The Build Your Own Thermal Camera for Under $50 Challenge

I am CoronaCameraMan (Level 2 Infrared Tech) and have sold infrared cameras for over 10 years.

The described technique of taking a grid of measurements from an area is similar to the wobble-mirror type of infrared camera which used a spot or line scan detector. The "spot size" or grid squares of your temperature detector is dependent upon distance to the surface.

A modern infrared cooled or uncooled detector is a wafer whose surface is cut into a grid of pixels that are individually connected to amplifier circuits and each pixel temperature is displayed as a color from a palette such as black cold to white hot in 40 steps is as many as a human eye can detect. So the temp range of the image is relative to the shade of grey.

but how are these non contact thermometers make infrared cameras like the ones you see on tv

50 dollars is a challenge.

Are you planning to use any focusing?

When I was about 10, I made a dish (ash-tray) in metalwork class by hammering an annealed brass disc. Some years later, when I (regrettably) started smoking, I noticed how effectively it reflected back the heat from the tip of a cigarette. You could feel the heat on the back of your hand from a cigarette held about a foot away.

Maybe something along these lines (with a properly designed parabaloid reflector) would give you better resolution?

[Edit - I was thinking of the longer-range work, like heat leakage from a house, rather than e.g. a PCB assemby]

This is a very interesting project.

Do you have an idea of your thermometer time constant?

Your device may have substantial averaging.

Melexis makes thermopile sensors, the MLX90247 is the cheapest at $20 in single quantity, but has an analog output. Time constant in 30ms.

I think the thermometers take about half a second to stabilize, but I would have to check on that. Its not really very fast for this application, but like I said its something you walk away from and come back later. You can always decrease the scan time by adding more thermometers. I will take a very slow to acquire thermal image over no thermal image.

This set with 3 bit and half display is really cheap,

I think I could buy from there fo our market if I were a businessman.

Like you, I have always been fascinated by FLIR cameras and wished I could buy or build one cheaply. I've thought about scanning, but the time involved with scanning, factoring in the response time of the sensor, sort of takes the fun out of it.

Here is an idea I've had but never got around to trying to implement it. Liquid crystal thermometers (thermochromic) are available with sensitivity as good as 0.1 C. Stabilizing the temperature so that the liquid crystal is just at the transition point would be a challenge. A lens for imaging would probably be too expensive (At 10 microns, I think they're made out of germanium). However, a copper mirror might work just fine.

Anybody think this might work?

I like your idea to use the liquid crystal thermometers. Looks like you would need to put it in some sort of temperature stabilised chamber that lets long wave IR pass into it yet hold it at the transition point. Holding it at this point doesn't sound easy but if you can do it, and if you can make a cheap lens then you would definitely have the scanning method beat.

While I agree that the scanning method is slow and less fun, I will take a very slow (un fun) and low resolution image over no image. A slow image can still be good for finding heat leaking out or hot parts on a circuit board. With the scanning method you should be able to get a decent resolution image if you are looking at something close. For the cost of the thermometers you could get 4 or more and collect data from all at once and then assemble them into one image as this would speed up the scan 4X.

Including some sort of visible laser "spot" could be a nice add-on idea too. Consider a web cam that could take a reference image of the scene and also relate the thermal data collected to a point on the image. This would make the thermal map created from the raw data much easier to interpret, and also would take slop in the system into account by placing the data right where there laser spot is actually pointed, rather than where the motors are supposed to be. It would also allow a user to direct the scan to an area of interest.

(I have no idea if a typical laser spot would be visible to a low-res web cam.)

I had thought about taking a picture to relate the thermal image to but just to be used side by side or as a rough overlay. The idea of having a web cam and a laser capture where the laser is pointed to correlate to the picture is a great addition.

Another person mentioned using mirrors rather than moving the sensor itself. What's your thoughts on this idea?

It seems like this could affect the ease of swapping in another sensor, since the sensor is fixed. (So, replacing a cheap device with one that has a laser spot might only mean changing the mounting, which would be "below" the mirror, and so have less impact.)

On the other hand, using a mirror could introduce error and reduce sensitivity, and maybe mounting the sensor to the moving platform would be easier than I image.

KISS seems like a good principle...

Using a mirror could work and would be nice for swapping out the sensor, but is not as easy as an off the shelf mirror. Most mirrors the reflective part is on the back side of the mirror so the image is going through the glass. In the thermal IR range we will be looking at, glass is opaque. The small diagonal secondary mirror used for astronomy in a reflecting telescope have the reflective side on the front and can be purchased for about $20 and I think would work. Other choices might be a really shiny flat polished piece of metal.

When I took apart the gun type hand held thermometer with the laser, the sensor part is about the size of the keychain thermometer. The sensor is connected to the circuit board with a ribbon cable so mounting the sensor part and moving it around shouldn't be too difficult. What I didn't see on the gun thermometer was an easy way to interface with it, although I am sure that could be worked out, even if it meant reading directly off the sensor.

At this wavelength, you do not need a very fine finish mirror.

A military thermal imager I worked on used a machined/polished aluminum mirror. The visible finish was not all that great.

If you have access to a lathe and polishing compound, you can go a long way for this application. Your mirror may need to be parabolic to colimate (focus to infinity) your sensor field of view.

Most modern digital cameras are more sensitive to IR than they are to normal light (well almost!). Some come with a low light IR setup, so taking pictures at night might be a good idea,,,,,

You can checkout any remote control by looking at the IR LED through the camera.

Is there a filter that would cut out normal light and only allow IR through that could be placed on any digital camera or CamCorder?

Near infrared imaging and thermal imaging are 2 different things, while both being out of the visible spectrum.

There are some web sites that show how to remove the NIR cutoff filter from digital cameras CCDs to provide ''night vision''.

Getting an IR filter is no problem, but I don't think CCDs are sensitive enough beyond near infrared (NIR) for this app (heat loss from buildings etc.) There's some interesting stuff about it on this site.

yes any lens that cant see through human eyes even if you loook in the sun can be used as infrared on any camera with or with out night vision. if you see a big fire and the smoke is to black to see the flames you will see hthe flame through the black smoke with infrared lens

to make a thermal camera you just have to use a simple theory with the ordinary camera that is you have to fix the lens with a thermal ccd sensor thats all.

sunny w

nasa.

Notes: Aluminum reflects heat well (which is why it's used in cooking)....and aluminum foil is cheap....

Secondly, look at using an old flatbed scanner....avoid having to build your own motor based scanning.

Hope to see one working...

-Donald

Maybe you could use a laser light emmitted and cadmium sulfide photocell to detect its light, then measure the changes in the timing between when the laser is activated to when the photocell detects the light. If you set up an array of these you could then take a few samples every second and find the small differences in timing, then build a color map with software that represents these timings. Because as we all know temperature effects the speed of light, that should also go for the speed of light reflected on a surface. If you wanted to scale this up, you could use an optical switch with fiber optics to create a bitmap for the lazer, then you could use one sulfide photocell as the receptor. the smaller you could make the laser points, the more fidelity you would have in the image.

"Because as we all know temperature effects the speed of light ..."

Which particular "we" would that be ?

So what happened with this? Was it a failure??

I have been researching on this, and the FLIR are extremely sensitive (search youtube videos). I'd say the $50 mark to be utopic and rather that the challenge would be to actually build a working equivalent prototype. I don't think the 5000$ plus price tag is random.

Currently i've been looking at modified webcam's and other similar sources (like wiimote) but I'm not sure they have the requiered sensitivity. I'm not sure on the color detection, but monochrome would be less demanding on calculation, while still providing a base/grid for color coding.

This is also towards a computer/laptop interfaceable solution, where the processing and visualization can occur.

EDIT: i've just found this page, which states thermal cameras use a special sensor:

http://www.maxmax.com/axrayircameras.htm

From your link: "If you want to see heat, you need a thermal camera. You cannot see heat with an NIR camera that has a CCD or CMOS image sensor."

That's why this thread started.

Thermal imaging (as in checking for heat loss from buildings) is not the same as near- or even medium-IR imaging.

Here are come comments on this topic:

http://forums.makezine.com/comments.php?DiscussionID=1950

This seems interesting. http://www.rapidonline.com/Electronic-Components/Sensors/Thermal-Sensors/TPA81-Thermopile-array-sensor/77927/kw/78-0792?source=googleps&utm_source=googleps

Any thoughts?

It is indeed interesting - but at £50 for the sensor (which would need to have a servo motor & other mechanics added to make an image scanner) - that's well over budget.

After a quick flick through, it looks as if the choice (according to that blog) is between very expensive "proper job" thermal imaging systems, or cheap'n'cheerful diy systems - one entry I spotted suggested:

"... start with one of the inexpensive IR Thermometers that Harbor Freight is selling. Figure out how to get a signal out of the thing, maybe put a pinhole over the sensor to tighten up the dot size, and then build some sort of mechanical scanner ..."

That's more-or-less what this thread is about.

Sounds great.

Although i don't discard the possibility of outsourcing and building a MUFPA thermal imaging camera, even if it requires a new thread, that would the the easiest part. ;)

Hi ,

Do you have any implementation about your ideas?

or have any test results?

The best and quickest answer may lie in a liquid crystal based version (like Guest's idea in post #7). About 5 years ago I tried to pull the materials for this together after thinking about it for 15, and if someone would take the last step where I dropped the ball there could be a payoff.
An outline of the ideal:
2 small "drums" stretched with ultra thin plastic or other material (Mylar from low-voltage capacitors? see below for other materials/suppliers). These materials have extremely low thermal mass (mass that would otherwise slow the response of the thing).
With the 2 drum faces almost or nearly touching, introduce a small droplet of liquid crystals to the edge of the gap. Capillary action fills the gap and produces a target 'screen'. Seal or otherwise protect the edge.
Either a lens or a paraboloid can be used to focus on the liquid crystals screen. Using plastic infrared Fresnel lenses might push down the cost some as compared with a germanium lens (with a resolution loss, but "Minimal Absorption Loss in the 8-14 micron Region", see below).
And I think that a silicon wafer (polished or not) might block out much or all of the visible light whatever you use for focus optics (available on Ebay).
frankd20 brought up something that made sense to me too (in #8). Making a little cold box with Peltier devices should be cheap though. I bought a stack of 11 for cheap on Ebay. Could also use one to keep things dry in there since you can get frost to condense and then flash it off by reversing the current (teensy air-lock?). You could even have the main ones cycle the temp of the chamber in response to the contrast of the image (blah-blah-blah).
Depending on the optical effects in the target, reflectance or transmission should get you something to point a cheap vid cam at. (I think there is a polarization effect in there somewhere, maybe even more sensitive than the reflectance effect per degree)
"Are we there yet?"
RR

Thermal Liquid Crystal Paints
(capsules seemed too massive for what I wanted)
As low as: 68 - 77°F (20 - 25°C)
http://scientificsonline.com/product.asp_Q_pn_E_3053484
"Water soluble paints contain microencapsulated nontoxic cholesteryl ester-based liquid crystals"

Liquid Crystal Sheets
(I think some of the cheapo "fish tank" thermometers can just be cut and squeezed-out, not these. and too much thermal mass here for anything but stand-in)
http://scientificsonline.com/product.asp_Q_pn_E_3072374
"3 to 5 micron sized crystals dispersed within a polymer matrix"

silicon wafers on ebay:
http://shop.ebay.com/?_from=R40&_trksid=p3907.m38.l1313&_nkw=silicon+wafer&_sacat=See-All-Categories

ultralight films:
(used for indoor model planes... click "Links" at left for suppliers)
http://www.indoorduration.com/IndoorDurationFrame.htm

Infrared Fresnel Lenses
(should never have sold my set...)
http://www.edmundoptics.com/onlinecatalog/displayproduct.cfm?productID=2042

Oh-yeah, might have to silver,lampblack, and/or otherwise change one half of the drum-set if light transmission turns out not to be the right sort of thing for reading the screen...

RR

I am now working on a similar project, which aims to propose a feasible and low cost method to realize the thermal imaging monitoring of object with under about 100 centi-degree without physical touching with it.

fistly, I believe thermal liquid cystal is a possilbe material, but i denified myself, because the liquid crystal is not sensitive to IR, it may not be used in un-touchlly temperature monitoring.

In addition, there is another material which is called up-conversion material, using this material the IR can be converted into visible light. I think this is a good material if it can be used or modified more or less according to the application.

If I have such material, I may have a test on it.

I know its been a while since anyone posted here but I may have a solution to one of your problems. I work with Texas Law enforcement on a daily basis as part of my work and we are always looking for low cost and durable night vision solutions. We were recently approached by a vendor that is working to get their basic uncooled units down to around $500.00 U.S. From what I understand their thermal imager uses an array of the same sensors that are used in baby thermometers, the kind they stick in the ear. Called infrared medical thermometers. Thats about all I could get out of him since he was a sales rep and not the engineer. But their systems come in at about $1500 cheaper than the competitors. Course in their case your paying for the "law enforcement mil-spec package"; lat/long, laser designators, PTZ mounts, yadda yadda... no actual quote on the thermal core itself. Hope this helps.

Happy new year 2012 to you all.

Just a thought that might (or not) help.

If you take a transistor (silicon or germanium) and expose the semiconducter part, it is sensitive to a range of wavelengths. Also, a reverse biased collector/base junction is the most sensitive as a temperature transducer. More than 30 years since I played with that sort of thing so what I say is from memory.

I like the idea of using the cheap keychain IR thermometer for the signal to reduce the spot and speed it up. Take the output signal and store it in a matrix that mataches a pair of mirror scanners out of an old junk supermarket scanner (X & Y ) or other set of scanning mirrors run off or a couple ramp signals. ( the orginal thermal cameras only had one pixel and were scanned saving the image to a buffer then asigning a color to each level in the buffer coming out with a great looking thermal image.