Too often, the first thought of standby power comes a few seconds after we have lost several thousand words of a word-processing project, or an enormous spreadsheet, due to a power failure.

Right away, we are entering the controversy of words. UPS stands for uninterruptible power system. In this section, we will talk about standby power systems.

SPS Building Blocks

It comes as no shock to most readers that backup power must come from batteries. But batteries are only capable of putting out direct current, and computers run off of AC power. Right away, it can be seen that some method of converting DC into AC must be used if we wish to make battery power useful. and , if we drain the power from a battery, we must have some way of recharging that battery. The obvious answer is a battery charger.

Hold on now. Anyone who has ever charged a battery knows that you must have DC power to charge a battery. This means that some rectification must take place to change the AC input current into DC.

To summarize the path that current must flow in an SPS, let’s look at ill. 11-1. We see two current paths controlled by a switch. The normal path for current is directly through the unit. However, some current is diverted through the charger to keep the battery fully charged. We have shown the charger converting a portion of the AC power into DC in order to accomplish this. This might leave one with the impression that some straight-line value of DC is available at this point. Actually this charger is small, and fairly inexpensive in most units. The charger’s output will have a large AC component in the form of ripple, and it will probably be pulsing instead of a steady current flow.

ill. 11-1. The functional block diagram of a standby power system.

If the power goes out, or for some products, drops below a pre-determined voltage level, the inverter is turned on and the switch activates the path from the battery through the inverter. We will spend some time looking at these components as the section progresses. By now, it's clear why this is a standby system. The inverter is basically cold until the utility power fails. The switch must be activated very fast so the load does not experience the power failure.

Transfer Time

One of the battles that rages between rival manufacturers of SPSs is who has the fastest transfer time. The more legitimate makers advertise and achieve transfer times in the 4- to 10-millisecond range. Others that advertise 4-millisecond transfer times are lucky to experience 15- to 20-millisecond times. Still, there is a class of manufacturers that claim transfer times of under 1 millisecond, or even in the nanoseconds.

Earlier in the guide, we dealt with the length of time that a PC can stand to loose power. Anything beyond that time limit and the unit goes down. This is why the successful switching on of the inverter—before the internal ride- through of the PC’S power supply is extinguished—is so important. All the batteries in the world won’t help if they don’t come up on line fast enough.

This is a particular concern in the minicomputer market. But, why should this be included in a guide about personal computers? The simple fact is that many low-end computers made by the larger commercial vendors can be powered by 1000 - 2000-watt SPSs. Many of the readers of this guide may have had occasion to apply these product lessons to machines of this variety, since computer power requirements fall with these physical sizes. Most of the major computer manufacturers have models that require only about 1500 watts, but must have an excellent electrical environment. Many of these computers, along with some of their fully loaded PC brethren, can't tolerate any transfer time. In these cases, an SPS is not the correct product.

Let’s examine first those products that claim to transfer in the nanosecond range. This would appear to be a highly desirable thing. After all, the faster a product switches, the less likely the computer will be affected. But the vast majority of PC installations don’t need switching that fast. Earlier in the guide, we discussed an envelope of time and magnitude, inside of which events that don’t affect reliable computing could take place. Super-fast switching must respond to the events occurring in this envelope. This means that the SPS would be activated often, over and over again, while most SPSs would ignore the event. This cycling on and off would, on the face of it, lead to excessive battery wear, lowered inverter life, and other annoyances, such as warning buzzers constantly going off. Believers in this super-fast switching system have failed to show any reliable data to substantiate their claims, or answer any of the healthy skepticism that surrounds their claims.

We are then left with those products that claim a 4-millisecond transfer time. This is plenty fast enough if the claims of the manufacturer are actually true. Most products will switch inside 10 milliseconds, under worst case conditions. and , 4 milliseconds is usually a best case. But how is one to know? it's not unusual to find specifications that can't be verified under even the most optimal circumstances. In recent years, there have been a number of articles written in the major computer magazines that compare SPSs. Transfer time is one of the normal benchmarks of these tests. But not every manufacturer agrees to buy advertising in that particular issue so their product may not have been included in the testing.

Possibly, the reason many call their product a UPS is apparent. Since the PC can tolerate loss of power for as long as 30 milliseconds, any standby unit that is switched on in time to keep the system up and running might be considered uninterruptible. This point is technically debatable, but it makes marketing sense not to cloud the issue with the facts.

How switching times are measured is important too. Some manufacturers include detection time, or the time it takes to know when it’s time to switch, in their transfer times. Others don't . The proper way to specify transfer time is to include detection time in the total.

Another gray area is the point along the sine wave that is used to sense that power is actually lost. Clearly this variable can make a difference of several milliseconds. What is important in all of this, then, is to choose a reputable manufacturer rather than the fastest transfer time. As we stated earlier, we don want a unit so sensitive that it's constantly reacting to minor power fluctuations. A good design is a compromise between fast reaction and sensitivity.

Making Waves

When the power goes off and batteries become the power source, an inverter must convert the DC power from the batteries into AC power that is usable by the power supply. With this in mind, let’s take a step back and consider the general market conditions. This is the PC market. it's highly competitive. Every dollar of cost put into a product risks sales volume. The consumer is not particularly well educated about the technological differences between products, and , ordinarily, won’t pay for features they can’t see or touch. it's into this type of environment that we probe into the heart of any UPS or SPS—the inverter. The inverter is the most complicated component of the device. It must take DC power and convert it into AC. If ever there was money to be saved, it's here.

ill. 11.2. These are popular alternative waveshapes to a sine wave. The inverters that produced these waveshapes are considerably cheaper than one that produces a sine wave. (A) Square wave. (B) Rectangular wave. (C) Step wave.

Not every inverter puts out the same kind of waveform. For the concerned user, this is the vitally important factor, although it may come as somewhat of a shock. Most users may have assumed that AC is AC, and the output of the inverter would naturally be a sine wave. This is true for only a few of the units on the market. Most SPSs put out square-wave waveshapes, or rectangular-wave AC waveforms. Some put out a waveshape called a step uxwe. Others put out a waveshape called a modified square uxive. ill. 11-2 shows the first three of these while the photos in Figs. 11-3 and 11-4 show the modified square wave and a sine wave, respectively.

ill. 11-3. The “modified” square wave from an SPS.

ill. 11-4. The sine wave produced by another manufacturer’s SPS.

Obviously, these waveshapes deserve some comment. But before we do that, let’s go back to something that came up in Section 6, when we discussed power supplies. We talked about the peak voltage and RMS voltage of a sine wave. For a square wave, these two values are one and the same (ill. 11-5). The job for a design engineer is to decide what the most reasonable compromise voltage will be for a square wave when given the fact that the equipment is designed to accept a sine wave.

This is especially critical when considering the needs of a switching power supply. Some experts claim that a switching power supply requires 148-V AC peak voltage during the conduction period of the switch in order to maintain a regulated output. On the other hand, some loads are sensitive to rms voltages. This results in a compromise voltage that may be around 140-V AC.

As ill. 11-5 shows, there are harmonics in a square wave and , at any point, a square wave contains more energy than a sine wave. These factors cause an additional heating of filter elements and some kinds of AC motors. This means that power supplies, printer motors, and some disk drives must digest this additional energy.

ill. 11.5. A square wave as compared to a sine wave.

The rectangular wave is an answer to some of these problems. Less energy is contained at any point under the curve in a rectangular wave than there is in a square wave. The trade-off for this is maybe just a bit more noise output. We will address noise later in this section.

Another potential problem with square-wave inverters is that as the battery is drained, the output voltage tends to fall. In an attempt to reduce costs, many manufacturers don't regulate their output. This may cause problems in meeting the peak-voltage demands of switching power supplies and other equipment that is peak-voltage sensitive.

Let’s not deceive ourselves, however. These low-cost alternatives do work. When the power fails, even the modified square wave will keep the computer running. For short outages, we might not care if our computer’s power supply is stressed and overheated. These designs allow the manufacturer to offer a popular product at lower prices, in a market where that is all that matters. There are significant performance trade-offs that must be considered.

The sine wave is obviously the best way to go. it's exactly what the PC was designed to anticipate. it's a fundamental, with no harmonic content. It will cause no performance problems. Today’s sine-wave outputs are generated with a pulse-width-modulated inverter. PWM inverters are not new to us. They are part of switching power supplies. When properly filtered, a PWM inverter puts out a sine wave like the one shown in ill. 11-4. Even here we see some slight distortion, but this is negligible when compared to the waveform shown in ill. 11-3.

Noise

An inverter is a very fast switch. and like switching power supplies and other electronic switches, it's the source of noise at the output of a UPS or an SPS. ill. 11-6 shows the noise envelope of a PWM inverter. While this may look frightening, it's on the order of 10-volts peak-to-peak. This noise is too small to be of any concern.

ill. 11-6. The noise envelope of a PWM inverter.

As we stated earlier, the noise output of cheaper inverters can be significant. To illustrate this, look at ill. 11-7. This is the noise output of the unit that produced the modified square wave in ill. 11-3. This doesn’t look as bad as the noise from the PWM we just saw. Just a few little spikes. Unfortunately, this unit has a noise envelope just like the PWM inverter does. It’s just that we can’t see it with the gain of the scope turned so far down. The little spikes that are shown in the photograph are impulses greater than 350 volts. That’s right, 350 volts, being repeated every cycle. Is that what we want to feed our machine, even for the short durations when utility power has failed?

Obviously, every inverter puts out some noise. But the cheaper units are cheaper because of the kinds of technological sacrifices that are made. Will the PC run while being bombarded with noise, harmonic distortion, and energy it can’t use? The simple fact is that it will in many cases. But wait until you hook up that modem or connect your computer to a network. It’s hard to relate that unexpected investment in a power supply to some vague problem with the battery back-up unit.

The point, here, is that these products may be purchased for legitimate reasons. Let’s analyze those reasons for a moment. We need back-up power because our data is valuable, and our time is valuable. Isn’t our investment in equipment valuable? Why not spend a few dollars more, or look a little deeper before following bad power with worse? A sine-wave output, with little noise output, should be the goal every user should try to accommodate, given their own personal financial constraints. Furthermore, it remains to be seen that worse is always less expensive. In many cases, a fine product is available at very competitive prices.

ill. 11-7. This modified square-wave inverter puts out noise impulses that are greater than 350 volts.

Synchronicity

The problems with low-end SPSs don’t just end with switching times and waveforms. There is a sticky little problem called synchronization. To illustrate this, let’s look at ill. 11-8. This photo will take some explanation. Since this is a time exposure of about one second, we see three distinct waveforms in this picture. From a time standpoint, the first waveform traced across the screen of the scope is the “modified” square wave of the same SPS output we saw in ill. 11-3. The second trace is the faint “S”-shaped trace. The third looks a little like a bright “S.”

The faint “S”-shaped trace and the bright “S”-shaped trace are the waveforms of the utility power after the SPS has switched from batteries back to the line. Here we assume that power has gone off for some period of time, and the batteries and inverter have powered the load. Then, the utility power is restored, and the unit switches from the inverter back to the utility line. The faint “5” trace is the actual first cycle that swept the screen after the switch was activated. During this cycle, the oscilloscope synchronized itself with the line and thus we have the bright “S” trace which is shifted away from the first “S” trace. So let’s ignore the bright trace and deal with only those events which occurred just before and just after the switching action.

ill. 11.8. This photo shows the switching of an SPS that does not sync to the line.

Notice that in the lower left-hand side of the picture (ill. 11-8), bordered by all the traces, are some faint markings. This is the actual switching, captured in process. it's at that moment that the retransfer takes place. We can see directly below this the faint tracing of the first cycle of utility power. We can see that the direction of this waveform is up. As it climbs, we see the previous “modified” square-wave waveform move down. Despite the crude shapes, we can see that these two waveshapes are virtually 1800 out of phase.

It is this phenomenon that the photo is supposed to illustrate. If the waveform prior to switching and that one which occurs after switching are not synchronized, the computer may be ready for more power moving in one direction, and will get it for a long period (say one and one half cycles) in entirely the wrong direction. This could have the same effect as an outage and could cause the system to crash.

Other possible consequences of non-synchronized switching are changes in the operational frequency of those devices that need a relatively stable frequency source as a reference. (The reference is for clocks and motors, as well as other sensitive timing circuits.) The photo shows that the power which was delivered to the PC was moving in exactly the opposite direction during the same cyclical slice of time.

The best engineered units on the market have circuitry that sense the phase relationship of the new utility power. It then shifts the inverter output to match the phase angle of the incoming power, prior to retransferring. This not only eliminates the problems we have discussed but it protects against an important event.

When power is restored after an outage, devices are waiting for it and greedily gobble up current as soon as it's available. This enormous demand of inrush current forms a virtual short circuit across the power source. This situation results in a transient pulse of immense proportions. Power engineers have long known that it's this impulse, which propagates downs the line, that causes most of the actual hardware damage, rather than the outage itself. Obviously, this transient pulse will not impact the PC that is connected to an SPS with synchronized switching, since it will be busy phase-matching, leaving the PC still on inverter power while this destructive event is taking place.

Brownouts

This section takes us back to our original discussion of transfer to batteries. There is a point at which every SPS must decide when to go to batteries. This is an extremely important design-and-buying consideration. Does the device wait until there is a total power outage, or does it transfer at some predetermined low-voltage state, regardless of whether the power is really going off?

A product that advertises a 4- to 10-millisecond transfer time may actually wait until the voltage has dropped to an extraordinarily low level before sensing an outage. This additional time adds up to an unacceptably long overall transfer time. What happens during a brownout when the supply voltage sags for more than a few cycles? Don’t we need some sort of backup then?

The more sophisticated units are designed with this problem in mind. They have a preset voltage level, say 103 volts, that causes the switch to battery power. This makes the unit a voltage regulator in effect. However, a word of caution is in order. There are many commercial sites where the voltage may sag for extended periods at regular intervals every day. The danger is that the SPS will transfer, and the voltage would not reach a high enough level to trigger retransfer before the batteries are drained. it's also important to remind ourselves that the switching power supply will run fine under most brownout conditions, unless they become severe. With this in mind, a compromise must be reached between switching time and switching level to satisfy all conditions. The most reasonable approach is to design the SPS with the capability of being switched to higher or lower transfer points according to the dictates of the site itself.

Batteries

Standby power systems come with an infinite variety of battery times. But before we consider why this is so, let’s understand what an SPS is really for. Since the invention of the UPS, they have been used for two basic purposes. The first is to ride through short-term outages. The second is to have time to do an orderly shutdown during long-term outages.

These are precisely the considerations that a buyer must have in mind before making a decision on one SPS or another. A unit that advertises 10 to 20 minutes of battery time will be heavy and large. But, depending on the needs of the site, it may be the best possible choice. However, longer battery times are desirable in those locations where brownouts are common. Battery time is critical if a particular processing job, say an accounting function, must be completed before shutdown can occur. In another section, we will talk about networks. If an SPS is supporting a file server, sufficient battery time must be available for every user to finish their job, close files, and log off the system.

Short battery time, on the order of five minutes or less, is available in the smaller sleeker units. Often these machines are designed to fit under a monitor and on top of the CPU. These short-battery-time units have advantages. They allow for better cosmetics in the design of the package, as the SPSs are not big, bulky boxes. They are attractive low-profile units. But their basic function is to give you enough time to save files and shut down the equipment. In most cases, this is adequate.

The vast majority of power outages are of the under-a-few-seconds-in duration variety. The ability to ride through those events is the primary utility of an SPS. But when it comes to longer outages, there is yet another battery consideration—end voltage. End voltage is that voltage which the cell of the battery reaches, where any further discharge will result in damage to the cell. Most SPSs have no provision to shut down the inverter when the end voltage is reached. As a result, it's common to find batteries that have given their life and must be replaced.

In other words, during a long outage, the SPS manufacturer must assume that the alarm on the unit has warned you of the loss of power and . that you have taken appropriate steps. As the outage is prolonged to near the rated backup length, the batteries become drained of their charge. At the point where the end voltage of the cell is reached, the manufacturer must provide a circuit that senses this and shuts off the inverter before the cells are damaged. Again, you get what you pay for. Reputable manufacturers are aware of this need and build this safety feature into their units. The cost differential is nowhere near the cost of a new set of batteries to replace those that have been damaged due to a total discharge beyond the cells’ end voltage.

Battery Times Again

Before we move on, there is another important point a buyer must keep in mind. This market, the PC market, is extremely price competitive. This forces a cutting of corners and outright deception. Batteries are no exception.

Batteries don’t fit well into the PC market anyway. They are big, bulky, heavy, and they don’t have the high-tech image of a turbo board or a laser printer. it's for this reason that manufacturers build smaller and smaller boxes. The look of the package is important in making the sale. it's for these reasons that some vendors stretch the truth about battery times.

Often a manufacturer will reason that most people will not fully load their SPSs. They reason that the typical user will only load an SPS to about 60% to 70% of the unit’s rated load. So a 1000-watt SPS will probably only be loaded to 600 or 700 watts. This is why 500-VA units are so popular. Most users only need 360 VA, but by the time they add in a margin of safety, they are shopping for a 500-VA product.

Some unscrupulous manufacturers have guessed this to be the case. They advertise a 10-minute battery, but then size their battery for about 65% load. By doing this, they can get away with a smaller, lighter cabinet that is a sleeker marketing package, and at a lower cost. There is no way it will deliver 10 minutes at full-rated load, but who is going to know?

The Ideal SPS

The ideal SPS for personal computers will vary somewhat from one application or site to another. However, what we have laid out in this section suggests some basic guidelines. We can say the ideal SPS will have the following qualities:

1. Transfer time that always switches fast enough so that the PC never experiences an outage.

2. Retransfer is done in-phase with the new utility power so the PC is not affected by inrush transients or by sudden changes in frequency.

3. An inverter waveform that is a sine wave with little distortion.

4. A battery time that will carry the full load for the rated time.

5. Battery protection, to keep the batteries from reaching excessive discharge levels, thereby damaging the batteries.

Power Conditioning

Many ads for SPSs claim that their product conditions power. This “power conditioning” is usually in the form of some sort of surge suppressor built into the unit. ill. 11-9 shows a photo of the input and output of an SPS. Notice that the white wire from the power cable enters the unit and immediately is routed to the output receptacles. Nothing touches the neutral. No surge suppressor can be seen. The only surge suppression in this unit is an MOV and a couple of capacitors connected between the line and ground. What’s even more ridiculous is that these components are mounted on the circuit board that contains the inverter section and all the other circuitry.

ill. 11-9. This SPS boasts surge suppression. Notice that the white wire (neutral) enters and leaves the unit, and is never touched by any form of surge suppression.

This is the height of folly. First of all, this arrangement is difficult to classify as surge suppression, and surely not “conditioned” power. Secondly, diverting energy on the same circuit board that contains sensitive circuitry jeopardizes every component mounted on the board. Marketing hype is one thing. Being competitive is another. There is no way that adequate power conditioning can be built into an SPS and stay competitive in the market. As buyers, we should know this and not deceive ourselves. it's for this reason that we highly recommend the setup shown in ill. 11-10.

ill. 11-10. The ideal electrical arrangement for a PC.

This setup combines the benefit of standby power with the low-impedance isolation-transformer line conditioner we described in an earlier chap ter. The line conditioner will fully condition the power from the utility, as well as eliminate the noise output from the SPS when it provides power. Care must be taken when selecting the SPS. If the unit puts out great quantities of noise and harmonics, like the “modified” square wave we saw earlier, it will cause the capacitors across the secondary of the line conditioner to burn up. With the proper SPS and a power conditioner, we have an ideal electrical environment for a PC.

But, ill. 11-10 suggests something to us that applies throughout this guide and to the largest computer installation. Putting a transformer between a UPS or SPS and the computer is never a bad idea. The normal- mode rejection, along with common-mode elimination through neutral-to- ground bonding, gives the advantages of clean power along with continuous power.

Recently, one company introduced a novel design. The SPS is placed between the secondary of a low-impedance transformer and the neutral-to ground bonding point. This also gives the advantages shown in ill. 11-10— all in one cabinet.

There are other ways of accomplishing this same thing, as we will see in the next section.

NEXT: UPS Designs

PREV: Power-Line Conditioners

All related articles   Top of Page   Home