In case you haven't noticed, power-line conditioning is a growth industry. There seem to be two basic reasons for this. The first is fear, whether founded
Publish date:


Jul 1, 1998 12:00 PM, Bill Whitlock

In case you haven't noticed, power-line conditioning is a growth industry.There seem to be two basic reasons for this. The first is fear, whetherfounded in reality or not, of catastrophic equipment damage. The second isthe pursuit of the reduction of noise and/or data errors in electronicsystems. My research convinces me that there is no world-wide deteriorationof power quality, but factions of the power protection industry arecreating paranoia about equipment damage and misleading information abouthow the power line causes system noise problems. Often, equipmentprotection and noise reduction are at crossed purposes unless the propertypes and locations are chosen for power-conditioning devices. For example,depending on its location within a system, a surge suppressor may actuallydestroy signal or data interfaces when a surge occurs. Power-conditioningdevices should not simply be randomly installed at every convenientlocation.

Characterizing the power lineA number of organizations conduct and document power quality surveys inNorth America. Among them are the National Power Laboratory, the CanadianElectrical Association and the Electrical Power Research Institute. Becausethese groups do not sell protection devices, I think their data isreliable. Survey results, of course, are statistical, classifyingdisturbances into several categories [ref 1].

A transient disturbance, such as a spike, is typically defined as onelasting less than 10 ms or about a half-cycle of the normal 60 Hz voltage.Recent surveys suggest that 95% of all locations in North America canexpect as many as 19 such transients per year at levels over 200 V. Fear ofthese spikes has probably sold millions of surge-suppressor devices. Infact, since the 1960s, researchers have reported a gradual decrease in theseverity of transient surges, probably because the widespread use ofsuppressors has made it nearly impossible to make measurements of actualtransient voltages.

Although this may sound like good news, the bad news is that high transientvoltages have now been converted to high transient currents [ref 2]. Theauthors of this reference also recently performed tests indicating thatvery few ordinary light bulbs will survive us range surges of 1,200 V to1,500 V. If the spike occurs at the proper time in the 60 Hz cycle, aslittle as 800 V triggers an internal flashover, which is fatal to the bulb.Because this kind of bulb failure is rare, with the vast majority failingas expected at the end of their designed service life, they concluded thatsurges over 1,500 V are quite rare. The same authors performed real-worldtests to demonstrate how lengths of typical branch circuit wiring limitfast spike currents.

Momentary disturbances-sags, swells or brief interruptions-are typicallydefined as lasting more than 10 ms but less than three seconds. Surveyresults predict less than six swell or over-voltage events (serviceentrance voltage up to 150 V for 10 ms or more) and fewer than 16 sagevents (service entrance voltage under 84 V for 15 ms or more) per year.

A steady-state disturbance-an over-voltage, under-voltage or outage-laststhree seconds or more. Survey data predicts fewer than ten under-voltage oroutage events per year (service entrance voltage under 96 V for threeseconds or more), and the majority of these tend to be complete outages.The data suggests no sustained over-voltage events (150 V or more).

Ideally, the power line voltage would be a pure, undistorted 60 Hz sinewave. Any other voltage at a different frequency is referred to as noise.Even if pristine power existed at the generating power station, the linesdelivering it have resistance (and inductance), and the non-linear loads(don't draw current smoothly) are connected to the line, thus resulting ina distorted waveform containing new frequencies which are multiples of 60Hz. This harmonic distortion commonly reaches 3% to 5%. Additionalhigh-frequency components are added by arcing switches or relays and loadssuch as switching power supplies, television or computer monitors andindustrial equipment. In addition, because an elevated power line makes anexcellent antenna, there may be many RF signals, too. This collective noisegenerally has a voltage level much lower than the line voltage and doesn'tpose an equipment-damaging threat.

Equipment protectionThe majority of electronic equipment can easily withstand brief transientover-voltages in the 1,500 V to 2,000 V range without damage ordegradation. Therefore, surge suppression is a form of insurance against arare but potentially destructive event. High-voltage transients are mostlikely to arrive on incoming utility power lines caused by lightning orneighbors who use heavy industrial equipment. Placing the suppressor at thepower entry panel is not only good common sense, but it also has theimportant benefit of eliminating high transient voltages in the building'sground wiring during transient events. In the absence of suppression at theentry panel, any suppressor on a branch circuit outlet may produce veryhigh transient voltages between various grounds within the facility. Ifsignal or data-interface cables link equipment connected to these grounds,interfaces may be damaged or destroyed. For the same reason, any otherlines that come from the outside world (telephone lines, CATV, DSS or otherantennas or even outdoor loudspeakers) should have surge suppressors toearth ground (preferably the same one as used at the power entrance) toprevent high surge currents from flowing into the building's safety groundwiring. If surge suppression is necessary at an outlet, consider aseries-mode suppressor, which does not dump high transient currents intothe safety ground wiring.

Most equipment can also cope with sags or swells, as long as they do notlast too long. Power quality people use the term "ride-through" to describethe ability of equipment to operate properly during a momentary period oflow line voltage. Equipment manufacturers sometimes call it "hold-up" time,and a typical specification might be 20 ms. An uninterruptable power supply(or UPS) is often used to increase ride-through time for criticalequipment, such as computers. A UPS relies on a battery with enough energyto supply power for some specified time. A UPS or voltage regulator may berequired to cope with swells, but they are a relatively rare problem.Obviously, a UPS or back-up generator of some kind must be used forextended outages.

Power line noise and safety groundingOutlets have an additional safety ground (grounding) conductor. Because ofthis, noise can exist as line to neutral (called normal-mode,differential-mode or transverse-mode noise) and/or neutral to safety ground(called common-mode noise). (See Figure 1.) Normal-mode noise is generallynot a problem in electronic systems-only the equipment's power supply seesit. Common-mode noise (or 60 Hz hum voltage for that matter) is zero at thepower service entry panel because neutral and safety ground must be bondedat that point. However, common-mode noise builds up as we follow branchcircuit wiring away from the entry panel. (See Figure 2.)

Two mechanisms create this common-mode noise. First is the voltage dropcaused by load current flow in the neutral conductor. Second is the flow ofleakage currents in the safety ground conductor. These currents flow fromline to safety ground inside each piece of equipment having a three-prongplug, flowing through the power transformer's interwinding capacitancesand/or the EMI line filter (if used) capacitances. (See Figure 3.) Becauseof these two mechanisms, it is virtually impossible to keep all of the(chassis) grounds of physically distributed electronic devices at the samevoltage.

Noise entry into electronicsMost noise problems blamed on the power line are actually caused by theseinter-chassis voltage differences in real-world systems, causingsignificant power line noise currents to flow in any wire connecting twodevices. These currents will flow even if both devices have two-prong plugs(ungrounded or floating), as shown in Figure 4. Currents up to 1 mA arecommon between floating devices, and currents up to 100 mA are commonbetween grounded devices. Because of common-impedance coupling in theinterconnect cables, unbalanced interfaces have no inherent ability tosuppress the effects of these noise currents, making them particularlyvulnerable to hum and power line noise [ref 3]. Unfortunately, consumeraudio, standard 75 V coaxial video and RS-232 computer data interfaces usesuch unbalanced interfaces. Artifacts include hum, buzz, clicks or pops inaudio systems, slowly rolling dark or light bars, bands of specks, orherringbone patterns in video systems and mysterious data errors, lock-upsor crashes in data systems. Of course, using high-performance balancedinterfaces throughout a system in the first place will greatly reduce, andin many cases eliminate, the need for other noise reduction measures.Theoretically, balanced interfaces can completely eliminate the effects ofpower-line noise currents, but those in most audio equipment don't performas well as their specifications may indicate when the equipment isconnected to a real- world system [ref 4][ref 5].

System noise reductionAt times, the safety and system performance objectives of grounding mayseem at odds with each other. It is very important to remember that theprimary purpose of grounding is safety. The safety purpose of groundingmust never be compromised for the sake of noise reduction.

In general, noise problems in any system become more severe when the powerconnections for the various system devices are more distant from eachother. Even if all system devices are powered from the same branch circuit,greater lengths of safety ground wire or conduit between devices willgenerally increase noise. Interconnections between devices powered fromdifferent branch circuits are likely to be even noisier. Obviously,powering every system device from, for example, a single outlet strip willtend to reduce noise problems.

For unbalanced interfaces, ground- noise coupling is an inherent problem,which generally worsens as cables get longer. Noise-reduction strategiesfall into two basic categories, one of which is the reduction of theshield's impedance, which can be done by keeping the cable as shortpossible, using cable with a heavily braided shield rather than foil anddrain wire and maintaining good low-resistance connections at each end.

A second strategy involves reduction of the ground noise current flowing inthe shield. One obvious way to do this is by avoiding unnecessary groundconnections. Adding additional connections from a piece of equipment tosafety ground, earth ground, rack ground or other special grounds willtypically only increase circulating current in unbalanced cables. In mostsituations, noise current in the shield will be much less if floatingequipment is simply left floating. Remember that if a piece of equipmenthas a three-prong plug, it must remain grounded for safety.

Another, and often the best, way to reduce shield current flow to virtuallyzero is to use a ground isolator in the signal line. Ground isolators aregenerally available as either passive coupling transformers, which breakboth shield and signal connections while coupling the signal magnetically,or active differential amplifier devices that electronically subtractground noise from the signal. Signal-line ground isolators, like balancedinterfaces, can often reduce ground noise by 40 dB to 100 dB and virtuallyeliminate the effects of RF interference.

It is intuitively appealing to think that some sort of operation on thepower line itself or special ground can solve the noise problem. Outrageousclaims are made for the abilities of power isolation transformers,symmetrical power (or balanced power, implying the properties of balancedsignal interfaces) or isolated grounding (or technical grounding) to reducesystem noise coupled from the power line. To quote from reference 6: "Someof the more creative grounding arrangements are devised in the name ofnoise reduction, but they often ignore the basic principles of electricity,such as electricity follows paths of least impedance; electricity flows incomplete paths, and electricity flows because there is a potentialdifference. Further, when trying to reduce the effects of noise, thefundamentals of noise coupling are sometimes ignored."

In theory, a power isolation transformer can completely stop coupling ofcommon-mode power line noise into the safety ground wiring. In reality,even a Faraday (electrostatic) shielded isolation transformer will havesubstantial capacitance from its primary winding to the shield, which isconnected to the safety ground wiring. Remember, the equipment's own powertransformer capacitances would otherwise couple noise currents from thepower line (through the chassis) to the safety ground wiring. Further, asshown in Figure 5, the safety ground must be carried through to theequipment for it to remain safe and legal. Thus, for an isolationtransformer used at an outlet, low-frequency (hum) noise coupling will belowered only if the isolation transformer's total primary capacitances areless than those in the equipment. In most cases the hum reduction is lessthan 20 dB (1:10).

At higher frequencies, the reduction is much less because the impedance ofthe safety ground wiring becomes so high that nearly all line noise iscoupled to it through either the isolation transformer or the equipment'sown power transformer. Specifications for line isolation transformers andRFI filters tend to be misleading because they assume a zero impedanceground connection to the Faraday shield. Power-line isolation transformersand RF filters are commonly tested on a large metal ground plane, whichserves as a zero impedance connection for the Faraday shield and the testequipment. Claims like "60 dB of noise attenuation at 1 MHz" may be validunder these lab conditions, but not in a real-world application involvingpower wiring and conduit. For these transformers or filters to be effectiveat high frequencies, they must either be installed at the power entry panelwhere a low-impedance (very short) connection is possible to earth groundand power common, or all system devices must be powered from the isolationtransformer or filter with no interface cables connecting to any devicepowered elsewhere.

Symmetrical or balanced power is a variation on the isolation transformerin which the transformer secondary is tapped, producing equal butoppositely phased 60 VAC at the line and neutral pin of the outlet. (SeeFigure 6.) The basic idea is that the equal and opposite voltage swingswill cancel the net power line current coupled to safety ground through theequipment's chassis. Unfortunately, real-world equipment does not haveneatly matched capacitances from line to chassis, which would otherwisemake this scheme very effective. In typical real equipment, the capacitancefrom one side of the line to chassis may be three or four times that of theother side. This translates to a 10 dB to 12 dB hum reduction. In fact,some recent ad copy touts a 6 dB reduction in system noise floor afterinstalling such equipment with a list price over $7,000. However, this costmay be justifiable, for example, in a video duplication facility where avery large number of interfaces would otherwise require signal isolationdevices to increase signal-to-noise from an unacceptable 35 dB to anacceptable 45 dB.

Because wires of any size, including the safety ground wiring, have such ahigh impedance at radio frequencies, it is generally impractical to shortout RFI with them. This also limits the ability of filters on the powerline to stop RF circulating currents in the ground system, which can coupleinto signal or data paths. Ultimately, the burden of RF rejection falls oneach piece of equipment. In fact, the European EMC directive now requiresthat any equipment sold in Europe be tested and classified for itssusceptibility to RF interference. Whether the RF energy enters theinterconnect cables via antenna effects or results from coupling from thepower line, the equipment's design will determine its immunity. Sadly, theperformance of much commercial equipment degrades when such interference iscoupled to its input. Symptoms can range from actual detection of radio, CBor television signals (heard as music, voices or buzz in the case oftelevision signals) to much more subtle distortions, often described as aveiled or grainy quality in reproduced audio. Video, RF and data systemscan exhibit a wide range of symptoms.

In summary, do not install power-conditioning devices everywhere and hopefor the best. At best, noise improvements may be disappointing, and atworst, system equipment may be at unnecessary risk of damage. Methodicalsystem troubleshooting often reveals only a few problem interfaces that canthen be treated relatively inexpensively with signal path ground isolators.Science and reason can triumph over trial and error.

1. D. Dorr, T. Gruzs, and J. Stanislawski, Interpreting Recent PowerQuality Surveys to Define the Electrical Environment, Power QualityAssurance Magazine OnLine,

2. F. Martzloff, A. Manscor, and D. Nastasi, Reality Checks for SurgeStandards, Power Quality Assurance Magazine On-Line,

3. B. Whitlock, Hum and Buzz in Unbalanced Interconnect Systems, JensenApplication Note AN004.

4. B. Whitlock, Balanced Lines in Audio-Fact, Fiction, and Transformers,Journal of the Audio Engineering Society, June, 1995.

5. B. Whitlock, A New Balanced Audio Input Circuit for Maximum Common-ModeRejection in Real-World Environments, Audio Engineering Society ConventionPreprint #4372, November, 1996.

6. T. Gruzs, The How's and Why's of Isolated Grounding, Power QualityAssurance Magazine On-Line,