Every engineer who works with optical receivers eventually meets the same contradiction: the changes that make a photodetector faster often make it less efficient, and the changes that improve sensitivity often slow it down. A serious high speed photodetector design​ program is therefore less about maximizing one number and more about managing a linked set of physical constraints.
The most fundamental metric is 3 dB bandwidth. In a PIN-style InGaAs device, the total response is governed by the RC time constant and the carrier transit time. Reduce junction capacitance and the detector gets faster. But capacitance is tied to active-area size and depletion width. Shrink the active area and the device becomes more sensitive to fiber misalignment. Thin the depletion region and transit time improves, but optical absorption drops. This is the classic bandwidth–efficiency trade-off that defines modern InGaAs photodetector​ development.
NEON’s FCPD product family shows how this trade-off is handled in a commercial module. The device offers 1 GHz, 2.5 GHz, or customized versions up to 8 GHz, with responsivity of 0.90 A/W or higher at 1310 nm and 1550 nm under +5 V bias. That level of responsivity is not accidental. It comes from an InGaAs absorption layer thick enough to capture most incident photons in the 1000–1650 nm window, but engineered so the depletion field remains strong enough to sweep photogenerated carriers out quickly. The result is a detector that is fast enough for many S-band and C-band analog tasks, yet efficient enough to keep the downstream TIA relaxed.
Dark current is the second hidden constraint. Even with no light, leakage paths inside the depletion region generate current. That current becomes shot noise, raises the noise equivalent power, and degrades receiver sensitivity. For NEON’s modules, low dark current is not only a chip issue; it is also a packaging issue. Hermetic sealing, clean fiber facets, controlled soldering temperature, and ESD protection all prevent the slow degradation that turns a good wafer into a noisy field unit.
The third constraint is electrical packaging. Above a few gigahertz, the bond wire is no longer innocent. Approximately 1 nH/mm of parasitic inductance is enough to disturb impedance matching and roll off the very bandwidth the chip was designed to deliver. That is why NEON’s microwave-class products emphasize 50-ohm matched outputs, SMA or SMP connectors, incorporated Bias-T, and compact hybrid packages. The photodiode may be capable of 18 GHz, but the module only delivers that performance if the RF exit path preserves it.
Advanced structures solve part of the problem at the device level. UTC-PD and MUTC-PD architectures use electron-dominated transport to reduce space-charge screening, enabling much higher bandwidth than conventional vertical PIN devices. Waveguide-coupled detectors go further by separating the optical path from the electrical collection path: light travels horizontally through a long absorption region for high quantum efficiency, while the vertical junction stays thin to keep transit time short. These ideas explain why next-generation receivers increasingly look different from a simple top-illuminated diode.
Still, no architecture removes trade-offs completely. An amplified photodetector​ such as the AMPD-S improves recovered signal strength and simplifies link budget, but the amplifier contributes its own noise figure and compression behavior. A wider bandwidth amplifier may flatten the frequency response less well. A higher gain stage may saturate earlier. The designer must choose gain, bandwidth, saturation power, and noise together, not separately.
Measurement discipline closes the loop. Bandwidth should not be inferred only from DC responsivity. Lightwave component analyzers, optical heterodyne techniques, and femtosecond-pulse time-domain measurements each reveal different behaviors: small-signal response, very-high-frequency limits, and impulse ringing caused by impedance discontinuities. A credible high speed photodetector design​ reports all of them.
The practical lesson is simple. Do not select a detector by bandwidth alone. Ask what modulation format will be used, how much optical power is available, how much dark current the system can tolerate, and whether the RF output will see a well-matched 50-ohm load. A well-balanced InGaAs receiver is not the fastest device on the bench; it is the device that keeps its promised response across temperature, power, and frequency.