Traditionally, demands for successive generations of high-speed networking technologies have originated at the backbone and long-haul level, to fulfil the bandwidth demands of carriers operating large core networks. Today, the requirement for increased speed at the metro and access levels is at least equal to the needs in the core, and is driving demand for 100 Gigabit Ethernet (100 GbE) equipment as well as optical transport equipment for long-haul transmissions. Over the shorter metropolitan and access distances, applications such as video-on-demand have whetted consumer appetite for high-definition entertainment. At the same time, storage area networking over distances from 3 km to 10 km is increasing bandwidth requirements, and the growing amount of data exchanged via peer-to-peer data networks is driving demand for increased Ethernet speeds at internet exchange points.
These trends are generating demand for faster bit rates than the current 10 Gbit/s connections. In response, the IEEE 802.3ba Task Force is defining the service interface and physical layer specifications for 40 GbE and 100 GbE on the client side — over shorter distances within the data centre or in local area networks. Meanwhile, the ITU's Telecommunication Standardization Sector (ITU-T) is working on defining the transport mechanism for 100 Gbit/s over long-haul networks by extending the G.872 standard for the Optical Transport Hierarchy (OTH) to ODU4. According to some of the world’s major network providers, the industry needs to be in a position to deliver 100 Gbit/s over existing optical fibres by 2010 to be able to satisfy demand.
As with previous Ethernet standards, the physical layer specifications allow for transmission over different media, including singlemode and multimode fibre, copper cable, and copper backplane. For distances up to 100 m over multimode fibre, transmission will be over 4 or 10 lanes over parallel fibres, while transmission over singlemode fibre up to 10km or 40km will require multiple channels at 10 or 25 Gbit/s using coarse WDM over a single fibre.
This move to higher data rates changes the landscape for engineers charged with testing 100 Gbit/s optical network equipment. The tests required, and the techniques used to perform those tests, vary somewhat between short-range and long-range transmission paths. In the same way, new requirements for test and measurement solutions differ for the two main areas of application: long-haul transmission and short-range data connections.
A variety of approaches in long haul
The challenges for long-haul fibre transmission are much tougher than in short-range applications, and hence these have been the focus of the most concentrated industry effort from test equipment vendors and others.
Long-haul optical transmission is more complex than for short-reach, data-centre applications because carriers want to use the higher data rates without installing new fibres. However, fibre impairments such as chromatic dispersion (CD) and polarization-mode dispersion (PMD) mean that some installed fibres cannot guarantee transmission quality data rates above 10 Gbit/s using ordinary on/off modulation.
Measures such as forward error correction (FEC) and PMD compensation have been proposed to overcome such impairments. Another option for improving the signal detection capability is to design the equipment with a higher optical signal-to-noise ratio (OSNR). (Higher-speed, shorter pulses are more susceptible to interference between similar wavelengths in the same fibre because they contain higher frequencies, and so tend to have worse OSNR.) This requirement for a higher OSNR would in turn call for higher operating power, which would potentially add to the cost of the equipment, while also increasing the effect of non-linearities in the fibres, and reducing energy efficiency.
But that's not all — there is a further challenge with the implementation of 100 Gbit/s equipment on existing fibres: some of these fibres might be unable to sustain the 50 GHz wavelength spacings commonly used in 10 Gbit/s transmission systems. Operators and carriers are stating this is a mandatory requirement for manufacturers to comply with.
New modulation techniques such as differential quadrature phase-shift keying (DQPSK) or dual-polarization quadrature phase-shift keying (DP-QPSK) help to overcome these challenges. DQPSK decreases the symbol rate to half the data rate, by using two half-rate modulated signals for each of the two I and Q inputs. DP-QPSK generates a symbol rate one-quarter the data rate, by applying polarization multiplexing, on top of phase modulation, to the I and Q signals.
These lower data rates promise a reduction in the impairments from CD, PMD and OSNR. At the same time, the optical transmitter is "always on", which eliminates the high-frequencies generated by the on/off transitions found in non-return-to-zero (NRZ) and return-to-zero (RZ) modulation, which reduces the impact of fibre non-linearities.
From an equipment designer's point of view, these different approaches to high-data-rate transmission result in various trade-offs in terms of increased circuit complexity. Hence, there remains some uncertainty over the type of equipment and modulation schemes to be used in 100 Gbit/s optical transport equipment for carrier-class applications.
Despite a lack of consensus on which modulation scheme is best for 100 Gbit/s transport, the proposed approaches do have several factors in common. Multi-channel parallel signals are used, and this creates the need to control and verify new parameters, such as inter-channel skew and synchronization. Plus, modulation pre-coding and de-coding become a new item to be verified and qualified in this next generation equipment. Finally, all of the modulation schemes under consideration share the characteristic that they use phase to encode information, not only amplitude.
The introduction of these new modulation schemes requires a new generation of optical test equipment, supporting maximum flexibility in terms of signal architectures, simulation/emulation of new equipment capabilities, and the ability to evolve to meet future testing requirements, which will include phase analysis, not only direct detection.
Interoperability requirements
While designers of long-range equipment are most concerned with the successful use of existing fibre, in the case of short-range data connections interoperability is the key issue. All of the architectures currently under consideration by standards committees are based on parallel transmission and management of multiple signals.
Just as network equipment manufacturers have developed a range of suitable optical module form factors at 10 Gbit/s (Xenpak, XFP, SFP+), they will adopt new optical modules to enable multiplexing and demultiplexing of multiple channels at 10 Gbit/s into 40 Gbit/s and 100 Gbit/s for optical transmission.
Notable challenges that equipment manufacturers and optical module makers will confront include control and tolerance of lane skew and signal jitter, due to factors such as transmission path propagation delay, fibre length, pattern length on the printed circuit board, cable length, and IC propagation time delay.
The new CFP module, for example, is a 100 Gbit/s pluggable module, where the PMA section de-multiplexes the 10 x 10Gbit/s lanes into 20 x 5Gbit/s lanes and then re-multiplexes them into 4 x 25Gbit/s. No skew control is performed in the CFP, so it is up to the MLD module in the network hardware to control and compensate for any existing skew between channels.
Another requirement of the higher data rates is that crosstalk must be measured with greater precision. Crosstalk generated by different channels inside active transmission components is strongly dependent on synchronization and skew between the channels.
High-speed test challenges
The introduction of the new transmission techniques to deliver 100 Gbit/s data rates changes the requirements for equipment testing so significantly that existing 10Gbit/s bit-error rate testers (BERT) will not be suitable.
There are several new testing challenges in the 100 Gbit/s environment. First, test engineers must be able to support multiple modulation schemes and test multiple channels and configurations. What's more, the test equipment must offer sufficient flexibility and scalability to support these different configurations, including some still to be defined by standards bodies. The need to generate and analyse multiple channels requires the BERT to be able to offer high flexibility in configuration, clock synchronization across multiple lanes, and pattern synchronization on all channels.
Indeed, in order to test and simulate the new modulation schemes, complex patterns for the modulator and pre-coder are required. An electrical signal generated by a new generation BERT (better described as a Signal Quality Analyzer) should not only be compliant with a pure pseudo-random bit sequence (PRBS) standard, but also adapted to emulate the pre-coding/decoding capability that new equipment will use to transform an NRZ/RZ signal into a phase-modulated signal.
It makes life a lot easier for the test engineer if this functionality is implemented as hardware conversion: the PRBS signal is automatically converted into a phase-coded signal at the transmitter of the BERT and, vice versa, recovering the original PRBS signal from the received modulated one.
Ideally, the BERT must also offer a high-resolution programming capability for user-defined patterns and support as many modulation formats as possible. Up to 1-bit resolution programmability per channel provides a convenient way to optimize programmable patterns when testing 40 Gbit/s and higher bit rates.
Another factor to consider is that channel synchronization and channel skew tolerance tests become more difficult at higher data rates. Channel synchronization and skew effect have an impact on the signal quality after optical modulation, as well as on the efficiency of multiple-wavelength transmission, and on accurate transmission through high-speed circuits.
High-speed signals show a tendency to skew even with only a small difference in signal-path lengths. A 20 Gbit/s signal, for example, has a 1-bit cycle of 50 ps, which is equivalent to a path length of around 1 cm. For adequate testing of modulation and demodulation, the tester must be able to guarantee correct synchronization of bit-pattern timing at the generator side, as well as providing the means to assess skew tolerance at the receive side.
As previously mentioned, crosstalk generated by different channels inside active transmission components is strongly dependent on synchronization and skew between the channels. It is important, therefore, for a BERT to guarantee pattern synchronization, and allow generation/compensation of at least +/-1 bit of skew between channels, to determine the robustness of the modulation/demodulation schemes and help to solve interoperability issues at the physical level.
The requirement for higher input powers by early prototypes of new components and devices also places more demanding specifications on test equipment. Tests must allow the devices under test to receive a greater input power than is necessary for today's 10 Gbit/s signals.
To avoid using an external amplifier (which can complicate the interpretation of test results), a BERT should be capable of generating high-quality, high-amplitude signal waveforms with a wide range of possible adjustments, including for amplitude, cross point and threshold. Generation of electrical output up to 3.5 Vpp from the BERT allows the instrument to directly drive 25 Gbit/s EMLs (electro-absorption modulated lasers) and run reliable evaluation of 4 x 25 Gbit/s channel aggregation (see figure 1).
Future test requirements
As the standards process progresses and equipment manufacturers pursue innovations, a new generation of testing equipment can be expected in the signal analysis R&D environment.
Test equipment manufacturers thus must attach great importance to:
- hardware and software flexibility;
- quick reconfigurability;
- provision of a very high number of different testing functionalities on the same platform.
The concept of equipment flexibility operates in two dimensions: in space (supporting multiple channels, multiple plug-in cards, stackable mainframes); and in time (easy to upgrade to future needs and new testing requirements).
In a market which shows renewed emphasis on standards and a desire to avoid excessive diversification, flexibility is the only way to cope with the rapid technical changes that evolving standards bring.
Thus, next-generation BERTs should build on a flexible platform, enable multi-channel testing, provide for bolt-on functions implemented in software, and combine different suites of hardware and software modules to enable switching among different testing capabilities. They must also integrate multiple techniques such as jitter tolerance on top of signal analysis and multi-channel skew testing.