Ping False triggers on noise

T Chap

I have two pings in a noisey envronment ( restaurant ) as an alternative to a passive IR motions sensor. The area of detection needs to be controller better than the PIR style sensor. In this case, there is a loop that cycles constantly, firing ping 1 then 2. The distance can be preset per sensor in the gadget software setup menu. This morning there was some nailgun activity ( air ) and each time the gun went off, the Ping triggered for a detection below the preset threshold set in the Propeller. I assume that since the cycles are so close together, that any other noise that shows up is false triggering the device. I wanted to see if this effect made sense to anyone else, the first test is to slow down the cycle and see if that helps. Any other advice appreciated.

Phil Pilgrim (PhiPi)

It makes perfect sense that a nailgun would be detected as an echo, since a loud impulse like that contains many frequencies. A reasonable solution would be to take several readings in succession and use the median value (not the average, or "mean") to filter out the errant data.

-Phil

T Chap

AHA! Good one.

T Chap

The median of a finite list of numbers can be found by arranging all the observations from lowest value to highest value and picking the middle one

I don't know how median applies when the results will only be 1 or 0?

Say there are 5 samples from the Ping, whereas a 1 represents a positive detection of an object below a preset distance:
Ex. with noise interference:
1 = 0
2 = 1
3 = 0
4 = 0
5 = 1

The median would then be to arrange the numbers from lowest to highest: 00011, then pick the middle number which is 0?

Or:
Ex. with noise interference:
1 = 1
2 = 1
3 = 0
4 = 0
5 = 1

00111, in which case the median is a 1, even though false triggers. So why not just test a number of Pings, and make a rule that the must ALL be true?

Phil Pilgrim (PhiPi)

Apply the median calculation to the distance readings first, then make your 1/0 decision from the result.

Checking for coincident readings from multiple sensors is also valid.

-Phil

T Chap

Phil, for SPIN is this all hard code or is there some simple trick or operation for median? Kind of tricky to get a concept sorted out to load in 5 samples, compare all, the select the lowest to highest values.

Phil Pilgrim (PhiPi)

Here's some PBASIC code for computing a running median:

http://forums.parallax.com/showthread.php?103778

It shouldn't be too hard to convert to Spin.

-Phil