Slashdot Mirror
Radar/Wireless Transmitter on a Chip
dganapa writes "Researchers at the California Institute of Technology, headed by Dr. Ali Hajimiri, have developed a low-cost radar system on a silicon chip. The entire system has been designed from the ground up on silicon, thus leading to reduced cost as well as robustness in response to design variations and changes in environment. The chip runs at a staggering speed of24 GHz (enabling it to transfer data as fast as the main network of the Internet) and can soon lift wireless, high-frequency communication to a whole new level. The radar as such is not as powerful as a conventional radar but because of its cost-effectiveness, a number of them can be coupled together to perform really well. A related NY Times article is here. A recent article from Slashdot shows that radar technology is increasingly being implemented in the automobile industry. This current chip is sure to be much more successful than its predecessors as far as the automobile industry is concerned, but whether or not its processing speed will become important in the computer industry remains to be seen."
* The chip could serve as the brains inside a robot capable of vacuuming your house. While such appliances now exist, a vacuum using Hajimiri's chip as its brain would clean without constantly bumping into everything, have the sense to stay out of your way, and never suck up the family cat.
Not really. The radar might reflect off the cat or your leg, but would pass right through wooden furniture and walls. A radar-equipped vacuum cleaner would still bump into stuff.
* A chip the size of a thumbnail could be placed on the roof of your house, replacing the bulky satellite dish or the cable connections for your DSL. Your picture could be sharper, and your downloads lightning fast.
Wrong on size. Satellite dishes are big to both help collect enough RF energy to get a clean signal and to pinpoint on a single satellite. Without the needed collecting area and beam-forming span of the antenna, the signal would be weak and overlaid with signals from other satellites in orbit.
Two wrongs don't make a right, but three lefts do.
Is the frequency band at 24 GHz actually licensed for automotive radar systems?
According to this press release it's not licensed in parts of Europe.
And in the US, there is only a temporary license.
I haven't found an unbiased summary yet - the referenced press release is from a working group of companies in the automotive industry.
This summary says that the frequence is reserved for radio astronomy and similar users.
A related NY Times article is here.
"You should never doubt what nobody is sure about." -- Willy Wonka
Just don't be surprised one the police finds a way to screw you over for a few more bucks by using passive radar to determine your speed. If you dont want the police to charge you for speeding, stop going over the limit
Lets see the most common American radar bands..
a nds.htm
X Band operates on ~10.5Ghz
K Band operates on ~22.4Ghz
Ka Band operates on ~34-35Ghz
(source: http://www.snooper-uk.com/radar_laser_speedtrap_b
The article states the frequency being used of is 24Ghz, so the only possibly problem might be with K band detectors.
I dont think they would put both in the same band anyway.. wouldn't that interfere with the radar guns themselves?
I'm not completely sure, but this site says that BellSouth's backbone could download the library of congress in 126 seconds- so it's gotta be pretty fast.
I found the NYTimes article dumbed things down a little too much. Basically, this is a press release by a fairly young professor about a ISSCC paper to be presented next week.
CMOS is getting fast enough (could be SiGe BiCMOS chip but probably CMOS) to allow for amplifiers and ADC (analog-to-digital) that work in the radar (~25GHz on up) range & also allows for million gate DSPs and digital logic on the same chip. The analog front-end is running around 24GHz which gives a 1/4 wavelength around 3mm (antennas are implemented as PCB traces off-chip). This is an analog GHz signal where the transistors are amplifying a tiny GHz signal using analog amplifiers. Digital clock speeds are completely different. Digital is like switching completely from off to on (ie. 0 to 5V -- in reality try 2V or 3.3V). This is like a uV signal being amplified to be later converted to a digital signal with a more reasonable bandwidth that a digital CPU could handle (like your overclocked Pentium).
The parallel analog antennas & blocks which allows for parallel ADC of 8 channels.. 8 parallel radar antennas. By using parallel processing you can use the information gained by the other channels to improve your ADC or have each channel only need to work at 1/8 of the total speed. Also, having 8 antennas allows phased arrays where you can control the beam and allows you to scan the beam or block out other signals (much like cell towers can focus in on one cell signal, and why your 802.11 router has two antennas). So, depending on how much bandwidth the ADCs need & how fast the DSP is running is really the 'digital' GHz part of the chip. So the digital processing is probably a more reasonable 100's of MHz (though hard to compare DSP speed to CPU speed). The processed digital waveform can be sent high-speed off chip, or to on-chip CPU to be used to disable your cruise-control and hit the brakes for you.
Why do you care? Well by using straight CMOS the radar system can be made on one chip and not need 'exotic' GaAs/SiGe/InP (BJTs of traditional radar systems) and when the automotive chips get down to sub-$5 they will show up in every car. Also doing it this way, much smaller power is involved and you don't need circuits that look like your microwave oven waveguides.
Well, looks like your math is right. But the resoultion of a radar is mainly determined by its bandwidth, not the carrier frequency. i.e. Shorter pulse = larger bandwidth = higher bandwidth.
Funny. The article I read says the antennas are on the PCB. What's new is they used Si, instead of more exotic materials.
Both Cadillac and Jaguar sell vehicles with Radar-based Adaptive Cruise Control, which will brake for you if needed.
Check out the Cadillac XLR.
Don't know; Don't care; Don't ask
As someone who volunteers at his car club's high-speed driver education events and has attended one of the events as a student- um, no.
First, braking is NOT always the best choice. When you're doing 60 and a moose jumps out in front of you, you STEER, not BRAKE. Why? Because under about 200 feet, you're never going to stop in time but you probably can change lanes. Simple physics tell you why- it's a lot easier to accelerate a car enough to move 10 feet to the side than it is to bring the whole thing to a stop.
Second, when said moose jumps out in front of you, steering while braking is exactly what causes many accidents, because you unbalance the car, shift a huge amount of weight to one corner tire, which becomes drastically deformed under the weight and becomes nearly useless; meanwhile, there's next to no weight on any of the other tires, and they're useless too. Your tires have what is called a "friction circle"; draw an X-Y axis, now a circle centered. That describes how much acceleration your tire can accomplish in any one direction. Notice that there's less of any one particular axis when you're doing both? Your tires always stop better when you're not trying to steer, and vise-versa. Both controls should ALWAYS remain under control of the driver so the system doesn't try to do something while you're doing something else.
Third, proper driver education is a lot cheaper(just one $200-300 event, depending on the club, will teach you quite a bit about how to handle your car properly) in the long run.
Your friends who have been in accidents need to analyze WHY they got into the accidents they did. I'm guessing an automatic braking system would not have "fixed" any of this, but better attentiveness, good judgment, and proper knowledge of how to handle their car would have.
Please help metamoderate.
This will be vitally important to the development of consumer robotics devices and the mitiuraization of existing devices. One of the big problems now with small robots is that they have limited choices for environmental perception. Ultrasound has a limited range and can easily be interfered with, Infared has the same limitations, and optics which is the ideal solution requires a large amount of processing for shape recognition. Ussing radar, longer range, lower interference sensing devices can/will be incorportated into robotic devices. Also, this is a huge boon for small remote controlled autonomous air vehicales, as it will allow them to have the sensing abilities of their larger cousins such as the predator. Whoo hoo!
It is indeed possible for a device to generate radio frequency with a wavelength greater than the devices physical size. A typical AM station generates a signal at a wavelength 600 meters to about 200 meters. Most AM stations do not have antennas this long and their transmitter boxes certainly aren't this size.
Frequencies can theorectically be generated with any size circuitry. Im pretty sure this circuit does so using the so called Phase Locked Loop (PLL) circuit, possibly mixing 2 or more together to get the very very very very high frequency by addition. This circuit does not require wire coils (often of relatively large size) to resonate at these "really" high frequencies. There would need to be a filtering step (or two or three) and I can see how this circuit would be hard to miniturize, but I guess they have done it!
Typically for a radio signal to be radiated you need at least a half wavelength antenna but even this can be cheated at. In the microwave region where this device is working at, signals are best radiated using a "dish" type antenna. This chip no doubt does not come with this dish. It simply generates the rf at 24 GHz.
ZBeat
What other people think of me is none of my business
ONE Of these chips can. It's a phased array (aka synthetic aperature) radar on a single chip.
Probably. Researchers were doing this with hydrophones picking up background noise 5 years ago (sorry, I don't have a link, I read it in Scientific American)
It's all about the power with radar. So, it's unlikely that a chip that actually manages to get in the right bandwidth to work as radar with the available power it has available is going to have much output below that optimum. I would bet they are using whatever frequency that sits on top of the bell curve, and are happy to have it.
...
Transmit; listen; figure out the difference between what you heard and what you should have heard if it went on indefinitely (ie no relfection). Repeat, very quickly.
The listening part is already at it's limit as to finding small reflections, though, they're already a very, very small fraction of your transmitting power. That's where all the computing is taking place, where you put the software resources.
You mentioned harmonics; I think you misunderstand them a bit. They don't go both ways from the original frequency. You must listen at the same frequency as you transmit; if you listen at a harmonic above that frequency you might hear something at a much reduced level; if you listen at anything below your fundamental frequency (the transmit one) you hear
Nothing.
There is no such thing as a harmonic below the fundamental.
Lowering the frequency in the transmitter means you need more power and it probably won't fit on a chip. If somehow you did try it with an existing chip, all that happens is you get even smaller levels of bounced signal power (you're transmitting less level because you're below the transmitter's optimum) which means even more difficulty listening for reflections.
By the way, diffusing reflections and therefore making an even smaller percentage of them bounce back to the reciever pretty much sums up the whole working theory behind stealth. If you think about it, each attempt to reduce your available reflected signal numbers and strength is like building stealth into your radar. That's like deliberately building bugs into your debugger.
Since virtually all (1) radar systems that can see more than a few metres still require large vacuum tube transmitters to work at all (power, power, power), I'd say this chip is pretty much state-of-the-art and I'd bet they're doing all they can with what we know how to make and what we physically can make right now.
(1) I'd say all (period) but I'm not privy to everything and governments do keep secrets. Perhaps some automotive-types that watch 10 feet for a parked car might be solid state, but so far as the ones I know about, they all still use a small transmitting tube. Solid State transmitters are coming; but this story is really about a breakthrough in making a SS radar at all.
First off, to scan the image, the transceiver's antenna would have to be scanned - physically moved around - at the same speed as the desired refresh rate of the image.
At the bottom of the article he mentions that these things are used as a phased array. You don't have to mechanically scan the antenna. This allows you very rapid switching. As for rates, lets say that you want a 1k by 1k image at 10 Hz. That's a pixel rate of 10 MHz, or 100 ns dwell time per pixel. That gives you a 50-ft range. Probably not sufficient, but close. You could very likely have multiple T/R beams simultaneously, as long as you are clever about sidelobe supression.
That being said, you are right about issues with gain and calibration. It's unclear from the post what frequency this thing works at, but they did mention 24 GHz. If that's the radar frequency, then won't you have some significant reflections off of people? This is really mm-wave radar...
By the way, I'd tought that passive imaging mm sensors were the next big thing; there is work on them for aircraft, I know. Maybe you could adapt this silicaon technology to that and make mega-pixel imaging mm arrays..?
Human genome = 3 billion base pairs = 6 GBit. Windows + Office = 20 Gbit. Which is more impressive?
In the microwave region where this device is working at, signals are best radiated using a "dish" type antenna. This chip no doubt does not come with this dish.
The article specifically states that the chip implements a phased-array of antennas. And that those antennas are actually physically on the chip itself.
This is one reason why this solution will be cheap to implement - it does ALL the RF work for you, you simply connect "a computer" to the resultant datastream and interpret it how you like.
Visit CryptoGnome in his home.