| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
Brian, See below.
John
-----Original Message-----
The important initial step in our process was to set the project objectives. These objectives and criteria framed a very viable and valuable project, and were carefully considered and ratified by the larger 802.3 body. Early Task Force work and consensus solidified the direction we are taking. 1. The project objectives specify operation over Improved FR-4. This objective establishes some very key guidelines: a)
First, it means that the project is dealing with
newly designed backplanes, as there is no prevalence of Improved FR-4 in legacy
backplanes.
Brian - I don't totally agree with this assessment. There was ongoing debate over how good a material would be, which drove the requirement for defining "improved FR'4". But there was an assumption that discontinuities would be gone, which quite frankly is a needed assumption in order to move forward with developing an informative channel model, as I have yet to see any reasonable proposals for trying to develop even a reasonable channel model where we deal with all the crap that stubs cause. However, this channel model is still informative. There are still a lot of people that would like to see if a stub effect can be tolerated, and in reality, when I talk to people, they don't talk about total elimination of a stub, they talk about zones of counterboring, which means getting stuck with some stub. But I also agree that some amount of reasonability is necessary, as i think there is probably very small chance of making a channel run that has a deep null starting at 2 to 3 G work. That is why I think case 6 is a reasonable case to start with. If we can't get that working, pulling nulls in even lower in frequency would most likely have even less chance. (But I will say that I will defer to the Gods of signal conditioning when relevant analysis proves this wrong!)
2. The directive we were given was to look at the channel model as a bilateral Contract between backplane and signaling. It was directed that the signaling must perform to at least under the line, and the backplanes must perform to at least above the line. (In this email, I will not deal with the over-simplicity of our present model approach) The bilateral nature of the contract means that if signaling is asked to provide margin to the contractual infractions of the backplane, the backplane will be asked to provide margin to the contractual infractions of the signaling. The net effect is mutually-contributed, balanced margin. 3. The directive given to the 802.3ap members was to unilaterally define the channel prior to any signaling considerations. The rationale was that this provided a more deterministic, linear progression through the project, which is admirable. The direction was that in the subsequent signaling evaluations, we could expect that any problems that such channel presented to the signaling could be dealt with by easing up on the channel at that later time. 4. The Channel Ad Hoc directive was to find the worst, reasonable channel and solidify the contractual line. And now, initial signaling evaluations have been performed on the channel. In what I believe surprised a number of people, the results show that the channel model does not require relief. All the signaling proved able to hit the target, even the most simple, available and low power. Instead of being ecstatic for the industry, it seemed this frustrated a number of people. Brian, I am not sure if you are including me in this statement. I was surprised, but in a very good way, and was baffled by the opposition to these models, since in general there were good results on all of them. I think the Case #6 is an excellent candidate to determine whether or not it is reasonable to even assume that the stub effect can be addressed, and results show it can, but this will come at the expense of additional power (my assumption that more DFE and FFE taps requires more power.) So now the other signaling groups also need to do the same thing. But this all goes into influencing a matrix of parameters that we need to specify, and then the vote will take care of it. I guess that is really the answer to the weighing scheme, right? Making sure that the right parameters are called out, so that the relevant information is provided for each individual to weigh all the factors and make an informed decision.
5. It would seem that the Task Force has done quite well in meeting objectives. Since we have been successful in meeting objectives, and our timeline says that signaling decisions are the next item, it is appropriate to be focusing on requirements consistent with our objectives. BrianS |