[time-nuts] question for expert time guys

[Hal Murray]

rickharold at gmail.com said:
> I would like to triangulate a position of a device which moves using 3 fixed
> positions devices of known location. The idea is to have these operate on
> 915mhz or 434mhz or 2.4ghz or appropriate frequency. 

> I'd like a range of 150 feet or better and accuracy of 3 feet or better. 

> I know the time accuracy is the key to count time =  feet, 1ns. 

> How to make it "inexpensive" is key. how inexpensive, very ;-)

I think it's going to be tough to do it at low cost.

The low cost transmit/receive chip sets you can get for 915 MHz are low 
bandwidth.  That will turn into noise on the on/off transition that you need 
for timing.

I'd suggest getting a pair of demo boards and running some experiments.

The first simple sanity check would be to wire them up to a couple of uProcs 
and send messages and see if they work at your required range.  Play around 
to learn the error rate vs baud rate.


> -device response ASAP on different frequency

Not with low cost.  But you don't need to use a second frequency.

My straw man would be something like this:

Only one station transmits at a time.
Wire up the transmit and receive lines to counter/timers.
On both the transmit and recv side, grab the time on each data-level change 
and average them to figure out when the packet started.

Look at the NTP protocol.  It collects 4 time stamps.  From that, you can 
compute the time of flight.  The time at the remote server drops out.

The sequence would be something like this:
  
fixed->remote: long timing packet.
  (with lots of 1/0 transitions to feed data to the timer hardware)
remote->fixed: long timing packet.
remote-> fixed: packet with 2 time stamps
  first from the receive side, second from the transmit side

I'd put a CRC on the packets as a sanity check.

Handwave time:
  Assume the counter/timers run at 100 MHz.  That's 10 ns.
  So you'll have to do lots of averaging to get below 3 ns.
  But that's assuming the RF units work well and that the averaging works.
  You could build a (much) faster counter/timer in a FPGA.
  I'm not sure that will help.

Get a pair (or 2 pairs) of units and see how well they work.

I'd expect FSK to work slightly better than ASK (on/off), but
I'm not enough of an RF geek to turn than into numbers.

I'll say more if that will help.  We should probably take it off list.

------------

One probably crazy idea...

Setup a directional antenna with a motor to aim it.  Scan for the best signal.


-- 
These are my opinions.  I hate spam.