When $f(t)-f(t-\tau)$ is already known, how to determine $f(t)$?

73 Views Asked by At

As an example, we want to measure $f(t)$, but it's hard to get it directly. What we can do is to measure $f(t)-f(t-\tau)$. How to use the information we have to find $f(t)$?

The only method I know is, when $\tau$ is small enough, $$f(t)-f(t-\tau) \approx \tau f'(t)$$ then integrate it. But I still feel this is not good enough.

Motivation

We need to measure the instant frequency of a tuneable laser source (or the instant phase $\phi(t))$, and through swept-wavelength interferometer we get the beat signal $\cos(\phi(t)-\phi(t-\tau))$, where $\tau$ is the delay of two arms, so I'm struggling with dealing this issue. If any solutions here, please feel welcome to show your wisdom.

1

There are 1 best solutions below

2
On

The transfer function of the measurement device is

$$ H_M (s) = 1 - e^{- \tau s} $$

Note that if the delay $\tau$ is "sufficiently small", then $H_M (s) \approx \tau s$, which is a scaled differentiator.

One would like to find a "compensator" that undoes the measurement. This compensator could have the following transfer function

$$ H_C (s) = \frac{1}{H_M (s)} = \frac{1}{1 - e^{- \tau s}} $$

Again, note that if the delay $\tau$ is "sufficiently small", then $H_C (s) \approx \frac{1}{\tau s}$, which is a scaled integrator.

The aforementioned compensator's could be implemented by the following

$$ y (t) - y (t - \tau) = u (t) $$

where $u$ is the compensator's input and $y$ is the compensator's output. Do you have access to a delay line? How accurate is your measurement of $\tau$?