Economy thermal imager?

I'm with John on this one. The first CCDs were largely in the visible with a strong peak sensitivity in the *near* IR and with a very big push to get them out to much longer wavelengths on somewhat exotic materials. I knew an astronomy imaging group that were given new military chips to test because they could turn one into a fully working prototype camera way faster and cheaper than the approved contractors.

Rough graphs of the typical sensitivity for silcon CCDs with front and back thinned window construction are online at:

Their response does not extend much beyond 1000nm which is still an order of magnitude short of the roughly ~10um IR wavelengths needed for thermal imaging at ambient temperatures.

Eventually they did get longer wavelength CCDs working, and they stopped at a particular point. As the man said "what we have is good enough to see what *we* need to see". Astronomers were a bit disappointed that after that they had to pay for their own chip R&D. It didn't stop terahertz sensors eventually being made though. Strangely the terahertz image of my favourite object Cass A appears to have been removed from the web.

That only gets you near infra red. The thermal band for things in the temperature range 0-100C is much more tricky and generally involves exotic doped materials, germanium lenses and cunning optical design since the emissions from the casing start to be almost as bright as the target. Some form of multistage thermoelectric cooling is usually employed or LN2.

Regards, Martin Brown

Our first imagers came out in '86, and they were LN2 cooled 16 color, 4 frame per second devices that sold for $95k each.

If the imaging plane on them was something other than CCD, I was unaware. I suppose it could have been a micro-array of bolometers.

So now you find a site that shows the spectral response of silicon, and draw a conclusion about the combination.

John

CCD detector + germanium lens = zero response.

John

I have a bit of experience with the radiometry of thermal imaging, both in the uncooled (pyroelectric) and cryogenic (InSb) ends.

The biggest, irreducible problems with thermal detectors are their thermal mass, which slows them down, and the fluctuations in thermal conductivity, which makes them noisy.

A thermal conductor has fluctuations exactly like Johnson noise in a resistor, and for the same reason--classical equipartition of energy applied to the heat capacity of the pixel (Even the formula is almost the same--noise power flux per hertz is sqrt(4kT**2/R_th).) That means that the power in the fluctuations goes as T, so a room temperature pixel is 4 times noisier than one at 77K for the same thermal conductance. However, the thermal resistance of materials tends to become very large at low temperature, and so does that of vacuum (the effective thermal conductance of vacuum goes like T**3--it's d/dT of the total thermal radiation, which goes as T**$).

Thus thermal detectors become really dramatically better at lower temperatures. Interestingly, it's a big win to increase the insulation, even though that slows down the response--the slowdown is due to 'bass boost' rather than 'treble cut', so you get more signal everywhere, as well as lower fluctuation noise. You just have to filter afterwards to crispen up the temporal response a bit.

Quantum detectors, e.g. Si, Ge, InSb, and HgCdTe (pronounced 'mercadtell' and sometimes abbreviated as MCT), are limited by their band gaps--there has to be a filled electron level around h*nu below the conduction band, or else nothing happens when you shine light on it. To see IR radiation of room-temperature objects, you have to go out to at least the 3-5 um band (InSb), and it's easier in the 8-14 um band (HgCdTe). The radiometry in the 3-5 um band is getting really difficult at 300K, because there's an exponential falloff of photon flux on the blue side of the radiation peak. The good news about 3-5 um is that there's a lot more contrast there--small temperature differences give rise to bigger intensity shifts, another consequence of that exponential falloff. You just have to have low enough noise.

There are cryogenic silicon-based detectors that are sensitive in the very long wave infrared (out to 100 microns or further), but they're extrinsic photoconductors--the relevant band gap is then between the impurity levels and the conduction band, so you have to run them at very very low temperatures so that the carriers freeze out. No silicon junction device (e.g. a photodiode, CCD, or CMOS imager) can work beyond about 1.1 microns, because at room temperature the impurity states are all ionized, so there aren't any valence electrons available that are closer to the conduction band than that.

On the other hand, people have made very low sensitivity IR detectors, good enough for measuring the power of CO2 lasers for instance, by forward biasing a photodiode and looking at the change in the forward voltage caused by the laser heating.

I'm expecting to start work soon on free-space IR detection in both bands, using modified versions of my antenna-coupled tunnel junction gizmos. They're sort of halfway between the thermal and quantum devices--they work by internal photoemission of electrons from one metal across the junction to the other metal. If the metals are sufficiently different, I think you can make them into zero bias detectors.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net

If you look at the black body spectrum, the power density drops quite quickly at wavelengths below the black body peak wavelength. The cheap IR sensors measure the amplitude around 1-2 um and if you have a idea of the spectral distribution in that area, you should be able to (gu)estimate where the spectral peak is and from the Wien law what the temperature is.

Paul

Archimedes' Lever wrote in news: snipped-for-privacy@4ax.com:

No idea what was in the thing I first saw, but it was a lot earlier than

1986, more like 1978 or so. I was a young kid, taken to the Science Museum in London. The imaging wasn't great, just contoured moving blobsfo false colour, but shapes were well defined, the refresh rate better than most simple webcams. It could see an electric iron, and it could see me vaguely, and my hand fairly well when I waved it open-palmed in front of the camera. I don't remember the camera being especially bulky. It was definitely picking up emissions, this wasn't reflection of near IR light. Maybe the mechanism was described, but at that age I wouldn't know.

Lostgallifreyan wrote in news:Xns9C6AB76B6DCE2zoodlewurdle@216.196.109.145:

Actually I think I remember it being based on thermocouples, some kind of fine array of them.

You need to specify *silicon* CCD + germanium lens = zero response.

QWIP arrays work from 1um down to around 12um from the open literature. Even has a Wiki entry.

Earlier systems were more like bolometers using doped germanium or PbS. Infrared "the new astronomy" was hot in 1975 and got hotter as the sensors became ever more sensitive and better resolution.

Regards, Martin Brown

OK, but the Fisher-Price toy was a silicon CCD.

Incidentally, Andor is an OEM customer of ours! Nice folks.

John

You + logic = water + oil.

This one could see the remnant of heat your hand would leave on an ambient formica desktop after ten seconds. It could still see the differentials nearly a half hour later. It had a 0.1°C sensitivity.

So do you think that a germanium lens would make the FP toy into a thermal imager?

Yes or no?

John

A window does not "make" anything into anything. Windows are used for filtration, just like say a "bandpass filter".

The camera has optics already. All it would need is a windows of the appropriate spectral range for the desired intended use.

I'll take that for a weasel-out-of-the-question.

John