The UV spectrum is the high energy neighbor of visible light. While we cannot see it, the sun produces the full UV spectrum and UV-A/B (280-400 nm) passes through the atmosphere, making its effect known on the surface. If you ever had a sunburn than you have felt the effects of UV-B radiation. The burns themselves can be anything from an annoyance to painful, but overtime the repeated damage to the DNA in your skins cells can lead cancer. It is this same mechanism that makes UV-B light a cancer risk that makes UV-C light an effective germicide.

All three of these breakout boards are designed around the GUVA-S12SD gallium nitride photodiode. The response curve of this photodiode is shown on the left. We can see that it is blind to visible light and responsive in the UV-A/B and part of the UV-C range. Critically for our application, it is responsive to 254 nm light, which our lamp advertises to be. Any sensor board built around this diode should be responsive to a UV-C light, as the three shown above are. However, upon further examination it is clear that the Proto and Waveshare boards are insufficient for our application. All three of these boards contain op-amps designed to boost the weak currents coming from the photodiode. The Proto and Waveshare use two to boost the response from the relatively weak UV-B intensities from the sun. Our UV lamps will produce UV light intensities far greater than that of the sun, and so these boards will be maxed out at best, damaged at worst.

So we have identified a sensor, the GUVA-S12SD photodiode on the adafruit breakout board, that can detect and respond to UV-C light. However their is a critical problem, the sensor is also responsive to UV-A/B, not UV-C exclusively. This means the sensor board alone will not be able to tell us whether our lamp is UV-C, the key property of a germicidal lamp. The solution to this problem is inspired from the lamp itself. UV-C lamps must use special glass for the bulbs as standard soda-lime glass will block UV-C. Most lamps are built using titanium doped quartz or other similar materials which allow for the full UV spectrum to transmit. If we can find a glass that allows UV-A/B to transmit, but blocks UV-C we can use this to select UV-C and measure its intensity. A long search finally turned up a candidate. SCHOTT glass manufactures produce a high quality borosilicate glass and conveniently cut it into cheap and reasonably sized microscope cover slides. The transmission curve for this glass shown on the left cuts out right at the beginning of the UV-C spectrum. The price, size, and optical properties make this the perfect glass for use in our sensor design.

As we can see from the plot on the left small doses of UV-C, on the order of a mJ/cm^2, are needed to eliminate the virus when directly exposed. Combined with information from both the adafruit and GUVA-S12SD datasheets, we can use our light measurements to estimate a lamp intensity of ~8 mW/cm^2 at 10 cm below the center of the lamp. This means our lamp should be able to effectively eliminate the virus in under a second. However, masks provide extra cover for the virus, and studies done on disinfecting masks from influenza virus have shown that roughly 1 J/cm^2 is needed to disinfect a mask. For our light, that should take roughly 125 seconds, but we will set our disinfection cycle to 300 seconds, covering for errors in our calculation, lamp degradation, and the likely reduced intensities the ends of masks will experience.