Wednesday, August 02, 2006

Ultrasonic Radar Improvements?

Jeff has once again shown us how some detailed math can do really cool things. He has even sparked my dread of math into a desire to use the math knowledge I have for things other than Math Tests...

He has written an article that describes how to use the NXT display to take range data and convert it into a radar screen style display. He's even provided false readings in a data file so you can try out this coolness without even having a US hooked up. He also discusses converting hte program to read from the US and not the data file.

This got me thinking (always a hazard with all these ideas flying).
Does the US need a certain amount of time on station to send and recieve each signal? I know from experience that it can recived data on the move but how good is it. I also know that each send/recieve cycle takes a finite amount of time (cant seem to relocate that data right now - commnet it if you have it)
Thus should a radar set studder from angle to angle allow 3 returns (incase you catch one durring hte shift).

So I tried to do some web research and looked at the Lego description.

It showed this diagram which labels one of the eyes as a reciever and the other a sender. If this is accurate then perhaps there is also a perfered direction of rotation. If you spin right (assuming the sender is right as in the diagram) then the reciever will be comming into the path of the signal. if you went the other way it might turn far enough away (if you try to do fast scans) that the return broad sides the reciever fairly weakly.

Just curious but has any one tried disabling (plugging up or unhooking) the sender and reciever from different sensors and see what happens when the sender and reciever are seperated but allowed to communicate. (RIS experiment?) It might be interesting especially if you interact with other robots to be able to play marko polo.

Lots of neat stuff emerging - Have fun every one


At August 04, 2006 6:25 PM , Blogger Brian Davis said...

> Does the US need a certain amount of time
> on station to send and receive each signal?

Yes, but it's fairly rapid. Sound travels around 300 m/s, and for a target 2 m away that's a time lag of just 0.01 seconds. If the US rotates at once a second (!), in 0.01 seconds it would swing through just about 5 degrees. Since the sensor has a rather wide field of view (certainly wider than 10 degrees), and in this example the rotation is *really* fast, there's little need to "pause" the rotation... or worry about it's direction.

Using a sender from one US and a receiver from another will be unlikely to get anything, most of the time. The problem is they're not syncronized, and trying to measure the time delay between they two would be very very challenging using NXT-G and the standard firmware. Again, for an object 1 m away the time delay is 0.007 seconds, or about 7 ms. Most NXT-G blocks take only slightly shorter than this to execute. While not in theory impossible (after all, the timers have better resolution, being run by the firmware), sub-one-millisecond time resolution is beyond the standard environment, corresponding to a minimum measurable distance of about 15 cm - closer than that, and the time delay is simply too short to measure within the standard environment.

Brian Davis

At August 07, 2006 9:54 PM , Blogger Drew Stevenson said...

I was looking for information about how often it broadcasts I assimue it waits for returns for a while between broadcasts. I understand the time the signal is in the air but was interested more in the system repeate time. (like how often it actually takes a reading of distance.)

At August 08, 2006 9:08 PM , Anonymous Guy Ziv said...

The I2C command 0x2, 0x40, 0xNN sets the delay time between bursts between about 12.5 ms NN=0 and 3 s NN=0xf.


Post a Comment

Subscribe to Post Comments [Atom]

Links to this post:

Create a Link

<< Home