Photodetector
Measures the relative intensity incident on its active surface.
Open in the canvas →In the real world
A photodetector is a semiconductor photodiode: a p-n (or p-i-n) junction that absorbs a photon and, if the photon carries enough energy to cross the material's bandgap, promotes an electron into the conduction band. The junction's built-in field sweeps that electron and the hole it left behind apart before they can recombine, and the result is a photocurrent proportional to the incident optical power. How efficiently that conversion happens is the quantum efficiency η — the fraction of incident photons that produce a collected charge carrier — and how much current comes out per watt of light in is the responsivity R:
Sensor material sets the usable colours
η is not a constant — it is a function of wavelength set by the semiconductor's bandgap, and it is the reason a detector has to be chosen for the wavelength it needs to see rather than assumed to work everywhere. A photon below the bandgap energy simply cannot promote an electron, however bright the beam: η drops to zero at a sharp cutoff wavelength, not a gentle roll-off.
Silicon is the default choice for anything visible or near-infrared. Its 1.12 eV bandgap gives it a cutoff around 1100 nm, and a typical commercial Si photodiode's responsivity climbs from a few tenths of an A/W in the visible to a peak near 0.5–0.6 A/W around 900–1000 nm, right before that cutoff[1]. It covers essentially every laser wavelength in this app's own palette below 1064 nm.
InGaAs (indium gallium arsenide) is the standard choice once a setup reaches into the short-wave infrared — the telecom bands around 1310 and 1550 nm, or Er-doped fiber sources. Its smaller ~0.75 eV bandgap pushes the cutoff out to roughly 1.7 µm, with peak responsivity around 0.9–1.0 A/W near 1550 nm[1] — silicon is completely blind out there; those photons simply don't carry enough energy to cross its wider gap.
Germanium was the original short-wave infrared material, and is still around: a smaller ~0.67 eV bandgap stretches its cutoff out to roughly 1.8 µm, past even InGaAs, with peak responsivity around 0.7–0.8 A/W near 1.5–1.6 µm[1]. What displaced it from most telecom and low-light work is dark current — the reverse-bias leakage current a photodiode carries with no light at all, which competes directly with a weak real signal. Germanium's is roughly two to three orders of magnitude higher than InGaAs at the same reverse bias and room temperature[1], so germanium detectors usually need cooling to be useful for anything faint, while InGaAs does not. It remains a cheaper option where that noise floor doesn't matter. Beyond these three, extended-range InGaAs and HgCdTe push further into the mid-infrared at the cost of even more dark current and, for HgCdTe, mandatory cooling — but silicon, InGaAs, and germanium between them cover the overwhelming majority of laboratory optics.
Frequency response
A photodiode also cannot follow an arbitrarily fast amplitude modulation. Its junction behaves as a capacitor C_j discharging through a load resistance R_L, and that RC time constant — together with how long a photo-generated carrier takes to drift across the depletion region — sets a 3 dB electrical bandwidth beyond which the output can no longer track the optical signal:
A large-area photodiode built for power metering — like the power meter in this palette — trades bandwidth for active area and sensitivity, and is typically limited to the kHz range or slower. A telecom-grade InGaAs photodiode with a 250 µm active area, by contrast, is built for the opposite trade and can exceed 10 GHz[2] — fast enough to demodulate a digital data stream, but far too small and insensitive to usefully catch a divergent free-space beam. Choosing a photodetector for a real setup means picking a point on that speed-versus-area curve, not just a material.
In OpticalSetup
The photodetector reports a qualitative relative signal — the sum of every ray's power reaching its front face, in arbitrary units — plus the wavelength or detected spectral band, polarization state, and spot extent of whatever light arrives, all read directly off the traced rays. If the arriving light is pulsed, it also reports the accumulated GDD, optical path delay, and arrival spread, the same as every other instrument in the Detectors category. This is genuinely useful for seeing whether light reaches a given point, roughly how strong it is relative to other configurations, and what its spectral or polarization content is — regardless of what real sensor a lab bench would need there.
The oscilloscope, and why it has a floor
Pulsed light turns the readout into a scope trace: the pulse train against time, with any chopper or modulator envelope behind it. The window defaults to two periods of the slowest thing on the beam, and Time interval and Time offset override it.
What that trace shows is the train convolved with the detector's own response, set by Response time and defaulting to 1 ns. This is not decoration. Two pulses closer together than the response merge into one bump, exactly as they would on a bench, and a detector slower than the pulse spacing stops resolving the train at all and reads the flat average instead — put a 15 ns response on an 80 MHz train and the trace goes level, which is what such a detector really outputs. The window will not zoom below five response times either, because nothing there is anything the instrument could have seen.
Sync puts several detectors on one axis and one time origin, so light that took a longer route is drawn where it actually arrives. That is where the limit bites hardest, and where it is most worth understanding: an arm 50 mm longer delays its pulses by 167 ps, and a 1 ns photodiode cannot see that — the readout names the delay but marks it unresolved, because the geometry knows it and the instrument does not. Give the detector a 50 ps response, as a small fibre-coupled diode really has, and the same shift is resolved. This is precisely why timing two ultrashort pulse trains against each other is a job for an autocorrelator rather than a photodiode and a scope: the autocorrelator sidesteps detector speed entirely by using the pulses themselves as the clock.
The reported signal is not calibrated to any real unit, and there is no concept of sensor material at all: a photodetector in this app reads every wavelength in its traced light with equal weight, whether that light is 405 nm (well inside silicon's range) or 1550 nm (which silicon cannot detect at all and would need InGaAs). There is no responsivity curve, no bandgap cutoff, and no way to configure or even see which material is assumed. The response time shapes the time trace and sets how finely the axis can be read, but it is not a filter on anything else: the relative signal is still an instantaneous sum, so a beam modulated far beyond what the configured response could follow still reads at full strength as a single number, and there is no roll-off, no 3 dB point, and no phase response. Nor is the response tied to the active area, though on a real device those trade against each other directly. Dark current, noise-equivalent power, and saturation are likewise not modeled here; the PMT variant is the only detector in this category with any qualitative gain/saturation behavior at all. Treat every reading as relative and instantaneous, never as a prediction of what a specific real sensor would output.
Related components
References
- Thorlabs — Photodiode Tutorial (responsivity vs. wavelength for Si, Ge, and InGaAs detectors)
- RP Photonics Encyclopedia — Photodiodes