@windytan

1,000 Followers
39 Following
181 Posts
Curious experiments with sound, waveforms, and computers. Slow account.
Project blog: Absorptionshttps://www.windytan.com
YouTube: Windytan's waveshttps://www.youtube.com/@windytan
Languages🇫🇮🇬🇧🇸🇪(🇪🇪🇫🇷🇩🇪)
You can sort of see the 5 repeats in the audio spectrogram (left) but especially the cepstrum (right - which shows echoes and repeating structure) has a sharp peak at 1878 ms, just a couple of bins wide

Viral shortform video is often augmented with subtle glued-on sound effects, but here's an interesting one - the background noise was "looped" for around 9 extra seconds to drive home the wow effect :)

https://www.youtube.com/shorts/7UZetfRezzI

This How Dangerous 😳 Cave #shorts #ytviral #facts #factshorts #like #trending

YouTube
Have you ever had a déjà vu where you also catch yourself predicting what happens next? Of course the prediction will be wrong, but there is some sense of what *should* happen for the déjà vu sensation to continue.
Fellow myopics, do you ever stare into a faraway LED just to appreciate the details in your eyes' point spread functions?

Update: One of the places that had this ultrasonic beep from their PA system has now stopped using it!

There's now a much quieter tone at a much higher frequency, outside anyone's hearing range for sure, which only shortly appears every now and then. Maybe they upgraded their sound system?

It's amazing how noise resistant LoRa is. It took me a while to realize that there can be plenty of perfectly decodable LoRa traffic even though the recording spectrograms look empty.

Here's a complete Meshtastic GPS position message (~50 chirp symbols each 250 kHz wide) buried in noise and some kind of smart meter traffic. I got a correct CRC after just 1 parity flip!

@windytan when you feed only part of a CW into the FFT you get the Fourier transform of a rectangle, which is a sinc of frequency proportional to the rectangle width. The symmetry that happens in time is best explained by thinking that feeding a CW that fills more than 1/2 of the FFT is equivalent to feeding a CW for the whole FFT duration plus a negative rectangle that occupies the beginning less than 1/2 of the FFT duration.

@windytan I also noticed the same pattern before. Still couldn't figure out the theoretical reason behind it, but I recognize the pattern kind of resembles an aliased sine wave

sin(x^2 + y^2)

x,y integers.

that I played a bit with before:

https://krasjet.com/eidos/cinoma/

The input audio in this case is a feedback loop constructed with two mic/spk pairs.

Whenever I run sine beeps through a home-made spectrum visualizer, and I use the square window or the Tukey window, I get these weird-looking vortices...

They stay the same size and shape regardless of FFT size, but they get bigger with bigger window overlap.

It might be a bug because I don't see it in other software. But it's pretty

Redsea v1.2 can now receive little pictures sent via RDS2 by some radio stations - usually these would be station logos meant for car radios. For now, this feature has very much an "engineer's UI" :)

Here's some logos our contributors have seen so far...

https://github.com/windytan/redsea/releases/tag/v1.2.0

Release Redsea 1.2.0 · windytan/redsea

RDS2 streams + RFT support Redsea can now receive the extra data streams 1, 2, and 3 introduced in RDS 2 (#32, #127, #102) . It is 3–4x more work for the CPU so this feature is not enabled for MPX ...

GitHub