IS PING))) the right answer

macman

I am working on a BoE powered 4WD Garden Rover, in a recent post in the Robotics forum it was suggested that instead of using the IR sensors outdoors that the Ping sensors would be a better choice. I have been looking through the the spec sheet and it would appear that the Ping unit has a narrow field of 'view' and that the height at which the sensor is placed also plays an important part in determining an effective range. So my question well two questions in fact are does anyone have any test data that correlates height placement with distance and if the planned width of the rover is 10" would three sensors be the right amount. At this point I am only looking to be able to avoid obstacles in its path.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Regards Nigel

Success is the ability to go from one failure to another with no loss of enthusiasm.
Winston Churchill

Ken Gracey

Macman:

Did you see our test data in the Ping documentation?

http://www.parallax.com/dl/docs/prod/acc/PingDocs.pdf

Both tests were conducted at 40" above the ground.· I plan on using the Ping on a 3.5" tall robot, but I think the effective range will be about two feet considering the presence of the ground will shorten the effective range. 10" high should be enough for a meter or more based on the picture of the "cone" I'm seeing in the documentation. Give it a try - if it doesn't work return the sensors to Parallax. We'll consider adding some additional test data to our documentation for lower heights - it'll be on our list but it will take a few weeks.

Sincerely,

Ken Gracey
Parallax, Inc.

macman

Hi Ken,
Thanks very much for the response, yes I had seen the new PingDocs file and to be totally honest I couldn't really make alot of sense from them but what do I know. The distances you quote in your e-mail are a lot better than I expected coming down from 40" to 3.5" I thought the range reduction would be much greater.
Given your experience would you be able to ballpark what kind of distance/range I would need for effective navigation, the bot should perform at around 43" a second, I had hoped to mount the sensors 6" high. Also looking at that data showing the span of detection do you think three front mounted sensors would be sufficient?

Thanks
Nigel

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Regards Nigel

Success is the ability to go from one failure to another with no loss of enthusiasm.
Winston Churchill

macman

I have re read my post and just wanted to say that there was no slight intended reference the Ping Data, itwas just that I personally truly could not make sense of the results.

Regards Nigel

Doug Pientak

When looking at the charts in the Ping documentation the area inside the detection cone is where the Ping detected the object. The area outside the cone Ping did not detect the object. This might help you understand the charts. Feel free to as me if you have more questions.

When you mount the Ping close to the ground as Ken stated you can see a reduction in range. But you may also see very little reduction as well. It really depends on the environment. If you're robot is on carpet, in a very large room with no other obstacles the range is affected very little. But if you are in a room with hardwood floors, the echo will bounce off the floor and depending on distance the colliding sound waves could cancel each other out. This would causing Ping to not detect the target at that distance. On the other hand with the right distance the sound waves could even amplify the signal making Ping more sensitive. The size and shape of the target object also plays a big roll is how Ping will respond.

So the tests we did were really to characterize the sensor, not how the sensor would react to different environments.

So after all that I think 2 front mounted sensors on each side would give you good results, but three would be even better.

Hope this helps,
Doug Pientak
Parallax Inc.

Ken Gracey

Nigel,

Doug designed the Ping))) module. If the SRF04 is used on robots and the Ping))) has a narrower angle, you should have success. You could always angle it slightly upward if the carpet poses a problem.

Ken

daviddrake

We have been using several Ping modules for a ranging application. We observed that at some times nearby objects are detected, vs. larger and more interesting objects some distance away (to a maximum of about 10 feet). We put some cloth over the Ping as an attenuator and found it more reliably detected the distant objects. It seems the sensitivity of the Ping causes it to trigger on the "closest" object vs. the "largest" object. We are thinking about an adaptive attenuator for the Ping to adjust for this problem. Our application may differ from yours of course. We found the new Ping to be pretty good, with stable measurements. We would love to see >40Khz signals with a narrower beamwidth and longer range . The price is right.

David Drake

macman

Hi Doug,
I have read and re-read the spec sheet and I do have questions about the data. One feature I could not work out was why a 3.9" cylinder produced a wider cone than the 12" x 12" board? I am not sure if you wanted me to contact you directly or via the forum? I am very interested in using the Ping sensors and really would like to understand the data more fully should I start a new topic reference Ping Data or e-mail you directly.

Many Thanks
Nigel

macman

Don't want to be a nusiance here in repeating my last post on this subject but after reading through the newer Ping Docs I still cannot not work out why the 3.9" wide cylinder in Test 1 p.3 produced a wider cone than the 12" x 12" board in Test 2 p3? I am very interested in using the Ping sensors and really would like to understand the data more fully any input anyone?

Regards
Nigel Emms

Jon Williams

Have you ever seen the F-117A Stealth Fighter? It has faceted surfaces that deflect the radar signal AWAY instead of back to the source. The same thing happens with the board when it's place wide in the sensor field. The can (a cylinder) presents a small surface that faces the Ping))) -- no matter where it's placed in the field.

And do note that the docs state: "target positioned parallel to backplane of sensor."· Why did we test this way?· Because in the real world, walls don't reorient themselves to face a sensor.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Jon Williams
Applications Engineer, Parallax
Dallas, TX· USA

Post Edited (Jon Williams) : 4/19/2005 12:56:23 PM GMT

Paul Baker

Nigel, here is is my interpretation of the data for the PING))) test 1 and test 2, my grad thesis involved acoustic communication so I have some knowledge but wouldn't catagorize it as expert.

Sound waves interact with objects much the way light does, reflecting off surfaces according to the angle of incidence and reflection in which the perpendicular of the surface bisects the two angles:

· ____
··· /|\
·· / | \

First the easier test 2, a 1' square board mouted perpendicular:
since the angle of difference between the normal angle (the vector perpendicular·to the target's surface) and the sound is 0 degrees, the sound will reflect directly back to the source. This will produce a reflection of relatively high intensity without much spreading of the signal.

Now for the first test involving the pole:
looking down on the target we have this profile:

   _
  / \
 |   |
  \_/

Sound which hits the target dead center will get reflected straight back just like the other example. Sound which still hits the target but is off center will be reflected but in a different direction (if it hits to the right of center it will be reflected rightward). But sound when reflected acts like a muted speaker, ie it re-radiates back out as if·the target was producing the sound and the waves look like ripples on a pond. So sound striking off center will still be reflected back to the source, but the angle·appears to be coming from a point further off center than it is actually coming from. This is why the curved surface appears larger to the PING))).

I know there was quite a bit of hand waving in the explaination, but pictorially drawing the example would be too time consuming.

macman

Hey thanks guys that helps alot, I really could not see how the cylinderical object made a larger cone but I get it now I think.

John reference your remark "And do note that the docs state: "target positioned parallel to backplane of sensor."· Why did we test this way?· Because in the real world, walls don't reorient themselves to face a sensor".

If the target was parallal to the sensor backplane would that not make it face the sensor? I am not trying to be picky just making sure I am still not getting 'It'.

If the sensor was directly in front then the cone would be narrow as in diagram in test 2, if the board was offset to the right would that not make the cone more 'P' shaped?

Here is hoping,
Thanks Nigel

Jon Williams

The test board will only face the Ping when it's directly in front of it. The farther off-axis it's moved, the more it bounces the sonar pulse away from the Ping instead of back to it. Notice that the board provides a much longer range than the can when it's directly in front of the Ping? This is because it provides a much bigger cross section than the can, so more energy is reflected back (again, only when ON AXIS).

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Jon Williams
Applications Engineer, Parallax
Dallas, TX· USA