Hi all! This topic supposed to be focused on waterfall design, which was started at Gr-satnogs new waterfall testing
I was trying to find any design docs or discussions re waterfall, but found nothing. Can we recap it here?
- Why do we need waterfall? Currently I see several issues with that:
- raspberry pi spends some CPU/RAM doing this
- waterfall could be created from audio uploaded to the server. Moreover it could be created on-demand on server.
- waterfall created based on original audio not the audio which will be used for decoding data. Am I missing something here?
.png? Currently it gives ~ 56Gb overhead on storage.
The main reason is the excellent debug capabilities it provides. Was there a satellite transmission in there? Terrestrial noise? Drifted satellite or old TLEs? See our docs for identifying issues with waterfalls.
This is true now, but @surligas is working on a much less cpu intensive way to do it.
This cannot be done as our audio filters are most of the times around 5kHz and the waterfall is representing 50kHz bandwidth around the center frequency.[quote=“dernasherbrezon, post:1, topic:1500”]
waterfall created based on original audio not the audio which will be used for decoding data. Am I missing something here?
Waterfalls are created by decimated and filtered IQ. Not by audio or any other decoding.[quote=“dernasherbrezon, post:1, topic:1500”]
Why .png? Currently it gives ~ 56Gb overhead on storage.
That is a good question. We could explore jpg. Although 56Gb is not such an overhead compared to almost 1Tb of audio we are currently seeing
Thank you @pierros! Now it makes more sense.
The only question which I have now is the image size: currently fft size is 1024, while image width is 800.Which means adjacent pixels will be merged together. Am I right?
This depends on your hardware settings. Some SDR we run by default on 1024 some others on different settings. 800px is just a compromise (that we can revisit) for display purposes.