DIY pinhole thermal camera

[frenky]

Very nice work: http://www.ribbonfarm.com/2016/05/12/artem-vs-predator/?_utm_source=1-2-2 

It's amazing how good images he was able to record with his setup...

[Fraser]

An interesting project with great perseverance displayed by the chap who built it. I have seen earlier stages of this build previously.

Interestingly a commercial camera operating on a similar principle was the Inframetrics Thermasnap (aka Snapshot 525). The Thermasnap contained a Honeywell core that consisted of a vertical linear array of thermal detectors that were swept across the output of the optical block. A single image takes around 1.5 seconds. The array is attached to a small linear slide assembly with a drive mechanism that looks similar to a floppy disk drive.

The images that the Thermasnap produces are very good, but it can only image non moving objects or scenes. It was designed to be a more affordable thermal camera solution.

http://www.x20.org/product/snapshot-525-radiometer/

http://www.infrared1.com/ThermaSnap.html

No surprises, I own one of these unusual Inframetrics cameras

Fraser

[tomas123]

This are infrared images but not thermal images (8.000 to 14.000 nm).
Nevertheless a great work!

[frenky]

He is using 3.4 um InAs photodiode.
Quick google search shows that this can be used for measuring temperatures.

In guided missile technology the 3–5 µm portion of this band is the atmospheric window in which the homing heads of passive IR 'heat seeking' missiles are designed to work, homing on to the Infrared signature of the target aircraft, typically the jet engine exhaust plume. This region is also known as thermal infrared.

https://en.wikipedia.org/wiki/Infrared#Commonly_used_sub-division_scheme

[frenky]

An interesting project with great perseverance displayed by the chap who built it. I have seen earlier stages of this build previously.

Interestingly a commercial camera operating on a similar principle was the Inframetrics Thermasnap (aka Snapshot 525). The Thermasnap contained a Honeywell core that consisted of a vertical linear array of thermal detectors that were swept across the output of the optical block. A single image takes around 1.5 seconds. The array is attached to a small linear slide assembly with a drive mechanism that looks similar to a floppy disk drive.

The images that the Thermasnap produces are very good, but it can only image non moving objects or scenes. It was designed to be a more affordable thermal camera solution.

http://www.x20.org/product/snapshot-525-radiometer/

http://www.infrared1.com/ThermaSnap.html

No surprises, I own one of these unusual Inframetrics cameras

Fraser

Interesting. 

It would be strange if you did not have it in your collection.

[Artlav]

This are infrared images but not thermal images (8.000 to 14.000 nm).

More like IR-B/MWIR from here -


As i said in the article, Near Infrared from the silicon graph is really easy to get to - all you need is to do some surgery on a regular camera.
Going below is where the tricky stuff starts.

The images that the Thermasnap produces are very good, but it can only image non moving objects or scenes. It was designed to be a more affordable thermal camera solution.

I can't seem to find what the resolution is. Do you happen to know?

My rig is, theoretically, 2400x1200, but tough luck getting it focused to that limit.
The frame time would be about 30 minutes at full resolution, so i usually do 5 minute snaps at a much lower X one.

Anyway, while their concept is faster, it relies on an even more exotic part - a thermal linear CCD, so i haven't really considered making something like that.

[Fraser]

The ThermaSnap uses a Honeywell 128 pixel vertical linear array and IIRC it produces a 128 X 128 pixel image. I would need to check to be certain though.

Fraser

[CatalinaWOW]

Certainly a tour de force.  But some base assumptions limit the performance.

One assumption - suitable optics are expensive.  Mostly true, but with some significant loopholes.  Reflective optics work fine, as long as the mirrors are first surface.  Sensitivity is proportional to the area of the optics.  Think of the improvement using a 6 cm parabolic reflector in place of the the roughly 1mm pinhole.  That sensitivity could be traded for a faster scan rate.  Another approach would be to use NaCl optics.  Transmission is fine.  Large crystals of NaCl are relatively easy to make or buy.  The downside is that you have to keep things very dry around them.  It works well enough that Germany during WWII used this approach.

[Artlav]

One assumption - suitable optics are expensive. 

That was the assumption at the start of the project, which later got lifted by the availability of cheap ZnSe lenses from China.
Basically i wasn't able to get any meaningful results in MWIR with a pinhole and no liquid nitrogen, so i had to get a lens of some sort.

Another approach would be to use NaCl optics.

To grow a big enough chunk of table salt and then polish it optically precise down to fractions of a millimeter of curvature turned out to be a little more than what i could swallow, even without the hygroscopicity issues.

[gardner]

There are a lot of low cost IR thermometers around.  I don't know what type of sensor they use but this job is one I find on Mouser.

http://ca.mouser.com/ProductDetail/Melexis/MLX90615SSG-DAA-000-TU/

It looks like it could be read repeatedly, at least a few dozen times/sec.  You could make a strip of 8 or 10 near the focus of a cheap ZnSe lens and sweep out a band of pixels scan-wise.  You could make a megapixel image over the course of a couple of hours.

[Ben321]

VERY interesting project. Instead of large gain, maybe try longer exposures? Gigaohm resistors are VERY rare. I hadn't even heard of them until I read the article you linked to in here. I've seen Megaohm resistors though.

And also, instead of reading the photodiode with an A/D converter chip and a microcontroller and SD card, why not read the photodiode with a regular PC with its soundcard's microphone port? The microphone port has a DC voltage superimposed on its signal pin (which is used to power the pre-amp built into electret microphones), and this can be used to read a photodiode or photoresistor type sensor. The more your sensor conducts the more current flows. Since the DC voltage on the microphone port is superimposed onto the signal line via a pull-up resistor, an increase in current being drawn by the port will result in a decrease in voltage on the signal line. This change in voltage will pass through the ac-coupling capacitor (dc blocker) that couples the microphone port to the signal amplifier on the sound card. A changing brightness level will get a signal into the computer, but a constant brightness level won't. To compensate for the effect, one can then integrate the samples of the recorded "sound" using any software that allows this (I use GoldWave for my audio experiments). This should be able to restore the original signal.