Power meter
Reports absolute optical power by carrying each source’s configured watts through everything that attenuated it on the way here.
Open in the canvas →In the real world
A photodetector reports a photocurrent; a power meter reports watts. The difference is calibration: a power meter's sensor has a known, measured relationship between what it outputs and the optical power that produced it, so the console can show an absolute number instead of an arbitrary one. Every commercial power meter is really two parts — a sensor head and a console that knows how to read it — and the sensor is where the real design tradeoff lives.
A photodiode sensor is the same physical device as a plain photodetector, just factory-calibrated: its responsivity R(λ) is measured at each wavelength, so the console can recover power from photocurrent. That calibration is the whole catch — R(λ) is not flat, exactly as on the photodetector page, so the meter has to be told which wavelength it's reading. Set the wrong one and the number is wrong by the ratio of the two responsivities, silently.
A thermal sensor — a thermopile, or a pyroelectric detector for single pulses — sidesteps that problem entirely. Incident light is absorbed by a black coating and converted to heat, and the sensor reads the resulting temperature rise (or, for a pyroelectric, the heat pulse from one shot). Absorption into heat is, to good approximation, the same process at every wavelength, so a thermal sensor's calibration holds across a broad spectral range with no wavelength setting to get wrong[1]. The tradeoff is speed: a thermopile takes seconds to reach thermal equilibrium, against microseconds for a photodiode, and needs more power to produce a measurable temperature rise at all — which is why thermal sensors dominate at higher powers and photodiode sensors dominate at low ones.
Either sensor has a hard damage threshold. A photodiode sensor can saturate or be burned out by too much continuous power — or, just as easily, by the instantaneous peak power of a pulsed beam whose average power looks perfectly safe. A thermal sensor's coating can be scorched by a tightly focused beam even within its rated average-power range. Every real power meter publishes a maximum power (and often a maximum power density) that the reading itself gives no warning of approaching.
In OpticalSetup
The power meter measures the same relative ray weight every detector in this palette does. On its own — selected in the canvas, read from its inspector panel — that relative number is all it shows, identical to a plain photodetector. The Watts reading appears once it drives a detector screen ("Connect to a detector screen" in its inspector does this in one click).
That figure is built per source. Every source launches rays whose weights sum to one, and each interaction along the way scales that weight by what it actually transmits — a beamsplitter's ratio, a filter's transmission, an aperture that clips part of the beam, a chopper's duty cycle, a nonlinear crystal's conversion efficiency. The weight that survives to the sensor face is therefore the whole source-to-detector efficiency chain in one number, and multiplying it by that source's Average power (W) gives the watts it delivered here. Several sources landing on the same meter simply add:
Attribution follows the light through wavelength changes too. When a specimen fluoresces, the emission is new light at a new colour, but its power is still a fraction of the laser that pumped it — so it is charged to that laser, not to the specimen. In the PMT's fluorescence example the meter would read the pump power times the roughly 2×10⁻³ that survives excitation focusing, conversion efficiency, collection solid angle, and the emission filter.
If some of the light arriving carries no power rating at all — the point source has no Average power field — the screen reports the rated contribution and marks the reading + unrated source, because that number is then a floor rather than the total. When nothing arriving is rated, it falls back to showing relative weight.
The conversion is wavelength-flat: one watt of 400 nm and one watt of 1550 nm read identically, so the element behaves like an idealized broadband thermal sensor no matter which real sensor type you have in mind, and there is no way to select one or to get the wavelength-setting error that a real photodiode meter punishes you for. Power is average power only — a pulsed and a CW source of the same average read the same, with no peak-power figure and no notion of a pulsed beam damaging a sensor a CW beam of equal average power would not.
There is no damage threshold, no saturation, no noise floor, and no response time, so nothing distinguishes a thermopile's seconds-long settling from a photodiode's microseconds. The watts are exact arithmetic on the traced efficiencies rather than a measurement: they inherit every idealization upstream of them — hard-edged filter passbands, flat per-surface transmission instead of Fresnel losses, no scatter and no absorption that the tracer was not told about — so treat the number as what this idealized bench delivers, not as what a real one would.