EEVblog Electronics Community Forum
Products => Thermal Imaging => Topic started by: Lambda on February 17, 2021, 12:29:05 pm
-
Hello.
I was wandering myself about the following subject:
-What is, in a thermal imager/camera available in market (non cooled micro-bolometre sensor), the contribution to the background signal or noise of the surrounding material, elements constituting the imager/camera by itself?
More precisely:
If we consider an objective made from a Germanium lens (for instance), settled in a cylindrical housing, with the sensor at the focal plane of the lens, and that the housing is at ambient temperature. Does this housing, being in proximity of the sensor, will not radiates thermal radiation, from its internal surface, which could be caught also by the sensor, in addition to the image of a target normally focalised on the sensor?
Would it not add a certain amount of noise to the picture, washing the contrast? For instance like with a classic VIS camera being not well baffled against external parasitic light?
If it is the case, can we imagine an improvement by thinking as following?
- on a commercial unit, thermal core, to simply cool down at least the housing of the lens system with a Peltier based system, for instance, to drop down the temperature with a delta of, let' s say, few dozen degres C from ambient temperature (we would just have to be sure to not create condensation in the housing of the lens).
- using a specific material with a very polished internal surface for the housing of the lens, with the lowest emissivity as possible....
It is something reachable for hobbyists, i think.
About cooling, astronom amateurs build often something similar for directly cooling CCD/CMOS sensor for their camera and decreasing thermal noise....This analogy is off course just for illustrating the level of feasibility of such modification in a DIY spirit, nothing more....
Here the idea would be just to cool down the immediate environment of the sensor, and limit parasitic thermal emission toward the surface of the sensor.
In this idea, the sensor itself would not be cooled for avoiding, perhaps, misadjustment during the periodic calibration steps, or simply messing with the genuine with factory calibration (shutterless or not), or simply avoiding to risk to damage the sensor by itself.
In advance, thank you for your teaching on this point....
Cheers.
Stéphane
-
For a system with an f/1 lens, the housing temperature variation changes the signal by about 3x as much as the scene.
To see why, think of being a pixel gazing upwards into the optics with a simple single element f/1 lens. Your response is simply averaging the hemisphere of incoming radiation.
The scene is only coming through a small part of that - a 57° (1 radian) or a 23.5° half-angle cone although favourably positioned.
The optical analogy is difficult, as there is no thermal 'black' - 'no signal' is really 'grey'.
In the simplest approach the flag is assumed to be the same temperature as the lens body, so the difference between flag closed and flag open is the scene. When this is close enough to the truth you see artefacts like a 'porthole' with brighter edges in the image (ie hot lens) etc.
Cooling electronics will help with noise, but how to stop that kind of thermal gradient becoming shading / false signals..... There are some detectors with an internal cold shield to achieve what you suggest, preventing the sensor seeing the housing at all, and as energy in the 8-14 falls steeply below 10°C the scene becomes more dominant.
Bill
-
Microbolometers measure a temperature variation against a reference.
That reference being made of many things like the silicon die temperature itself, the infrared flux not coming from the scene as you described, voltage references, etc.
Your goal is to make all these constant, not necessarily minimal, because you just subtract them --> that's the easiest way I can see a hobbyist getting more performance.
Alternatively you can constantly "measure" them and update your subtraction --> that's the advent of TECless cameras for example. That's harder to implement as an aftermarket hack.
-
Bill W, Bap2703,
First of all, thank you for your explanation (Sorry, a bit long to react on this thread i was... :-[ )
... When this is close enough to the truth you see artefacts like a 'porthole' with brighter edges in the image (ie hot lens) etc.
...
Bill, i recently did the following constatation in this sense. I firstly observe, from my balcony, a urban landscape during a cold evening with my camera just pulled out from my warm living-room.
- Housing of the lens and body of the camera was simply "mild/fresh" when touching it: the image provided by my system was normally good, except indeed this "porthole" effect, like a brighter "haloish" washing out a bit the peripheral zone surrounding the central area of the image. A bit like a vigneting but on a "clear version...
then,
- I let the whole camera cooling down few minutes on my balcony, and i came back to repeat my observation. The bright "haloish" "porthole" disappeared and the image what with a normal distributed brightness, according the landscape.
Therefore, indeed, cooling (even just) slightly down the immediate environment of the sensor, would not harm, and keeping constant the whole temperature of this surrounding, would help to maximize the SNR provided by the sensor.
Thank you!
Best regards
Stéphane
-
It has always amazed me to think that LWIR imaging is rather like trying to take a normal photo through a white-hot lens.
The other thing that bakes my noodle is the fact that an uncooled LWIR sensor can see things cooler than itself. Logically, of course, I know the answer but it's still amazing.
-
Therefore, indeed, cooling (even just) slightly down the immediate environment of the sensor, would not harm, and keeping constant the whole temperature of this surrounding, would help to maximize the SNR provided by the sensor.
I thought you were talking about the temporal noise.
What you see is spatial noise and if you want to achieve better image quality there's a lot even a hobbyist can do, provided you do your own image processing.
The easiest is to do the NUC including your lens. Going crazy you can do that for several lens;chip temperatures.
You can also try to guess the fixed pattern noise if you're looking at a changing scene.
-
Hello bap2703, yes indeed i was focused on the spatial noise, and following your suggestion, it makes me reminding that one thing i practice in astronomy is live-stacking for increasing the SNR of my resulting image. It works well and it should be easily applied for images coming from a thermal camera...
For the NUC modification, i remembered also this very interesting thread.
https://www.eevblog.com/forum/thermal-imaging/seek-thermal-hardware-modification-to-improve-image-quality/ (https://www.eevblog.com/forum/thermal-imaging/seek-thermal-hardware-modification-to-improve-image-quality/)
and i remembered also your sentence:
Beware using 'noise' as 'spatial noise'.
To reduce the residual spatial noise you need to stabilize the temperature, not cool without any control.
I will investigate in this direction.
Best regards.
Stéphane