| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
All,
Rich Mellitz and I were reviewing the channels being discussed, and were reviewing the crosstalk aggressors. We discovered a potential issue with the FEXT data submitted for the ATCA channels. In Tyco’s data it is assumed that the two FEXT aggressors are equivalent, while in the Intel data the FEXT data was generated where all aggressors came from the same line card. For a real ATCA system the two above scenarios are incorrect. In the ATCA system channels are based on the 8 differential pair from a single line card to the fabric. Thus, if you look at the fabric a single Rx pair would be hit on one side from an aggressor originating from the same line card and another aggressor originating from a different line card. From initial simulations Rich has done, the value of FEXT in a net where a single connector is shared is significantly lower than a case where two connectors are shared. Therefore, one of the FEXT aggressors is probably less than what was originally anticipated. Tyco is currently modifying line cards to provide test data that shows this effect. Hence, we should gain some further margin for ATCA applications by reducing one of the 3 contributors.
John D'Ambrosia