = dx (1.35)
4 x=0 (a/x)2/3 − 1
s
Z a
1 − 34 (x/a)2/3 2
= (a/x)1/3 dx (1.36)
x=0 1 − (x/a)2/3 3

## The latter part is modified as

2 2
(a/x)1/3 dx = (x/a)−1/3 d(x/a)a = d(x/a)2/3 a (1.37)
3 3
so, r
Z a
1 3 1
L=a 1 − (x/a)2/3 × p d(x/a)2/3 (1.38)
4 x=0 4 1 − (x/a)2/3

## The following step is tricky. √ 1

suggests to express d(x/a)2/3 as
1−(x/a)2/3
−d(1 − (x/a)2/3 ) so that
1
p −d(1−(x/a)2/3 ) = −(1−(x/a)2/3 )−1/2 d(1−(x/a)2/3 ) = −2d(1−(x/a)2/3 )1/2
1− (x/a)2/3
(1.39)
Finally, we change of variable
q
1 − (x/a)2/3 ≡ z (x/a)2/3 = 1 − z 2 (1.40)

7
Summarize,
Z 0
r
1 3
L = a 1 − (x/a)2/3 − 2dz (1.41)
4 z=1 4
Z 1 r
3
= 2a 1 − (1 − z 2 )dz (1.42)
4
Z 1z=0√
= a 1 + 3z 2 dz (1.43)
z=0

2 √
L = a 4 + √ ln(2 + 3) (1.44)
3
Remark. In my opinion, the only principle is to keep changing of variable
and finally you will reach the result.

## 1.11 spring 2006 (1)

The smallest distance occurs when the tangent of y1 = x2 has the same slope
as y2 = 2x − 4 which is 2.

## so we know the point on y1 = x2 is (1, 1). We want to know the distance

between (1, 1) and y2 = 2x−4. One way is to find a line normal to y2 = 2x−4
and it passes (1, 1). Since the slope product of two perpendicular lines is -1,
we can model the normal line directly as

x + 2y|(1,1) = 3 (1.46)

## This line intersects y2 = 2x − 4 at√(11/5, 2/5). As a result, the distance

between (1, 1) and (11/5, 2/5) is 3/ 5.

## 1.12 spring 2006(2)

This kind of coupling differential equation has a systematic way to solve
them. We are given

## x0 + 2x = y → (d/dt + 2)x = y (1.47)

y 0 + 2y = x → (d/dt + 2)y = x (1.48)

8
The differential operator (d/dt + 2) transforms x, y to each other. So we
apply it to either x or y again to setup a single variable DE.

## D2 x = (d/dt + 2)2 x = (d/dt + 2)y = x (1.49)

or x00 + 4x0 + 4x = x (1.50)
or x00 + 4x0 + 3x = 0 (1.51)

## Similar to problem (a) of the sample, it is a second order DE without the

driven source/ inhomogeneous solution. The general solution thus reduces
to only possess homogeneous solution.

## Since y(0) = 1 = A + A, A = 1/2. Finally,

1 £ −t ¤ 1 £ −t ¤
x(t) = e − e−3t y(t) = e + e−3t (1.55)
2 2

## 1.13 Fall 2005 (a)

Let’s use identity relation to expand e−2x . We are asked to expand at x = 1.
We therefore construct e−2(x−1)−2 .
1 2 1
e−z = 1 − z + z − z 3 + ... z ≡ 2y (1.56)
2! 3!
1 1
e−2y = 1 − 2y + (2y)2 − (2y)3 + ... y ≡x−1 (1.57)
2! 3!
1 1
e−2(x−1) = 1 − 2(x − 1) + (2(x − 1))2 − (2(x − 1))3 + ... (1.58)
· 2! 3! ¸
1 1
e−2x = e−2(x−1)−2 −2 2 3
= e × 1 − 2(x − 1) + (2(x − 1)) − (2(x − 1)) (1.59)+ ...
2! 3!

9
1.14 Fall 2005 (b)
If it is cut by z-axis, each slice is a √rectangle box. Use x as a parameter,
the width of the rectangle is 2y = 2 a2 − x2 , the height √ is xh/a, and the
thinkness is dx. Each rectangle box has volume dV = dx×2 a2 − x2 ×xh/a:
Z a Z a √ 2
V = dV = 2h/a dxx a2 − x2 = ha2 (1.60)
x=0 x=0 3

## 1.15 Fall 2005 (c)

This is an implicit function connecting x, y and z. The question is to ask the
’partial’ derivative of ∂z/∂x, and so y is fixed as a constant. Simply partially
differentiate the given relation and organize what we need, and all done.
∂ ∂
[sin(2x − z) + cos(y − 2z)] = 1=0 (1.61)
∂x · ¸ · ∂x ¸
∂z ∂z
cos(2x − z) 2 − − sin(y − 2z) −2 =0 (1.62)
∂x ∂x
∂z 2cos(2x − z)
= (1.63)
∂x cos(2x − z) − sin(y − 2z)

10
1.16 spring 2005 (1)
Separation of variable:
Z
dy dx
= ln(x) = y −2 dy = −y −1 + C (1.64)
y2 x

## Boundary condition (x, y) = (1, 1) gives C = 1, so

1
ln(x) = − + 1 (1.65)
y

## 1.17 spring 2005 (2)

Similar to problem (b) of the sample,

 
1 √1 1
√ √ √
A[v1 , v2 , v3 ] = A  0 2 − 2  = [λ1 v1 , λ2 v2 , λ3 v3 ] = [1v1 , (1+ 2)v2 , (1− 2)v3 ]
−1 1 1

11
Probability
Electromagnetism
Circuit
Linear systems
Device
Computer Architecture
& systems
Digital Logic