"GBIC" Review of MSA Requirements list

The group reviewed the document:

The following items were discussed as part of this activity.

Note that it may be only moderately important to include some of these items in the requirements list, but critically important to include these items in the standard specification for the components, connectors, and guide-rails. The discussion also mentioned the possibility that additional comments and requirements would be identified as people studied these modules.

Items 10, 11, and 12,

were felt to require some clarification.

The intent appeared to be to specify a maximum pull that could be applied to the module before it was yanked out of the socket, always without damaging the socket. The maximum pull appeared to be 130 newtons. Then a separate and lower force would be associated with the pull on the connector that would unlatch the cable connector from the module. The intent was that it could be pulled with a force of 40 newtons without disconnecting either the circuit or the cable connector, but that some force between 40 and 130 newtons would pop the cable connector off the module. This was a case where the requirement was less important than the actual specified range of values that compliant MSA modules must meet.

Item 12,

appears to specify a connector robustness spec, but the specified value may not reflect an actual requirement unless cable stiffness and the length of any semi-rigid strain reliefs are also considered. It was felt this requirement was not well defined.

Item 13,

specifies a plug/unplug cycle requirement supporting a minimum of 100 cycles. This was considered adequate.

Item 14,

specifies Z-Axis insertion and removal forces under 20 N. Actual measurements show about 15 N during insertion, which is harder because it compresses the kick-out spring. Some felt a number on the order of 10 N would be preferred, although it was felt that low removal forces were more important than low insertion forces.

Item 34,

specifies that the module uses only 3.3 volts. This was considered acceptable.

Item 37,

specifies only that operating environments need to be specified. Discussion indicated that, while most components were in a relatively benign office environment, many were placed on servers that were kept in a closet somewhere. Such environments may be prone to short-term or long-term high temperature operation. Ambient temperatures as high as 45 or more were mentioned, with air temperatures in the air-stream crossing the component as high as 75 degrees, though most agreed that represented the worst case component temperatures expected. Power density and air flow are the two major issues related to cooling. As a result, it was felt that a maximum module power needed to be specified as well as a minimum set of ambient conditions that all components needed to be able to tolerate.

New item A) Connectors supported.

The minimum set of the specific connectors that need to be supported is not specified in the document. The list should include at least:

Optical: LC, VF-45, MT-RJ, MU

Copper: TBD separable connector, non-separable module

New item B) Labeling requirements

The minimum set of labeling required should be specified. This includes the requirement for transmission mode color coding (beige or black for multi-mode, blue for single-mode). It may also include laser safety labeling, laser type, serial numbers, transmission rates, company identifiers, and other items.

New item C) GBIC compatibility

The electrical protocol and electronic levels and supply currents must be GBIC compatible, with the exception that 100 ohm differential impedance is expected for the high frequency signals. The protocol must include GBIC compliant timings as well as signal definitions. As part of this discussion, the valid MOD_DEF signals for MSA removable components was discussed. At present, it appears that all MSA modules use MOD_DEF = 4.

New item D) GBIC compatibility requires powered signals

The LOS and XMIT FAULT signals should use powered drivers. If a power signal is not available internal to the module, the LOS and XMIT FAULT should not simply ground out the signals terminated by the host side of the connector, since mistaken indications of proper operation would be provided.

New item E) Large addressability

More than 50 modules may co-reside on a switch back plane. A mechanism to address the modules would be useful, perhaps involving a module-select line of some sort. If all modules are assumed to be MOD_DEF = 4, there is a MOD_DEF signal which can be converted to a select signal.