As the need for speed took fiber optic transmission beyond 10 Gbps per channel, long-range optical transponders have moved beyond the Morse code-like on-off keying (OOK) used in fiber optic communications for decades. More sophisticated modulation and demodulation schemes are commonly used in modern wireless equipment where radio frequency bandwidth is scarce, but had not been widely applied to optical transmission where bandwidth is more plentiful. For example, a 50 MHz channel bandwidth is relatively wide-band for licensed microwave radio, while 50 GHz channel spacing in DWDM optical transmission is relatively common.

OOK is cheap and works well on standard single-mode fiber up to speeds around 10 Gbps, providing little commercial demand for more efficient optical modulation to this point. At 10 Gbps, fiber impairments like dispersion come into play, and fiber characterization testing and dispersion compensation are needed to ensure reliable operation. At higher data rates dispersion sensitivity increases, and more precise dispersion compensation is required. Quadrupling the data rate using the same modulation would reduce dispersion tolerance by sixteen times. 40 Gbps optics would not be as widely deployed if network operators had to re-engineer their DWDM networks again, as was the case when moving from 2.5 to 10 Gbps.

Coherent optical receivers had been a hot topic of research in the 1970s and 1980s when system reach was limited by signal attenuation in the fiber. A coherent receiver is less noisy and can reliably detect a weaker signal, thereby extending the distance between network elements. Once optical amplifiers were developed, they became the preferred solution to the attenuation problem, and work on coherent optics was largely abandoned.

As data rates have moved into the Gigabits per second range, optical signal-to-noise ratio (OSNR) has become an important limiting factor. Although an amplifier can make a weak optical signal stronger, it also amplifies all of the spurious noise carried along with the signal. So an amplifier cannot improve the OSNR—in fact each amplifier tends to degrade it. Coherent receivers became a hot research topic again when their better signal-to-noise performance was needed to increase throughput without using additional optical bandwidth. Advances in technology over the last 30 years also made deployment of coherent optics much more practical.

Coherent receivers use a local reference laser for pre-processing in the optical domain (in a transceiver, the transmit laser can be taped to provide the receiver reference). The coherent receiver mixes the received signal with the reference using an optical hybrid. Photodetectors then convert the resulting optical signals to the electrical domain. These analog signals are digitized using analog-to-digital converters (ADC) and then demodulated like the signal in a digital radio receiver using digital signal processing (DSP).

Dispersion compensation can also be built into the DSP algorithm of the coherent detector, eliminating the need for extra dispersion compensating hardware on each span in the optical network. By moving the dispersion compensation from the optical to the electrical domain, the receiver can adapt itself to the level of dispersion present in the as-built network. This allows network operators to deploy 40 Gbps waves on an existing DWDM system alongside in-service 10 Gbps waves, without requiring disruptive changes to the network.

But in order to make 40 Gbps transponders a complete drop-in for 10 Gbps DWDM systems, the transponders also need to push four times as many bits through the standard 50 GHz wide channel. OOK requires much better OSNR and dispersion to operate at a higher data rate in the same bandwidth, making it a non-starter. More spectrally efficient modulation was required; otherwise the optical signal bandwidth would need to span several DWDM channels anyway. Higher-order modulation schemes can encode more than one bit at a time (as OOK does). This allows a higher data throughput with the same bandwidth, dispersion, and OSNR.

An OOK transmitter can be demodulated with a non-coherent detector—one that only looks at the signal amplitude. In each symbol interval this direct detection scheme requires only a binary choice between laser on or off to decode a single bit to either a 1 or 0 state. With higher-order modulation schemes that encode two (or more) bits per symbol using carrier phase angle, a coherent optical detector is advantageous. It allows the receiver to split the incoming signal into two independent phase components that can be processed in parallel. With two parallel streams, the combined data rate can be double the rate that the ADC and DSP components can handle. This is important, because components that could process a full 40 Gbps payload would be near the bleeding edge of integrated circuit technology.

DWDM transponders operating at 40 Gbps are now commonly available and widely deployed because they quadruple DWDM capacity at a price point below 4 x 10 Gbps. Unfortunately, there were no standards for single wavelength DWDM at 40 Gbps. Each optical vendor developed proprietary implementations using different modulation schemes, so don’t expect Ethernet type inter-vendor interoperability. Compatible transponders from the same vendor must be used on either end of a link. But since the DWDM channels are standardized, these transponders can operate as “alien wavelengths” on an existing DWDM system from another vendor.

In the nearer term, coherent optics will be used where optical bandwidth is valuable, as it is in amplified DWDM systems. As the price/performance of ADC and DSP continue to improve with Moore’s Law, and as related optical components get mass produced, coherent detection will become less expensive and more commonly deployed. Even with simple OOK, coherent detection can compensate for fiber impairments that limit reach with direct detection. Going into the Terabit per second realm, coherent optics will be required for the more complex encoding schemes needed to pump more bits through fiber impairments without using more optical bandwidth.

Doug Haluza,

CTO, Metro|NS

Ed note, this is the first post in a series. The next post is here.

Share