Shutterless NUC using Thermopile Array

24 April, 2019

Summary
This method uses a low-resolution, low-cost thermopile module (multi-pixel array, optics, and electronics) in tandem with a shutterless infrared camera to provide non-uniformity correction and absolute temperature error correction. Eliminating the need for a shutter would reduce the height from the Infrared package by several mm as well as significantly improving thermographic accuracy. Reducing the camera height would remove a significant barrier to integration in mobile devices. Also, improved thermography would address a key end-user desire. Removing the shutter also helps remove a transient heat-source in the Infrared camera that also leads to thermographic instability.

The Infrared and thermopile module would be located side-by-side and be aligned to give roughly the same field of view. Essentially, the thermopile array provides a very low-resolution (8×8) but thermographically accurate image while the Infrared camera provides a much higher resolution (120 x 160) thermal image that is susceptible to non-uniformity drift. Since the non-uniformity errors are typically very low frequency, a low-resolution image can be used to estimate them. Image processing would allow the low-resolution data from the thermopile to correct for per-pixel offset errors in real-time.

Problem Solved
Infrared cameras have at least two major challenges for integration in to mobile phones:
1) module height and
2) thermographic accuracy.
Currently, a shutter is required for thermographic accuracy but adds approximately 1-2mm of height to module. The shutter adds non-uniform heating to the infrared camera when it is actuated, further degrading thermographic accuracy. Finally, even with the shutter, the thermographic accuracy is still +/- 5C which is not good enough for many high-volume mobile applications. Thermopile arrays are known for having very good stability and accuracy but low sensitivity and large pixel sizes relative to bolometers. They are typically used in sensing applications as opposed to imaging applications due to the low resolution.
Mounting an infrared camera and thermopile module side-by-side would allow the device controller to use the low-resolution image from the thermopile to estimate the non-uniformity error of the infrared camera and provide a pixel-by-pixel correction map. This map would be updated in real-time, which would provide thermographic accuracy to every pixel in the infrared image. Thermopile modules are available in 8×8 and even 32×32 resolution, which should be sufficient for estimating the non-uniformity errors. Thermopiles are cheap enough that there would likely be a minimal cost difference between the thermopile module and shutter at high volume. Also, reliability of the thermopile sensor, with no moving parts, should be superior to a mechanical shutter. Finally, it may be possible to eliminate other costly aspects to the Infrared camera design such as thermistors and the traces/electronics supporting them.

Description of Solution
A brief overview on how to implement the concept for a shutterless NUC using Thermopile Array
– A shutterless infrared camera is mounted next to a multi-pixel thermopile module (including optics)
– The two imagers are aligned so that the field of view is similar, and the images can be registered to each other
– The low-resolution thermopile module is used to estimate the non-uniformity error of the Infrared
– The Infrared image can be down-sampled to resolution of the thermopile array
– The differences in the down-sampled image and thermopile image are assumed to be error in the Infrared imager due to pixel non-uniformities
– The difference map can be fit to function and then applied as a correction map to the Infrared image
Estimate of non-uniformity:
– The non-uniformity can be fit to various functions in order to create a pixel-by-pixel correction map.
– In this case, the low-resolution image has been fit to a thin-plate spline using the Matlab curve fitting toolbox.
– The correction map can be directly subtracted from the Infrared image.


By Jordan Hall