grounding motor case

I don't have an answer to your final question, but I wonder whether using a shielded cable on your motors, rather than just a gnd wire, would help. I suspect that adding the gnd wire just created an additional inductive loop, which acted like an even better transmitting-antenna.

And I'm not surprised that high-gain analog amps might pick up high-freq noise. You might try putting the sonar ckt, in entirety, into a conductive box. Could use either aluminum, ABS/plastic w/additional shielding added internal to the box [you can use aluminized duct tape for this], or an ABS/EMI-case. Last is found at digikey.com ... p/n SR111E, SE131E, SR151E.

- dan michaels

I don't have an answer to your final question, but I wonder whether using a shielded cable on your motors, rather than just a gnd wire, would help. I suspect that adding the gnd wire just created an additional inductive loop, which acted like an even better transmitting-antenna.

And I'm not surprised that high-gain analog amps might pick up high-freq noise. You might try putting the sonar ckt, in entirety, into a conductive box. Could use either aluminum, ABS/plastic w/additional shielding added internal to the box [you can use aluminized duct tape for this], or an ABS/EMI-case. Last is found at digikey.com ... p/n SR111E, SR131E, SR151E.

- dan michaels

Well. I suspected something like that. I think I'll build a new enclosure for the sonars; the current one is plastic and nylon, (quite stylish!) but not sufficiently shielded.

thanks, dpa

You are at the point where you need good and proper grounding practices.

Bypass caps on the motor Star grounding Twisted motor leads, shielded to chassis ground Chassis ground connected to system ground

Luck

More luck.

Howdy,

I think I have done all the above as you suggest. The chassis ground to system ground is problematic as the chassis is nylon and fiberglass; only the motor mounts are aluminium. Here's a picture:

and here's one with the sonar array mounted:

It seems to work pretty well with the aluminium shield mounted between the sonar array and the front motor. I was hoping that grounding the motor cases would let me remove the shield, but that does not seem to be the case.

Here's a video (12M mpg) of the robot navigating through a garden on the campus here at SMU. The robot is attempting to drive in a straight line, 100 feet out and back, repeatedly, using it's sonar array to navigate around any obstacles:

thank for the input, dpa

Should have recognized the initials....

Anyhow, DPA, some of my best friends include a roll of Teflon film, 0.010" and some 0.005" copper foil. You can build a dround plane out of that, or even a box. Tie that to the ground of the sonar board.

BTW, which R/C car components are you using, if you don't mind me asking. Doing one of those is on my list...

Mike

So, can I just line the inside of the little plastic sonar housing that I have? How does that work? Does the foil have a sticky side, like tape?

The suspension components on the robot were scavenged from one and a half Traxxas E-MAXX monster trucks with some after-market parts. Really nicely engineered stuff.

dpa

"dpa" wrote in news:1120604947.923482.273980 @o13g2000cwo.googlegroups.com:

So, it is programmed to travel 100 feet and then come back, or are you using mapping that will have it travel as far as it has to in order to reach a pre determined straight line 100 feet?

Joe

I don't know about BluEP's foil, but the aluminized duct tape I mentioned certainly has stickum backing. I originally used it to shield both the insides of ABS cases before I discovered there were such things as ABS-EMI cases [these have a metal grid inside the plastic], and also to lay down ground planes under certain chips on some of my original pcb designs. In both situations, I was careful to gnd the foil side. You can't solder it, but you can put a metal screw through it. It worked great in both situations. Basically, I had a situation where analog amps and digital crktry both resided on the same pcb, and digital noise was being radiated thru the magical ether over to the analog part. I found thru experiemtnation that this noise went away by using a gnd'ed aluminum plate under the pcb or over top the pcb, and this led to the shields as described above.

So you just might try putting some shielding inside your current box, as a starter.

- dan michaels

Joe, it's traveling to a point 100 feet ahead of where it was pointing when last reset. With all the obstacles in the way it actually travels more than 100 feet to reach the target, but if there were no obstacles it would just travel 100 feet in a straight line, or as straight a line as possible, and return, like in this pre-sonar video:

When obstacles interrupt the path the robot tries to find the shortest straight-line path to the target, if that makes sense. In other words, it doesn't try to re-acquire the original path. it finds a new one.

regards, dpa

Well, a new data point. I wrapped the bottom half of the sonar array in food-type aluminium foil and clipped a ground lead to it and it seemed to reduce the noise considerably. So perhaps a tin-foil lined box will indeed be all that's needed.

thanks for the suggestion, dpa

Here is the ASCII version of what I have done in the past...

-------------------------------- PCB ttttttttttttttttttttttttttttttttttttttttttttttttt Teflon cccccccccccccccccccccccccccc copper foil

I then use desolder braid to join the copper to the ground plane.

Mike

Congratulations!

-- D. Jay Newman

thanks!

"dpa" wrote in news:1120624623.866337.166910 @z14g2000cwz.googlegroups.com:

What are you using to tell the robot where that 100' target is, and how does it know when it arrives?

Thanks,

Joe

The robot tracks its position in cartesian space, x and y in inches and theta in radians, using a combination of wheel odometery and inertial measurements, based on an origin of 0,0,0 where ever it was when it was last reset.

It can follow a list of waypoints, and there are about a dozen default waypoint lists in flash ram on the robot that I can set with a button array and LCD display. One of the waypoint lists has only two waypoints in it: {0,1200} and {0,0}, which is the one that is active in this video.

So the robot tries to drive to a point where X == 0 and Y == 1200, which is 100 feet ahead of where ever it was pointing when reset, and then tries to drive to a point where X == 0 and Y == 0, which is where it was when last reset.(with some tolerance allowed). So for example, if you block it's path back to the origin, or chase it away such that Y < 0, it will approach the origin from the rear.

There is another flag which is set that tells the robot to do this repeatedly, otherwise it will stop when it gets to the end of the list.

By way of contrast, here is a video of the robot driving a waypoint list with four points: {0,1200} {1200,1200} {1200,0} {0,0}. So it drives clockwise around a 100 foot square, like Borenstein's UMBmark test (9M mpg):

This one is "jBot in Pied Piper Mode."

regards, dpa

That's great! Follow the leader :-)

-Brian

kids love robots dpa

Howdy

Ok, some progress to report. The sonar cage was disassembled and the inside lined with metal foil, as advised. An aluminium cover was added to the back of the box, and I grounded the whole thing back to the power supply. Then I re-ran my noise tests.

The end result was... no noise! So I removed the offending metal plate from the front of the robot, and still no noise. It's actually quieter than it was before, probably because it was still picking up some noise from the rear motor, which was not shielded, and now it's shielded all around.

So, thanks for the help, guys! It worked!

Here's a video of the new-and-improved sonar array navigating the robot through the woods at a campsite near my house. The navigation is done with wheel odometry+inertial measurement (no GPS) and seems to accumulate an error of about 1% over the course of the test. This is the first leg of that test (10 Meg):

The robot is navigating to a point 200 feet away through the woods, and back to where it started. I have marked the starting point on the ground by scraping an "X" with the heel of my shoe, so as to have a reference point to measure the return position.

The robot ran that course 4 times without stopping, out and back through the trees and picnic tables, for a total trip of more 200*2*4 = 1600 feet, (plus maneuvering around obstacles), more than 1/3 of a mile, with a total accumulated position error of about 20 feet, or about 1.25%.

Interestingly, almost all of the error was in the X postion (lateral), and the robot was only off a foot or so in the Y direction. Also of interest is the fact that the error on the first leg (8 feet) was about twice as large as the error on the subsequent legs (4 feet). I have noticed this before, that the navigation accuracy seems to go up when the robot is in motion. That would make sense with a GPS, which is not being used, but harder to understand for the odometry + IMU navigation being tested.

For those with the bandwidth (and patience) here is the complete test, all four legs (50 Meg):

Thanks again for the helpful input.

best regards, dpa