UV camera project build question

ProfHankD wrote:

petrochemist wrote:

Transmission data according to a Perkin Elmer Lambda 35 research spectrometer:

900-1100nm 87 to 93% transmission (interference between surface reflections has made this area less precise)

800nm 82% transmitted

750nm 41.7% transmitted

700nm 0.23% transmitted

600nm 4.1% transmitted

500nm 66% transmitted

400nm 53% transmitted

350nm 18% transmitted

300nm to 200nm less than 0.02% transmitted

Nothing surprising -- that matches the published spec well -- except wow, it sure climbs fast after 740nm! They say 0% at 740nm, and judging by 700nm, I believe them.

Actually it climbs rapidly after 705nm. IIRC it was up to more than 30% by 740nm

I'm quite surprised by this, although it creeping up wouldn't have been a big surprise. Then again, you're certainly right that organic dyes rarely block NIR -- that's what got Sony's NightShot into trouble.

From the areas under the spectra around 10% of the normalised transmitted light would be UV, visible light mainly blues & green s account for about 30% & NIR about 60%

Taking into account camera sensor response will boost the visible portion at the cost of both UV & IR.

The UV transmission is better than most filters, but both IR & visible are HUGE.

Fair enough.

It might still be possible for the OP to computationally separate UV out using RGB CFA sensitivity differences, but it's pretty clear noise level will be HUGE.

I don't think theres enough data available for that. The ratios of the RGB channels are already used for getting colour information. This is how the eye works & cameras too. All the visible spectrum gives some signal in each of the channels.

Here's one of the on-line spectra for the Bayer filters https://static5.olympus-lifescience.com/data/olympusmicro/primer/digitalimaging/images/cmos/cmoschipsfigure4.jpg?rev=88D6