Falling objects and suggestions on sensors to use

Hello Everyone, I’m trying to figure out if is possible to build an app which retrieves data from Arduino. To get deeper in the problem, please imagine a vertical 2d-plan with 5 rows and 8 columns. Each of the squares determined in this way has an object in it. Question: can I remotely know on my smartphone if one of those objects pushed from the back falls down? Which kind of sensors should I use? I tend to exclude ultrasonic ones because of a particular condition that would lead to echoes - being the vertical plan inserted in a sort of box. Thank you in advance for the help.

Detect object presence using a light sensor, either ambient light or beam break type.

The smart phone part is a completely different problem.

Th3Cg: ...please imagine a vertical 2d-plan with 5 rows and 8 columns. Each of the squares determined in this way has an object in it. Question: can I remotely know on my smartphone if one of those objects pushed from the back falls down?

Th3Cg: Which kind of sensors should I use?

Believe it or not, the answer depends to an incredible extent on what the cryptic 'objects' are, their physical/electrical properties, etc.

Hope this drawing can make things clearer…

Falling objects.pdf (856 KB)

The same drawing in .jpg

Th3Cg:
objects pushed from the back

How does this happen?

dougp: How does this happen?

Mechanically via motors which push objects

Are the motors under the control of the device which is to do the monitoring?

no. they are existing and I prefer not to work on them

I'd second the break beam/thru beam idea raised in reply #1. More wiring and power consumption but, no worries about getting a reflection from odd shapes and as long as the objects aren't transparent, essentially foolproof.

Th3Cg:
The same drawing in .jpg
3599eed9c3a1dc0483048d4ad6175920ae88e46d.jpg

Place a light sensor under the object.

dougp: I'd second the break beam/thru beam idea raised in reply #1. More wiring and power consumption but, no worries about getting a reflection from odd shapes and as long as the objects aren't transparent, essentially foolproof.

So, are you suggesting to place one sensor for each row OR column? If so, would those sensors be able to determine the distance of every object as soon as it begins its falling trajectory? In this way the app could easily say "object A1 fallen" (please refer to the sketch attached to previous posts), then "A2", "C3" and so on. Moreover, what if a longer range is needed? (let's say up to 100 cm)

wvmarle: Place a light sensor under the object.

I could, but 12 sensors would be needed in this case. Do you think it should be possible having the same result with less sensors?

Th3Cg: So, are you suggesting to place one sensor for each row OR column?

No, one for each object.

Th3Cg: If so, would those sensors be able to determine the distance of every object as soon as it begins its falling trajectory?

No, the sensor detects the presence or absence of an object in the cubicle only.

Nothing was said in the initial post about distance.

The point is: can I know [u]which[/u] object falls, every time that an object falls? Maybe we could use a sort of trilateration/triangulation to decrease the number of sensors. Am I wrong?

A 3x4 grid of sensors over the front openings would work. Look up "matrix keyboard" for ideas on how this applies to detecting a large number of buttons with the fewest number of wires.

With a little bit of math, you could have just one horizontal sensor across the bottom and 3 "column" sensors. Work out how long it took between the two sensors triggering and you could estimate how far the object fell before it hit the bottom sensor. This would be an absolute bear to tune it properly as the column sensors will see the object starting to tip while it's still supported by the shelf.

For the effort involved in tuning the alternatives, 12 sensors looks like a great solution. They could be wired as a matrix, but you would still have 12 individual light beams for the objects to sit in front of.

Are the objects all unique?

If you would use a grid of break beam sensors crisscrossing the house, you could detect the presence of no more than three objects at the same time before they start overlapping and hiding each other, but you could tell "there's an object in that compartment". This works in a situation where the compartments are normally empty, and you just have to measure which one the ball is thrown through or so.

In your case the compartments are normally filled, and I can't think of a way to tell which one is empty as the rest is full, without simply having a sensor at each compartment.

If it's used in a well lit environment, a simple light sensor would do LDRs are small and dirt cheap, maybe you can make a digital signal of them by connecting them to a Schmidt trigger, and connect that in turn to a shift register which you can easily read without using too many pins. I guess there are shift registers with a Schmidt trigger as input. You may need a trim pot in combination with your LDR to adjust at which brightness level it shifts (or if you're confident you know the values a fixed resistor will of course do just as well), but all in all it's a cheap solution and not exactly hard to build. The LDR is a great sensor for dark/light sensing.

MorganS: With a little bit of math, you could have just one horizontal sensor across the bottom and 3 "column" sensors. Work out how long it took between the two sensors triggering and you could estimate how far the object fell before it hit the bottom sensor. This would be an absolute bear to tune it properly as the column sensors will see the object starting to tip while it's still supported by the shelf.

Seems definitely the correct way to act. Should I use IR "column" sensors + 1 vibration sensor at the bottom? Please consider the size of the grid (height 120cm * width 80cm).

What a choice... Fewer (but probably more expensive overall) sensors + a mighty complex bit of software to try to trace back what happened, or a slightly greater number of sensors that simply tell the sketch "I've got a piece here!", and also can instantly tell when one is leaving it's perch. I think I'm a tad less desperate to bring down the number of sensors.