At 640x480 it's gonna nuke your pocketbook. Until prices REALLY start coming down for LWIR technology, I don't think there is going to be any truly affordable 640x480 thermal imagers.
With respect, US$4000 may not be a huge amount of money to everyone, particularly if you have a way of making money from your investment. Yes, it's not a trivial sum for an average individual: you could probably buy a reasonable used car for that money, but when you compare it with the cost of 640x480 imagers a very short time ago, it has never been more affordable.
I fully expect that 640x480 imagers will become relatively commonplace relatively soon - and the cost will continue to tumble as the non-recoverable engineering expenditure is amortised. I also expect higher resolution sensors to start emerging - at a premium price to start with - but eventually becoming more affordable.
The 384 x 288 Therm-App debuted at about US$2000 - that's 1.45 US cents per pixel. The 640 x 480 version, at US$4000, is 1.3 cents per pixel. By extension, a 1024 x 768 at 1.2 cents per pixel would be about US$9500 today, though the optics would necessarily more expensive as discussed below.
My knowledge of optics is insufficiently detailed to guess whether it will be physically possible to reduce the pixel spacing on imagers very much below one wavelength (say for the sake of this argument 14µm); this has a knock-on effect on the minimum size of the sensor and (probably) the ROIC. Someone may find a way of miniaturising the ROIC and making the necessary connections to the sensing elements, but we'll have to see. As for the sensing pixels themselves, correct me if I'm wrong I believe the minimum dimensions are essentially fixed by the wavelengths in use, rather than any fundamental limit of process technology itself. If, say, one could make a pixel one-half wavelength square then, all other things being equal, a 1280x960 pixel array would be the same size and cost in real terms as today's 14µm sensors. Higher resolutions will necessarily involve physically larger arrays and ROICs, and optics with larger image circles (=bigger, more expensive lenses).
It may be informative to compare for a moment LWIR sensors with good quality 35mm-format visible light cameras. The 45Mpix Nikon D850 has 8256 x 5504 pixels on the 35.9 x 23.9mm active area of its sensor. That's about 230 pixels per mm, or roughly 4.5µm each. They're sensing visible light with a wavelength of roughly 450nm (0.45µm) (blue light), so each pixel is about 10 wavelengths x 10 wavelengths - that's 100 square wavelengths. Today's uncooled LWIR pixels are about 1 wavelength square. If LWIR sensor designers had that sort of luxury, a 640 x 480 pixel sensor would measure about 89mm x 67mm (3.5" x 2.7") ... It might possibly be quite sensitive, but the optics would be frighteningly expensive.
With respect, US$4000 may not be a huge amount of money to everyone, particularly if you have a way of making money from your investment. Yes, it's not a trivial sum for an average individual: you could probably buy a reasonable used car for that money, but when you compare it with the cost of 640x480 imagers a very short time ago, it has never been more affordable.
I fully expect that 640x480 imagers will become relatively commonplace relatively soon - and the cost will continue to tumble as the non-recoverable engineering expenditure is amortised. I also expect higher resolution sensors to start emerging - at a premium price to start with - but eventually becoming more affordable.
The 384 x 288 Therm-App debuted at about US$2000 - that's 1.45 US cents per pixel. The 640 x 480 version, at US$4000, is 1.3 cents per pixel. By extension, a 1024 x 768 at 1.2 cents per pixel would be about US$9500 today, though the optics would necessarily more expensive as discussed below.
My knowledge of optics is insufficiently detailed to guess whether it will be physically possible to reduce the pixel spacing on imagers very much below one wavelength (say for the sake of this argument 14µm); this has a knock-on effect on the minimum size of the sensor and (probably) the ROIC. Someone may find a way of miniaturising the ROIC and making the necessary connections to the sensing elements, but we'll have to see. As for the sensing pixels themselves, correct me if I'm wrong I believe the minimum dimensions are essentially fixed by the wavelengths in use, rather than any fundamental limit of process technology itself. If, say, one could make a pixel one-half wavelength square then, all other things being equal, a 1280x960 pixel array would be the same size and cost in real terms as today's 14µm sensors. Higher resolutions will necessarily involve physically larger arrays and ROICs, and optics with larger image circles (=bigger, more expensive lenses).
It may be informative to compare for a moment LWIR sensors with good quality 35mm-format visible light cameras. The 45Mpix Nikon D850 has 8256 x 5504 pixels on the 35.9 x 23.9mm active area of its sensor. That's about 230 pixels per mm, or roughly 4.5µm each. They're sensing visible light with a wavelength of roughly 450nm (0.45µm) (blue light), so each pixel is about 10 wavelengths x 10 wavelengths - that's 100 square wavelengths. Today's uncooled LWIR pixels are about 1 wavelength square. If LWIR sensor designers had that sort of luxury, a 640 x 480 pixel sensor would measure about 89mm x 67mm (3.5" x 2.7") ... It might possibly be quite sensitive, but the optics would be frighteningly expensive.
High resolution thermal imaging is a comfort and somewhat pricey, but like having any other type of tech, like the top iphone or top macbook pro. Not everyone buys the full spec'd version, but there are more people than expected buying those which really don't need it, it's just for the sake of having a good thing. Same thing about cars, etc. If I buy a thermal imager, it's just for hobby, so guess there are more people alike.
The thing with high resolution thermal imagers is thay they need big lenses and sensors, because infrared is longer wavelenght than visible. They could do just demultiplexing like hyperspectral imaging systems do, but instead of being bulky the objective it would be the body and they require a sheer amount of bandwidth to the computer.
Corning Hyperspectral Imaging Sensor Products