DIY Thermal Imaging

[jaxbird]

Really nice! Did you measured the response time of the thermometer after each measure? (maybe you can hack the thermometer board and read the raw values directly from the sensor?)

Thanks

I did have a quick look at reading the sensor directly, but it's very high impedance, so just trying to measure how to read it already radically change the output, even just extending the sensor wires seem to change the output. The unit also appears to hold some calibration. Not sure if it's really calibrated (as it's so cheap) but it does appear to give fairly accurate readings close to it's spec of something like +/- 1C accuracy and 0.1C resolution, as long as what is being measured is larger than the spot size.

I'd be interested if anyone got experience in sampling one of these sensors directly. Not sure what the settle time is, but I believe it could drastically improve resolution/speed.

[tom66]

What would be required to get a say... 1000:1 spot ratio? Very long tube? Different sensor?  Which would allow for 10mm pixel size @ 1m distance (quite usable for basic imaging)

I would be interested in what type of sensor these use. Wonder what response time is?

[echen1024]

A relative newbie to MCU's, so I have a few questions:

1: Can you post the code? I might try to port it to Python and use it on a Raspberry pi.
2. Is the servo absolutely necessary? I think I should just be able to use the sensor and scan it.

Thanks.

[lewis]

Very interesting project, keep up the good work. 

[jaxbird]

A small experiment on the resolution and precision vs distance. All scans at what I would call medium speed and covering a smaller area than the previous scans, to speed up the process.

I took a small 1/8W resistor, 47R I think, and hooked it up to a power supply to create a small heat source. Approximately 2x4mm in size.



Then I did a few scans at different distances (ignore the backgrounds, not at a stable temperature)

First at ~50mm distance, reaching 54.7C, that is probably very close to the actual temperature.



And at ~100mm, it's down by ~5C



150mm, down by ~13C



And finally at 250mm distance it's still clearly visible, but down by 18C.



The resistor is visible out to about 500mm, after that it's no longer visible.

[jaxbird]

What would be required to get a say... 1000:1 spot ratio? Very long tube? Different sensor?  Which would allow for 10mm pixel size @ 1m distance (quite usable for basic imaging)

I would be interested in what type of sensor these use. Wonder what response time is?

100:1 should do it ..  Based on my resistor experiment I believe that it will easily pick up an isolated 10x10mm heat source at 1m distance with the current ~12:1 ratio, as long as the delta temperature is more than ~5-10C from the surroundings, but it will report a reading lower than the actual temperature when the source is significantly smaller than the spot size.

From what I can gather (pure speculation, I'm no expert in this) the ratio is easy to calculate based on the diameter and length: (my unit says 12:1 so that is where I got the number from)

The focus part of the construction appear very simple like this:



with a small fresnel lens at the bottom, so based on diameter and length it should be possible to get a fairly accurate ratio. Do note there are some small ribs inside the tube, most likely to prevent reflections from making it all the way to the lens.

[MacAttak]

The ribs are baffles - same thing used in telescopes. Yes, to reduce reflections (which add noise). The inner edges of the baffles need to be machined properly - burrs or other irregularities will just cause more harm than good. The alternative is to just use a good blackbody coating on the inner surface of the tube (or at least something highly absorptive of the wavelengths you are collecting).

Good baffling helps produce higher contrast and sharper measurements.

[MatCat]

To get better line scan resolution I recommend either a good digital servo like those used on helicopter tails, or a good resolution stepper motor .  Looks awesome though, making me contemplating a trip to harbor freight for one of those temp guns! (I actually have one but I don't want to take it apart hehe).

[edavid]

The ribs are baffles - same thing used in telescopes. Yes, to reduce reflections (which add noise). The inner edges of the baffles need to be machined properly - burrs or other irregularities will just cause more harm than good. The alternative is to just use a good blackbody coating on the inner surface of the tube (or at least something highly absorptive of the wavelengths you are collecting).

Regular black won't work, since it re-emits IR.  Is there a paint that will absorb IR in the passband (but not anything above that) and emit it at even lower frequencies?

[MacAttak]

These folks make a number of products specifically designed for IR absorption. No idea if you can actually get any of it, nor how expensive it might be. But it sure looks like the bee's knee's when it comes to IR and optical sensor stray light control.

http://www.acktar.com/


edit:
Actually, it looks like you can buy some sample coated foil for about $100 or so: http://www.acktar.com/category/Sample_Order
They don't seem to offer any samples of the coating in liquid form (one of the sample sets includes metal that has been pre-coated)

[jaxbird]

The ribs are baffles - same thing used in telescopes. Yes, to reduce reflections (which add noise). The inner edges of the baffles need to be machined properly - burrs or other irregularities will just cause more harm than good. The alternative is to just use a good blackbody coating on the inner surface of the tube (or at least something highly absorptive of the wavelengths you are collecting).

Good baffling helps produce higher contrast and sharper measurements.

Thanks for detailed explanation .. I was just guessing based on common sense, also it makes sense that the actual surface of the baffles should be irregular to make the noise as minor and random as possible. Anyway, this is from a <$15 device, so I don't really expect any attention to that level of detail

I guess some kind of fairly random open foam would work as well inside the focusing tube.

[jaxbird]

To get better line scan resolution I recommend either a good digital servo like those used on helicopter tails, or a good resolution stepper motor .  Looks awesome though, making me contemplating a trip to harbor freight for one of those temp guns! (I actually have one but I don't want to take it apart hehe).

Yeah the slack in the servo gears did cause some problems, easily to see if you look in the first couple of pictures, but I found that if you only scan from one side the accuracy is much improved, as in scan from x to x + 100, then instead of scanning on the way back from x + 100 to x, just move the servo back to x and restart scan at x to x + 100.

It's just kind of a fun little project, but actually somewhat useful for thermal imaging of where you don't need great resolution and you got a few minutes to do a scan.

[jaxbird]

These folks make a number of products specifically designed for IR absorption. No idea if you can actually get any of it, nor how expensive it might be. But it sure looks like the bee's knee's when it comes to IR and optical sensor stray light control.

http://www.acktar.com/


edit:
Actually, it looks like you can buy some sample coated foil for about $100 or so: http://www.acktar.com/category/Sample_Order
They don't seem to offer any samples of the coating in liquid form (one of the sample sets includes metal that has been pre-coated)

Thanks man .. that is interesting, but the $100 kind of ruins the very low budget of this experiment ..

That money would probably be better spent on just buying a precision sensor.

I looked around a bit, and this is actually not a new idea, someone even made a business out of it: http://www.cheap-thermocam.net/

They recommend this sensor: http://www.futureelectronics.com/en/Technologies/Product.aspx?ProductID=MLX90614ESFDCIMELEXIS3003055

I might just get one of those, it's got digital and PWM output, surely a lot easier to deal with than the elaborate display hacking I did for this one