Strong response to UV radiation despite severe attenuation Recently, I performed experiments to characterise the ultraviolet-A response in a smartphone camera (with the lens still attached). This question focuses on Figure 2 of my paper "Characterization of a Smartphone Camera's Response to Ultraviolet A Radiation" - please note, I am only posting this link to show the figure.
In figure 2, the transmission of UVA through the lens drops significantly at wavelengths of less than about 370nm, not shown in the paper is that several lenses were tested with similar results.  However, with the lens in place, UVA radiation from a monochromator at wavelengths as low as 320nm still cause saturation of the silicon-based CMOS image sensor despite the heavy attenuation at wavelengths lower than ~370nm.
What is reason for heavily attenuated UV radiation at 320nm causing as strong response as for far less attenuated UV radiation at 380nm?
 A: What happens if you do the same tests without the lens in place ?
It is entirely possible that the sensor is not responding to the 320 nm UV, but to a longer wavelength fluorescence due to one or more of the lens elements, absorbing the 320 nm photons, and fluorescing at a longer wavelength that the sensor responds to.
A good many optical glasses are known to fluoresce, so it is important to check that an output signal is still at the same wavelength as the input signal.   And a first requirement of a fluorescence output, is to have a strong absorption of the input signal.
The well known set of Schott glass sharp cutoff long wave pass filters, are famous for fluorescing.  If for example, you try to suppress say a 4416 blue He-Cd laser beam with say an RG645 Schott red filter glass, the blue laser wavelength will be attenuated by 4-6 orders of magnitude, but the energy transmission is much higher, but will be at a longer wavelength.  So the usual Beer's law, exponential absorption with thickness, might apply, but that doesn't mean the energy transmission follows Beer's law at all.
So run your UV line through the lens, and then run the output through another monochromator, and see if it is still 320 nm or something much longer.
A: I had one other thought.  If you are unable to get pixel filter samples to test them, you could try looking at the RAW pixel data for R, G and B.  Higher end digital cameras allow you to save in RAW and not JPG format.  While I don't know what your test setup is, I assume you're basing the saturation readings on saved images, not real time hardware CMOS information.
Anyway, you might try seeing if either the R, G or B pixels saturate more or less than the other over the frequency of UV input.  You could plot the saturation vs color vs frequency and you might see that one of the filters is fluorescent.  It may even be possible to improve your measurements of UVA using separate RGB measurements.
Edit:  It could also be possible that the tight spacing between pixel filters causes the fluorescent response to leak into the neighboring pixel and you might not see much difference between the color channels...I don't know.
