For 10Gbit/s data rates, there are a variety of transceiver MSAs, as follows:

• The 300-pin MSA converts between a 10Gbit/s serial optical signal and 16 parallel 622Mbit/s electrical signals, and currently accounts for most shipments.

• The XENPAK MSA was co-founded in March 2001 by Agere Systems and has more than 25 member companies. It provides a smaller form factor, since it uses four channels running at 3.125Gigabit/s on the electrical side. Users have limited development efforts due to the market slump, but demand for smaller modules is now picking up.  But, although XENPAK is designed for the heat dissipation necessary with high-power, long-reach telecom lasers, it also  provides the standard for shorter-reach 10 Gbit Ethernet (10GbE) transceivers, for which it is larger than desired. MSAs have  been proposed that are compatible with XENPAK’s four-wire 10 Gbit attachment unit interface (XAUI) and 70-pin electrical connector, but with smaller form factors for space-constrained applications to move to 10Gbit/s.  The XPAK MSA group (www.xpak.org) was formed in March 2002 by Intel, Infineon Technologies and Picolight. In August XPAK made available Revision 2.0 ( “build-to” specification).  In September membership grew to 21.

• The X2 MSA (www.x2msa.org) was formed July 2002 by component suppliers Agere Systems and Agilent Technologies and subsequently supported by vendors JDS Uniphase,  Mitsubishi Electric, NEC, OpNext, Optillion and Tyco Electronics. Shipping from first-half 2003, XPAK and X2 transceivers are initially focused on shorter-reach 10km links (comprising 80% of 10 GbE port applications) and second-generation applications that do not need XENPAK’s thermal capacity (though the heat sink can be adapted to different 10Gbit/s applications). So, although OEMs wanting to launch products immediately are going with XENPAK, it is expected that XPAK and X2 will grow faster than XENPAK implementations.

• The XFP (10 Gigabit Small Form-Factor Pluggable) serial interface module group (www.xfpmsa.org), was founded in March 2002 by 10 networking, system, optical module, semiconductor and connector companies, including Broadcom, Brocade, Emulex, Finisar,  JDSU, Velio, Maxim Integrated Products, ONI Systems, Sumitomo Electric,  ICS,  and Tyco Electronics. Unlike XENPAK, X2 and XPAK, XFP has a 10Gbit/s serial electrical interface (XFI) that converts serial 9.95-10.7Gbit/s electrical signals into external serial 9.95-10.7Gbit/s optical or electrical signals. This eliminates  mux/demux serial-to-parallel conversion logic chips inside the module and allows the serial 10Gbit/s physical-layer IC (PHY) to be moved on to the PCB (away from optics generated heat) and everything up to the XFI serial interface to be integrated into the CMOS media-access controller chip.
                                                                                                                                                                            

 
XFP component integration compared with a traditional
10G tranponder. Graphic courtesy of Essient.

The effects on the module are to reduce:

• Size (at 30 pins, a fifth the footprint of a parallel-interface MSA, enables higher-density boards with up to 16 ports in a 19” rack);

• Power dissipation (to one half) - Half of XENPAK’s 6W consumption can be for serial-to-parallel conversion, reckons Christian Urricariet, director of optical product marketing of Finisar, which is developing modules for XFP, but not XPAK and X2.
  
• Cost (due to fewer components, materials and higher circuit integration at 10Gbit/s).

Also, because XFP is not protocol dependent, multiple market segments will benefit from the aggregate volume.  The XFP spec was available for final review in January, and final adoption in February.  “XFP products compliant with this spec are being shipped by several companies,” said chairman Bob Snively, principal engineer at Brocade. The XFP group has now grown to over 80 companies.

Extended reach at low power

The first XFP module demonstrations were given at September’s ECOC 2002 by Luminent, when Finisar also started to ship its first 10km-reach XFP module.  The 10km-reach market could be lucrative for the use of 1310nm lasers, especially for storage area networking, once this has reached 10Gbit/s speeds and its 10Gbit/s Fibre Channel protocol has achieved standardisation.
 
However, March’s OFC 2003 saw demonstrations of XFP modules with a reach extended from the standard 10km to 40km (10GBASE-ER/EW spec 10 Gigabit Ethernet and IR-2 spec SONET OC-192).  This enables more cost-effective 10Gbit/s optical links for metropolitan-area networks. For telecoms, 40km reach usually needs a 1550nm laser, says Steve Joiner, marketing director of Ignis Optics, but at OFC 2003 Ignis demonstrated a 1310nm 40km-reach XFP module.

“There are some advantages in using 1310nm rather than 1550nm,” he says. One of these is lower power consumption and hence possible elimination of cooling components and a reduction in module size and cost for a given reach.

For example, Finisar also demonstrated 40km-reach XFP modules that can use either a cooled 3.5W 1550nm electro-absorption modulated laser or a 2.5W 1310nm distributed feedback laser.

Meanwhile, at the first Physical and Link Layer interoperability demonstration, sponsored by the Optical Internetworking Forum, IC-maker Applied Micro Circuits Corp (AMCC) demonstrated use of electronic compensation technology in combination with both 1550nm and 1310nm DFB lasers to extend reach of low-cost optics to more than 60km.

XFP announcements have also come even from XPAK-adherents Intel and Infineon.  At OFC 2003 Infineon’s Wireline Coms unit gave live interoperability demos of its 10km-reach 1310nm 10G XFP modules, which consume less than 2.5W, it says.  “Even some systems due to sample aren’t locked into an MSA yet,” says Bob Zona, senior product marketing manager for Intel’s optical platform division.

Opto integration

An alternative to using uncooled 1310nm lasers to reduce module size is to retain 1550nm lasers but to use optoelectronics integration to reduce module size and power consumption while maintaining reach. Compared to XENPAK, XPAK and X2, XFP’s elimination of silicon electronic components opens up greater possibilities for employing integration of the optoelectronic functions.

Scottish opto component developer Essient Photonics - formed in February 2002 and recently gone into liquidation – said that, for 1310nm solutions limited to 10km reach and power Level 4 (4.5W), XFP’s 900mm2 package allows an area for components of about 400mm2. In contrast, intermediate reach (IR) 1550nm modules have difficulty meeting XFP’s power and size constraints.

Essient claimed that, by applying its InP technology to integrate both high-speed electronics (resonant tunnelling diodes) and optoelectronics (waveguide structures) on a single chip, its electro-absorption modulator (EAM) enables improvements in manufacturability, component count and size (to 350mm2), power consumption, simplicity of optical and electrical interfaces, reliability and cost, while extending 1550nm reach to 40km and 80km.

A high DC optical extinction ratio of >20dB is obtained with an optoelectronic interface drive voltage of about 100mV, compared with 5V for a LiNbO3 modulator and >2V for a conventional InP-based EAM.  This eliminates the need for high-frequency, high-power InP-based external transistor circuits  and makes signal levels compatible with output signals coming directly from the CMOS silicon serialiser/deserialiser chips. Driver circuitry is a significant source of heat in a module, using bulky heat sinks.

Essient’s EAM-10G EA Modulator/Detector which integrates O/E and E/O functions - was announced at NFOEC 2002 show. Essient claimed it enabled a reduction in power consumption of more than 70% - to XFP-spec power of 2.5W (compared to 10W for traditional transponders). By 2005, Essient aimed to integrate of components on a platform, giving size reduction and 1.5W power.

Next-gen adaptation

Developments are prompting OEMs to look at moving directly to XFP a generation early. “You might as well start designing for it now…We have some people moving directly from 300-pin to XFP,” says Christian Urricariet, Finisar’s director of product marketing for 10 Gbit/s optics.  “Even very traditional telecom OEMs are starting to implement XFP in next-generation  products.”

For example,  AMCC and Broadcom have produced adapter cards that allow an XFP module to be plugged into the slot for a 300-pin MSA, intended for testing, so that OEMs can begin working with XFP without waiting for new boards to be completed.  AMCC also demonstrated its PHY with XFP evaluation system, which verifies the interoperability and performance of its 10G PHY devices in an XFP application using customers’ existing 300-pin MSA slots for maximum flexibility.

At OFC 2003 TriQuint Semiconductor  introduced both cooled and uncooled 10Gbit/s Transmit Optical Sub-Assembly (TOSA) and Receive Optical Sub-Assembly (ROSA) optical engines to support all reach versions of both X2 and XFP modules, from its recently expanded main site for assembly & test in Matamoros, Mexico.

“We are leveraging our ability to make our own optical components, chips, and transceiver modules,” says Glen Riley,vice president and general manager of TriQuint Optoelectronics (acquired from Agere Systems). Flexible manufacturing platforms allow volume cost advantages in material sourcing for X2 and XFP, he says.  The objectives of such developments are to accelerate adoption of XFP modules and demonstrate how it can be used to lower the costs of fibre-optic communications in the metro arena.




Optical modules: good news and bad

Over the last 3 years, optical modules have slowly been growing in popularity, according to In-Stat/MDR.  The high-tech market research firm reports that, if it had not been for the recent economic downturn, optical modules would be the product of choice for most optical networking equipment.

But though they afford a benefit to both those that make optical networking equipment and those that buy it, growth of optical modules has been stifled (specifically in the 10Gbit/s line rate) by an unwillingness amongst companies to try something new during long periods of uncertainty.

However, Optical Modules: Building Blocks for the Future reports that from 2002 to 2007 the market is expected to grow from just over  $1bn to roughly $3.5 billion with a  CAGR of 30.5%.

“Fortunately, as  the economic recovery starts to take hold in the second half of  2003, the benefits of using optical modules will merge with a renewed optimism expected about the networking equipment industry, and we will see real traction from these products,” says Eric Mantion, a Senior Analyst with In-Stat/MDR.

“The  companies that make optical networking equipment will choose to use optical modules as they reduce the complexity and difficulty normally associated with creating platforms that can operate at speeds up to ten billion bits per second.”

In-Stat/MDR notes the optical modules benefits :

• Since networking equipment vendor will be purchasing modules from reputable providers, there will be less of a concern about quality issues.
 
• Modules can be tested before they are mounted on the optical networking board, reducing the timeneeded to diagnose component failures by quality departments.

• As optical module makers refine and automate the process of making modules, associated prices are expected to decline due to better  yields, thus making optical networking equipment more affordable.

• For the end-user, some of the pluggable modules that are emerging will help enterprises and service providers reduce their operating expenses since fewer replacement boards need to be kept on hand in lieu of stock of the much cheaper optical modules.

But there is a risk to optical modules at the 10Gbit/s line rate. Today, there are no fewer than 5 different Multi-Source Agreements (MSAs) at various stages of development.

Almost 100 companies are involved in all of these MSAs, with a third of them active in more than one MSA.  In-Stat/MDR feels that if there is not at least some paring down in the number of MSAs, then the market at 10Gbit/s may be too fragmented and there will not be sufficient volumes for any one of the MSAs to drop prices enough for optical modules to be a true success.

On the  bright side, there are a large number of mature companies that  are involved with this process; therefore, unity is likely to prevail over fragmentation. By the end of the forecast period, In-Stat/MDR expects that there will only be two main MSAs left standing.


 The report is priced at $2,495.
Contact: Eric McKeighan.  Email: emckeighan@reedbusiness.com or Web http://www.instat.com/abstracts/ IN0301116NT.asp