davecb@yunexus.YorkU.CA (David Collier-Brown) (05/13/91)
Something struck me when I was wiring the console ports of a set of machines to a terminal server: why was I connecting the terminal server via optoisolators? Rooms full of terminals are exactley as likely to cause ground-loops as rooms full of cpu's... would not terminal servers have ground-loop protection already? Can someone confirm or deny this, so I can avoid spending money and laying glass-fibre all over my machine room. --dave -- David Collier-Brown, | davecb@Nexus.YorkU.CA | lethe!dave 72 Abitibi Ave., | Willowdale, Ontario, | Today's featured dish: CANADA. 416-223-8968 | Sun-dried alligator.
bruce@camb.com (Barton F. Bruce) (05/14/91)
> Something struck me when I was wiring the console ports of a set of > machines to a terminal server: why was I connecting the terminal server > via optoisolators? DEC does push their credibility gap a bit wider with their VCS recomendations. They are giving you a configuration that is supposed to be VERY conservative and safe and reliable. Needlessly so as you have noticed. Many have safely ignored their suggestions. Many have found that their fiber got knocked about too much and IT proved to be a very weak link when it BREAKS. If you are in an environment where ground potential is SO different between say another cluster on the other side of your arc furnace in your steel mill and where you are, use fiber for your ENET, and stick the remote console ports on a terminal server over at the other cluster. Don't use fiber for console ports themselves, especially in a typical computer room or average building. Just ignore DEC.
kla@kampi.hut.fi (Kimmo Laaksonen) (05/14/91)
In article <22713@yunexus.YorkU.CA> davecb@yunexus.YorkU.CA (David Collier-Brown) writes: > Something struck me when I was wiring the console ports of a set of > machines to a terminal server: why was I connecting the terminal server > via optoisolators? > Rooms full of terminals are exactley as likely to cause ground-loops as > rooms full of cpu's... would not terminal servers have ground-loop > protection already? No, the way RS232 is defined and implented you'll have the SIGNAL grounds tied together at whatever box (terminal server, port selector &c) you connect your devices to. However, that's not the real reason for ground loops. If the signal grounds were all "floating", the box signal ground could be connected to a proper (external) ground, and you'd have a single common ground, which is the proper way. Unfortunately in the power supplies the signal ground of a device is usually tied to the chassis (cabinet), which is tied to the PROTECTIVE ground. So each device, including the common box, has its own external ground. The ground loops commencing depend on how "close" the external grounds are to each other. In general I wouldn't worry about devices in a computer room, if the power wiring is properly done with a good, separate, protective ground from a single grounding point to all devices. A mix of power feeds from different phases may cause some ground loops, because the phases are usually not evenly loaded, but proper protective grounding should eliminate most of that. You can probably hire a company to check your power feeds (balance of load between phases, the quality of grounding). Ours was checked by a company selling UPS'es with a rebate if we chose their product (which we did). However, when connecting widely distributed terminals to a terminal server, ground loops are likely to occur. A simple rule is that the further away the terminals are from each other (different floors or wings of a building, and - heaven forbid - in separate buildings altogether), the more likely ground loops are, because the power feeds will come from sources "groundingwise" further apart. On the other hand, a room full of terminals with a terminal server, is not likely to have ground loops, since very probably the power feed comes from a common source, and all devices are groundingwise very close to each other. A guide book from a Finnish company selling terminal servers suggests that separate "data processing" power feeds with clearly marked outlets should be installed in new office buildings with one common DP power feed for about 10 - 20 rooms (and their terminal server and other common devices like a shared printer). Another note: in old power wiring the protective ground may be "faked" by connecting it (at some point along the power wiring) to the power return ("zero") wire. Such a ground is constantly floating with respect to a proper ground, depending on the load of the power feed. Tens of volts, sometimes even hazardous levels, will exist between, say, a chassis and a radiator (which is grounded through the plumbing). Beware of the shock! I've also seen thin Ethernet segments utterly unusable, when a T-connector in a PC was in contact with the chassis. The Ethernet was properly grounded at one end, but the PC's power came from a fake grounded outlet! ------------------------------------------------------------------------ Kimmo Laaksonen Internet: kla@hut.fi Helsinki University of Technology uucp: ..!mcsun!tut!santra!kla Computing Centre BITNET: KLA@FINGATE Otakaari 1, SF-02150 Espoo, Finland phone: int'l 358 0 451 4308