Looking for help/info on photo diode project

[bigmark1972]

I want to build a sensor similar to the image below. I am only using this image to help represent the concept.

Imagine a second 'curtain assembly' place in front of the first and rotated 90 degrees.

My objective is to determine the position of a part approx 1/4" sphere as it falls through the two assembly's.

I realize a simple light curtain can not do this for me as it views the light beams as one single beam.

I am wanting to be able to determine which of the light streams on each the x and y plane have been broken by the object. This should let me calculate the position of the part.

I would estimate the distance of each side of the two sensors to be under 5 feet and possibly under 3 feet. I am wondering if I could use low cost UV photo diodes or phototransistors and uv led's as the light source to accomplish this?

I would like to keep costs down if possible, also the parts are moving quite fast so I need it to have a fast response time also.

I'd appreciate any input or if anyone knows a better solution please chime in.

I think a laser sensor would work but would be expensive. I am also wondering if a machine vision system could accomplish this but the parts are moving very fast and these are also very expensive.

I thank anyone who can help.

 

[MikeHalloran]

Random thoughts..

To detect the passage of a 1/4" diameter sphere, the individual photosensors you will need must be on ~1/4" centers at most, and probably a little closer than that. So you need a fair number of individual sensors to cover a 3'x 3' square. They have to be pretty well collimated to respond only to objects directly along their axis, or you will have to figure out which sensor is the correct one if several respond.

Separating the detection planes along the flight path also allows you to measure the objects' velocity, which may or may not be of any use to you. They will have to be separated some distance anyway to prevent anomalous response from reflections, assuming the sphere is specular.

UV sources, sensors, and optics tend to be expensive. IR parts are more affordable. Visible sources are easier to align ... and alignment will be a major pita even then.

This all depends to some extent on how fast 'very fast' actually is, i.e., how long it would take your parts to pass through a sensing plane of zero thickness. Most mechanical systems are glacially slow, when compared to even 'slow' electronic systems ... which introduces another option; scanned sensors. E.g., the rotating polygon mirrors used in laser printers could sweep a collimated sensor's effective axis through an arc at a speed that might be sufficient to be reliably occluded by your parts. You could use, say, two fixed sensors at two corners of your square, with linear light sources on opposing edges ... or you could scan two laser beams and use arrays of small photosensors (that wouldn't need to be collimated).



Mike Halloran
Pembroke Pines, FL, USA

 

[itsmoked]

Reminds me of a system I had to design and build to measure square footage of uneven pieces of leather flying across an optical array.

Try this on for fun.

You take one of those curtains make it perpendicular to the falling spheres.

Place another below the first at an angle, say 30 degrees. The time between a ball breaking the first plane followed by the second plane will be a function of which side it is falling thru. Understand?

Now because so far you only have one axis of position you add a third curtain at the same angle and perpendicular to the other angled this will give you a position in the perpendicular plane.

You have the position now in both axes what else could you possibly desire?(well never-mind)

Now the above scenario works with constant/fixed velocity spheres passing thru. If the spheres aren't constant velocity you just add another curtain before the first one. You then use the first two to accurately measure the particular sphere's velocity.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[logbook]

3 foot at 0.25 inch pitch = 3x12x4= 144 sensor pairs
and you need two.

What you are proposing as a "cheap system" requires 288 sensors and a >300 pin FPGA to do the digits. That is not cheap!

This means MUXing to reduce the pin count. So the $1000 question is what do you mean by "quite fast".

To get a photodiode to work over 3 feet it would probably need to be strobed to improve the signal to noise ratio since the signal will be really small.

It is not clear if a 0.25 inch sphere would actually be able to break a beam at 3 feet unless you use a laser diode. In this case the cost just increaesed by a factor of 100. Try one beam first to see if you can get the scheme to work!

 

[itsmoked]

Yeah, you could do it logbook. Well, you would need beams spaced narrower than the sphere for sure. As just a grazing sphere edge ain't gonna cut it. You would need mechanical collimation built into your design, which would mean some space out on the edges. to pull off the collimation.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[OperaHouse]

Just got the movie FIREWALL from the library and Harrison Ford ripped the scan sensor out of a fax machine and taped it to a crt to copy data. That would be a cheap sensor, but the optics to make it look at 3-5 feet would be a pain. A cheap laser level puts out a nice line scan. With a dark chamber you could use a CCD camera.

 

[richs]

Hiya-

You mentioned:
"My objective is to determine the position of a part approx 1/4" sphere as it falls through the two assembly's. "

Have you considered "funneling" these parts ahead of the sensors? Seems like it's a mechanical question that begs a little sheet metal.

If you can funnel it down to like an 8.5" X 8.5" field, then indeed you could use the optical sensors from a pair of cheap scanners. I believe that they would be fast enough to get the information that you would need.

Collminating the light source might also be an issue as itsmoked mentioned.

Hope this helps.

Cheers,

Rich S.

 

[MacGyverS2000]

I feel like that old TV show Name That Tune...

I can do it with one laser diode. Granted, I'd still need 144 detectors and 144 mirrors, but it's doable. sinc it's a visible laser, alignment would take some time but could be done with the naked eye.

mark, give us some more details on the actual application. Our answers may vary heavily based upon the actual circumstances of use.


Dan - Owner

http://www.Hi-TecDesigns.com

 

[itsmoked]

Agreed, more info!

How fast are these balls moving?

Why do you want to know the positions, as the solution may be part of the 'system'.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[DaveScott]

I think you might want to consider using fluorescent tubes, and a linear position sensor, or photodiode array, with an appropriate lens in between. It's not perfect, but it would be effective.

See

http://sales.hamamatsu.com/en/produ...ition-sensitive-detectors/one-dimensional.php

and explore some of the other sensors as well.

Alternatively, buy a load of laser pointers, and photodiodes.

 

[MacGyverS2000]

The ball would have to fall almost directly next to the sensor array to make any accurate reading if you use a fluorescent tube...

Dan - Owner

http://www.Hi-TecDesigns.com

 

[2dye4]

My take on it
Scan both the emitters and detectors in sequence so you can be sure that the ball can fully block one pair. Also this eliminates the need to precisely align the two as the beam is detectable even at an angle.
Make the detectors have a binary output with a single select line that is scanned in sequence along with the emitters. Select the emitter IR LED's so that they are high power type that don't roll off in their intensity until somewhere in the 2 amp range. Then drive them very quickly. On time 100 uS. For 100us X 144 = overall rate of 14.4 ms per full scan. Drive them at 2 Amps for the 100us and scan the reciever array in sync with the transmit array. Take care in location that reflections are not possible from surrounding equipment.
A difficult job i think.
They sell these things you know, but I am sure they are more than $1k so you may have more time than money.

 

[bigmark1972]

Thanks for all of the input, and for some very interesting suggestions for me to consider and I appreciate it greatly.

It is hard for me to get into much detail on this as I have signed a letter of disclosure and I am already dancing on the edge of the agreement.

I appreciate all of the intelligence and know how this forum has to offer.

I will say the object would be moving in excess of 2000'/sec

I would imagine this most likely will exceed the response time of a photo-diode system.

I guess the rule of if it were easy, someone would be doing it applies here.

Thanks again to all I really enjoy figuring this type of thing out, and I will continue to seek a solution it's good for the mind.

 

[2dye4]

Bigmark

I think that speed does kind of change the game.
At this point I would think the thing would leave a sound profile wave that could be used to detect the xy coordinate of the ball as it passes through an enclosed ring.
Position microphones in a triangle with each one having a DSP processor to detect the onset of the pressure wave. Then through triangulation you could detect location.

Good luck.

 

[richs]

Hiya-

Yep. 2000'/s does change the game. Under $1000. I'm thinking nope.

Old adage applies: cheap/fast/good. Pick two.

Best of luck though and tell us the results that you can.

Cheers,

Rich S.

 

[bigmark1972]

I'm not limited to 1k but I do think an important element of any good design is achieving a good balance between price and performance.

I'll report back if I figure anything out.

That microphone triangulation theory has merit, but I wonder the performance implications of the object being supersonic. I imagine that would be a challenging but possible solution.

Possibly a sonar, radar or even ultrasonic triangulation could also work....

 

[itsmoked]

So the army is finally tired of having paper targets..

You could use metal detectors. The strength of the detection shows the distance from the detector. Might need a few layers to pin one down.

Alternatively, Detect the objects approach (easy). Then set the system up so an area following the detector is poorly lit. Hammer the object with a strobe light taking a single still shot of each one as they pass thru. You then have very little processing to do, to nail its X,Y position in the plane. You have the time before the next detection to figure it out even. You could do it all with a simple micro, a camera, a strobe, and a light curtain.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[DaveScott]

Ah, with that kind of speed, and small target size, I would suggest a line laser (not expensive), and a cut-down fresnel lens to convert the diverging light back to parallel. That's the transmitting part finished.
The receiving part could be a long line of individual photodiodes, as the object would be in the field of view for about 1x10-5 seconds, and photodiode rise times are in the order of nano-seconds.

Trouble is, is there a fresnel lens that big?

 

[MacGyverS2000]

They have Fresnel lenses on the order of 5' square... plenty big to trim down. They're typically found as novelty science items, always being shown setting tarred roads on fire in the mid-day sun.


Dan - Owner

http://www.Hi-TecDesigns.com

 

[MikeHalloran]

Wow, 2000fps is fast for any projectile, let alone a sphere.





Mike Halloran
Pembroke Pines, FL, USA