Ask the Experts

Power requirements for computer equipment have changed over the years. The majority of computer equipment has no need for isolated ground (IG) receptacles and the additional installation and wiring headaches associated with them.

Welcome to EC&M's monthly forum where power quality experts Mark McGranaghan, vice president of Electrotek Concepts, and Mike Lowenstein, president of Harmonics Limited, address your PQ problems and concerns.


I design low-voltage data cabling systems for computer networks, primarily for K-12 schools. I'm not an engineer, but I recommend that the design team incorporate TVSS devices, dedicated computer circuit panels (no motors), k-rated transformers, and 200% neutrals. I advise them not to design with isolated ground (IG) power. Most of the engineers I work with have been doing things the same way for years, and they continue to design with IG wiring. I think they're wrong to do so, as this was the design parameter of 25 years ago, when mainframes were the only systems around and they were typically located within 40 ft of the main power source. I will only make an exception if a traditional PBX system equipment manufacturer requires the use for warranty purposes, but this is rarely an issue. What do you think?

McGranaghan's answer: Power requirements for computer equipment have changed over the years. I agree that the majority of computer equipment has no need for isolated ground (IG) receptacles and the additional installation and wiring headaches associated with them. The advantage of IG receptacles is that they can provide some isolation from common mode transients caused by other loads on the system. Computer power supplies generally have good transient protection and immunity from these transients anyway, so they're unlikely to see any benefit even if the IG systems are wired correctly. As you mention, the possibility of miswiring reduces the potential benefits even further.

My advice for supplying computer power supply loads is similar to yours:

TVSS devices. Install them close to the supply transformer so they don't introduce additional transients into the ground system. One TVSS at the supply point will generally suffice for a facility. A computer actually serves as a good transient suppressor because of the large capacitor in its power supply.

Dedicated computer circuit panels (no motors). I like this idea becausemotors can create voltage sags during starting, which can affect some electronic loads, depending on the design of the power supply. Feeding these loads from a separate panel helps avoid this problem.

K-rated transformers. It's a good practice to use this type of transformer for transformers that feed power supply loads. A k-factor of 7 is sufficient for virtually any system with computer loads.

200% neutrals. This is definitely a good idea for these systems. It's a little conservative, but it will most certainly keep you out of trouble.


How much should you derate the efficiencies of a k-rated transformer or a standard 480-120/208V, delta-wye transformer that feeds single-phase, nonlinear loads? IEEE says the efficiency drops significantly. All k-rated transformers are rated for efficiency with a linear load, yet they're used in nonlinear load applications. How efficient are they in the real world? Also, how important is it to meet the voltage distortion limits in IEEE 519, which recommends the Emerald Book limits of 5% maximum voltage distortion for general office environments and 3% maximum for hospitals and airportapplications? If you exceed these limits, what happens to the power quality? Are there any papers to support your answer? How important is the choice of steel in the manufacture of a transformer? Are some materials more efficient at no load, such as in the core of a transformer?

McGranaghan's answer: Different components of the transformer losses respond differently to the presence of harmonics. One component of transformer loss, eddy current loss, is particularly sensitive to the frequency of the current. These losses go up with the square of the frequency so that eddy current losses associated with fifth harmonic currents are 25 times as high as those associated with fundamental frequency currents. ANSI/IEEE Standard C57.110 also provides some guidance for this calculation.

This effect is the same for all transformers, k-rated or otherwise. But eddy current losses are often much smaller for k-rated transformers, meaning that this component will be a smaller portion of the total losses.

Your question on voltage distortion limits is a separate matter altogether. It's currently causing extensive discussion among the members of the task force working on the next revision to IEEE 519-1992. The current standard recommends a voltage distortion limit of 5% for all voltages 69kV and lower. IEC standards specify compatibility levels of 8% for voltage distortion on low-voltage systems 600V and lower. The equipment used in Europe and elsewhere in the world isn't much different from that used in North America, so the loads must be able to handle harmonic distortion levels up to 8% in many cases.

We're considering two sets of voltage distortion limits for the next revision to the IEEE standard and are likely to have limits for low-voltage systems on the order of 8%. Most loads can handle this without problems. Computer power supplies are unaffected by harmonic distortion and can in fact operate fine in the presence of very high levels of distortion.

Lowenstein's answer: IEEE Standard C57.110-1998, “IEEE Recommended Practice for Establishing Transformer Capability When Supplying Non-sinusoidal Load Currents,” addresses transformer capability and suggests derating factors when nonlinear loads are served. The document provides methods for conservatively evaluating the feasibility of applying non-sinusoidal currents to existing transformers and clarifies the requirements for specifying new transformers to supply non-sinusoidal loads.

The standard gives the basis for the k-rating system and provides equations for calculating losses due to eddy currents, winding I2R heat dissipation, and other stray losses caused by harmonic currents. It also discusses the effects different materials like special steels have on transformer losses.

However, the document lacks information on overall transformer efficiencies. I suspect that it would be possible to make the loss calculations for a transformer carrying only 60-Hz current and then repeat the loss calculations for a given harmonic current spectrum. You could then use the difference in losses to estimate the efficiency change.

A published paper by Key and Lai, “Costs and Benefits of Harmonic Current Reduction For Switch-mode Power Supplies in a Commercial Office Building,” Trans. Ind. App. 32, No. 5, Sept./Oct., 1996, estimates that the energy losses in a distribution transformer serving nonlinear loads are more than double the losses for the same transformer serving 60-Hz loads. We have carried out a number of measurements on both a standard transformer and a k-13 transformer serving nonlinear loads, and found that the transformer losses decrease by about 2% to 3% when harmonic currents are removed from the system. So there is some evidence to suggest that transformer efficiency could be reduced by as much as 3% to 4% below the 60-Hz efficiency when nonlinear loads are served.

Office loads, particularly computers, aren't as sensitive as they once were to harmonic voltage distortion, and published articles question some manufacturers' insistence on harmonic limits that are far below anything that would affect their equipment. In many cases the very operation of multiple computer loads causes voltage distortion of more than 5% on the distribution system — yet all the loads continue to operate.

IEEE Standard 1100 is currently under revision, and the new edition should contain language addressing the fact that in some cases a high-voltage distortion level doesn't cause problems.

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.