Thread Links | Date Links | ||||
---|---|---|---|---|---|
Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Chris, I totally understand your perception regarding the ATCA connectors. Tyco Electronics made the decision several years ago to conservatively market the Z-PACK HM-Zd connector as a 3.125 to 6.25 Gb/s connector, but also has been quite active in pursuing the concept of the Active Interconnect, which essentially is that the system performance is a combination of the active and passive components, to explore the limits of performance of this connector. Since that time there have been a number of demonstrations with NRZ, PAM-4, and Duo-binary that show the ATCA connector can support 10 Gb/s with a proper implementation. For your reference see the following presentations on the IEEE 802.3ap website-
* Per Tyco channels from non-ATCA platforms n hoppin_01_0104 (PAM-4 testing and simulation) http://grouper.ieee.org/groups/802/3/bladesg/public/jan04/hoppin_01_0104.pdf n hoppin_01_0304 (PAM-4 testing) http://grouper.ieee.org/groups/802/3/bladesg/public/mar04/hoppin_01_0304.pdf n dambrosia_01_0504 (PAM-4 testing) http://grouper.ieee.org/groups/802/3/ap/public/may04/dambrosia_01_0504.pdf n Sinsky_01_0704 (duobinary testing) http://grouper.ieee.org/groups/802/3/ap/public/jul04/sinsky_01_0704.pdf * Other ATCA channels n seemann_01_0504 (NRZ testing) - http://grouper.ieee.org/groups/802/3/ap/public/jul04/seemann_01_0704.pdf * Per ATCA channels defined in dambrosia_01_0904 n sinsky_01_0904 (duobinary testing and simulation) - http://grouper.ieee.org/groups/802/3/ap/public/sep04/sinsky_01_0904.pdf n abler_01_0904 (NRZ simulation) - http://grouper.ieee.org/groups/802/3/ap/public/sep04/abler_01_0904.pdf n lui_01_0904 (NRZ simulation) - http://grouper.ieee.org/groups/802/3/ap/public/sep04/liu_01_0904.pdf
Also, you can check out a presentation we did almost 3 years ago at DesignCon - http://www.tycoelectronics.com/prodimages/DC2002_Tyco_Gennum.pdf
So I am trying to identify the limits of the implementation that people are concerned about. If you look at Case #4 of my presentation http://grouper.ieee.org/groups/802/3/ap/public/sep04/dambrosia_01_0904.pdf these are ATCA implementations (full mesh - 6" LC's +20" BP, dual star - 6" LC +10" BP). Cases #4 and 5 are well above the proposed channel model, but don't include significant stub effects. Case #6 fails the channel model after 6 GHz due to stub effect. Same connector - different implementation. If one assumes stub reduction via backdrilling, this channel will easily pass as well.
So Tyco does feel that there is a forward path for ATCA, and will continue to support that belief.
John D'Ambrosia
-----Original Message-----
Hi,
I am also a fence sitter here. Just a few comments from a market perspective. Current ATCA blade implementations use GbE on the fabric interface (according to PICMG 3.1). It would be a plus for P802.3ap based ATCA hub boards (switches) if they would (beside talking to P802.3ap compliant node boards) operate with existing node blade designs based on 1000BASE-BX. It would help al lot to bring P802.3ap based products much faster to market as not each node blade design would be required to be adopted to the new backplane standard.
So any technology decision you are making here should be done also with respect to backward compatibility to existing ATCA fabric technologies if possible. I understand that current ATCA backplanes may not be suitable for 10G and it would be acceptable to replace those by others with suitable material. ATCA connectors where initially not designed to cope with 10G. Obviously it would be desirable to keep this connector type for backward compliancy reasons. I appreciate any effort you are doing here to reuse current ATCA connectors for 10G. Chris Engels, Strategic Marketing Manager Motorola, Inc. Embedded Communications Computing Group Tel: +49 (0)89 608 14-235 Fax: +49 (0)89 608 14-276 Lilienthal Str. 15 D-85579 Neubiberg/Muenchen email: Christian.Engels@motorola.com http://www.forcecomputers.com
|