Just as the gold standard for multimeters and other instrumentation likely comes in a yellow package of some sort, there is a similar household name for thermal imaging. But, if they’re known for anything other than the highest quality thermal cameras, it’s excessively high price. There are other options around but if you want to make sure that the finished product has some sort of quality control you might want to consider building your own thermal imaging device like [Ruslan] has done here.
The pocket-sized thermal camera is built around a MLX90640 sensor from Melexis which can be obtained on its own, but can also be paired with an STM32F446 board with a USB connection in order to easily connect it to a computer. For that, [Ruslan] paired it with an ESP32 board with a companion screen, so that the entire package could be assembled together with a battery and still maintain its sleek shape. The data coming from the thermal imagining sensor does need some post-processing in order to display useful images, but this is well within the capabilities of the STM32 and ESP32.
With an operating time on battery of over eight hours and a weight under 100 grams, this could be just the thing for someone looking for a thermal camera who doesn’t want to give up an arm and a leg to one of the industry giants. If you’re looking for something even simpler, we’ve seen a thermal camera based on a Raspberry Pi that delivers its images over the network instead of on its own screen.
A problem I’ve noticed with my cheap low-resolution thermal camera is that its design makes it easy to miss small hotspots. The way the image is interpolated over the screen might lead you to believe that you’re sensing data for the entire field of view, just at a low resolution, averaged for each pixel. What really seems to be going on is that each sensor “pixel” is only responding to a tiny area (just a fraction of the pixel’s logical area), and then that reading is interpolated over the whole pixel. Thus if the rays traced through the pixels don’t happen to hit the hotspot, it will be missed. At least, something like seemed to be happening at the time.
Interesting. Useful to know. I have not noticed that on mine (yet).
It implies that we could so a sort of “superresolution” by pixel-shift dithering it around.
Must try that.
There is always the “InfiRay T3S” 384×288 resolution @ 25 fps
got a phone with the next in line infiray at only 256×192 at 25fps…
That thing work so great…
but the phone is a brick, lol
“InfiRay T3S” – ooof $750 bucks sans phone.
That will drop, once they release newer models with higher specs. Or it might make ITAR allow higher spec US items be sold outside the US which would force the price to drop. Nothing drops price point like a flooded market.
the lower spec infiray I have on my phone, which is only 256×192 at 25fps is definitely much cheaper, a whole phone with one inside cost only 300$… if you do not have specific needs, going the 750$ model is definitely too expensive.
There’s a “name brand” thermal imager where I work and I’ve tried to used it to spot shorted componenets and boards, but oddly it does a horrible job. We also have an asian handhelp thermal imager, but the problems are the same. Lighting and reflections are everywhere.
I’ve had great luck with the small lipstick sized electronic thermometers that cost under $10 USD from China. They can be slowly swept over a board and on degrees F they can resolve temperature going up or down until the hot spots are found.
BTW I tried using a fingers, but stopped after I ended up with a Xilinx logo branded into my finger for a week…
Thats the ultimate fanboy body monification, did you post and tag xilinx? :)
yep, I decided decades ago to NEVER use my finger ever again to test for hot spots..
Lick it and flick it, before committing to a long touch. Works for testing mains voltage as well :)
That’s like universal advice, always true ;)
Username checked out
If you can stand the thermal shock, cold spray (upsidedown canned air)
I don’t know your situation, but many thermal imagers are far focus devices. We have an imager with a zoom macro lens on it and it’s absolutely essential for finding hotspots on boards. I also have a FLIR with a 3d printed lens adapter and some ZnSe lenses. That thing’s so good you can see what part of an atmega328 die is hot through the packaging on a QFP, but only because of the extra optic.
I would agreed with the ‘far focus’ issue on infrared imaging devices, as the electromagnetic waves in the infrared spectrum (BTW: even as low as the FM radio frequencies [read: microwaves]) are longer than what humans see. Even in the early days of RADAR an object could not be evaluated within an 1/8 of a mile from the site.
You could increase resolution mechanically. Use a slightly flexible mount for the camera that allows the camera’s aim to change with slight movements to the camera’s board. Attach something like a pager motor parallel to the camera’s board aligned with the optical axis. This motor will shake the camera in a controlled way. Use a hall effect sensor to detect the counterweight position. Stitch the frames together. Frame rate is 64Hz, so the ‘pager motor’ would need to spin around 8Hz. A large counterweight and stepper motor might work.
Sounds like “IBIS from hell” :)
IBIS?
Unless I’m mistaken, these very things are all over Aliexpress and the usual sources. What might be unique here is the software.
I just got an 256×192 25 FPS USB-C thermal cam from aliexpress for <$250, which I believe is a very reasonable price for what you get.
Not very hacky – I was sure it will only work with the manufacturer-supplied Android application, but well well well.
When plugged into a laptop, it's shows up as a regular UVC video device and some smarter-than-me guys (on eevblog forums) have already hacked together a proper ffplay line to get the same image I see in the mobile app. Should last me years, until 1080p thermals will become affordable.
Reasonable? It’s a STONKING GREAT price for something that was previously unobtainable at *any* price. (thanks, FLIR and your ITAR-backed monopoly! Die in a fire!)
Oh god pyroelectric detectors suck.
Better bets are getting an IRAY microbolometer like the T2Pro or the HT-301 for higher res. I am the proud owner of 28 thermal cameras, all aside from the Lepton 1 are 320*240 30hz minimum. My best scores are a pair of VGA cores at $500 each and a 328*288 50hz handheld unit at $75.
The hz rating is the real weakpoint for what I want these things for. I want thermal overlay on NVGs or thermal overlay on rifle optics. Both are available and both are vastly cheaper then they were 5-10 years ago. But their still not exactly cheap. You can get a good gen3 NVG monocle and a thermal overlay for 5-6k now which is just nuts.
Beautiful work, [Ruslan]! Much more complete than my attempt back in 2021: https://hackaday.io/project/181230-simple-thermal-camera
Thanks for sharing.