jqj@cornell.UUCP (08/21/84)
From: jqj (J Q Johnson) Cornell Engineering is in the process of designing a large Ethernet configuration spanning 10 buildings on the Cornell campus. Our plan is the traditional 500m backbone, with repeaters in each building connecting to an in-building segment. Concern has recently been raised about the differences in ground reference potential between buildings. Does anyone out there know whether differences in ground reference between two different controller/transceivers on an Ethernet are likely to cause degradation in the network? What is the maximum allowable difference in ground reference? My Blue Book does not discuss this issue directly, though some of the discussion (notably sects. 7.6.3, 7.4.3 and 7.2.4) leads me to believe that a moderate difference in ground might be acceptable. Anyone out there have experience or advice?
cak@CS-Arthur (Christopher A Kent) (08/22/84)
As I understand it, ground voltage differences should not be a problem, IF you follow the recommendation to only ground the shield of the ether coax at one point. The transceiver power supply and signals between the xcvr and host are isolated from the cable, so there is no possibility for interference there. If you ground the shield more than once, you'll almost certainly have troubles. Note that some transceivers have portions of their cases connected to the shield ground, so nudging them into contact with a local ground can cause ground loops that are VERY hard to find... Cheers, chris
sunny@sun.uucp (Sunny Kirsten) (08/23/84)
LAN Electrically conductive Ethernets should be grounded in one and ONLY one location. 3Com or Xerox could provide you with a copy of the "Blue Book" Ethernet 2.0 Specification. In cases of properly installed ethernet equipment being "blown-out" by electrical storms, replace key links in the network with the Ungermann/Bass-Seicor-FiberLan fiber-optic Ethernet to achieve electrical isolation of now two different electrical Ethernets, each of which may/must be grounded at exactly one point (end). -- {ucbvax|decvax|ihnp4}!sun!sunny (Sunny Kirsten of Sun Microsystems)