How fine can single photon detectors made to be?

There is some confusion of what "single photon" really means in the context of single-photon detection.

• When applied to photon counters, single photon means means the precision rather than a quantity of photons. Assuming Poissonian statistics the relative error in counting photons decreases as $1/sqrt{N}$, where $N$ is the number of photons. That is the number of photons can be counted with precision up to a single photon, provided we have much more than one photon to detect.
• Alternatively, one may talk about detecting very weak fields, potentially containing very few or even one photon. In this case the device may be sufficiently sensitive to produce a signal, if only one photon is present. But, as should be clear from the previous point, we may err by at least one photon in this case.
• Finally, it is important to note that in both cases above we discussed only the energy of photons. Photons however may also differ by their polarization and direction (or other quantum numbers, if in a cavity). Some problems, such as studying Compton scattering, require detecting single photon of a particular direction, which requires compromising on the precision in photon counting or sensitivity.

It is important to understand that the wave energy is quantised. The wave equation merely tells you where that quantum of energy might be found. When it is found, that event is localised and we say that we have detected a photon.

This quantisation of light waves as particle-like photons was proposed by Einstein to explain the photoelectric effect, for which he received his Nobel prize.

Very low levels of light take the form of sporadic quanta, each travelling as its own little wave. This causes what is known as "shot noise" in the detector, which fires sporadically as each such wave comes along and its quantum of energy gets absorbed.