0

I'm building a device that must find its own position on a field (400x400m). Since there will be lot of devices on the field I could not send them their position individually and I want them to be passive (regarding signal emission).

My current plan is to setup 3 rf emitters on the field one master and two slaves. The master on one side emits a signal every second. On the opposite corners are the two slaves emitters, they each send an echo signal when they get master signal. In the device I need to measure the time between master pike signal and slaves echoes. This makes 2 durations. Then send these duration to a micro-controller that will compute position.

I don't need a big accuracy, 1m will be enough. But this means a timing accuracy of 3ns wich is quite a lot. It must be cheap, let say less than $50 each for 20 devices (prototype), and less than $2 each for 10.000 pc (production batch).

What kind of cheap component can measure a time gap with such an accuracy ?

I'm also interested if you know any other solution to this cheap positioning problem.

---------- Edit

I'm not asking for any design service. I explained the context to help other understand the needs. I think my question is quite simple for someone experienced.

Let's ask it another way : Given two impulse signals, how to mesure the time gap between them with 2-3ns accuracy ? I'm thinking of a crystal and a counter. Does it look ok, or plain crazy ? The microcontroller could calibrate itself using the master signal, so if the crystal frequency is not precisely set it's not a problem providing it is constant over time.

bokan
  • 301
  • 2
  • 9
  • 3
    This is not a design service ! What (commercially available) solutions have you found ? Even if it doesn't meet all your requirements it gives us a starting point. But my guess is: this will not be possible at the cost level you want. Not even close. Maybe you have a chance if you multiply the dollars by 10. And then I did not mention development cost at all. – Bimpelrekkie Oct 19 '15 at 09:35
  • 1
    This IS a design service, within reason. Mention commercial possibilities and people will vote to close faster than the speed of write (under 10 ns from trigger pulse here). BUT a lot more what is REAL:LY wanted and why you want what uou want is needed to allow hal;f a chance of a good answer. – Russell McMahon Oct 19 '15 at 09:52
  • "No shopping questions" should mean that "where to buy cheap X" is disallowed, while "what should I shop for to solve this problem" ought to be allowed. – pjc50 Oct 19 '15 at 10:26
  • 1
    A similar question: http://electronics.stackexchange.com/questions/122926/very-high-speed-counter-around-1-5-ghz-to-2-ghz – pjc50 Oct 19 '15 at 10:36

3 Answers3

3

This IS a design service, within reason, AND people will close your answer before starting to think (under 5 ns) if you mention commercial alternatives. So, maybe you can't win :-).

A fast enough comparator (they exist) will allow time of an "edge" relative to a local clock to be turned into a pulse.

A local clock and counter will allow clocks between start time and trigger time to be determined.

People make short range time of flight ICs made for cellphone gesture detection (believe it or not) with times approaching what you want.

You should conceptually look at as many alternatives as you can think of. One or more may even work.

Rotary scanning of a field of sensors with a number of transmitters may be good enough.

You may be able to do computed tomagraphic position detection by seeing which targets are blocked or affect the signal along a number of paths.

You say there are "many" devices but you need to be more specific about quantity and blocking characteristic. If there are eg 100 each 1 metre square and 5 metres tall "there may be problems" with some systems. If they are say 300mm x 300mm and 200mm tall so you can "look down on" the field fromm an angle you may be able to use cameras or barcode or similar pattern reading or ...

You say you want them to be passive wrt emission but do not say why or how close to a "MUST" your "want" is. IR LEDs each coded could be very helpful. Transponding LED tx's could be VERY helpfil (Poll and respond. Resonant circuits in each device that respond to whatever could be helpful but does this violate your radiation objection if passive - or active?

A lot more flesh and blood will help with a good answer.

Russell McMahon
  • 147,325
  • 18
  • 210
  • 386
  • Do you mean it's forbiden to name any commercial component ? even naming the type of component will help. The device is to be worn by public during concerts. – bokan Oct 19 '15 at 10:15
  • There are teardowns of existing concert bracelets eg http://jelmertiete.com/2014/11/17/TML-bracelet-teardown/ - apparently that one does location in the other way, by having active bracelets and triangulation in the base stations. – pjc50 Oct 19 '15 at 10:55
  • @pjc50 Thank you again ! ;) I didn't knew about Tomorowland Bracelet, only Xylobands ones (coldplay). Anyway, all those bracelets don't know precisely where they are so they all light up the same way. – bokan Oct 19 '15 at 11:31
  • I want them to be passive because there could be 10 thousands of them. I thought about all the solutions you're talking but I don't see how to do them. – bokan Oct 19 '15 at 13:11
1

Does it look ok, or plain crazy ? The microcontroller could calibrate itself using the master signal, so if the crystal frequency is not precisely set it's not a problem providing it is constant over time.

It's plain crazy using your method I'm afraid - a slave receiver would have to decode the master message with an MCU and the time taken to lock into the signal could be several tens if not hundreds of cycles of the transmitted carrier. If the carrier is 1 GHz then somewhere between 10ns and 100ns (or longer) would be taken up just by the locking-in process. That's a big gap in time to issue a slave response.

Also, one cycle of an MCU might be +/- 20nsec error. You cannot guarantee that the MCU operating frequency is somehow synchronized to the master transmission hence the uncertainty error.

I think you are miles off getting 3ns accuracy or even resolution.

BTW it's called Trilateration and not triangulation

Andy aka
  • 434,556
  • 28
  • 351
  • 777
  • If the signal is a simple spike, do I need a MCU to detect it ? – bokan Oct 19 '15 at 12:12
  • 1
    How do you prevent false readings - there is zero integrity in a simple spike and you might contravene emission laws. Also how would the device recognize a spike from a master or a slave? – Andy aka Oct 19 '15 at 12:15
  • The +/- 20nsec quantization can be overcome in principle with *sub-threshold signal detection*, aka *[stochastic resonance](https://en.wikipedia.org/wiki/Stochastic_resonance)*, if the clocks are drifting slightly at random. Like in [this paper](http://xxx.lanl.gov/abs/physics/0201053) where they measured the speed of light with only *ping* (on Ethernet cables). – maxy Oct 19 '15 at 12:31
  • @maxy please do consider contributing an answer. – Andy aka Oct 19 '15 at 12:38
  • @Andyaka I could use two different frequencies for master and slaves. Or even use timing information, every second the master sends a spike. Slaves answer just after each one its turn, odd seconds for a slave, even seconds for the other... could be a way to do it. Regarding reliablility, this is not a life saving device, I don't care if there's some glitches. So do it still looks so crazy ? (ho and sorry for the triangulation, it's the french version of trilateration) – bokan Oct 19 '15 at 13:04
  • 1
    I have pointed out what I see as major problems and you can take it or leave as you please. Using two different frequencies is abandoning the spike idea - I'm not going to get involved in an evolving development of ideas that I fundamentally believe will result in disappointment unless you throw money at it. Try doing some more research on it and come back when you have a better proposal. – Andy aka Oct 19 '15 at 13:09
  • I have very little knowledge about radio. Thank you for your time handy. – bokan Oct 19 '15 at 13:13
  • "I have very little knowledge about radio" is not an ideal qualification for building a radio-based location system... – pjc50 Oct 19 '15 at 13:26
  • @pjc50 I'm building the whole system and localisation is just a part of it. I will hire people who know how to do the radio part in the end. I'm just asking about what's possible to do here. – bokan Oct 19 '15 at 17:18
1

Check out the amazing TDC7200 Time-to-Digital Converter...

http://www.ti.com/product/tdc7200

It can measure time with ~55 picosecond resolution (!) and costs about $2 at 1K quantity.

bigjosh
  • 9,888
  • 29
  • 48
  • What kind of clock do you plug to this chip ? – bokan Oct 21 '15 at 20:57
  • You can use pretty much any 1-16MHz clock, but you will better results with a faster clock, so no reason not to use a super-cheap 16Mhz one. – bigjosh Oct 22 '15 at 00:25
  • If you are going to pair this stopwatch with a cheap MCU like an AVR-8, then you can use the clock output pin from the MCU to directly drive the stopwatch. – bigjosh Oct 22 '15 at 00:35
  • Ok I see, it's the clock for digital signal output. I thought it was a clock for timing and it requires a very very fast clock, with 55ps period. – bokan Oct 22 '15 at 10:00