I am looking to build a circuit to detect extremely low levels of light (like a very very dark room). This circuit will be connected to a microcontroller like Arduino, so it needs an interface to communicate the readings. Arduino has an ADC present, but it can also receive data over i2c, uart, spi. I was wondering which component to use to detect such low amounts of light. Basically I'd like something more sensitive than a simple photoresistor, but the element should not be too expensive i.e. a photo multiplier.
You should try a specification for "extremely low levels of light", e.g. which light source do you consider. In simple terms, in a dark room, there's no light at all, respectively nothing to measure, even by a single photon counting multiplier.
With a light source specification, you can start to calculate sensor irradiance and sensitivity.
A good quality, small solar cell might do the job nicely.
Consider buying a cheap solar calculator (if you do
not have several laying about already) and harvest
the cells from it. Often they are 4 in series and can
give you more than enough voltage to (say) energize
the base of an NPN for some more current gain and
with a pullup resistor to the uC supply, you've got a
pretty clean and non-threatening signal at the
collector. A high value base shunt resistor can set
an illumination threshold.
Now if you wanted quantitative measurement, I'd
throw that solar cell in photocurrent (shorted) mode
at a transimpedance amplifier with an appropriate
scaling feedback resistor. You could use a single
one for this, you will hold the cell voltage to zero
(virtual ground) with the TIA.
I'm sorry, perhaps I should have been more specific with the lightness... when I said dark room I didn't mean a completely black room. I meant a room with very few weak light sources i.e. your living room when you turn off all the lights at night. There are still some electronics that emit light via their status led's i.e. a modem, a TV, a switch, etc. So if you wake up at night, having turned off all the main lights and blocked all the windows you'll notice that these devices still provide sufficient lighting for you to barely see in the room. This is the amount of light I wish to detect as opposed to a complete blackout.
I would expect irradiance amounts down to a nW/cm² order of magnitude, which gives still detectable counts with the said TSL25911. You need good photodiodes and amplifiers with pA input current and GOhm resistors to achieve similar sensitivity with a discrete setup. Solar cells might suffer from too low parallel resistance, I have no data about it.