Society of Robots - Robot Forum
Electronics => Electronics => Topic started by: airman00 on April 07, 2009, 10:37:56 AM
-
Here's my idea for RF distance measurement that uses the Time of Flight method
HARDWARE: Two transceivers - transceiver A and transceiver B
OPERATION: Transceiver A sends out a signal and starts a timer. Transceiver B receives A's signal and immediately replies with another signal. Transceiver A stops its timer once it receives the reply from B. Take the time , divide it by two and you have the time it took to go one way.
Distance = Velocity * Time . Only problem is that the velocity for radio transmission is 300,000 km/s, aka the speed of light.
SYSTEM REQUIREMENTS: I need to design a system that operates on a 2 mile (~3.2 km) range , and I want a resolution of about .1 mile (~.16 km). I have already found RF transceivers that can support the 2 mile range I need.
DISTANCE = VELOCITY * TIME
.16 km = 300,000 km/s * TIME
TIME = 5.33333333 × 10 -7(thats 10 to the -7) seconds
TIME = 0.533333333 microseconds or 533.333333 nanoseconds
So that means I can have an error tolerance of at most 533 nanoseconds
Would a standard AVR timer be able to handle this much precision?
Can someone point out any useful links or things to read?
Thanks in advance
EDIT: I found this - http://www.eecs.berkeley.edu/~slanzise/research/rf_tof_ranging_dist.pdf (http://www.eecs.berkeley.edu/~slanzise/research/rf_tof_ranging_dist.pdf)
-
This all depends on the processor crystal clock, interrupt response times, and software - these are all several orders of magnitude too slow for what you attempt to do.
(In my opinion)
-
AVRfreaks.net replies:
You are limited to what ever the system clock rate is. If it is a 20MHz processor running at 20MHz, you have 5ns. That is pretty much max for any of the standard AVRS, I think. XMegas are reported to be spec'd for something like 32MHz.
I think my idea is really doable!
-
The inconsistencies in the system clock would be a huge source of error. referring to variations in the duty cycle of a clock which was not designed for high precision.
interrupts in the computer would throw things off.
as my flight instructor was fond of saying... You go ahead. I'll watch.
-
I made a similar post about this on AVRfreaks
Read about it here: http://www.avrfreaks.net/index.php?name=PNphpBB2&file=viewtopic&t=77431 (http://www.avrfreaks.net/index.php?name=PNphpBB2&file=viewtopic&t=77431)
-
Your other post only talked about timing, not about sending signals, not about keeping two systems in sync. Like kd5kfl's instructor said...
-
not about keeping two systems in sync.
they don't have to be in sync. Think about it...
OPERATION: Transceiver A sends out a signal and starts a timer. Transceiver B receives A's signal and immediately replies with another signal. Transceiver A stops its timer once it receives the reply from B. Take the time , divide it by two and you have the time it took to go one way.
Transceiver B is constantly waiting and immediately sends a reply to the Transceiver A that receives , so Transceiver A and B do not have to be synced. Obviously you would have to subtract how long it takes to send out the RF signal out UART or whatever.....
-
not about keeping two systems in sync.
they don't have to be in sync. Think about it...
OPERATION: Transceiver A sends out a signal and starts a timer. Transceiver B receives A's signal and immediately replies with another signal. Transceiver A stops its timer once it receives the reply from B. Take the time , divide it by two and you have the time it took to go one way.
Transceiver B is constantly waiting and immediately sends a reply to the Transceiver A that receives , so Transceiver A and B do not have to be synced. Obviously you would have to subtract how long it takes to send out the RF signal out UART or whatever.....
RF signal out of a UART?? The point that is being made is you are talking about speed of light measurements using a $2 processor running at 20 MHz using a crystal with 200 parts per million accuracy. You have to account for the variability of a signal happening at any state of the processor operation. The AVR's variability makes this effort impractical.
-
RF signal out of a UART?? The point that is being made is you are talking about speed of light measurements using a $2 processor running at 20 MHz using a crystal with 200 parts per million accuracy. You have to account for the variability of a signal happening at any state of the processor operation. The AVR's variability makes this effort impractical.
I think I understand now, thanks GearMotion
But still ,from what I understand, I will be able to get my +-.1 mile (~.16 km) accuracy. What do you think GearMotion?
Thanks
-
But still ,from what I understand, I will be able to get my +-.1 mile (~.16 km) accuracy. What do you think GearMotion?
Frankly I'd take the problem out of the domain of the microcontroller. A microcontroller (the stationary one - point A) can start and then gather the measured results, but you are better off making some external hardware to handle this.
Might be easier to explain using LEDs. Point A has a red LED triggered by a microcontroller. That same trigger also trips a high frequency counter. Point B has a red LED detector that trips a blue LED. No micro involved, so you keep the number of variables reduced.
Point A then gets a signal back from point B with a blue LED detector. This signal stops the counter. The counter value is read by the the microcontroller.
Would you be able to replace these ideal LED devices/detectors with real-world devices? What is the frequency of your radio? Or put a different way, what is the period of your carrier frequency? Is it a variable? Can you detect a signal with zero delay? How many periods of your radio frequency signal does it take to establish detection? Is that reliable, consistent?
-
Hi,
I think my idea is really doable!
So, in other words you didn't read the other quite recent threads we had on the subject??
You confuse timer ticks, resolution and precision.
I know it isn't doable!
(What you describe and the way you wanna do it, but hey, it's your time and money, so feel free to waste some of each).
-
So, in other words you didn't read the other quite recent threads we had on the subject??
I did .
Here's what I found : http://www.societyofrobots.com/robotforum/index.php?topic=5923.0 (http://www.societyofrobots.com/robotforum/index.php?topic=5923.0)
There was another thread, but I cant find it right now.
From the AVRfreaks thread it seems like I can get relatively good precision. I would get an error of +-.1 mile, which is acceptable to me.