How DEMON works ?

Dr.Sid · 11-30-06, 05:34 AM

Any idea how DEMON works ? I mean in real life .. technically. What kind of signal processing does it use ?
Noise demodulation says pretty nothing to me.

My analysis:

Each blade is source of noise. The amount of noise received will vary as the blade travels the circle. Each blade will generate noise signal with amplitude peaks, mostly 1 distinct maximum per turn.
You get several similar phase-shifted signals summed together. For each blade one.

You roughly know the frequency, so you can demodulate the noise, by watching amplitude of the noise. Standard 'radio' demodulation, consisting off cutting minus half-wave and filtering for low frequency would do. Then you can do normal frequency analysis in 1-100 Hz range.

But this would only give you blade frequency. Not blade count, not turn frequency.
It would be hard to find which peaks come from the same blade. Especially sub's blades would be made very precisely, so there should be no distinct blade signature.

Only thing I can think of is dopler. Blades on one side will make slightly different sound then blades on the other side, because of relative speed. This would vary heavily with listener aspect, but there should be some difference almost always.

So .. any idea ?

11-30-06, 05:34 AM	#1
Dr.Sid The Old Man Join Date: May 2005 Location: Czech Republic Posts: 1,458 Downloads: 6 Uploads: 0	How DEMON works ? Any idea how DEMON works ? I mean in real life .. technically. What kind of signal processing does it use ? Noise demodulation says pretty nothing to me. My analysis: Each blade is source of noise. The amount of noise received will vary as the blade travels the circle. Each blade will generate noise signal with amplitude peaks, mostly 1 distinct maximum per turn. You get several similar phase-shifted signals summed together. For each blade one. You roughly know the frequency, so you can demodulate the noise, by watching amplitude of the noise. Standard 'radio' demodulation, consisting off cutting minus half-wave and filtering for low frequency would do. Then you can do normal frequency analysis in 1-100 Hz range. But this would only give you blade frequency. Not blade count, not turn frequency. It would be hard to find which peaks come from the same blade. Especially sub's blades would be made very precisely, so there should be no distinct blade signature. Only thing I can think of is dopler. Blades on one side will make slightly different sound then blades on the other side, because of relative speed. This would vary heavily with listener aspect, but there should be some difference almost always. So .. any idea ?