Timeline for How to detect decrease of a digital signal?
Current License: CC BY-SA 4.0
17 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Nov 6 at 16:28 | vote | accept | Steve | ||
| Jun 2, 2023 at 6:45 | comment | added | Steve | @DanBoschen, Peter K. I am sorry I have overlooked that I have wrong time units on the time axis. I have just corrected my mistake. | |
| Jun 2, 2023 at 6:44 | history | edited | Steve | CC BY-SA 4.0 |
edited body
|
| Jun 1, 2023 at 21:37 | answer | added | Peter K.♦ | timeline score: 3 | |
| Jun 1, 2023 at 20:48 | comment | added | Peter K.♦ | Also wondering at the sampling period? If it's 1ms, as stated, then there'd only be two samples in the graph plotted which goes from 0 to 2.4ms in the $x$ axis. | |
| Jun 1, 2023 at 11:32 | comment | added | Dan Boschen | Your time constant does not look like 10 ms in the filtered result shown in the graphic: It looks like you settle within 0.1 ms which would be a time constant closer to 20 ps. Distributing the filtering before and after the derivative may be a better strategy. Have you evaluated the spectrum of the noise? Another filter may be more appropriate based on what that looks like. To proceed with a more detailed answer, I would like to see what the typical noise spectrum looks like and a criteria for what the minimum duration is for the $-1000V s^{-1}$ for triggering an actual event. | |
| Jun 1, 2023 at 5:01 | history | edited | Steve | CC BY-SA 4.0 |
added 5 characters in body
|
| May 31, 2023 at 21:23 | comment | added | robert bristow-johnson | Yup. What @Ash said. There is a fast $O(\log(N))$ median alg, but it's difficult. Also consider weighting functions like a sliding Hann window that can also be done efficiently but it's also a bitch. | |
| May 31, 2023 at 20:51 | comment | added | Ash | You could also try running $x$ through a median filter prior to estimating its derivative. Order statistic filters are far more robust to outliers than your exponential moving average. | |
| May 31, 2023 at 20:47 | comment | added | robert bristow-johnson | Hay, if you're gonna evaluate the derivative with a two sample $\Delta t$ then assign it to the sample in the middle. It should be $y_2[n-1] = x[n]-x[n-2]$. Also you should be running a sliding max or maybe a sliding r.m.s. of the derivative, so you have something to compare to. | |
| May 31, 2023 at 20:01 | history | edited | Steve | CC BY-SA 4.0 |
added 108 characters in body
|
| May 31, 2023 at 13:02 | comment | added | Steve | @Ben thank you for your reaction. Please can you tell me your opinion regarding the proposed method? | |
| May 31, 2023 at 11:58 | comment | added | Ben | FIY, your post describes a specific case of the filtered derivative. | |
| May 31, 2023 at 11:41 | comment | added | Ben | isy.gitlab-pages.liu.se/fs/en/courses/TSFS06/PDFs/… Based on what you wrote, I recommend either CUSUM algorithm or filtered derivative | |
| May 31, 2023 at 11:39 | comment | added | Steve | @Hilmar thank you for your reaction. Let's say I will have a precise requirement regarding the decrease e.g. $-1000\,\mathrm{V}\cdot\mathrm{s}^{-1}$. Can you tell me your opinion regarding the proposed detection method. | |
| May 31, 2023 at 9:03 | comment | added | Hilmar | You need a crisp definition of what exactly "decrease" means within the context of your specific application. As stupid as it sounds: that's actually the most difficult part. Once you have a good set of requirements, the algorithm typically just follows along. | |
| May 31, 2023 at 8:32 | history | asked | Steve | CC BY-SA 4.0 |