I honestly do not know what the story is behind the noise injection but it is clear that it is a function that has the ability to degrade the image noise performance, so to me that has to be an engineering solution to a marketing requirement to artificially create a performance difference between the various cameras within a particular series and between ranges of camera. This is the only scenario that makes any sense as degrading a products performance is not normally something a manufacturer does without a very good reason, Be it longevity of operation or a specific requirement to meet a well defined objective. I remember talking to the VP of consumer products at FLIR about why the FLIR One G2 App did not offer a non MSX image option. He said the marketing team were very aware that the Seek Thermal competing camera had a higher resolution that could appear more detailed than that of the FLIR One G2. They had many meetings and insisted that MSX be permanently engaged to make the image appear more detailed than it truly was. They did not want the user to see the raw thermal image of the Lepton 160 x 120 pixel core. The marketing team have a lot of power in an organisation as they help to ensure good profits, along with the bean counting accountants of course
Many a product has been ruined by the combined efforts of the Marketing team and the Accountants.
I had another look at the figures I provided above and had a thought. It does not fit all cases, but could it be that the “Target Noise” is in fact a number that the camera should achieve after noise injection ? That is to say, in the case of the E4, the Target Noise setting is 135mK ands the stated camera specification is <150mK. With the expensive P620 the Target Noise is set to 60mK and the specification for that camera is <65mK.
This makes some sense until you look at the E60. That has a Target Noise of 35mK and a stated NETD of <50mK? 35mK is high performance from a microbolometer so why do we see improvement in the image when the Target Noise generator is disabled or set to 0mK ?
The E8 does not fit the theory either as why would FLIR set its Target Noise to 5mK…. An NETD that the microbolometer could never achieve !I would suggest that a FLIR microbolometer would normally produce an NETD of between 40mK and 70mK in its factory released state (not tuned for highest NETD performance).
As I said in the post above …. It is all a bit weird !
Fraser