Forum: GNU Radio Passive radar using my hacked dual coherent channel rtl_sdr dongle

Announcement (2017-05-07): is now read-only since I unfortunately do not have the time to support and maintain the forum any more. Please see and for other Rails- und Ruby-related community platforms.
Juha Vierinen (Guest)
on 2013-09-26 00:24
(Received via mailing list)
Attachment: passive-000001.png (100 KB)
I hooked up my dual coherent channel rtl_sdr dongle into two of our
radar antennas that we have here (log-periodic antennas pointing North
South). After fine tuning gains and signal levels, I acquired some FM
data (WWLI 105.1 MHz). To my surprise, I got passive radar echoes from
airplanes when I analyzed the data. Passive radar is one of the most
challenging types of radar because the requirements for dynamic range
larger than those for eg., pulsed radar, I really didn't expect much
my test.

I expect the $16 system to also be able to see specular meteor echoes
(typically stronger than airplanes) and ionospheric field aligned
irregularity echoes. Not bad for something that required about
investment in a digital receivers ten years ago, and a $2k investment

Visually, the rtl_sdr dongle system has a little bit more noise than the
usrp n200 with a tvrx2. And also, there is the hassle of aligning
So I would still use the usrp if given a choice, as it does perform
and requires no hacking.

Here is the passive radar video I just made using my $16 dual channel
rtl_sdr (the timestamp is broken, but I just did the measurement an hour

For comparison, here is an example passive radar video of airplanes and
specular meteor echoes done using a usrp n200 + tvrx2. This was earlier,
the targets aren't the same:

Does anyone know who manufactures these dongles, or who designed it?

PS. I ordered a seven port usb hub, in anticipation of adding more
channels (to do interferometry and maybe imaging). It seems that every
I browse ebay, I end up ordering a few dongles.

M Dammer (Guest)
on 2013-09-26 15:01
(Received via mailing list)
This is awesome. How did you analyze the data ? Is the software/source
(both grc setup and data analysis) available ?
Juha Vierinen (Guest)
on 2013-09-27 03:22
(Received via mailing list)

I'm just using a simple script to dump data to disk. I then have another
program that analyzes the files. It is not too complicated. The
equation is a convolution with the assumption that the target is
over ~20 ms and that the ground clutter is stationary over ~1 second.

My code is a mess at the moment and it depends on other unpublished
libraries that I have written, which are also a mess.  I might release
once I have cleaned it up a bit.

Jared Clements (Guest)
on 2013-09-27 05:28
(Received via mailing list)
Since they're clocked together I would assume that your alignment would
consist of interpolation on the primary FM signal, what do you do to
it later?  If your code is too messy to release can you share a block


Marcus D. Leech (Guest)
on 2013-09-27 20:17
(Received via mailing list)
On 09/26/2013 11:27 PM, Jared Clements wrote:
I just re-did my tests with a pair of synched dongles, and instead of
inject RF externally, I just tuned it close to the 2nd harmonic of the

That produced quite stable, near-perfect, correlation, once I adjusted
the phase to peak up the correlator output.  This puzzles me greatly.

This implies that both dongles are properly phase-locked to the master
clock, and that the leakage  is being coherently downconverted, and
   that no samples are being dropped.

So, my phase-coherence issues are with my external RF injection.  Which
I just plain don't understand....

Marcus Leech
Principal Investigator
Shirleys Bay Radio Astronomy Consortium
Juha Vierinen (Guest)
on 2013-09-27 21:50
(Received via mailing list)

I wrote up a few blog posts about the tests I did this week:

PS. I tried adding a third dongle to run with the same master clock. I
didn't do it correctly this time either by using an active signal
as that would increase the total cost :). I'll hopefully have some more
report next week.

This topic is locked and can not be replied to.