SURGE PROTECTION REVISITED 
by CHARLES P. KOONTZ, 
LAN TIMES, May 1990 
 
RESEARCH SHOWS SURGE PROTECTORS COULD BE CAUSING INTERMITTENT 
NETWORK FAILURES 
 
If this were a news story, it would have a panic-inducing headline 
like "Power Surge Protection Kills LANs."  Has your LAN ever crashed 
for no apparent reason?  Recent research reveals one more possible 
cause of intermittent network failure.  Surge protectors may 
contribute to LAN crashes by diverting surge pulses to ground 
thereby contaminating the reference used by data cabling. 
 
If you're like other conscientious LAN supervisors, you've already 
installed a dedicated power line for your server, you've checked the 
grounding to avoid problems with static electricity, all your PCs or 
workstations are protected by anti-surge devices, and all your 
software is widely used and accepted as appropriate for your network 
operating system.  In spite of your best efforts, every now and then 
your network hangs for no apparent reason, and you have to reset 
your file server.  The resulting downtime has everybody, including 
your boss, breathing down your neck and bemoaning the loss of the 
morning's work.  If Murphy's Law holds true, some downtime happens 
at the least opportune time.  One recent survey indicates that LAN 
failures can cost nearly $3.5 million annually. 
 
In 1988, research engineers at the National Bureau of Standards 
(now the National Institute of Standard and Technology, or NIST) 
accidentally found that conventional surge protectors could harm 
interconnected computing devices by surging the ground and 
contaminating datalines.  During an experiment in a newly 
constructed building, they introduced surges into the powerlines. 
After the experiment, they found damaged components attached to 
datalines between computers and printers.  The damage was caused by 
coupling between the data lines and the surges diverted to ground. 
This suggests that diverting a surge to ground with a surge 
protector may be one more power problem that can cause network 
difficulties. 
 
Powerline surges are packets of energy which propagate through 
powerlines like shock waves.  They are analogous to the vibration 
energy in water lines when the washer shuts off its water supply. 
 
Surges may be caused by factors outside the building.  Lighting 
induces large currents in powerline wiring.  Utility load switching 
and power factor corrections may also cause irregularities in the 
quality of power.  The Institute of Electrical and Electronics 
Engineers (IEEE) standard for external surges in 110-volt circuits 
is 6,000 volts and 3,000 amperes for 50 microseconds.  Major 
appliances usually ignore such surges, but delicate electronic 
devices usually sit up and pay attention. 
 
Equipment within the building may also cause surges as the machines 
switched on or off.  The magnitude of the surge is not necessarily 
proportional to the size of the change in load.  The type of power- 
switching devices in the building seems to have more impact on the 
magnitude of the surge than does the change in load.  These surges 
can easily reach 1,500 volts, but generally for much shorter 
duration than those associated with external surges.  A rise time 
of 5 nanoseconds and a duration of 50 nanoseconds are typical.  

While these internal surges lack the energy to cause massive 
physical damage, their short rise time permits easy inductive or 
magnetic coupling through nearby circuitry.  If the coupling occurs 
in datalines, errors may result.  A dedicated powerline will prevent 
internal surges, but has no effect on surges from outside the 
building.  The conventional wisdom regarding surges has been to 
shunt them to ground rather than permit them to pass through 
delicate electronics.  Surges are assumed to disappear harmlessly 
into ground. This is the crux of the problem - network datalines 
use the ground circuit for voltage reference.  Surges diverted to 
ground may not disappear harmlessly, but may have an adverse effect 
on the interconnecting datalines.  The surge-induced differences in 
ground reference voltages between the two computers will couple 
into the dataline circuitry and cause surge currents to flow in the 
data lines as shown in the accompanying figure. 
 
Because of dataline vulnerability, it may be best not to use shunt- 
type surge protectors which divert surges to ground.  The critical 
reference ground should be kept clean. Shunt components such as 
metal oxide varistors (MOVs), avalanche diodes, and gas tubes 
returned to safety ground all have the potential to contaminate the 
reference ground.  Unfortunately, most power conditioning devices 
contain these devices.  Many uninterruptible power supplies (UPSes) 
also contain them to protect their own circuitry. 
 
Modems may suffer more than other computer peripherals because they 
are connected to two different ground systems -- the local building 
ground and the telephone system ground.  A surge directed into the 
powerline ground could create a large voltage difference between 
the powerline ground used as a reference for the computer and the 
telephone line, which is referenced to the telephone system ground. 
 
All this implies that networks should only employ surge protectors 
that do not shunt surges to ground.  Interviews with the original 
researcher at NIST, a UPS manufacturer, and a manufacturer of MOV 
surge protectors support the idea that surges diverted to ground 
could have an adverse impact on networks depending on other 
grounding factors notably the quality of the surge protector.  It 
may be that this is an application of inappropriate technology.  If 
your LAN suffers from frequent unexplained crashes (this assumes 
that you have checked everything else -- static, valid grounding, 
perfect hardware, decent power quality, OS/application software 
compatibility, error-free operators, etc. -- you might want to try 
a better surge protectors.  The best way to judge a surge protector 
is to examine results of ANSI C62.41 testing on the device.  
Remember that dealing with power problems is a little like hang 
gliding: never fly higher than you're willing to fall.  One close 
lightning strike with poor surge protection could ruin your entire 
day. 
 
An alternative would be to find a surge protection device designed 
specifically to solve this problem.  As a minimum, surge protectors 
should not shunt the surge to ground, but rather to neutral at the 
point of use (next to the PC).  Although the ground and neutral are 
connected in the electrical service entrance for the building, the 
distances involved should help isolate the neutral conductor from 
the ground conductor because surges, like water follow the path of 
least resistance. 
 
The ideal surge protector would disconnect the powerline for the 
short duration of the surge (40 nanoseconds to 50 microseconds) and 
reconnect it immediately thereafter. "Immediately" is the problem. 
Today's technology has no power switch which will operate fast 
enough in a nanosecond environment.  An acceptable alternative 
would be a circuit that presents a high impedance to the surge and a 
low impedance to the power wave, while protecting the integrity of 
the ground circuit.  It should also contain no degrading components 
like MOVs. 
 
If power conditioning devices contaminate the reference ground by 
introducing surges, then it may be wise to remove such devices from 
a network or to replace them with something better.  This statement 
may cause some turmoil for network managers.  It should not be 
misconstrued as an indictment of surge protection devices, but 
rather as a call for additional research and development of better 
power conditioning devices designed specifically for network 
applications. 
 
Sidebar A--> Evaluating Surge Protectors 

User of  MOV-base surge protectors should be aware that small MOVs 
wear out more and more with each surge or spike.  This degradation 
is significate in MOVs smaller than 14 mm.  The more frequent the 
occurrence of surges and the more serve the surge, the shorter the 
life of the MOV.  They sacrifice themselves in the line duty while 
protecting the associated equipment.  If your surge protectors have 
been in use for awhile (6 months is a reasonable time), the MOVs may 
be incapable of proper performance.  You may want to consider 
replacing the surge protector, or, if you're good with hardware and 
a soldering iron you can replace the MOVs.  
 
Good surge protectors contain more than MOVs.  Look for hybrid surge 
protection that includes MOVs, active tracking filters, and safety 
fuses.  If it doesn't have UL or CSA certification for a transient 
voltage surge suppression device, don't buy it.  If a device has a 
UL certification as a temporary power tap, it means UL has a high 
opinion of it as an extension cord not as a surge protector.  
Somewhere on the product look for the UL 1449 clamping voltage.  The 
device should also provide normal and common mode protection. - CK   

 
Sidebar B--> MOVs and CLAMPING VOLTAGE 
 
MOVs contain a metal oxide paste that functions as a variable 
resistor where resistance is proportional to voltage.  The rating 
applied to an MOV that describe its surge protection characteristic 
is clamping voltage.  The MOV clamping voltage contributes to the 
surge protector's overall suppressed voltage rating. 
 
The clamping voltage is the voltage at which the MOV begins to do 
its job.  Intuition would indicate that the ideal clamping voltage 
would be as just above the normal AC line voltage.  This may not be 
the case.  The varistor is intended to survive repeated surges.  As 
the varistor ages, its clamping voltage decreases and it may begin 
clamping at normal voltage levels.  This can lead to a process 
called thermal runaway.  This process has resulted in at least one 
fire. 
 
The suppressed voltage rating for a surge protector is the actual 
amount of voltage permitted through the device during standard 
tests such as these conducted by Underwriter's Laboratories. - CK 

