Recognizing Power Problems

In this section, we will define power problems for what they are and will explore their causes. At the outset, it's important that we recognize the truth about utility power. it's a product that has been around in the same basic form for years. The technology of generating, transmitting, and distributing power has remained frozen in time throughout the computer revolution. Is that bad? No. The alternative would be to have a utility bill each month that rivaled the cost of a PC.

Public utilities bring an amazingly clean product to our home and offices. it's economical and reliable. The plain and simple fact is that modern electronic equipment simply will not run on it. We will find out why in this and later sections. The point here is to acknowledge the fact that we will never solve computer power problems by attacking the electrical utilities.

In fact, many of the more enlightened companies in the field have launched programs to assist customers in solving their problems. These efforts range from mere public relations to actually selling computer power products.

The greatest exposure the power-distribution system has comes from birds, weather, errant motorists, and other factors that are mostly beyond the control of the utility companies. If the companies built redundancy into their system or adopted more defensive installation procedures, the net effect would be dramatic rate increases. Even if the utilities were willing to do it, computer users would still find themselves facing an alarming array of factions that are opposed to higher energy costs.

So for the foreseeable future, we will have to put up with some frequent (and some infrequent) perturbations on the power line. What are these strange occurrences? Where do they come from?

Definitions

Definitions are important. it's sometimes hard to understand this left brain penchant we all have of naming things—defining them in minute detail. Take olives, e.g.,. To one person, Greek olives and Italian olives might be similar in many ways. They are both dark colored and salty tasting. But ignoring the fine distinctions might get the man at the Italian deli a little hot under the collar. The differences are like night and day to him.

Power problems have definitions but not every source defines every event in exactly the same terms. One of the biggest differences is in the matter of time. Some events are distinguished from another only by the fleeting nature of the event. Not all experts agree on these event durations. The definitions that we give here are an amalgamation from a number of sources. They are arranged in order of duration with the swiftest events listed first.

Impulses

An impulse is a disturbance of the voltage waveform for a duration of less than roughly 1 millisecond. An impulse may be additive or subtractive in nature (ill. 3-1) and it may have a ringing or oscillatory characteristic. Some slang words that might be associated with this precise term are: spike, notch, transient, whisker, and glitch.

ill. 3.1. Impulses on a sine wave.

A spike is an imprecise term for an additive or positive-going impulse, while a notch is generally a subtractive impulse. A transient is an often misused word that merely describes the non-repetitive nature of an impulse.

The slang term, whisker, refers to the thin line that an impulse of short duration traces on the screen of an oscilloscope. Glitch is a word used to describe almost anything.

Noise

Noise is a term used to describe repetitive, high-frequency impulses. We will see noise riding on a sine wave in a later section. Suffice it to say that noise occurs as impulses, and , in that way, is really a slightly different category or class of impulses. Noise, even though it's not a precise term, is an important concept which we will deal with in detail later.

Dropouts

A dropout is a total loss of power for a short period, from about 1 millisecond to 1 second (ill. 3-1). Utility switching is the normal cause of dropouts.

Surges

This is as good a time as any to run our razor-sharp rapier through a commonly held misconception. A surge is not an “impulse” or a “spike.” A surge is a transient condition where a series of peak sine-wave voltages exceed the set standard voltages for a period of time—from about one half cycle to possibly several hundred cycles (ill. 3-2). This might also be defined as from about 8 milliseconds to 2.5 seconds.

ill. 3-2. Sags and surges are short-term increases or decreases in voltage.

The controversial point to be made here is that a surge is not an impulse. and the computer product we normally hear mentioned, the Surge Suppressor or Surge Protector, may not suppress surges at all. These devices are designed to suppress impulses of a high magnitude. Perhaps this may seem like a fine point, but the truth is that there are electronic products that we will talk about which cure impulses and there are other products that cure surges or slow-moving over-voltage conditions. When it comes to shelling out hard-earned dollars, we don’t want to buy the wrong product for the wrong reasons.

Sags

A sag is identical to a surge except that the peak sine-wave voltage has decreased from the set standards for a period of time, rather than increasing. A couple of slang terms which you might hear associated with sags are “flickers” and “dips.”

A flicker refers to intensity variations in lights. Of course, incandescent electric lighting flickers 60 times a second, but our eyes can’t see it. In order for the eye to perceive a flicker, an event of several hundred milliseconds has to occur. Of course, as we will see later, this is an eternity for a computer. Countless times I’ve heard a DP manager say, “Oh, the power didn’t go out. The lights just “flickered’.” This may seem sort of ignorant in the context of a guide about power problems, but I have seen tens of thousands of dollars wasted, based on this kind of silly misinterpretation of the facts.

The term dip, which is slang for sag, usually relates to an event inside the building—an event which is not utility related. Dips are caused by heavy step load changes that demand large starting currents. More on this shortly.

Overvoltage and Undervoltage

The terms overvoltage and undervoltage refer to conditions like sags and surges that last longer than a few seconds.

Brownouts

A long-term undervoltage condition that is actually planned by the utility, or is the result of excessive loading, is called a brownout. There are many locations throughout the United States that experience chronic brownout conditions. These may be caused by seasonal use of air conditioning or by heavy loading during certain times of the day.

Blackouts

A blackout is a widespread loss of utility power, either accidental or planned.

Outages

An outage is a localized loss of power that usually lasts from 1 to 5 minutes.

Computer Tolerances

The simple fact is that power is very dirty. In an earlier photo, we saw what appeared to be a relatively clean sine wave. Using a special filter, we can look at the noise that might typically appear riding on that sine wave (ill. 3-3). Even if the magnitude of this noise is not particularly large, we should be shocked to see this much dirt on the line. What this means is that at any given time, a computer is seeing some unwanted signal as part of the power that appears on the power line. If this is true, there are evidently some power disturbances that won’t affect a computer. After all, we have just shown that there are always some disturbances on the line; just because they are there doesn’t mean we have a problem.

ill. 3-3. High-frequency noise.

On the other hand, there are obvious limits to this optimistic view. There are two major considerations in determining when a power problem becomes a real problem. Those two considerations are duration and magnitude. The longer an unwanted excursion lasts, the greater the chance that it will affect computer operations. and , the larger the event, in terms of peak energy, the greater the chance that it will affect computer operations.

A computer’s susceptibility also depends on what kind of power it was designed to expect. Most personal computers have power supplies designed to demand large amounts of current, but they are not too fussy about the level of voltage that appears at their terminals. Some older devices have altogether different power supplies that must see that voltage held to within a certain rigid range or problems will develop. We will cover this later in our section on power supplies.

ill. 3-4 makes this point graphically. We see an area (designated by the shading) where events can occur without causing problems for the computer. The area under the curve describes events that are so short in duration that their effect is not felt. As we trace the curve to the right along the “x” axis, we see that magnitude becomes a greater factor. The area under the curve (the cross-hatched area) is now bounded much more tightly than it was for shorter- duration events. What this means is that excursions above or below these longer-duration limits can't be tolerated.

There are two points to be made about the graph in ill. 3-4. This is an approximation of what one might expect to be design criteria for PC power tolerances. We can only approximate since there are so many manufacturers in the market that no firm standards are being followed. For larger systems, the criteria has been set by a group called the Computer Business Electronics Manufacturers Association. This group, CBEMA, has agreed on input power standards for larger systems. PCs are somewhat different from larger systems in that they are less critical about the power they receive.

ill. 3-4. A graph showing the estimated susceptibility profile for voltage variations and disturbances in PCs.

The second point to be made about the graph is that the power supplies in today’s PCs will pass certain kinds of noise directly into the computer. This is a change from earlier computer power-supply designs which were substantially immune to noise. But we will cover this in more detail later.

Survival Time

So what is the survival time of a PC? How long can it go without normal AC power before it crashes? For most PCs, the answer is about 20 to 40 milliseconds, or about 1 1/2 to 2 1/2 cycles. (Some lightly loaded PCs will ride through 3 to 4 cycles without power.) This is a wide variance when you consider that a mainframe will die after about 10 milliseconds. The reason for this is that in the mad rush to build the cheapest PC, many manufacturers succeeded in building a cheap power supply. Cheaper designs use smaller components which have less electrical capacity and , therefore, provide less ride-through capability. Better-built power supplies have a greater tolerance for power interruptions due to the energy-storing capability of the components.

This tolerance is also an indication of the power supply’s ability to provide for the instantaneous current demand of the computer during low line voltage conditions. In plainer language, we mean that the greater the energy-storage capacity of the power supply, mainly in the large filter capacitors, the greater will be its ability to provide large quantities of current, even when the line voltage may have dropped. We will explore this subject in detail later.

Power Quality

In 1984, Con Edison published a pamphlet entitled Electric Power and the Computer. The utility warned customers that transient voltage fluctuations are a recurring fact of life. Before we explore why this happens, let’s see how often it happens. Con Edison is very forthright in predicting how often a sag might occur that would affect computer operations. On a 120-volt system, they estimate sags:

• To 90 volts (25%) for 6 cycles (100 ms) once or twice a month.
• To 90 volts (2 5%) for 30 cycles (500 ms) once or twice a year.
• To 55 volts (54%) for 12 cycles (200 ms) once or twice a year.
• To 0 volts (100%) for 12 cycles (200 ms) very rarely.

That adds up to some 14 to 28 events every year that may cause an interruption of data processing.

A number of years ago, the Department of the Navy did a survey of power problems at six sites. Table 3-1 gives a summary of what was found.

Table 3.1. Power-Related Computer Failures

The Bell System Survey

From May 1977 through September 1979, the Bell System conducted one of the landmark surveys of power disturbances. Two Bell employees, Goldstein and Speranza, produced a study which was probably the most scientific and careful evaluation of the problem of power quality done to that time. Unfortunately, for those of us who are not mathematicians, they choose a statistical method called a Polya probability density function to describe their findings.

They, of course, had very good reasons for using this mathematical approach since Polya functions were used to create a set of tables that predicted, statistically, the number and duration of events that might be expected in the future. Had they merely been trying to show how bad the power was, we might find their results a little more intelligible.

Data was gathered from 24 Bell System data-processing sites for a total number of 270 monitoring months. At the end of that time, an analysis of the total number of disturbances measured showed that 87% were sags while only 0.7% were surges. Outages accounted for 4.7% of the total, and impulses represented 7.4%.

One thing that stood out as they assembled the data was that lightning- prone areas had no greater rate of impulses than did other areas. In fact, the power disturbance most often recorded during electrical storms was a sag. This is logical, however, since the power company places devices on the distribution line that momentarily divert lightning energy to ground by shorting the power line.

The study predicted that 50% of Bell sites could expect 25 sags of significance and 4 power failures annually.

The IBM Surveys

Perhaps the power-quality survey most-often referred to was the one under taken by George Allen and Donald Segall of IBM. This survey was done in 1972, but IBM has done two newer studies since then. Several sites were monitored, for a total of 3312 monitoring days. The results are summarized in Table 3-2.

Table 3-2. Summary of the IBM Survey

What do the results in Table 3-2 tell us? If we divide the total number of disturbances by the number of monitoring days, we get an alarming average figure. We can expect 2 events of one type or another every day!

IBM has since done two additional surveys; a small one in 1974 - 1975 and a comprehensive one in the 1980 - 1983 time frame. The more recent survey included 6827 days of monitoring and it covered 34 states, as well as Europe and Japan. This study showed that over the intervening years, power quality had deteriorated significantly.

The data collected indicated a wide disparity between different parts of the country. Even so, conditions on the whole were not favorable to reliable computer operations. One conclusion that has been reached from the survey is that an outage, or a low-voltage condition severe enough to interrupt computer operations, might be expected once a week. We could belabor the point by presenting a more detailed examination of power quality, but the facts given here speak for themselves. Utility power as it arrives at the building service entrance is often not compatible with on-going computer operations. While these studies may be interesting and compelling, most of us already knew that, for some time, a basic problem existed in one form or another.

Let me make one more point that might make the evidence a bit more dramatic where PCs are concerned. All of these studies were done at major data centers. These facilities are carefully engineered to give a large computer a very friendly electrical environment. If the power fed to mainframes is bad, imagine what the poor lowly personal computer has to put up with. We merely plug it into any convenient outlet, and it must take whatever comes down the line.

What causes these problems? Obviously, there are events that affect power once it leaves the generator. We will look at this issue from two perspectives— outside the building and inside the building.

Utility Grounding and Switching

Much of what was discovered during the surveys of power quality cited here had its genesis out on the power line. Strange things can happen to electricity between the generator and the user. Unintentional grounding, called ground faults, are perhaps the most common cause of sags in voltage. This unintentional ground connection can be caused by equipment failure, arcing, or animals that suddenly complete a path to ground. In desert communities, transformers typically display ground faults during rainy periods. The natural accumulation of dust inside the unit mixes with the moisture in the air to provide an unwanted path for current flow to ground. High winds can cause power lines to slap together. Arcing switches or arcing circuit breakers can flash over to ground.

Coastal areas are also subject to their own special kind of problems, due to the ravages of wind, salt, and spray. Then, there are the ever-present dangers of cars, kites, and ice. There are a multitude of causes that can cause ground faults. Note that ground faults don't always have to be direct shorts to affect the voltage relationships in a distribution system. Current leakage can occur through faulty components and this will affect the voltage. In the next section, we will devote some time to one of the most spectacular causes of ground faults—lightning.

All of these things that divert energy to ground will cause wide swings in power as it travels along the power lines. But a more subtle form of fluctuation is that caused by standing waves, which emanate up and down the line from the point of a fault. Standing waves are reflected back to their source when they encounter the end of the line. This reflection, back and forth, sets up an oscillation. The frequency of this ‘noise” is determined by the length of the line from the fault to the reflection point.

Utility switching causes this same effect. Of course, utility switching also causes a momentary interruption of power. There are many reasons why the utility might switch power lines. They might need to compensate for heavy loads on other parts of the distribution system, or switching might be necessary to bypass points of failure.

Another common reason for switching is power factor correction. Remember, we talked earlier about power factor as it related to power that was circulated but not used. Utility companies are not too excited about this since they charge their customers according to the wattage, or real power, used. Therefore, they will usually place a surcharge on the bills of those customers who have a bad power factor in their system. Often, utility companies will switch capacitors into the line to compensate for a lagging power factor. This switching causes several rather severe problems, not the least of which is a dropout of sufficient enough duration to bring down any sensitive equipment that is connected to the line.

All these disturbances, whether caused by switching or ground faults, can be coupled by the inductance and capacitance that exists between the power lines. We will look into these principles a little more later on.

Changing Loads (Outside the Building)

Some common power problems are heavy loads being switched on and off along the distribution system. Large motors, like those associated with building air conditioners, can cause damaging impulses when they are switched on and off. In fact, there are many reported cases where entire industrial neighbor hoods have been affected by the switching on of AC units atop one building.

Arc furnaces are another traditional culprit. They draw huge amounts of current while reflecting distortions back into the line. The list goes on, but suffice it to say that any sudden shifting of the load on the distribution line will affect the quality of power delivered to neighboring facilities—a changing load will create sags, impulses, and noise.

Good and Bad News (Inside the Building)

The good news is that most personal computers are plugged into those circuits that are the farthest from the building electrical service. What this means is that impulses generated outside the building have a lot of building wiring to travel through in order to reach the PC. The resistance and the inductance of the wires have a cushioning effect that reduces the shape and magnitude of incoming impulses. Also, there will usually be one or two additional trans formers placed in the line, especially in commercial buildings. The inductance of the windings in these transformers has a dampening effect on noise and impulses.

Every silver lining has a cloud, and here we find no exception. This same long path that can protect PCs can also cause a voltage drop. The voltage at the wall outlet might be significantly lower than the voltage at the building service. This may very well make the PC more vulnerable to sags. PCs use switching power supplies. They have a great tolerance for voltage swings, but, as a general rule, equipment will draw more current at lower operating voltages. This causes heating that might damage equipment. You can compensate for voltage drops by using transformers (with taps) to adjust for chronic low-voltage conditions. We will discuss the merits of transformers in greater detail in a later section.

Inrush Currents

As loads are turned on, they present an initial load to the power source that changes dramatically over a few seconds. Nearly all loads have power supplies with large filter capacitors or windings. These supplies drive motors or other energy-absorbing components that have an early heavy “demand” in order to reach some proper operating level.

Take a motor, for instance. A motor must produce an initial starting torque. The current demand of this initial inrush of power can be many times the steady-state running load of the device. This demand can last for as long as 10 seconds. An induction motor can easily draw five to ten times its running current for 3 seconds or more during start-up. To illustrate this, look at ill. 3-5. Here we see a portion of the current sine wave of a common electric hand drill as it's starting up. Each additional sine-wave trace shows a decreasing demand for power. Finally, after a few seconds, the current demand flattens out to the steady-state running voltage at a level much lower. The inrush current of this device is so large that the full sine-wave trace was too big to get all of it on the oscilloscope screen and thus show a clear detail.

ill. 3.5. The inrush current of an ordinary electric hand drill.

Switching power supplies, on the other hand, have a slightly different inrush characteristic. Most of the energy used by the power supply is held in a large capacitor. At turn-on, this large capacitor charges fully during the first half cycle. The inrush current during this brief period is large and can affect nearby equipment. Transformers have inrush needs that can be from five to fifteen times normal.

Inside a building, just across the hall or in the next office from a PC, there may be a device that cycles on and off. The inrush demands of the device can be such that it interferes with the current demand of the PC on your desk.

Step Load Changes

Inrush current could easily be classified as a step load change. But there are other events that produce step load changes that are not related to inrush currents. Elevators, e.g., are already powered up and ready to operate when we step into them. Yet the amount of current demanded to lift a car full of people several stories upward can affect the power available all over the building.

Step load changes are heavy loads that cycle on and off. Elevators, copiers, and even coffee makers and microwave ovens are examples. This constant shifting of loads off and on causes voltage sags, noise, and impulses of a large magnitude. Several studies have been done and results show that such things as an oil burner—located in a residence—can produce a starting transient of up to 2500 volts. Commercial facilities can expect to experience impulses, caused by devices going on or off the line, of a magnitude of anywhere from 300 to 2700 volts.

Off the line? Yes, turning off a device can sometimes create a larger disturbance than that which is created when it's turned on. Any device with windings in it can produce extreme voltage transients when they are turned off. A transformer, e.g., can produce an impulse in excess of ten times normal voltage when the power is switched off, especially when the load is a high-impedance load like power semiconductors. it's the collapsing field that induces this voltage into the secondary of the transformer.

Voltage impulses can be caused by switch arcing, fuses blowing, and breakers clearing. These aren’t strictly what we would call step load changes, but they are an important source of undesirable disturbances within a building.

A good way to visualize step load changes is to think of a manufacturing plant. In back, there are huge machines stamping, pressing, and grunting— cycling one piece at a time. Meanwhile, in the front office, there is a local area network for the company’s accounting department. The heavy cycling machinery may be placing such a high demand on building power that the computers may not see anything like a sine-wave voltage at all—just some distorted rise and fall of voltage waveforms, peppered with harmonic distortion, large impulses, and noise.

Power (Inside and Out)

There are plenty of gremlins to spoil our clean supply of pure repeating power sine waves. Outside the building, there are any number of events that can create perturbations, while inside the building we hold our breath every time a new piece of machinery is plugged into an outlet.

In succeeding sections, we will discuss much of what we have commented on in more detail, as it relates to specific product technologies aimed at providing a cure for power problems. But before we do that, a little more background is required. What about power supplies—the things that convert AC into DC so the computer can use electricity which is at logic levels? and , of course, there is noise. We have referred to it a time or two. We must look more closely at noise and its specific impact on PCs. We will also talk in more detail about lightning, and will have a section on lightning arresters, as if lightning could actually be arrested. We will discuss grounding. There are three wires attached to that socket in the wall outlet. One of them is coded green and it’s there for your health and much, much more.

NEXT: Voltage Transients

All related articles   Top of Page   Home