21June2004 802.3ap Back Plane Ethernet Channel Adhoc Meeting Minutes Participants: mike-lerer@worldnet.att.net ahealey@agere.com joel@force10networks.com william.r.peters@intel.com cathyl@lsil.com glen@vitesse.com charles_moore@agilent.com popescu@quaketech.com john.dambrosia@tycoelectronics.com brian.seemann@xilinx.com Richard.Wolcott@taec.toshiba.com jcain@cisco.com ------------------------------ Joel Goergen opened the meeting. Mike Lerer acted as secretary and kept minutes. Joel made the remark that the package of documentation associated with test cards was large (more than 50 files). His intention was to remove all confidentiality statements. Inadvertently, some may have been overlooked. If in reviewing the material a "confidentiality statement" is found please let the document author know. There appears to be some confusion associated with Group Delay and more work needs to be done. The purpose of this call is to develop a plan to validate the channel mask sets. John D’Ambrosia talked about new variable length line cards that he is developing for use with the Tyco XAUI and QuadRoute Backplanes. He is developing a new set of daughter cards with "Improved FR4" and he is having difficulty deciding on what "Improved FR4" means. Currently he is targeting two grades of Improved FR4. (One a somewhat better grade and the other a somewhat lower grade.) Data from these daughter cards and his existing back planes will be available after July. At present John does not have approval from his management to develop 3 connector mezzanine card configurations. Joel reviewed his presentation Goergen_01_0604 (updated to Goergen_01a_0604) Updates Page 6 should be 3.5 and 3.4 (not 2.5 and 2.4) Updates Page 8 sma on left should go to layer HS8 Updates Page 9 3rd column under LYR should be HS8 not HS1 Page 3 reviews potential PCB materials. Joel intends to build test cards with Isola IS620 and Regular N4000-13. Page 4 shows the definition of Improved FR4 adopted at the last IEEE meeting. Page 5 shows Joel’s recommended changes to the definition of Improved FR4 Joel bases the recommended changes on the Dk values of Isola 630 and N400-13 Isola at 10 Ghz has a Dk of 3.42 and at 20 GHz has a Dk of 3.3. N4000-13 at 10 Ghz has a Dk of 3.38-3.45 and at 20 GHz has a Dk of about 3.4 John D’Ambrosia notes that the temperature range needs to be increased since the temperatures listed are currently system inlet air temperatures. Page 6 summarizes Joel’s recommend changes to the definition of Improved FR4. Page 7 reviews the 2 connector test card outline. There is provision for aggressor tracks to be evaluated for crosstalk. The test cards have SMA means to saturate the connector with measured amounts of crosstalk. Page 8 reviews the 3 connector test card outline. Joel asked if anyone on the call has a problem with viewing Allegro and Gerber files. No one had a problem. The goal of the test card development is to demonstrate that the Channel Model mask set is feasible. Page 9 Reviews the Channel Lengths. Channel 1 thru Channel 15 gives the total range of representative trace lengths. General discussion established the need for the group to have more specific detail on the mezzanine card configurations, connector types, trace lengths, typical applications etc. Page 10 & 11 Review the PCB Stack details. Note, that Joel will release two databases, one with Isola 620 construction and one with N4000-13. The construction strategy is that all traces go from bottom to top of board and so no back drilling is required. Joel’s boards are expected to be available July 6 or 7th. He is willing to share the boards. He is building 5 panels of each material. He is hoping to have two of each material available. If interested in access to the boards contact Joel. Joel is seeking volunteers to measure S parameters and other characterization, to correlate their results with him. He also expects that the test card configurations can be used to experiment with various coding alternatives. The meeting is opened to comments. Jeff Cain & Charles Moore has issues beyond 10 Gigahertz to 20 Gigahertz. Equipment just doesn’t operate reliably. There was group discussion about changing the upper limit of the frequency band specified from 20 Gigahertz to 10 Gigahertz. This will be discussed further on the Friday Conference call. Charles Moore intends to talk about a pulse response approach to channel characterization. John D’Ambrosia again reviewed the details of the test cards he is building. Joel and John intend to present channel data at the next IEEE Meeting in July. The group is interested in receiving channel data from ATCA backplanes and platforms. Group Delay Joel asked whether anyone has evaluated group delay. No one has yet examined in detail. John D’Ambrosia has spent some time and can not yet validate the group delay mask with his real world data. Neither can Adam Healy match the mask with the real world. Comments: The group delay is scaled around 0. The first number is the zero point. Everything else is about this number. Adam Healey stated that as a part of the measurement procedure we need to specify the group delay aperture (Df). A smaller aperture gives a noisier measurement, a larger aperture loses detail. To Do Actions for the group Evaluate changing the top frequency of the mask sets to 10 Gigahertz instead of 20 Gigahertz. Joel will release fabrication database files, schematics, and Cadence files for his test boards for the group to review Jeff Cain will take one of Joel’s test boards for an independent evaluation.