Posted by BigDumbDinosaur
Search Keys: lightning,grounding
Referencing: How do you protect yourself against lightning?
Grounding is everything, especially if building B gets its power from building A. Also, a high quality UPS (not the door stops made by APC!) is essential to protect hardware from power line hits. In areas frequented by lightning strikes, equipment should be powered by a ferroresonant UPS, which has far greater capacity to withstand strikes than a line interactive or double conversion unit.
In the only instance where a client of mine lost a system to lightning it was subsequently determined that the fatal strike had occurred to the roof of a storage building that was receiving power from the main office building via a local service drop (not a drop from a utility pole, but one strung between the buildings). This storage building has a metal roof and siding, and therefore is a natural lightning rod.
Anyhow, the bozo who wired the storage building did not bond the circuit breaker panel to a ground rod as required by the NEC. So when lightning came calling the energy of the strike saturated the metal structure with current, which passed into the guy wire supporting the service drop and then into the neutral of the main building. Due to the intense amperage involved, the neutral side, which normally should be at or very near ground potential, was suddenly raised to a very high voltage and like Sherman's army marching through Atlanta -- or, perhaps, driver ants munching their way through an African forest, the massive current flow burned up almost everything in its path. The entire computer system -- UNIX server, terminals, a PC, modem and the APC UPS -- was ruined.
This, incidentally, was the last time I installed an APC UPS in any system. My subsequent post mortem indicated that the actual voltage rise experienced at the input to the UPS should have been survivable. APC refused to make good on their advertised 25,000 dollar system protection, claiming that it didn't cover direct lightning hits to a building -- a caveat that I couldn't find anywhere in their warranty statement. My opinion of their products has not changed since then: a client of a friend of mine lost a system to a power line strike, which destroyed his APC UPS.
While on the subject of power issues, a sneaky source of trouble can be an effect I refer to as ground plane potential imbalance -- GPPI for short. Suppose you have a server in one location and a peripheral device -- a serial interface terminal, for example -- in another location. Each location gets its power from a different breaker panel.
In order for the server and terminal to exchange data, the signal grounds of each device must be connected together. It is customary design practice for the signal ground in each device to be bonded to chassis ground at some point, which means signal ground at both ends will be bonded to earth ground via each device's line cord (assuming the power receptacle itself has been properly grounded).
The problem is that even though ground is ground, small amounts of current may flow within the ground plane, resulting in GPPI. This happens because most grounding in interior electrical systems is obtained through the network of interconnected EMT (thinwall conduit), and not all joints are as solidly bonded as they should be. Also, neutral imbalance can occur -- often due to poor planning in the electrical system layout, which can cause some current to flow through the ground plane because neutral is always bonded to ground in a breaker panel.
If two interconnected devices -- our server and terminal in this case -- are at opposite ends of a GPPI situation, some of that stray current may flow through the ground lead in the data cable, causing the nominal logic levels to vary at twice the line frequency. This sort of voltage variation is not accounted for in the design of the device and instability may occur for no apparent reason.
In cases where I suspect GPPI is making a data link act up I will insert a low value resistor into the signal ground connection at one end and measure for 60 Hz voltage across the resistor. Experience has shown that as little as 2 volts GPPI can affect the stability of a high speed RS-232 link.
Got something to add? Send me email.
More Articles by BigDumbDinosaur © 2011-04-28 BigDumbDinosaur
The history of the world teaches us that succession is dangerous and that the strong take what they want. It's not likely to be any different with Linux. (Tony Lawrence)