SignalForge - it's open source software, should it be in this section??

[LightningPhil]

It's been a while since looking on the forum. But seemed a sensible plave to ask for advice.

Have been making (Vibe coding and thus cheating massively), a web app that's ended up being called SignalForge. Here's a link to the github repo and github pages site:
Repo ==> https://github.com/LightningPhil/SignalForge
Site ==> https://lightningphil.github.io/SignalForge/

It's purpose evolved from a work need, where we have some really noisy signals. We're mesuring voltage and current right next to an arc. The noise was all very high frequency, and looked easy enough to control, but the rabbit hole deepened. Ended up needing the differential of the signals for purposes and that was unbeleivably noise.

So, lots of faff later, a web app popped into existance that can inport CSV files, then apply a pipeline of filters. I particularly like the Sav-Gol filter and implimenting it successively. This really smoothes out the differential and made the data useful enough for further processing.

There's a lot in the app. Wondering if anyone else needs something similar and if there's any feedback (good negative is welcomed  ).

[macaba]

Here is a comprehensive README.md file for your project.

First line of the readme! Vibe coding works better if you review the output for correctness.

With regards to features - averaging only works right up to the 1/f noise corner of your input data, then further averaging doesn't improve RMS noise. With that in mind, you might want to add a noise spectral density estimation plotting mode which will allow the user to see where the 1/f corner is, and use that to inform how much averaging, or what kind of averaging, to use.

[LightningPhil]

Haha   

I'd not read the readme for ages. But least I admited to massive cheating and vibe coding 

Anyhow, you're quite right it won't magically reduce RMS noise.  Though a spectral plot as you mention is nearly there. You can view things in the frequency domain and apply fft based filters. I've been trying to figure out how tom improve the spectral filters.

The origional use of the app was seeing through the HF noise that's all over some voltage and current signals. We were attempting to calculate the impedance of an arc as it varies over time for a discharge in a pulsed plasma thruster. Involves taking mesurements from within a vacuum chamber with quite high voltage and current signals. Most of the diagnostics are built in house to be as compact as possible. Risetimes are in the ns range and whole pulse is over and done with in ~150 ns.

Not expecting this to be the best tool ever. But have found it handy a few times for processing and making plots.

Will report back after meddling with the spectral plots..

[macaba]

Makes sense - this kind of quick app is right in the sweet spot for LLMs. They generally fall apart when looking at bigger or more rigorously coded apps. It's the modern day version of writing a quick LabVIEW VI for it!

If it struggles to find a good NSD solution, then point it at my app for inspiration, maybe there's equivalent JS libraries that can be substituted:
https://github.com/macaba/nsd

[KE5FX]

Actually looks pretty slick!  Would be nice to scrub zoomed views with the middle button.

[voltsandjolts]

Can you provide an example file to load?

Regards that noisy signal you want to measure; can you modulate it in some way, shift it to a less noisy part of the spectrum? Although with arcing, maybe the noise is everywhere