DPSK offers alternative high–speed signal modulation

Jul 1st, 2002

By Guy Sauvé

Differential phase–shift keying has emerged as a strong candidate for modulating 40–Gbit/s signals. The author believes that DPSK demodulators based on fused–fiber components offer a high–performance, low–cost design.

In 1844, Samuel Morse sent the first telegraphic message from Baltimore to Washington using the first version of on/off–keying (OOK) modulation, the Morse code. Nearly a century and a half later, OOK is no longer used in telecommunications with one major exception: in fiberoptic systems.

On–off keying is a very simple form of amplitude modulation. For most applications in the long–haul network, a continuous wave (CW) laser emits a signal processed by a modulator. For a "1," the modulator allows total transmission; for a "0," it extinguishes the signal (see Fig. 1). The main advantage of OOK is its simplicity in implementing the design of modulators and demodulators. While research projects have studied alternative methods of modulation for many years, fiberoptic communication systems have largely relied on OOK to transmit information, with the exception of RF–subcarrier amplitude modulation used in the cable TV industry.

Modulation techniques are in their infancy in the optics field compared to other telecommunication products, where the advances in modulation paralleled an increase in capacity. In radio transmission, wireless, or even simple modems, designers have innovated with new modulation for years. The modem industry has traditionally spearheaded the use of efficient modulation techniques, because the telephone company's main infrastructure consists of sharply band–limited voice–grade channels. Phase–shift keying (PSK) modulation was used as early as 1962, when the Bell 201 modem was introduced.

Many studies have demonstrated the advantages of using an alternative modulation to OOK. In particular, the advantages of using phase modulation for optical applications have been known since the mid–1980s. Two phase modulations schemes are worth mentioning: phase–shift keying (PSK) and differential phase–shift keying (DPSK). Phase–shift keying works on the principle of the use of a different phase for "0" and "1." A "1" signal is denoted by a phase φ1; a "0" signal is denoted by a phase φ1+ 180°.

The DPSK modulation signal is not the binary code itself, but a code that records changes in the binary stream. This way, the demodulator only needs to determine changes in the incoming signal phase. The PSK signal is converted to a DPSK signal with two rules (see Fig. 2): a "1" in the PSK signal is denoted by no change in the DPSK; a "0" in the PSK signal is denoted by a change in the DPSK signal.

Phase–shifted keying is not considered practical because it is extremely difficult to perform synchronization of the oscillator required for the reception. Differential phase–shifted keying has emerged as the type of phase modulation most likely to reach commercial deployment.

Several studies have compared the performance of DSPK and OOK. Two of them demonstrate that a system using DPSK has a maximum transmission distance much longer than one using OOK.1, 2 The studies determined that DPSK is improving range by 3 to 6 dB, an improvement linked to the physics of four–wave mixing (FWM). So far, however, the use of DPSK has been limited to laboratory experiments because the OOK was efficient enough and could be implemented at a lower cost.


As increasing data rates and reach drive development of a new generation of components, system designers face the challenges of minimizing the impact of impairments such as FWM. As a result, in recent months, research projects on alternative modulation have moved from the university laboratories to the telecommunication vendor's R&D departments. Deployment of new modulation methods has been unveiled. For instance, manufacturers have announced the roll out of systems based on return–to–zero (or so–called pseudo–solitons) as a building block of the next generation of 10–Gbit/s ultralong–haul systems. Differential PSK is another enabling technology that could help push the limits further, but the obstacles delaying its deployment have more to do with economics than science.

The implementation, starting in the 1980s, of the DPSK modulation protocol for air–traffic–control radio communications is worth mentioning because of its parallel with the fiberoptics industry. To minimize costs, the designers first experimented with simple binary modulation schemes such as frequency–shift keying or phase–shift keying. However, when they modeled the sensitivity of various schemes to interference, they found that DPSK had a clear performance advantage compared to alternate schemes. Even so, they faced an uphill battle to convince the Federal Aviation Administration and manufacturers of equipment to use DPSK modulation. The challenge was to demonstrate that inexpensive DPSK demodulators could be built. Eventually, they succeeded and DPSK was widely implemented.

A similar path is to be expected in the fiberoptics industry: now that laboratory experiments have shown the advantages of DPSK over other techniques, the challenge is to demonstrate that it can be implemented with minimal cost premium over OOK. To understand why DPSK is cost effective to implement, it helps to understand the technology behind the modulators and demodulators.

The use of phase–shifted keying for the modulation of the optical signal is not an issue. On/off–keying modulation is actually implemented using a form of PSK (see Fig. 3). The principle behind most of today's modulators is the electro–optic effect, which is the change in the index of refraction caused by an externally applied electric field. To implement OOK, the emission from a laser is split equally between the two branches of a Mach–Zehnder device. Voltage is applied to each path, creating a phase shift in one arm compared to the other. When the signals recombine, a total transmission occurs for a 0° net phase difference, and total extinction for a 180° difference between the two paths.

The challenge until now has been to produce an economical demodulator. Schematically, the demodulator can be built from a Mach–Zehnder interferometer where a delay is created in one branch. It is composed of a 3–dB coupler, a delay line, and a combiner. The delay line must create a delay, τ, equal to one bit length (see Fig. 4). The main challenge in the creation of a demodulator is the stringent accuracy required of the delay, τ.

Different methods could be used to create that delay, including changing the refractive index of the branches using the electro–optic effect, or using a fiber with a different refractive index in one branch. Another method is to create a path length difference with one branch longer than the other.

In theory, Mach–Zehnder DPSK modules could be built using various technologies, such as crystals, planar, or fused fiber. Fused fiber offers three advantages for DPSK demodulators: it is currently available, it can be built in volume at low cost, and it provides the best optical specifications.


The first DPSK demodulators to reach the market were made of fused fiber (see Fig. 4, bottom). Fused fiber is used to build Mach–Zehnder interferometers for interleaving devices, which require tight control on the delay line. Moving from interleaving products to DPSK demodulators is a natural evolution.

Most of the investment to bring fused fiber products to the market is in R&D; once the sequence of events is mastered to build a device, the devices can be manufactured rapidly and inexpensively due to the low cost of the fiber. In addition, fused fiber provides superior optical performance. For instance, polarization–dependent loss is a challenge for planar devices, but is very low for fused fiber (see table).

Another important aspect is the stability of the component. Fused fiber demodulators drift in temperature at a rate of approximately 10 pm/°C. A heater is needed to maintain the device at a constant temperature. Dynamic control is also required even if the demodulator is perfectly stable. The laser can drift away from the center wavelength because of temperature shift, aging, or other problems. The narrow–range dynamic adjustment can be performed in different ways, from temperature control to the electro–optic effect.

Guy Sauvé is marketing director of transmission products at ITF Optical Technologies, 175 Montpellier, St. Laurent, QC, Canada H4N 3L2. He can be reached at gsauve@itfoptical.com.


1. A. Smiljanic, H. Kobayashi, J.–K. Rhee, SPIE Proc. 3491, 507 (Dec. 1998).
2. A. H. Gnauck et al., Opt. Fiber Comm. Conf. 2002, Postdeadline Paper FC2, Anaheim, CA (2002).

More in High-Speed Networks