For the evaluation of the characteristics of a continuous linear system is known correlation function technique which is based essentially upon the following simple relation:
φyz(τ)=∫T0g(s)φmz(τ-s)ds where
g(s) is an impulse response,
m(t) is an
M-sequence input with a basic clock pulse interval Δ and a total pulse number
N in a period
T,
y(t) is an output and φ
yz(τ) is a cross correlation function between the output
y(t) and an appropriate function
z(t).
In the conventional correlation method,
z(t) is an
M-sequence signal, while, in the persent paper, it is proposed that
z(t) be
z(t)=NΣk=1{m(t)+c}δ(t-kΔ+Δ/2) where δ
(t) is the delta function and
c is an appropriate constant.
For processing the output data to evaluate an estimate of the impulse response is derived a simple computational algorithm which, by proper choice of the constant
c, is made to reduce to a minimum the effect of additive noises present at the
M-sequence input and at the measured output. Furthermore, an analysis of error introduced by non-ideal input transducer is also developed.
The error estimate is given when the above algorithm is applied to a second order system. The experimental results through an analog computer simulation are in good agreement with the calculated estimate.
抄録全体を表示