Can Phototransistor distinguish between colors?

[afink]

Can a circuit be designed using a Photodiode or Phototransistor to distinguish between the 3 different LED light sources (day and night) from a stoplight at a distance of 30 ft? The output would drive a relay for each color. I need to know when each color light turns on and off. This would be used to help collect data on traffic patterns at and given intersection using stand alone data logging equipment.

 

[MikeHalloran]

Banner might have something appropriate:

http://www.bannerengineering.com

They are, of course, not the only game in town for such things.

I'd expect some trouble aiming the detectors, and interference from sun behind the signals, and from wind moving the signals.

Alternatively, you could work with a video image instead:

http://www.ntnu.edu.tw/acad/epub/j47/mst471-4.htm

Simplest thing I can think of is to use a bucket truck to tape a simple detector to each lens of a signal.

Another idea that I might explore is tape acoustic sensors to the control box and listen for patterns in clicking relays. Newer controllers might be solid state, in which case I'd expect them to have interfaces that might be accessible with permission.



Mike Halloran
Pembroke Pines, FL, USA

 

[IRstuff]

Unless you're sure you're going to get LED lights, I'd expect a lot of overlap in the color pass bands for standard lights.

Additionally, unless you're willing to spend some money for extremely narrow field of view optics, the sun can be in the field of view of the detector, and no amount of filtering will screen out the sun. Even with the sun out of the field of view, glint from the light structure can give you false signals. 8" light at 30 ft would need something like a 2/3 deg field of view with ~1/6 deg angular aiming tolerance; expensive...

Using a camera would seem to be better, to me, since the same camera that collects the traffic data can also simultaneously collect the light signal data, thereby simplifying the synchronization of the data. Moreover, since the imagery is collected continuously and from a fixed location, the traffic lights can be localized in the image and some image processing can be readily applied to weed out false alarms from glint and solar background.

TTFN

 

[davesaudio]

suspect one could calibrate focused optics shining onto
three sensors each with coloured filter lenses to sense the resultant colour. The relative intesities should tell you something about the active colour?

 

[afink]

How much easer would it be if I only monitored the red LED and the sun was always behind or it was night?

 

[IRstuff]

But, don't you already have a traffic camera for this application?

In any case, you can, but the field of view calculations were for a single light anyway, so there's no gain there. You'd still need filter, and you'd still need some amount of smart thresholding. On a small intersection, particularly with LED lights, the lights facing the opposite direction can throw quite a bit of light across the intersection.

TTFN

 

[VEBill]

If this is a one-off study, then just capture the raw data (one or more light intensities versus time). Knowing the light timing, you should be able to overlay a clean timing signal after the fact.

If you're inventing a system, then you could use the often repeatable timing to help extract the noisy signal. Watch out for when the schedule changes. If the lights are 'intelligent', then this won't work.

If the lights simply cannot be distinguished from the sun and glint, then use the sound of crunching fenders to synchronize... ;-)

 

[IRstuff]

One reason I'm paranoid about using phototransistors is a previous bad experience with the like in an inadvertent daylight system.

One alternative is to turn down the device. A PIN diode has somewhat less gain and will only produce responses to pulsed signals. While the pulse duration of a signal light is too long to result in a clean pulse, the transient characteristic of the light signal is an addditional filtering method to weed out false alarms. Since you'll get a triangular pulse, that should be quite unique, compared to glints and direct sunlight.

Nonetheless, saturation is still a potential problem, so spectral filtering and narrow FOV is still a must.

TTFN

 

[itsmoked]

You could get 22 caliber type rifle scopes (cheap) and perhaps be able to greatly increase your SNL. Or really sharp focused reflectors out of flashlights. I have done the latter in several cases to great affect. Just make sure the flashlight when in flashlight mode casts a tight spot.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[IRstuff]

Depends on the size of the detector. An 80 um diameter detector would need optics with focal length of >=7.2 mm to subtend 1/2 of an 8" light at 30 ft. A larger detector would need a correspondingly longer focal length optic.

Marshall Electronics sells an 8 mm EFL lens for $37:

http://www.mars-cam.com/optical.html

TTFN