Seeing Through Walls 163
An anonymous reader writes "Researchers at MIT's Lincoln Lab have developed new radar technology that provides real-time video of what's going on behind solid walls. 'The researchers’ device is an unassuming array of antenna arranged into two rows — eight receiving elements on top, 13 transmitting ones below — and some computing equipment, all mounted onto a movable cart. But it has powerful implications for military operations, especially "urban combat situations," says Gregory Charvat, technical staff at Lincoln Lab and the leader of the project.' ... each time the waves hit the wall, the concrete blocks more than 99 percent of them from passing through. And that’s only half the battle: Once the waves bounce off any targets, they must pass back through the wall to reach the radar’s receivers — and again, 99 percent don’t make it. By the time it hits the receivers, the signal is reduced to about 0.0025 percent of its original strength. But according to Charvat, signal loss from the wall is not even the main challenge. "[Signal] amplifiers are cheap," he says. What has been difficult for through-wall radar systems is achieving the speed, resolution and range necessary to be useful in real time (PDF).'"
Re:So what if your standing IN FRONT of the wall? (Score:5, Informative)
Re:So what if your standing IN FRONT of the wall? (Score:5, Informative)
It's non-ionizing radiation so it's about as dangerous as your cellphone. This is an interesting and informative radiation chart https://www.xkcd.com/radiation/ [xkcd.com]
Re:So what if your standing IN FRONT of the wall? (Score:4, Informative)
Re:Pivot That View (Score:5, Informative)
Re:Cool (Score:5, Informative)
It was a linear phased array. It literally can't tell up from down. If you wanted to make it sense in 3-D, you would have to make the array 2-D. Stack a couple of these units, throw in a couple more GPUs for processing, re-tweak the algorithm for an additional dimension, they could probably have a 3D model working in a couple weeks.
The issue is that 3D really doesn't get you much. With the current 2D system, you can tell where someone is in a room, but its not like you can see any identifying features. All 3D would get you is a very rough estimate of height.
Re:What they don't tell you (Score:4, Informative)
It's called coherent integration gain. It's done entirely digitally in a modern radar such as this and can in theory allow you to detect pretty much any signal no matter how weak [there are practical limits of course...] The whole radar they've described probably has a BOM cost of less than $200,000. The real gotcha is labor to make it work, not the material cost. That'll cost millions [probably >$10Million, you could find out if you want to dig through some defense contracts and find the value of this one...] but so did your new iPhone 4S. The difference is that your iPhone 4S is going to have millions made this not so much. If the government wanted to build 100,000 of these, the cost would probably drop to around $50,000....
Here's the idea:
1) You transmit N identical pulses of radar waveform (probably an LFM or NLFM waveform for this application)
2) They bounce off the target and return to the radar
3) You receive them. They are WAY below the noise figure (say 50db). No amount of normal filtering will get them back. You have to analyze the noise for something that isn't "noise" like....
4) You use a matched filter that has a maximum output when the input signal is exactly the LFM you originally sampled to "pulse" compress the signal
5) If you're lucky the matched filter output has gotten you 20-30 db of gain because it's looking on a single pulse basis for the exact signal of interest. That 20-30db gain DOESN'T apply to the noise, because the noise won't match the matched filter [random vs determinisitic], therefore you've gained 20-30db of SNR.
6) Now remember you transmitted N pulses. Why not look for a signal across all of those? That's the next step. For this application they'd probably use Doppler processing. Turns out that if you do this properly you get gain on the desired signal equal to the number of pulses, so if you transmit thousands you can get that remaining 20-30db needed to make the signal >15db SNR which is the usual minimum for reliable detection in thermal noise.
It's really straight forward. The challenges here are not in that part of the design. That part is easy..
The challenges are:
1) Making it realtime (Coherent processing doesn't work when targets lose coherency that happens when they move "too quickly"). This limits the number of pulses you can use to make useable system
2) Dealing with the Dynamic range between the (very) STRONG wall return and the very weak internal targets. [Very expensive ADCs and RF amplifiers can help, they've also apparently added a doppler filtering step in analog which is interesting.... But fundamentally it's a pain]
3) Target classification. The military could care less how many TV and appliances you have. Unfortunately those will show up as targets behind the wall too...
4) Making it small enough and draw a reasonable enough amount of power to be vehicle mounted
===> If you fix #1 with more output power or a larger antenna you run into this problem.....
4) Having enough resolution to actually differentiate 2 separate targets. Without going into the details this becomes problematic for short range radars like this....because you want to see things that are on order 1ft x 1ft.. Radar is much better at seeing Planes and Tanks...