snipped-for-privacy@manx.misty.com (Don Klipstein) wrote in news: snipped-for-privacy@manx.misty.com:
Thankyou. I can see what I was missing now. The beam of a CW laser diverges more, fades over distance more than the sun's light does, and I wasn't taking that into account properly.
I don't know much about extracting signals from noise, but I tried it with sounds. 1KHz pulsed at 1Hz with 50% on-time, against a background of white noise. I could hear it down to -30 dB, but only down to -24dB reliably, which is lower than your -17 dB, so that should work. Shorter pulses would make it harder though. One thing I noticed was that it didn't matter if I filtered the mix or not, it had no effect on my ability to pick out the sine wave, it only made the test a little more comfortable to listen to. :) If anything though, the wideband background provided a better anchor for my perception of the narrowband one than the filtered mix did. Whether that effect would be the same for a mechanical monitor I don't know. It might have been due to a poor quality filter too.
What this leads me to ask is: could it be better to convert light signals to sounds or other representations to let human perception do the filtering, to take advantage of perception we haven't learned to model? Could that work better than doing it entirely by mechanism? I know that this must happen anyway with all measurements, but my point is that science tends to refine as much as possible before human interpretation is allowed, and this might not be the best way. A human can pick out the tune and the harmony in music, even in an instrument low in the sound mix. This is enhanced dramatically in a a two-channel stereo mix. No-one's modelled that and made a machine do it, the results so far have been a joke, so if human perception is allowed to have a greater share of detection, we might get better results.