Direct detection

An optical receiver architecture in which the photodetector responds to the optical power (intensity) of the received signal. The simplest and most widely deployed detection scheme in optical communication.

In direct detection, an incoming optical signal illuminates a photodetector that produces electrical current proportional to optical power:

iphoto(t)  =  R ∣Esignal(t)∣2  =  R Psignal(t),

where R is the photodetector responsivity. The detector responds only to intensity. Phase information of the optical signal is lost.

Modulation formats compatible with direct detection.

FormatEncodingReceiver
NRZ on-off keying (OOK)Amplitude (presence/absence)Single photodiode + decision threshold
Pulse amplitude modulation (PAM-N)Multilevel amplitudeSingle photodiode + multilevel decoder
Subcarrier-multiplexedRF subcarriers on optical carrierPhotodiode + RF demodulation
DuobinaryOptical filtering produces correlated waveformSingle photodiode

Noise sources in direct detection:

Noise typeSourceScaling
Shot noiseQuantum nature of photodetection∝Psignal
Thermal (Johnson) noiseReceiver electronicsConstant, independent of signal
Relative intensity noise (RIN)Laser source intensity fluctuations∝Psignal
Dark current shot noiseDetector reverse-bias dark currentConstant

Sensitivity. At the quantum limit (shot-noise-limited reception), direct detection requires ∼ 10 photons per bit for BER =10−9 in NRZ-OOK. Practical receivers, limited by thermal noise, typically require 100 – 10,000 photons per bit.

Strengths.

  • Simple: single photodiode, no local oscillator
  • No requirement for phase or polarization stability
  • Low-cost: workhorse of datacenter and short-reach telecom
  • Suitable for short-reach links at 100 – 200 Gb/s per lane with PAM4

Limitations.

  • The absolute optical phase and the second quadrature are lost, so QPSK and QAM need a local oscillator; differential phase formats (DPSK, DQPSK) can be received with a delay-line interferometer and balanced photodiodes, but spectral efficiency stays below that of coherent detection
  • Chromatic dispersion and PMD are only partly correctable in the electrical domain (equalizers), because the optical phase is lost at the photodiode
  • Lower practical sensitivity than coherent detection: with no local-oscillator gain, receiver thermal noise rather than shot noise usually sets the floor (the ideal quantum limits are similar, about 10 photons/bit for OOK and about 9 for homodyne BPSK at BER 10−9)
  • Difficult to scale much beyond 200 Gb/s per wavelength

Direct detection dominates short-reach datacenter optics. Coherent detection (with full optical field recovery, including phase and polarization) has displaced direct detection in ≥ 100 Gb/s long-haul transmission and is increasingly common in metro and access networks above 100 Gb/s per wavelength.

The contrast between direct detection (intensity only) and coherent detection (full optical field recovery via interference with a local oscillator) is the major architectural distinction in optical communication receivers.

New entries, monthly

The glossary grows every month. One short email lists what’s new, nothing else.