Cheap IR sensors for horizon detection.

[dydx]

Have a look at these: http://www.futurlec.com/PIR_D203B.shtml

These IR sensors appear to operate within the IR range specified here: http://www.ctie.monash.edu.au/hargrave/horizon_sensing_autopilot.pdf

They are alot cheaper than the MLX90247-ESF-DSA specified by paparrazi: http://parts.digikey.co.nz/1/1/225363-ic-sens-thermopile-w-therm-39-mlx90247-esf-dsa.html

Any thoughts on how these will perform for horizon sensing? I don't know much about thermopile/IR technology.

[Blue Sky]

I may be off the mark on this and the PDF doesn't specifically say, but I believe the internal IR senors are connected such that their outputs cancel out. The output from the FET is the differential response between the sensors. These sensors are designed for motion detection; not for the detection of ambient light levels. You might still be able to make them function, but for our purposes you would have to cover up one of the internal sensors. (Mask off half the window on the device.)
-Dave

[dmgoedde]

The response bandwidth is very important for horizon detection in a UAV application. It should be in the 20ms range (around 50Hz) like the Melexis and other brands of thermopiles I know of.

Top of page 3 seems to indicate response is VERY slow on the order of 1 second. If true, these won't work for small aircraft.

Three things you want in a thermopile for UAV App:
1) Response cutoff around 5 microns
2) fast response in the tens of milliseconds
3) large view angle, around 100 degree FOV

[rich smith]

"Cheapest" IR horizon detector is short wavelength IR LEDs (about a dime ea). Not quite as good as thermopiles but worked for my solar UAV project.

[small_rcer]

Explain further if you could please?

Jim H

[Blue Sky]

They are using a 1HZ optical chopper for test purposes.
If that represents the actual bandwidth of the sensor then it is indeed useless.
Perhaps it uses a thermalcouple as a sensor?
Electronic Goldmine has these for cheap, but no specs.
http://www.goldmine-elec-products.com/prodinfo.asp?number=G15801
I have some ideas on how they may be made to work and it has piqued my
interest as a hobbiest. I'll let you know what happens.
-Dave

[rich smith]

Explain further if you could please?
Jim H

If your asking for more info on LEDs the point is they also perform well as photovoltaic light sensors. You can verify this by connecting one to a meter set for 200mv. Point it at sky, ground, sun, and other objects to see voltage difference.

Even common 2 cent red ones can be used to level plane but IR types are a little more sensitive to longer wavelengths. Two of them fed into the +- inputs of an op amp give a reasonably good indication of position on that axis. Biggest issue is they also respond to visible light which degrades performance in this application. They also are a little too sensitive so may require some masking.

Thermopiles with their dielectric mirrors are more sensitive to longer wave IR (heat) than LED and pretty much ignore visible light so work better but cost much more than couple cents. The ones used for motion (PIR) have multiple elements behind one mirror and differential amp built in so I don't see how they could work for horizon sensing.

[Blue Sky]

"The ones used for motion (PIR) have multiple elements behind one mirror and differential amp built in so I don't see how they could work for horizon sensing."
I have some of the inexpensive Goldmine units on order and some ideas for their use
that I'll leave for the next exciting episode. First I'll have to see if they are fast enough to use at all.
-Dave

[rich smith]

They're much cheaper direct from china. You can buy 20 for the price of one FMA type. I'm really curious to see how they work out. Please keep us tuned in.

PS time constant for the ones I've used are in the order of 30ms which should be ok but that's not what worries me.


"The ones used for motion (PIR) have multiple elements behind one mirror and differential amp built in so I don't see how they could work for horizon sensing."
I have some of the inexpensive Goldmine units on order and some ideas for their use
that I'll leave for the next exciting episode. First I'll have to see if they are fast enough to use at all.
-Dave

[Blue Sky]

Got the sensors!
So here's the plan. The problem isn't the differential connection of the sensors, that is actually a good thing. The problem is that both of the internal sensors point in the same direction when we need them to point 180 degrees to each other. What we need are two mirrors set opposite to each other at 45 degrees to the surface of the sensor. This gives a reflected angle of 90 degrees for each mirror and a total for the two mirrors of 180 degrees. Just what we want! By shaping the mirror you can also change the field of view. A slight sigmoid shape should increase the field of view. Too clever by half? Perhaps. I have some lm358 op-amps that should be suitable for use with these guys. Here are a few pics. I'll play with 'em soon as I have the time. Feel free to try it on your own!
-Dave

[rich smith]

This would be an incredible discovery if it works. I've got thousands of these cheap sensors. I don't think you even need the op amp because it's built in. At least it is in mine which are TO-5 case unlike yours.


Got the sensors!
So here's the plan. The problem isn't the differential connection of the sensors, that is actually a good thing. The problem is that both of the internal sensors point in the same direction when we need them to point 180 degrees to each other. What we need are two mirrors set opposite to each other at 45 degrees to the surface of the sensor. This gives a reflected angle of 90 degrees for each mirror and a total for the two mirrors of 180 degrees. Just what we want! By shaping the mirror you can also change the field of view. A slight sigmoid shape should increase the field of view. Too clever by half? Perhaps. I have some lm358 op-amps that should be suitable for use with these guys. Here are a few pics. I'll play with 'em soon as I have the time. Feel free to try it on your own!
-Dave

[Blue Sky]

Well, here are the tentative results.
The differential circuit with the mirrors works!
The mid-range voltage out is about .5V.
Using my thumb as a heat source increases the voltage by 10's of millivolts
on one side and decreases it by 10's of millivolts on the other.
Response time was seriously slow, fully 1Hz for a full signal swing.
Waving my hand quickly, maybe 3Hz, resulted in a much attenuated output.
I hooked the output to a scope hoping maybe the lag was in my voltmeter but no dice.
Perhaps when used with heat sources more subtle than my thumb the response would be quicker. Maybe.
It's obvious it will take an amp with considerable gain to detect the subtle changes in output.
No free lunch, but perhaps it could still work out to be a cheap date.
Hopefully other similar sensors may be better suited to our cause.
-Dave

[rich smith]

Well, your results are encouraging enough for me to try with my TO5 type devices. They are much quicker in PIR apps than 1s so maybe better results. Thanks for a great idea.

[small_rcer]

An off the wall idea--

If these devices are so inexpensive, why not use 2 devices. Combine the input in a single device, and then use another alongside it pointing the way you want. Or could you just use the signal from one of the elements in the device and ignore the other. Using half of each of 2 very cheap items acting individually could be still cheaper than 1 very expensive device.

Eliminate the mirror device. Use 2 devices. Blank off the input to one of the elements. Maybe I have missed something in how they work, but maybe this idea will stimulate an alternate approach.

Jim H

[Blue Sky]

An off the wall idea--

If these devices are so inexpensive, why not use 2 devices. Combine the input in a single device, and then use another alongside it pointing the way you want. Or could you just use the signal from one of the elements in the device and ignore the other. Using half of each of 2 very cheap items acting individually could be still cheaper than 1 very expensive device.

Eliminate the mirror device. Use 2 devices. Blank off the input to one of the elements. Maybe I have missed something in how they work, but maybe this idea will stimulate an alternate approach.

Jim H
Might work if they are fast enough.
-Dave

[rich smith]

Great idea but I like the mirror trick. Lighter and cheaper still. Can't wait to try it in a plane.

An off the wall idea--

If these devices are so inexpensive, why not use 2 devices. Combine the input in a single device, and then use another alongside it pointing the way you want. Or could you just use the signal from one of the elements in the device and ignore the other. Using half of each of 2 very cheap items acting individually could be still cheaper than 1 very expensive device.

Eliminate the mirror device. Use 2 devices. Blank off the input to one of the elements. Maybe I have missed something in how they work, but maybe this idea will stimulate an alternate approach.

Jim H