indoor locating or GHz counter

Discussion in 'The Projects Forum' started by MrSingingClub, Nov 9, 2011.

  1. MrSingingClub

    Thread Starter New Member

    Nov 9, 2011

    sorry for the lot of text ;)

    I am thinking about a solution for locating people (me) in my house. The house will be newly built, i. e. provisioning for cabling etc. is easy.
    I am open to any suggestions.

    Basically I just want to know if someone is in a particular room of the house. The persons would also agree to carry around a small piece of electronics. Identifying them and having an accuracy of <1m would be fine.

    My concept so far:
    After googling I found that time-of-flight measurements appeal most to me. I would not have to synchronize clocks or anything.
    Imagine I distribute small senders (for no better name) in every room. They are all connected to PC hardware.
    Looking at a sender. This one sends out a short code (4-8bit) at some arbitrary carrier frequency (433MHz maybe). a Receiver (small object carried around) receives the signal and if the code matches its own it answers with a single pulse.
    The sender counts the time-of-flight from sending the first/or last bit to the answer pulse.
    Because of the short distances and the speed of the wave I have to count really fast like at GHz rate (1ns =0,3m). This seems to be the main challenge.
    I am no electronics engineer but only a physicist so I dont know about parasitic capacities and such. Therefore sorry for wrong nomenclature. This is why I am asking about the feasibility of the approach.

    My solution for a GHz single shot counter:
    There are very fast comparators with a transition delay time of 700ps. I cascade 4 of them and feed back the Not Q output of the last with the first input over an and-gate. The other input is my start trigger. When it goes high, I logical high is propagated along the cascade when it reaches the end the Not Q will pull the input of the first comparator down and send a logical low through the cascade. If the trigger signal pulls the and-gate low the counter will reset the cascade to 0 (cascading).
    The output of each comparator is put to an input of a 4bit D-Flipflop. At the stop trigger it will latch the state of the four outputs. This should produce an oddly coded 3bit counter with a takt rate of 1.428 GHz.
    Additionally, the Q output of the last comparator goes into an 8bit synchronous counter. Its takt rate would be 1.428GHz / 8=178.5MHz.
    How about that?

    Now if I send out a datagram I trigger the start with it. Then I watch the answer signal intensity and if it goes ove a threshold I latch the cascade and stop the second stage counter.
    I know there will be a lot of overhead time in comparision to the time-of-flight time. But this should be roughly constant.

    If it works I get in the best case a distance measurement with a resolution of 10cm with range depending on how many counters I put in.

    Do you think it could work? What other problem do arise?

    Thank you for your processing time ;)

    Last edited: Nov 9, 2011
  2. praondevou

    AAC Fanatic!

    Jul 9, 2011
    I don't think it's feasible.

    From the time you give the command to a transmitter (module or discrete) there are too many unknown time delays until you receive it back. These delays will also change with temperature.
    First there is a delay in the transmitter that will not be stable. Then when received it has to pass through at least a transceiver module that gives a response automatically, this may even be possible if you find one that has this option. But there will be an unpredictable time delay inside the transceiver, unpredictable meaning it could easily change a few ns, totally unacceptable in your case. Last but not least an additional unknown delay in the receiver before you get a signal that you could count.

    It works for GPS in a one-way fashion, with synchronized satellites and a huge distance to them.

    Then, even if you found a way/components to do this you probably won't have the instruments to measure anything in case something doesn't work out as expected.

    You would be better off with ultrasound, but it still won't be an easy project.
  3. MrSingingClub

    Thread Starter New Member

    Nov 9, 2011
    You are right. I think I could calibrate the (huge) constant delays that are in there but not the variing ones. Though my theoretical accuracy of 10cm would allow for a loss of an order of magnitude, i.e. 1m accuracy would be sufficient.
    I simply do not know enough about the details of radio comunication to see all the challenges.
    For the answering problem I thought I could do the following on receiver (the thing one carries around): I detect the signal rise of the incoming message and trigger a counter with a fixed initial count, when it reaches 0, I send my answer pulse. This can be inhibited by the logic that compares the address.
    With this I can bypass all the jitter that comes from analyzing the address.
    This would move the problem to detecting the incoming of the first signal bit.

    But you are right, ultrasound would make the timing easier. There are problems to that too. For example I read the the ultrasound bursts a strongly directional. I would need a receiver (the thing I carry around) that sends more or less in any direction. And how about having the receiver in my pockets, I guess it would damplen the signal to much.
    Do you have experiences with that?

    How about having synchronized clocks in my senders?
    They can be all connected by installed wire. I guess there is a protocol out there how to sync clocks. *Or* I could have a master clock, walk from sender to sender, and just set each sender to the counting state of the master clock by a short service wire connection.
    I then could send a signal over installed wire to measure individual delays and store it in the sender. From time to time I send the installed wire signal to resync the distributed clocks.
    With this setup the carry-on transmitter would only send a short id. The stations now only receive it, add the timer count, and push it to the PC hardware.
  4. Kermit2

    AAC Fanatic!

    Feb 5, 2010
    Google on how RFID is used, and consider having a multitude of detectors at various locations...
  5. wayneh


    Sep 9, 2010
    Here's an old-school method. Just cup your hands around your mouth and yell: Hey Maw! Are you in the kitchen?! :p

    Just kidding! How about IR imaging? I was just watching NCIS or something like that where they watch a site from above via satellite with night vision. If you could cover a 3D volume - your house - with 3 or more cameras, I think you could find and place warm bodies within that space using simple geometry.
  6. MrSingingClub

    Thread Starter New Member

    Nov 9, 2011
    IR cameras and image processing could be a bit expensive. I thought about installing kinect cameras. Then I could maybe issue commands via gestures.
  7. Kermit2

    AAC Fanatic!

    Feb 5, 2010