Numerically how to approximate exponential decay in a discrete signal

Click For Summary

Discussion Overview

The discussion revolves around methods to approximate exponential decay in a discrete signal, specifically using Laplace transform or Z-transform techniques. Participants explore various approaches to modeling and fitting data that follows an exponential decay pattern.

Discussion Character

  • Exploratory
  • Technical explanation
  • Debate/contested
  • Mathematical reasoning

Main Points Raised

  • One participant inquires about using Laplace transform methods to approximate a decay constant from a given vector of numbers representing an exponential function.
  • Another participant suggests plotting the logarithm of the function against time and finding an approximate line, indicating that this approach does not involve Laplace transform techniques.
  • Some participants express that regression methods are being sought but emphasize a preference for Laplace transform or pseudo-analytic methods instead.
  • There is a mention of using polynomial fitting for curve fitting, with a caution that without prior knowledge of the system, this approach may not yield correct results.
  • One participant notes that the data can be chaotic and emphasizes the need for a specific functional form, proposing a model that combines exponential decay with sinusoidal components.
  • A link to an external resource is provided, suggesting interest in methods for decomposing signals into exponentially decaying sinusoids.

Areas of Agreement / Disagreement

Participants express differing opinions on the appropriate methods for approximating exponential decay, with no consensus on the best approach. Some favor regression and polynomial fitting, while others advocate for Laplace transform techniques or specific functional forms.

Contextual Notes

Participants highlight the importance of assumptions and constraints in modeling, such as continuity and the nature of the data-generating system, which remain unspecified in the discussion.

cppIStough
Messages
24
Reaction score
2
Given a vector of numbers, say [exp(-a t) ] for t - [1, 2, 3, 4, 5] and choose maybe a = -2.4, how can I approximate -2.4 from using Laplace transform methods?

I know you can use regression for this, but I'd like to know the Laplace transform (or Z-transform since it is discrete) approach.
 
Physics news on Phys.org
Say given number sequence is f(t), plot t - log f(t) and find the approxmate line to connect the points and its tan. It is my idea, though Laplace transform plays no role here.
 
anuttarasammyak said:
Say given number sequence is f(t), plot t - log f(t) and find the approxmate line to connect the points and its tan. It is my idea, though Laplace transform plays no role here.
yea this is regression. was looking for laplace transform or some psuedo-analytic manner
 
This would be about statistics and curve fitting, I think. You'll have some basic assumptions as constraints for your model, things like continuity, that you haven't told us. Then I would just use a polynomial fit. It 's not that that's the correct answer, it will be just as likely to be wrong as other models. But since you haven't specified any prior knowledge of the nature of the system producing the data, I don't see a better approach.

Or, maybe I misunderstood and you KNOW that the system is ##e^{-at}##, in which case the answer is almost trivial.
 
DaveE said:
This would be about statistics and curve fitting, I think. You'll have some basic assumptions as constraints for your model, things like continuity, that you haven't told us. Then I would just use a polynomial fit. It 's not that that's the correct answer, it will be just as likely to be wrong as other models. But since you haven't specified any prior knowledge of the nature of the system producing the data, I don't see a better approach.

Or, maybe I misunderstood and you KNOW that the system is ##e^{-at}##, in which case the answer is almost trivial.
The data can be chaotic. Even curve fitting assumes a functional form (polynomial, which I cannot use, must be exponential decay and sinusoidal, so I think ##f(t) = A \exp(-\alpha t)\cos(2\pi f t + \phi)##.

I saw this post and thought there would be a nice implementation for extracting both the sinusoidal frequency and exponential decay:
https://dsp.stackexchange.com/quest...a-signal-into-exponentally-decaying-sinusoids
 

Similar threads

  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 3 ·
Replies
3
Views
4K
  • · Replies 1 ·
Replies
1
Views
3K
  • · Replies 8 ·
Replies
8
Views
4K
  • · Replies 3 ·
Replies
3
Views
1K
  • · Replies 4 ·
Replies
4
Views
3K
  • · Replies 36 ·
2
Replies
36
Views
8K
  • · Replies 65 ·
3
Replies
65
Views
8K
  • · Replies 1 ·
Replies
1
Views
4K
Replies
3
Views
3K