Up until last year, owners of thermal cameras were the envy of everyone. Luckily, with the launch of Seek Thermal, Flir One and Flir One V2 recently, that's no longer the case. Thermal cameras are now affordable and they're already being applied in a variety of hobbies. Our focus is on use in electronic design and debugging, so we won't be looking at dogs, deers, houses or make thermal selfies.
Seek Thermal XR is an improvement over the original version, in that it has adjustable focus. This comes with the disadvantage that is has a narrower field of view, but in our case, it doesn't matter. The smallest focusing distance appears to be somewhere around 12 cm, which allows us to get decent detail even for 0402 footprints, in spite of the low resolution (206 x 156). Macro mode was only possible with the original version if an additional lens was used.
As you may already know, there are two versions available, one for iOS and one for Android, which you can buy off of Amazon for about $300. The issue everyone is struggling with is that the Android version comes with a micro USB connector, which can't be rotated. This means that if your phone has the USB connector facing in the other direction than how it should be, the camera will be pointing towards you, like when you're taking a selfie, rather than pointing forward, as you'd like it to, when inspecting something.
At the moment, there are only apps for Android and iOS, but Seek Thermal has announced an SDK for early this year, which would allow the development of custom applications for all other platforms. "Early 2015" has gone by already and the release period was moved to later this year, so there's still no SDK out there. The good news is that this hasn't stopped anyone from writing their own software while talking directly to the hardware, so a bunch of applications have already popped up.
Personally, I prefer having it connected to the PC, because I can easily save photos and using it connected to a cable allows me to mount it somewhere so I can use it hands free.
Let's take some pictures!
This is probably what you're most interested about, so let's get to it. The following pictures are of a board I'm working on and they're taken with a custom application I wrote.
Distance: 12 cm; Approximate picture dimensions: 38mm x 28mm.
Upper left corner: QFN32 Package (0.5 mm pitch).
Bottom right corner: 1206 resistor.
I think it's important to note that you can distinguish the individual pads for the QFN package. Definitely better performance than I expected.
This picture is obtained by moving the camera farther away from the board.
Distance: 18 cm; Approximate picture dimensions: 65mm x 47mm.
If we move the camera even farther, we get the following image, which is still not bad and offers plenty of information.
Distance 29 cm.
As you can see it's very usable for electronics work, but there's something that I didn't mention until now: the images generated by the original application are worse! I believe this is due to the fact that the factory calibration is not properly done, particularly the gain calibration, which is used for flat field correction. I didn't put this hypothesis to the test yet (I might in the near future), but judging by the fact that I had to use an extra calibration frame (obtained by pointing the camera at an object that had the same temperature across its entire surface), I would say that that's a fair guess. Another possibility is that they're not compensating for the drift in gain, caused by the camera heating up, but again, I'm just speculating.
With that in mind, I think the original program is doing quite a good job, since it doesn't require any periodic, manually obtained extra calibration frame and if they ever decide to add this as an option, I'm sure they'll get much better images from their own cameras than any third party developer can.
For reference, this is the same picture as the first one, but taken from the phone:
That's it for now. I know many of you are planning to use this camera for repair jobs, so I'll follow up with more pictures as soon as I can get my hands on a broken TV.
An interesting thing I discovered while playing with this thermal camera is that it sends the value of each microbolometer directly, giving it full dynamic range, or at least it seems it does. What this means is that it can capture both a 300 degrees spot and a 30 degrees spot, in the same frame, without losing details. This might be true for all thermal cameras (and it makes sense if it is), but it was a surprise to me because I was expecting the same behaviour you get from visible light cameras.
It's obvious to me that it can be successfully used in this field, however, while the original application delivers decent results, I think you're better off using a custom application at the moment, which might require a bit more work to set it up, but can deliver higher quality images. My only hope is that they'll release the SDK soon enough and that it will offer enough flexibility to allow 3rd party developers to extract the best images possible from the sensor - it's clearly capable of more.