Measuring distance from an RF beacon

[taylaron]

Greetings, PF'ers

I've got a project where I'm trying to measure the distance from a stationary AM radio beacon transmitting a short coded pulse every 300ms or so to a non-stationary receiver. I'm dealing with a range of about 1 to 500 ft between points. I need to have at least 1'' accuracy. While the transmitter is easy to build, the receiver technology is a bit out of my expertise at the moment.

I expect I will need to purchase or fabricate an antenna to match my desired transmitter frequency, and then have the on-board electronics on the receiver measure the amplitude of the beacon pulse to calculate the distance. Knowing that the magnitude of the field drops with the square of the distance, the it should be a relatively simple process to take the voltage of the incoming signal and run it through a simple algorithm to calculate the distance, right? But how should I measure the voltage?

overcoming the signal reflection from nearby objects may be overcome by alternating frequencies between pulses. right? By the time the next pulse comes, the last pulse on that frequency should have dissipated sufficiently.

I have basic knowledge of radio technology and minimal knowledge of antenna theory. I will be using an Arduino Uno for the receiver microcontroller.

Thanks,
Taylaron

 

[berkeman]

Greetings, PF'ers

I've got a project where I'm trying to measure the distance from a stationary AM radio beacon transmitting a short coded pulse every 300ms or so to a non-stationary receiver. I'm dealing with a range of about 1 to 500 ft between points. I need to have at least 1'' accuracy. While the transmitter is easy to build, the receiver technology is a bit out of my expertise at the moment.

I expect I will need to purchase or fabricate an antenna to match my desired transmitter frequency, and then have the on-board electronics on the receiver measure the amplitude of the beacon pulse to calculate the distance. Knowing that the magnitude of the field drops with the square of the distance, the it should be a relatively simple process to take the voltage of the incoming signal and run it through a simple algorithm to calculate the distance, right? But how should I measure the voltage?

overcoming the signal reflection from nearby objects may be overcome by alternating frequencies between pulses. right? By the time the next pulse comes, the last pulse on that frequency should have dissipated sufficiently.

I have basic knowledge of radio technology and minimal knowledge of antenna theory. I will be using an Arduino Uno for the receiver microcontroller.

Thanks,
Taylaron

You have outlined a very challenging task!

Be sure that your transmitter is legal -- in the right bands, with low enough power and no splatter out of band.

And multi-path reflections do indeed become a problem -- even reflections from people can be a problem in some bands.

Are there any other modalities instead of RF that you can use? For example, use laser rangefinding or similar?

 

[taylaron]

Hey Berkeman, thanks for the reply,
I've got a robot on tracks going on my lawn that needs to know its precise location (within an inch). there would be 3 beacons and probably 3 antennas on the robot to triangulate the position.
Originally, I considered DGPS, but they cost about 10k for my requirements, laser range finder is out of the question due to line of sight issues and range of motion, differential image analysis is possible, but I don't know how much error accumulates over long distances. Wheel encoders are not accurate enough over very long distances. RFID tags on the lawn are unattractive, a hassle and the multiple sensors on the robot to detect where the tag passes under the robot costs about $50 a pop (per detector)
While I might rely on wheel encoders and a simple image analysis from a downward looking camera for short distance calculations, none of the above seem reliable and accurate enough for an overall solution. That is where RF comes in.

I like the radio beacons because they're relatively simple (seemingly), cheap (seemingly), and capable of long distances without much modification (while considering legality) and they can pass through objects fairly easily.

For the antenna, I expect to use a coil of tiny magnet wire with many, many loops because the more loops I have, the higher signal strength I get. Correct? Integer multiples of the wavelength/2 of course.

What are your thoughts? It looks like your a Ham Radio operator. I would really appreciate some advice.

Kind regards,
-Tay

 

[berkeman]

Hi Tay,

From my perspective, RF is not going to be able to meet your initial specs for the system. RF is just not reliable for finding distance in a general setting. Multipath issues are significant, and your 1" accuracy specification would be hard to meet even with no multipath, IMO.

In addition to my HAM radio background, I also was involved in an R&D investigation project looking at adding RF networking to my company's other control networking technologies (twisted pair networks, powerline networks, etc.). We did an extensive investigation into existing standards like BlueTooth and Zigbee, and into other potential RF transceiver options. In the end, we concluded that RF technology (with its current limitations and regulatory issues) was not reliable for control networking. We published the following whitepaper which outlines some of the problems for networking applications, but many of those problem would also apply to your proposed application:

http://www.google.com/url?q=http://...ds-cse&usg=AFQjCNEUnU-zi8ey2oy8f99I_uClJOUfoA

So instead, it sounds like a track-based system is your best and most reliable approach. You might put marks every 1/2" or so, with an absolute barcode number every foot or so. That way your robot would only need to go at most 6" if it ever got confused about its absolute position on the track. Or you could use a combination of little magnets in the track system, plus a barcode every so often to refresh/check the absolute position...

 

[berkeman]

BTW, I would use two different optical sensors for the single marks and the barcodes. Maybe reflective optical sensors that are aimed sideways on the insides of the track, to keep them out of reflected sunlight. They should also probably be modulated (like an IR remote control) and demodulated/filtered to give you better Signal/Noise ratio in the measurement. You can use a single optical sensor for the barcode, since you would have the robot just move over the barcode when it was reading it (so you don't need a CCD image of the barcode).

 

[skeptic2]

I concur with Berkeman, RF just isn't that precise. You're asking for 1 part in 6000 accuracy.

First of all 1/d^2 applies only to free space propagation. Typical land based propagation is closer to 1/d^3.5 to 1/d^4.

Second, everything nearby made of metal will be a source of multipath. Once I was testing the range of low power transmitters in an open field and discovered a three strand barbed wire fence, 50 ft. away was causing so much multipath interference I couldn't receive any of the transmitters at the multipath point, at well below maximum range.

By the way, the company I work for had the first DGPS transmitter in the US ( I'm told) and it was used to help guide the driverless Stanford U. car up Pikes Peak.
http://www.gpsworld.com/Transportation/Road/handling-limits-10284

 

[taylaron]

Thank you for your input everyone. I appreciate it.
Because RF simply travels too quickly for my required accuracy, why not turn back to ultrasonic beacons?

The US beacons would be equipped with a simiple radio that receive a signal for all of the beacons to start chirping. The robot, knowing when they were supposed to start chirping would listen for the chirps and measure the time it took for the chirps to reach the robot and calculate the distance based on the speed of sound.

Multi-path and line-of-sight would be an even more difficult problem with ultrasonics, but I can utilize a relatively infrequent rate of chirps to let the previous chirp dissapate into space as not to interfere with the next chirp's detection. After all, the robot is only interested in the first chirp it hears (from each beacon). The shortest distance between two points is a straight line... All subsequent chirps between the expected interval would then be assumed as the result of multi-path.

The robot can determine if a signal is from a beacon that is out of sight by measuring the continuity of the chirps it receives and the expected speed at which the robot is traveling.

line of sight can be overcome by using multiple beacons in areas where objects prevent line-of-sight interactions. The robot would still receive the chirps from non-line-of-sight beacons and when it calculates its distance from the beacon and the conflicting rates at which it calculates compared to the other beacons indicates its out of sight. The object causing the multi-path will likely come into and out of sight, causing discontinuity in measurement frequency.

Any glaring problems? The ultrasonic frequencies would obviously need to be well beyond human hearing.

Regards,

-Tay

 

[sophiecentaur]

Absorption of ultrasound frequencies would is quite high. At 30kHz it's about 0.9dB/m for 35% humidity - that would mean 90dB over 100m- a mechanism that wouldn't be a problem using Radio. This is in addition to the 1/r², spreading loss. Your receiver design would need some care, I expect. I don't know much about u/s transducers but the link power budget may not be trivial over these distances.
Hopefully someone can point you at suitable transmitting and receiving gear.

 

[vk6kro]

One problem would be that the robot would not know when the pulse was transmitted, so it wouldn't be able to calculate distances from the delay in hearing it.

I wonder if you would be able to count wheel rotations on the robot and also use an electronic compass to work out the distance traveled and the direction of travel, to work out the position of the robot.

You could limit the maths load by only letting the robot travel in north-south or east-west paths.

 

[sophiecentaur]

One inch of resolution would call for 100ps time resolution. This is non-trivial. You could think in terms of using RF with a frequency of, say 1GHz and use some interferometer system to detect phase changes but propagation effects would introduce a lot of uncertainty. You could, perhaps, calibrate the area first to compensate for this - if the clutter could be relied upon not to change.

Before starting to think about antenna design or even how to handle the signals you really need to examine the most critical aspects of the system design and that is the timing accuracy and signal propagation involved in any of the options. Just think "100ps!" to get the thing in perspective. The best time domain reflectometer for analysing transmission lines can only look at the distances within a coax connector, for example. Your scenario is far less amenable than a well behaved piece of 50Ohm cable.

 

[taylaron]

One inch of resolution would call for 100ps time resolution. This is non-trivial. You could think in terms of using RF with a frequency of, say 1GHz and use some interferometer system to detect phase changes but propagation effects would introduce a lot of uncertainty. You could, perhaps, calibrate the area first to compensate for this - if the clutter could be relied upon not to change.

Before starting to think about antenna design or even how to handle the signals you really need to examine the most critical aspects of the system design and that is the timing accuracy and signal propagation involved in any of the options. Just think "100ps!" to get the thing in perspective. The best time domain reflectometer for analysing transmission lines can only look at the distances within a coax connector, for example. Your scenario is far less amenable than a well behaved piece of 50Ohm cable.

How did you calculate 100ps time resolution? My robot is only traveling perhaps 60ft/min.

One problem would be that the robot would not know when the pulse was transmitted, so it wouldn't be able to calculate distances from the delay in hearing it.

I wonder if you would be able to count wheel rotations on the robot and also use an electronic compass to work out the distance traveled and the direction of travel, to work out the position of the robot.

You could limit the maths load by only letting the robot travel in north-south or east-west paths.

With both ultrasonic and radio methods, the robot would know when the initial pulse is transmitted because the robot itself would send the initial pulse, which instructs the beacons to start transmitting pulses. The robot would know how far away each beacon is when it starts because it measures the time it takes for the first pulse to reach the robot. The further the robot moves, the better defined the location of the beacons are given the distance to each beacon along numerous points along its path.

Again, multipath would be reduced by two methods.

1)When the robot moves away from the beacon, the amplitude of the ultrasonic pulse decreases. The decrease in amplitude should always be proportional to its estimated velocity. If the amplitude is not proportional, the robot will conclude that that pulse is the result of multipath interference.

3) The beacons would wait x milliseconds between pulses, allowing the previous pulses to dissipate enough so that the robot would not so easily confuse the old pulse with the new pulse.

Encoders on the wheels can be used, but slip is inevitable on gravel, grass, dirt, etc... so they cannot be relied upon for long distance calculations.

I will likely use a downward looking CCD camera to operate like an optical mouse for short distance measurements. This method has advantages over using encoders because the camera measures actual movement, not expected movement caused by rotation of the wheels which is susceptible to slip.

 

[sophiecentaur]

How did you calculate 100ps time resolution? My robot is only traveling perhaps 60ft/min.

If you are using EM waves, their arrival time is what tells you the position. It's not to do with the speed of your robot, it's the consequence of 3x10⁸ms^-1 that a distance resolution demands a resolution of the arrival times of the waves. For instance, the old fashioned Decca Navigation Hyperbolic system had, at best, a positional accuracy of a couple of hundred metres, if I remember correctly.

For ultrasound, this problem is much less (timing accuracy in microseconds instead) but the possible signal loss due to sound absorption would be worse at long range.

To make the best choice between the two approaches you need a spreadsheet with two columns - one for em and one for ultrasound and to calculate the achievable resolution for each, using likely power levels, noise performance figures for the respective receivers, transmitting and receiving transducers. etc..