Traditional sonar and radar use an echo-based approach, and have been amazingly successful in sensing and tracking vehicles and objects under water, through the air, or even in space. The principle is simple, even if successful execution is not: send out a known signal in a given direction (a ping), and “listen” for the return. The time it takes for that echo to return in a medium with known propagation delay gives the distance, while its magnitude gives some indication of the size and shape of the target.
There's one big problem with active radar/sonar: sending out that initial ping tells everyone in the area that you are there and where you are. That's why there is so much interest in passive radar/sonar. In this approach, the system is a good listener but only that, as detailed in an interesting update/assessment article in Military & Aerospace Electronics , “New frontiers in passive radar and sonar.”
In a passive system, the system using one of more sensors to try to capture all the sounds (sonar) or EM signal (radar) it can as they reflect off objects in the area, and then tries to make sense of it all. The sources for the radar signals can be known and unknown nearby commercial broadcast transmitters, base stations, other vehicles in the area, satellites, or almost anything transmitting; for sonar, there are many sound sources in the ocean (ships and whales, for example). Passive systems are not new, as basic passive towed-sonar arrays have been used for many years to listen for nearby submarines; what has changed is their increased deployment and the much-higher expectations placed on them.
In a basic passive radar system, the received extracts useful information from reflections whose energy is provided by nearby “transmitters of opportunity” (from the Adelaide Radar Research Centre at The University of Adelaide, Australia)
The interest in passive approach calls for a radical shift in the analog aspects of a radar/sonar system and a major “step up” in received signal analysis. The reason is that in a conventional design, the signal being sent in fully known (although it may have a complex shape or chirp) and therefore the general shape of the echo is known, although it is distorted by the reflection and the medium. It's what signal-theory academics call “partially known signal in partially known channel.” Instead, we have unknown signals in a partially or poorly known channel, especially as there is no way to do test pings to calibrate the channel. [For an equation-laded analysis of this situation, start with the classic text by H.L. Van Trees, “Detection, Estimation and Modulation Theory, Part I: Detection, Estimation, and Filtering Theory.”]
Passive systems need to develop a more-detailed sense of their surrounding environment, so they require many more sensors than active systems; “more sensors” also means more analog signal conditioning, front ends, and filters. In fact, filtering in passive systems is much more difficult since so much less is known about the received signals compared to the active approach. It's the opposite of the situation with LIDAR, a time-of-flight (ToF) optical image-sensing technique in widespread use (such as in autonomous vehicles), where a single light flash illuminates the surroundings to create a 3-D picture (LIDAR gives sight to autonomous vehicles).
Of course, all those passive sensors won't be of much good unless there's a way to make use of the flood of data they produce. Passive system needs a huge amount of processing power in the digital domain to even begin to create a picture of what's around them. It's hard enough when the source and targets are not moving, but becomes much more difficult when one or both are in motion. The signal-processing algorithms must have a strong emphasis on detecting correlations across channels, in addition to finding subtle changes within a given channel. (Interestingly, there's also significant overlap between passive radar/sonar algorithms and those of MIMO wireless systems.)
The move to passive radar/sonar means that the thinking about signal generation and sensing which has prevailed for decades must be shifted. The number of sensors increases, the analog front end has to deal with more unknowns, and the signal processing demands are much, much higher. The irony is that we normally associate “passive” with little or no power dissipation, yet this “passive” approach actually increases the dissipation needs significantly (especially if you discount the high-power, low-duty cycle transmit ping signal which active systems use but passive ones don't have).
Have you done any work on passive sonar or radar? What aspects of it gave you the largest analog signal challenges?