| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
My observation is that we have a lot of discussion about specific numbers (e.g. “34” vs. “28”), but it is unclear what these numbers mean. A lot of COM analysis has been done on “die-to-die” channels, but people seem to have different opinions on how these channels are constructed. We have not seen a proposal for a reference model for a module (PCB, AC coupling caps, package?) The die-to-die channel is a combination of host and module that come from different suppliers; its loss cannot even be a recommendation unless these two components are specified in some way. I think the discussion is shifted towards this “die-to-die loss” because of the (TBD) dB values in Figure 176E-2; but nothing in this figure is normative. In C2M we don’t specify loss, we specify signals and functionality. None of the loss values in the figure is measurable and the indicated test points aren’t accessible to make signal measurements. In comparison, the similar Figure 176D-2 does not include any loss values. Neither does Figure 179-2, which is related. I would like to focus our discussion on well-defined and clearly-owned loss values. My suggestion is to seek consensus on the loss allocation of the host channel as defined in 179A.4 of D1.1 (“consisting of controlled impedance PCB, device package, and host connector footprints”). This can be easily translated to TP0d-TP1a loss by adding the reference loss of a mated test fixture (currently TBD) minus the MCB PCB/via (2.7+0.8=3.5 dB, per Figure 179A-3). The host channel loss is not a specification by itself, but it will affect normative specifications, directly (module input test, and host input test calibration) and indirectly (limits for host output characteristics), so it is required to close some critical TBDs. Separately from that we should discuss loss allocation of the module channel (from the contacts on the paddle card to “TP1d”). We will also need a model based on expected module design; does it include package, and what type? These will affect normative specifications too. If we agree on these two components, then the die-to-die loss budget will be a direct result. I am planning a contribution on this topic in the next ad hoc. If people have suggestions for loss allocations for these two components, I would like to include them (multiple viewpoints!) BOTTOM LINE: proposals for loss allocations of host channel and module channel (as defined above) would be welcome. </Adee> From: Lusted, Kent C <kent.c.lusted@xxxxxxxxx> Dear Colleagues, As the Task Force Chair noted during the TF closing business, there will be an agenda item at the 1 August electrical ad hoc on the topic of “AUI C2M Loss Budget”. I personally observed much discussion and difference of opinion on the total loss number for AUI C2M during the Plenary week. I expect that there will be a large number of comments submitted against D1.1 on this subject during D1.1review. In anticipation of efficient resolution of comments, I would like to initiate the discussion here on the reflector before the electrical ad hoc meetings commence. Furthermore, I kindly ask each participant to think about solutions to this issue and ways to resolve it. Please reach out to participants that have a different viewpoint than yours and initiate a discussion towards consensus. With regards, -Kent Electrical Track Chair, IEEE P802.3dj Task Force To unsubscribe from the STDS-802-3-B400G list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-B400G&A=1 To unsubscribe from the STDS-802-3-B400G list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-B400G&A=1 |
Attachment:
smime.p7s
Description: S/MIME cryptographic signature