Looking at the market for open specification equipment, a trend emerges – the customer is demanding higher performance. This drive relates not only to the performance of the silicon, but to the interconnect, the fabric that links these blades and modules together.
The requirement to develop higher speed interconnects has put designers under increasing pressure to create solutions that meet more and more stringent requirements.
At the turn of the century, the single Gigabit Ethernet (GbE) interconnect was considered challenging.
However, by last year the standard performance was the 10GbE XAUI (four lanes of 2.5GbE combined to create a 10GbE channel), and now in 2009 the 40GbE systems (four lanes of 10GbE running at 12.5Gbit/s SERDES combined to create a 40GbE channel) are being released to the market by companies like Emerson Network Power.
Based on this development timeline, 100GbE systems are not far away.
Plug and play
While people expect devices to “plug and play” together, increases in backplane speeds can be a barrier. The higher the frequencies involved, the more the signals create problems that require solutions in layout, material and many other aspects.
These differences magnify – in terms of silicon, connectors and backplanes – with these higher-speed systems. As a result, the interoperability aspects of open solutions, where many vendors’ equipment can be combined, have to be considered carefully.
Recognising these developments as part of the natural evolution of the market, PICMG (PCI Industrial Computer Manufacturers Group) formed the Interconnect Channel Characterisation (ICC) committee in 2007 to develop a new specification. The specification is designed to establish ground rules and definitions for characterising the transmission, reflection and crosstalk performance of backplane and mezzanine carrier interconnects.
This information is being developed to help system designers and integrators to predict more effective solutions in the scope of increasing the capabilities of the interconnect.
By knowing the performance of a channel before designing or integrating a solution around it, they are given the capability to drive the system architecture. For example, the committees looked at the aspects of channel performance characterisation and subsequently created a common methodology to represent it in a standard form.
The initial framework of the specification was created to allow multiple solutions to be tested in a standard way. There are a number of variations of this framework, which relate to different combinations of mezzanine, carrier, blade, backplane, etc, with the channel definitions being constant for all.
Once created, this framework allowed the committee to focus on the other more crucial aspects of the problems. When the initial statement of work was written, it was expected that differential s-parameters would be used to define the channel performance. These measurements would be standardised by using one of the frameworks.
Aspects of the ongoing work have been focused on the test equipment required to provide an accurate and repeatable set of values to accurately reflect the characteristics of the channel.
Currently, the specification suggests a number of solutions in the area of testing fixtures and connectors. These solutions are included to provide consistency and standardisation of measurements, as well as the experience of the committee in terms of measuring these characteristics.
Of course, once you have the model/parameters for the channel there is a need to deal with the information gained. The specification examines the techniques and definitions associated with the analysis of the results, again to provide consistency, and aid the designers in using the solution.
High-performance channels
The work being carried out by the ICC committee represents a strong addition to the PICMG specification family. It has paved the way for the next generation of specifications with a baseline framework for defining new high-performance channels. The ICC committee is also working closely with the ATCA 3.1 R2.0 (Advanced Telecom Compute Architecture Ethernet dot specification) committee to provide suggestions and understand the requirements of the specifications.
The specification is in draft 0.59, with a prospective release date of late 2009.
Stuart Jamieson works for Emerson Network Power