Timeline for Ultrasonic signal detection
Current License: CC BY-SA 3.0
17 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 21, 2014 at 16:33 | comment | added | Tarin Ziyaee | @endolith Right exactly. Think of it as just a series of trial and errors, where we constantly check the score by cycling through various delays and frequencies until we get a peak. Granted it is somewhat brute force, but no reason not to use it - application dependent of course. :-) | |
| May 21, 2014 at 16:10 | comment | added | endolith | @user4619 Ah, I didn't realize you could so easily use the ambiguity function as part of a receiver, I thought it was just for analyzing the waveforms and needed to resample for the Doppler axis. | |
| May 17, 2014 at 16:51 | comment | added | Tarin Ziyaee | @endolith As Jim Clay described his method, he is basically computing what is known as the Ambiguity Function. That is, cross-corr results, with the second dimension corresponding to base frequency. This will then reveal the peak, and hence, since we know the original frequency, its doppler degree. | |
| May 17, 2014 at 4:04 | comment | added | Jim Clay | @endolith Yes, doppler is a problem. I handle that by correlating multiple times, shifting the frequency of the received signal each time by a different amount. This is easy to do if you are correlating in the frequency domain. | |
| May 16, 2014 at 17:56 | comment | added | endolith | Your sizes imply a MLS, right? Generalized PN sequences are not limited to 2^n-1 sizes? | |
| May 16, 2014 at 17:19 | comment | added | Tarin Ziyaee | @endolith Also, the degree of doppler matters of course. In light doppler carrier recovery can work to undo the effect. For heavier doppler cases, other methods can be used to compensate. Certainly the degree of doppler has to be determined first. | |
| May 16, 2014 at 17:09 | comment | added | Tarin Ziyaee | @endolith Frequency offsets using PN-codes will manifest themselves as a degradations of the cross-correlation peaks upon reception, VS affecting peak positions, like what linear chirps would do. Thus for TDOA purposes, the relative peak positions should not be affected. Eitherway, carrier-recovery methods can be applied to offset those effects. | |
| May 16, 2014 at 16:53 | comment | added | endolith | These are not Doppler-tolerant, though, are they? So when the phone is moving, the cross-correlation will disappear. (I've been thinking about the same problem lately for measuring path from speakers to microphones without making audible noise.) | |
| May 16, 2014 at 14:54 | comment | added | Tarin Ziyaee | @JimClay Glad to hear that. Im curious to see what the received xcorrs/signals look like now. That's great though. | |
| May 16, 2014 at 12:35 | comment | added | Jim Clay | I need to do some more testing, but it appears that a longer pn sequence will work. | |
| May 16, 2014 at 12:35 | history | bounty ended | Jim Clay | ||
| May 16, 2014 at 12:35 | vote | accept | Jim Clay | ||
| May 13, 2014 at 20:36 | comment | added | Tarin Ziyaee | @JimClay Excellent. Yes, for an application like this, PN-sequences are your friend. And also yes, use more than 1-bit if you can. Definitely post your results! (Please use @ to ping me so I know when you have something new). | |
| May 13, 2014 at 20:29 | comment | added | Jim Clay | I've decided to go a little overkill on the signal length by using an order 10 lfsr (1023 chips) to either prove or rule out that the problem is solvable by lengthening the signal. I'll post what happens. | |
| May 13, 2014 at 20:28 | comment | added | Jim Clay | Yes, I plan on trying longer sequences. I did not know that the circular autocorrelations of pn sequences were so nice- interesting. Unfortunately for my application it is the linear autocorrelation that matters. Regarding the preamble- the entire sequence is, in a way, a "preamble", in the sense that what makes a preamble useful is that it is a known data pattern. My entire signal is known a priori. | |
| May 13, 2014 at 18:05 | history | edited | Tarin Ziyaee | CC BY-SA 3.0 |
added 1814 characters in body
|
| May 13, 2014 at 15:20 | history | answered | Tarin Ziyaee | CC BY-SA 3.0 |