More info.  I found that the measurements are very low, in the range of -5 to 30, very roughly speaking, with the HW_GAIN set at 11  (0xb), reading the UV diode.

If my code decides that a measurement over 6 is daylight, then the sensor only detects daylight when the sensor is within about three feet of a south facing window, with the sensor pointed towards the window, and with a piece of PTFE plumbing tape against the sensor as a diffuser.  I was hoping for a better result, since I want to place the sensor farther from the window, and facing down from the ceiling, in a typical household room.  As it is, it would only work if the room was a sunroom or greenhouse, with significantly more windows than a typical room.

Can you explain the notation in the datasheet in the "operation" column of the ADCMUX row of the table for the ADCCONFIGx parameter?  The notation uses for example "D + 10" for the UV "optical function."  But the term "D" has no antecedent anywhere in the datasheet, as far as I know.

I assume "D" refers to a separate diode for UV. 

Does the "+10" mean that the mcu on the chip adds ten to some raw result?

Also, what is the interpretation of a negative result?  That the dark current is less than the light current (where you are measuring the diode in two ways, dark current and light current, and taking the difference) ?

Related to my other post about detecting daylight using UV indoors.

My application is: detect day versus night, indoors, in a room with south facing windows.  I am exploring the Si1133 to measure UVA.  I can put the detector within 10 feet of the windows, high and facing down so the detector gets reflected UV, and never gets direct sunlight, and gets only reflected artificial lights from typical room lighting of about 400 lux white, fluorescent light.  My understanding is the window glass will transmit about 50% of UVA,  that the sky will scatter significant UVA,  that night sky has no UVA, and that artificial lights produce negligible UVA.

Might that work?  Do you have any suggestions re diffuser, physical filters, parameters of the Si1133, and which set of diodes to use?  My plan is to read only the UV diode.

I want to estimate sun time by detecting sundown and sunset, despite noise from artificial lights.  I don't need accuracy.  I plan to take samples over many days.   I understand that the weather and season will have an effect.