TDOA estimation with USRP and GPSDO

Hi all,

I’ve been working with Cheng C.
(http://gnuradio.4.n7.nabble.com/How-good-can-USRPs-with-internal-GPSDOs-being-synchronized-td43502.html)
on a project where we try to estimate the TDOA (Time-difference of
arrival) of a signal using USRP-N210 equipped with a GPSDO board.

The setting is as follows: the two USRPs are set to start measuring at
some time in the future (time meaning GPS time obtained with the GPSDO).
We transmit a FM signal with a signal generator, and connect the signal
generator to the USRPs using identical-length cables. We then record the
data, and post-process it offline to estimate the TDOA. The TDOA should
be zero, as the cables have identical length.

In practice, we see that the TDOA varies between -100ns and -30ns. This
is within the margin given by GPS precision, so that’s okay. But when we
perform regular measurements (say every 20 seconds), we see that the
TDOA varies from -100ns to -30ns and back in a cyclic pattern. Where is
this regular pattern coming from, and why isn’t it zero-mean?

Any help appreciated!

Thanks,
Francois


Francois Quitin, Ph.D.

Research Fellow
INFINITUS - Infocomm Centre of Excellence
School of EEE, Nanyang Technological University
50 Nanyang Ave, S2-B4b-05
Singapore 639798
Phone: +65-8502-3690
Email: [email protected]