> -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] > On Behalf Of Reyk Floeter > Sent: Monday, December 18, 2006 11:22 AM > To: Jacob Yocom-Piatt > Cc: misc@openbsd.org > Subject: Re: wifi signal triangulation > > On Sun, Dec 17, 2006 at 12:09:12PM -0600, Jacob Yocom-Piatt wrote: > > only today have i tried out hostapd, it is quite neat. > while adding a 2nd AP to > > my network a thought occurred to me: if you had >3 APs that > were sufficiently > > spread out and had tightly synced clocks you could likely > triangulate the source > > of a wifi signal with a fair deal of accuracy. > > > > is this doable? > > > > yes > > but it needs some heavy math ;). you can get some results by using the > signal strength, but it is probably better if you also use the round > trip time and some low level information.
I'm curious about this, especially about the final triangulation resolution. The wifi signal propagates at the speed of light, 300k km/s, so to get a (relatively poor) distance resolution of 1 km, one would need to be able to reliably clock times smaller than (1 km) / (300k km/s) = 3 * 10^-6 s, or in other words, less than three microseconds. GSM does something similar - since GSM is using TDMA, the signal from a mobile terminal have to reach the base station during a specific timeframe slot. On the mobile terminal there is a parameter called TA (for Timing Advance) that shows the timing correction factor because of the distance to the BTS, and if I recall correctly, it is possible to get a 250m resolution out of TA. But GSM hardware is probably more suitable for this than regular PC hardware. > > once we implemented it with hostapd, a sql patch (to allow the central > hostapd sensor to log into a postgresql database), some gps > coordinates, and a hacked psql script to directly query the > triangulated results from the database. a guy from the ccc implemented > a php frontend to draw the station coodinates on an area map, but i > would prefer an implementation using svg and firefox without the need > of a server-side scripting language now ;). Do you happen to have a screen capture of the result? > > unfortunately, our code got lost after the experiment, but i may still > find the hostapdsql diff. > > reyk > Mitja
