Tom Székely, P.E., LEED AP

Home | Background and Experience | Services | Affiliations | Projects | Newsletter Archives | Newsletter Sources | Contact Us

Last Issue

October 22, 1997

Next Issue

The Great Third Harmonic Scare, or, the Birth of an Industry

Back around 1970, the then new edition of the triennially revised National Electric Code, (ANSI/NFPA 70) specified that where general purpose electric outlets were pro­vided, they should be figured at an ascribed load of 180 volt-amperes each. The pre­eminent trade magazine of the Electrical Contracting Industry, Electrical Construction and Maintenance, in its customary in-depth review of the Code changes, took it upon themselves to declare this meant each plug attachment point of a duplex receptacle be so figured, with the result that, after accounting for Code derating factors, no more than five such receptacles could be served from a standard 20 ampere branch circuit.


The outcry from the professional design community was loud and immediate, and the National Fire Protection Association issued a bulletin (incorporated in the body of the fol­lowing issue of the Code) clarifying "outlet" to mean a device on a single mounting strap. The dollar signs dancing before Contractors' eyes before the NFPA clarification took us back to the still used ten outlets per circuit, must have been disappointing as they faded.


When PC's first hit the corporate streets, a similar effect took hold as installers and PC vendors insisted on dedicated circuits for their equipment, an affectation which still afflicts certain electronic cash register manufacturers. With experience, the fashion seemed on the verge of fading, but it has had new life breathed into it by the shadowy demon of third harmonics.


Conventional (linear) power supplies, which have served us well from the days of the first electronic devices, regulated their output in response to a fluctuating load by dissipating heat through their series pass control ele­ments, much like lights used to be dimmed by inserting more of a rheostat's windings into a circuit, leaving less voltage available to cause a lamp filament to incandesce. In addition to being inefficient, however, (50% efficiency for a linear power supply is as good as it gets), more capacity begets sub­stantially more weight, as a transformer is the primary limiting factor in capacity.


Non-linear supplies are so named because they draw power from the line in discon­tinuous chunks by varying the on-time of their control elements much like a modem solid-state dimmer provides power to lights; the longer the duration of the on-time, the more area subtended by the chunk, and the brighter the lamp. Fourier analysis of such a waveform shows it to be rich in odd harmonics, particularly triplens (multiples of the third) harmonics, and the problem with multiples of third harmonics is that they pile up in the neutral conductor of a three phase system.


In fact, while balanced three-phase systems (virtually all commercial power) have their return currents cancel out in the neutral, where the branch circuits supply exclusively PC outlets, triplens harmonic pileup can cause the neutral to (in theory) carry twice as much as any phase.


The electrical construction industry answered the call by offering multiconductor cable with oversized neutral conductors, and "K-Rated" transformers to withstand the excess heat generated by all these harmonics. Pretty spooky. Now, if we put all this stuff on its own distribution subsystem, we've virtually doubled the copper and iron in the subsystem, and made the thing as close to damnfoolproof as is possible. Of course, the installed cost has gone up in direct propor­tion, but that's the price one pays for safety in the computer age.


Wait a minute. What's going on here? In the old days computers had to be on dedicated circuits because all the hash on general pur­pose circuits would find its way into the load side of the power supplies and scramble their poor little brains, right? Not so. The heavy iron in the early power supplies acted as a buffer, while the control elements of present-day switching power supplies act so fast that their (smaller) load side buffers don't notice any change.


The possibility of overloading the neutral to the point of starting afire by serving multiple computer outlets on multiwire branch cir­cuits may be a very real one, but not one that necessarily calls for superneutrals, or for dedicated computer circuits with one com­puter per circuit. To prove this, let's look at the real world as portrayed by those with something to sell.


The John Fluke Manufacturing Company, Inc. makes a line of very fine handheld digital electric multimeters, and offers a free 17 minute video entitled "Understanding Harmonics", and a brochure reprising that video entitled "In Tune With Power Har­monics". In these promotionals, they discuss an actual case where they discovered a common neutral carrying 15 amperes, serv­ing computer branch circuits carrying 7.8, 9.7, and 13.5 amperes. The fact that the loads were below the continuous 16 ampere rating of the nominally 20 ampere rated wire precluded any immediate danger, but one could suppose someone not knowing of the danger, and not having a meter capable of measuring harmonics in the neutral might add more computer outlets to the 7.8 or 9.7 ampere circuits and overload the neutral while thinking everything was fine. Fluke recommends pulling in extra neutrals, refraining from adding additional loads unless steps are take to reduce harmonics, and, of course, monitoring of the load cur­rents on a regular basis using true-rms measuring test equipment (which just hap­pens to be manufactured by Fluke).


Harmonics Limited makes passive filters to "fill-in" the harmonics needed by the power supply, effectively isolating the computer from the distribution system. In their pro­motional flyer on their 3rd-Out harmonic filter, they show the current drawn by sample PC's on dedicated circuits as 1.94, 1.25, and 1.97 amperes, and a common neu­tral current of 3.06 amperes. Though what they're selling is the reduction in phase cur­rents to 1.26, 0.85, and 1.26 amperes, and a neutral current reduction to 0.81 amperes, after installation of their filters, the pre-filter numbers are the hard data needed to make reasonable design decisions.


Some propositions. First, the preoccupation of both Engineers and Contractors with spare capacity, or what the owner will do with a system after it's in operation, is usually mis­placed at best. Circuits feeding rows of fluorescent fixtures in an office space are not candidates (especially in light of State and National energy codes) for additional load. Second, the belief that general purpose receptacle circuits will actually see anything close to the 180 VA ascribed to each outlet is so farfetched that utility companies routinely provide electric services one quarter to one third that which the Codes tell us to design for, after we have applied Code demand fac­tors to our calculations (although, when one considers the Code demand factor to be applied to office spaces is 100%, utility company sizing doesn't seem so surprising). Third, the worry that circuits reserved for special equipment such as Xerox machines, microwaves, or computer outlets will have additional loads connected to them is so removed from reality as to call into question whether the one having such a worry lives on the same planet as the rest of us. Lastly, the whole issue of conductor overloading only becomes operative with continuous loads; power dissipated is I2t (as in time).


The National Electric Code recognizes this in defining a continuous load as one which remains on for three hours or more. We in New York City (so far) labor under an Elec­tric Code drawing no such distinction.


So where does all this leave us? Between the reality that 230 and 300 VA computer power supplies rarely operate at rated capac­ity, as reflected by the unfiltered load currents recorded by Harmonics Limited in the paragraph before last, and the further reality of the extrapolation of a total neutral current of 9.18 amperes for three PC's on each phase (3 times the 3.06 ampere neutral current of the paragraph before last), it seems extremely difficult to make a cogent argu­ment against putting 3 PC's on a branch circuit, and impossible to argue carefully loaded branch circuits will yield a neutral full of third harmonics which can become overloaded if a phase goes down.


Positing (to be conservative) that the 3 PC's per circuit which give us the 9.18 ampere neutral harmonic current comprise 5.19 amperes (3 x the 1.97 amperes at the top of this column) per phase, a phase going down would inject 5.19 amperes unbalance current into the neutral while simultaneously remov­ing 3.06 amperes of harmonic current therefrom, for a total neutral current of 12.03 amperes, not a problem on a 15 or 20 ampere branch circuit.



Next Issue