I was reading up on how Avalanche Photodiodes(APD) are used to detect very low light levels, some are even used in "photon-counting" experiments (in Geiger mode). However, a survey of some commercial APDs such as this indicates that these sensors also have non-negligible dark currents for the large reverse voltages they operate in. The one in the link for example, has a nominal dark current of 50pA, in intensity terms it would be ~1pW (M=100, 0.5A/W sensitivity at 850nm). This is actually pretty high. Even if we treat this as an offset, say, for example in a charge integration based measurement, the noise in this bias current can mask any detection of ultra-low light levels(~1fW). So what techniques are used in signal conditioning? Any links or pointers would be greatly appreciated.
If background is a problem, there are photomultiplier tubes. But those are bulky and delicate vacuum tubes and require high-voltage power supplies (about 1000 volts).
For solid state detectors, one can use the avalanche photodiodes in an array, where each diode has low area and low dark current. When there is a small flash of light (for example from a particle in a scintillator material), several of the photodiodes in the array will detect a photon at the same time. So adding the output from all diodes together will create a peak that is above the dark current. This kind of "pixels" cannot be used for imaging. The device is called SiPM (silicon photomultiplier).
Another method is gating to only look at the photodiode in the fraction of a microsecond that one knows that there might be a signal (for example, when a laser pulse in the experiment is firing).