My friend Derek and I recently built a passive radar system as a university design project. It turned out to be a super fun and challenging project, so I am writing a series of blog posts covering the design of our system in case others find them interesting. In this initial post I’ll cover the basics of how passive radar works and go over some of our results. I’ll add links to later posts at the bottom of the page. The GitHub repo for the processing software is here.
Firstly, what does it mean for a radar system to be passive? Conventional radar systems are ‘active’ – that is, they have a dedicated transmitter as well as a receiver. Radio signals are emitted from the transmitter which then bounce off objects in the environment before being detected back at the receiver. These echoes can be analyzed to determine various properties of the targets, such as their location and velocity. Passive radars are different from active radars in that they don’t have a transmitter of their own. Instead, they take advantage of existing signals in the environment such as radio or TV broadcast.
A simple passive radar system has two separate receiver channels. The first is called the reference channel, which directly receives the illuminating signal coming from the transmitter. The second is the observation channel, which detects echoes of the illuminating signal that have bounced off radar radar targets. An idealized version of this process is shown in the animation below.
Notice that the target echo seen by the observation channel is delayed with respect to the reference signal because of the extra distance it has travelled. By measuring this delay we can determine how far away the target is. Additionally, the echo signal acquires a Doppler shift which is related to the speed of the target. By determining the frequency shift between the reference signal and the echo signal we can infer the target’s velocity.
Passive radars come in many different varieties. For example, systems using VHF or UHF band illumination such as FM radio or digital TV broadcast are commonly used to track airplanes. VHF passive radar has also been used to study the upper atmosphere, allowing the detection of meteor trails and other ionospheric disturbances. Additionally, WiFi and cellular LTE based passive radars have been developed to track targets at close range. Passive radars have even been implemented using signals from transmitters onboard satellites.
We used FM radio broadcast as an illuminating signal since it has high transmit power and is available almost everywhere. A radar’s range resolution is proportional to the bandwidth of the illuminating signal, so we looked for the local FM radio stations with the widest bandwidths. Since it is an analog transmission, the bandwidth of an FM radio channel depends on the content that is being transmitted. For example, a channel transmitting noisy rock music will have a wider bandwidth on average than a talk show channel (although all channels must remain inside their allotted 200kHz-wide band). We found three adjacent music channels that fit within the 2.4MHz bandwidth of our receiver, allowing us to digitize them all at the same time.
The next step was to go out and record some data. After a few unsuccessful attempts we finally obtained some good results. The key was to select a location that provided an uninterrupted line of sight in the direction of the targets but provided some shielding from the direct path signal emitted by the radio tower. This was especially important since we were located relatively close to the broadcast towers (about 7km away), so the strong direct path signal could easily overpower the target echoes we were trying to measure. We also found that it was important to optimize the orientation of the observation channel antenna and the receiver gain.
The video below shows a passive radar recording of some planes flying around the Halifax area. It is sped up by a factor of 30 (so the whole video is about 10 minutes of data.)
We used a flight tracking app to obtain the trajectories of planes in the area while we were recording the passive radar data. One such trajectory is overlain on the map below, showing the path of a Boeing 787-9 Dreamliner that was on its way from from Cancun to Istanbul. With knowledge of the transmitter and receiver locations as well as the plane’s trajectory, altitude, and speed, we did some geometry to derive the theoretical range and Doppler signatures for this plane.
Comparing the experimental results to the predicted results from the plane trajectory, there is a very good match! Due to the nature of the FM signal that was used, the Doppler resolution is more precise than the range resolution.
Note that we could not run this calculation in reverse – there is not enough information contained in the range and Doppler shift values to fully specify a target’s position and trajectory. However, the range values are enough to constrain the target’s position to an ellipse where the two foci are at the transmitter and receiver locations. The target can be further localized by adding additional receiver channels that allow angle-of-arrival information to be calculated.
I extracted the experimental passive radar data shown above manually, which was very annoying since I had to go through all the frames and click on the target location for each one. In the future I’d like to do target tracking automatically using a Kalman filter or something similar.
If you’d like to learn more about passive radar, make sure to check out the blog posts linked below where I go over the details of each part of our system: