and
N
2
imn
*
mn
e E
; H is a (NxN) diagonal
matrix composed with the complex frequency response functions at the discrete frequencies
( ) 1 N , , 2 , 1 , 0 m
T
2
m m
p
m
K
; and T
p
is the extended period. The generic term of H is
( ) ( ) [ ] { }
1
m
2
m m
2 i 1 k H
+ + sgn (4)
where k is the system stiffness;
; m
m
m
is the natural frequency, is the
damping ratio; and is the hysteretic damping factor. It is worthwhile to mention that the
complex frequency response function of Equation 4 can take into account viscous and
hysteretic damping. The discrete frequencies
m
must be considered in accordance with
Tables 1 and 2 where the symmetry of positive and negative frequencies is expressed.
Table 1. Discrete frequencies (N odd) Table 2. Discrete frequencies (N even)
m
m
m
m
0
1
2
0
2
0
1
2
0
2
(N1)/2
(N+1)/2
N2
N1
[(N1)/2]
[(N1)/2]
2

(N/2)1
(N/2)
(N/2)+1
N2
N1
[(N/2)1]
(N/2)
[(N/2)+1]
2

3 MATRIX FORMULATION OF TD DYNAMIC ANALYSIS
The TD abalysis is performed through the unitimpulse transfer function h(t) which is the
inverse Fourier Transform of the complexfrequency transfer function ( ) H . Thus, by
definition,
F. Venancio Filho, F.S. Barbosa and A.M. Claret
259.4
( ) ( ) ( )
d e P H
2
1
t h
t i
t
0
(5)
where ( ) P is the FT of the loading p(t).
The discrete form of Equation 5 is
( ) ( )
( )
( ). r s e H
2
h t t h
1 N
0 m
t t i
m sr r s
r s m
(6)
Taking into account that t
s
=st, t
r
= st, and m , and separating the positive and
negative exponents, Equation 6 transforms into
( ) ( ) . e H e
2
h t t h
N
2
irm
1 N
0 m
m
N
2
ism
sr r s
(7)
(NxN) matrices E* and E with generic terms
N
2
ism
*
sm
e E
and
N
2
irm
mr
e E
, respectively,
are now introduced. Considering these matrices and the (NxN) diagonal matrix H whose
generic term is given in Equation 4, the unitimpulse transfer matrix derives from Equation 7
as
E H E h
2
. (8)
Substituting now E* H E from the foregoing equation into Equation 1 and considering that
N
1
2
t
, leads to
p h v t (9)
which expresses the TD response.
The response in the generic time t
s
is, from Equation 9,
( ) ( ) ( ). t t h t p t t h
1 N
0 s
r s r s
(10)
Equation 10 is the discrete form of the convolution equation ( ) ( ) ( ) d t h p t v
t
0
.
Comparing now Equations 1 and 9 and taking into account that
N
2
t
one obtains
e h
t
1
. (11)
F. Venancio Filho, F.S. Barbosa and A.M. Claret
259.5
Equation 11 expresses the relation between the ImFT matrix (analysis in the FD domain) and
the unitimpulse transfer matrix (analysis in the TD domain).
4 CONVERGENCE ANALYSIS
The convergence analysis is presented along two lines. First, it is shown that, when N is
even, there is an imaginary term in the generic response v(t
n
) and then, proven that this term
tends to zero when N tends to infinity. In the sequel a proof is given that, with increasing N,
v(t
n
) converges to the real solution.
The generic response v(t
n
), in the generic time t
n
, is derived from Equation 1 as
( ) ( ) ( )
1 N
0 n
N
mn
2 i
n m
1 N
0 m
N
mn
2 i
n
e t p H e
N
1
t v
. (12)
Taking into account the discrete frequencies of Table 1 (N odd), the first summation
,
_
1 N
0 m
in
Equation 12 is a summation of pairs of complex conjugates whose generic pair is
( ) ( )
1 N
0 n
N
mn
2 i
n m
N
mn
2 i
m , m
e t p H e C
m
(13)
where the subscripts m and m correspond, respectively, to positive and negative frequencies
(Tables 1 and 2).
When N is even (Table 2) a complex term, C
N/2
, associated with the Nyquist frequency,
N/2
, is present in Equation 12. Substituting
2
N
for m in Equation 13 leads to
( ) ( )
1 N
0 n
n 2 / N 2 / N
n cos t p H n cos C . (14)
As N is even the summation in the RHS of Equation 7 can be developed as a summation of
pairs as follows:
( ) ( ) ( ) [ ] ( ) ( ) [ ]
( ) ( ) [ ] ( ) ( ) [ ] , t p t p t p t p
t p t p t p t p n cos t p
1 N 2 N 1 r r
1 N
0 n
3 2 1 0 n
+
+ +
+ + +
K
K
(15)
with r even. All the pairs in the previous equation can be expressed, like the generic one, as
( ) ( ) [ ] ( )
1
1
]
1
,
_
+
+
N
T
t p t p t p t p
p
r r 1 r r
. (16)
F. Venancio Filho, F.S. Barbosa and A.M. Claret
259.6
In the limit, when N tends to infinity, the following result is obtained:
( ) ( ) ( ) 0 t p t p
N
T
t p t p lim
r r
p
r r
N
1
1
]
1
,
_
+
. (17)
Therefore, when N tends to infinity, the central complex term given by Equation 14 tends to
zero.
Consider now the modulus of the terms of the generic pair of Equation 13
( ) ( )
1 N
0 n
n m m m
t p H C C . (18)
where, from Equation 4,
( )
( ) ( )
2 / 1
2
m
2
2
m
m
2 1
1
k
1
H
1
]
1
+ +
sgn
. (19)
The maximum value of ( )
m
H occurs only when one of the discrete frequencies
m
sgn +
. (20)
Taking into that ( )
1 N
0 n
n
t p , Equation 18, tends to zero, according to Equations 15, 16 and 17
and that ( )
m
H , Equation 20, is limited,
m
C and
m
C , Equation 18, tend also to zero.
Therefore v(t
n
), Equation 12, converges to the real solution.
5 CAUSALITY OF THE RESPONSE
The causality of the response is the property that the response at any time t is not influenced
by the excitation at all the times greater than t. With the consideration of Equations (1) and (9)
the causality property corresponds to the lower triangularity of matrices e and h. In order that
h be lower triangular the load p(t
r
) (r>s) should not contribute to the response at time t
s
with is
equivalent to h(t
s
t
r
)=0 for the r>s. To prove this consider Equation 7 with the exponents
gathered:
( ) ( )
( )
1 N
0 m
N
2
s r im
m sr r s
e H
2
h t t h
. (21)
F. Venancio Filho, F.S. Barbosa and A.M. Claret
259.7
The maximum absolute value of ( )
m
H occurs when, eventually,
m
. Then from
Equation 4:
( ) ( )
2 i k
1
H H
max
m
. (22)
Therefore the following inequality is verified:
( ) ( ) H H
m
. (23)
Introducing this inequality in Equation 21 the modulus
sr
h of h
sr
is expressed as
( )
1 N
0 m
m sr
Z H
2
h
. (24)
where
( ) s r e Z
N
2
) s r ( im
m
>
. (25)
Now the series
m
Z is a geometric one with ratio
( )
N
2
s r i
e Z
and it is proven that it is
absolutely convergent
6
, that is,
( ) 1 Z
Z 1
Z 1
Z Z Z 1
N
N 2
+ + + + L . (26)
On the other hand, from Equation 25, Z
N
=1. Then from Equation 26 the series
n
Z
converges to zero. In consequence, from Equation 25, the terms such as h
sr
(r>s) converge to
zero as N tends to infinity. This conclusion proves finally the causality of the response in the
TD and FD dynamic analysis expressed, respectively, by Equations 9 and 1.
The causality of the response is closely related to the period extension. When the extended
period tends to infinity a causual response is obtained
7
.
6 NUMERICAL EXAMPLES OF CONVERGENCE AND CAUSALITY
The response of a SDOF system to a short duration impulsive load with time duration
t
d
=0.275 sec whose peak is 15kN. The properties of the system are: natural frequency: =5.71
rd/sec; natural period: T=1.1sec; damping ratio =0.10. The response is calculated from
Equation 1 with an extended period given by s 07 . 8
605 . 4
T
p