UPS Designs

We named this section UPS Designs for a reason. We will be discussing not only what many people call “true on-line UPS,” but we will touch on several other designs that are arguably not on-line but are still not standby types.

We will start with on-line UPS designs since they are the products which really fit the definition of being uninterruptible. ill. 12-1 shows the block diagram of an on-line UPS. We can see that the incoming AC power is converted to DC by the rectifier/charger.

ill. 12-1. The block diagram of a true UPS.

When utility power fails or drops below a certain point, power is automatically pulled from the batteries. The physics of the way that the batteries float on the DC bus is such that no switching ever takes place. If the batteries need charging, they pull current from the bus. If the batteries are charged and the bus voltage is low, the batteries will deliver the needed current to the bus.

This AC-to-DC conversion process not only eliminates the need for any switching, it also has the facility of conditioning power at the same time. ill. 12-2 shows a photo of the output waveform of a UPS designed for PCs. We are feeding our 900-volt impulse into it. Notice the output shows practically no effect of the impulse and is a perfect sine wave.

ill. 12.2. The output waveform of a UPS, with a 900.volt impulse across the UPS’s input.

Since the inverter is on all of the time and is always providing power to the load, we call it an “on-line” UPS. This on-line design is the basic unit that powers not only PCs but huge mainframe computers. We should mention that UPSs for PCs will usually not have a bypass switch such as is normally found in larger units.

Why a UPS?

There are a couple of reasons why someone would want a UPS in a PC environment. Primarily, it's the uninterruptible nature of the power that the UPS delivers. There are a number of larger systems which can't tolerate switching that fall into the PC power range, but, they are not really PCs.

And, there are a number of PCs used in exotic systems which integrate special hard drives or other peripherals which are sensitive to any disruption of power.

The other reason, as we mentioned before, is the power-conditioning aspect of the AC-DC-AC conversion that takes place. This virtually eliminates any impulse coming down the line. To illustrate this, we have included the pictures shown in Figs. 12-3 and 12-4. ill. 12-3 shows the noise that was on the line before the insertion of the UPS. Both the common-mode and normal-mode signatures were virtually identical. ill. 12-4 shows the nor mal-mode and common-mode tracings of the output of the UPS.

ill. 12.3. Illustration of the noise on the line before the insertion of a UPS.

ill. 12.4. The normal-mode (top) and common-mode (bottom) output of a UPS.

ill. 12-4 shows something interesting about on-line UPSs. Notice how normal-mode noise has been eliminated. All that remains is some high frequency component of the PWM inverter. The common-mode output is an entirely different story. We see a noise envelope that has developed between neutral and ground, along with the noise that was on the line before. In some cases, we might observe an increase in common-mode noise due to a UPS. To make this point even more dramatic, let’s go back to ill. 12-2. What does the high-frequency noise output of the UPS look like with our 900-volt impulse being fed into it? ill. 12-5 shows us the answer.

ill. 12-5. The output of a UPS with a 900-volt impulse fed to the input. Normal-mode to common-mode conversion has taken place.

Clearly, the impulse has been stopped in its tracks as far as the normal- mode noise is concerned. This is significant since the impulse was being fed into the UPS from line to neutral. But inside the UPS, some coupling has taken place, and the impulse has been converted from normal-mode noise to common-mode noise. This new impulse is not as large as the one that created it, but it's large enough to cause significant problems for the PC downstream.

This raises an interesting issue with regard to UPSs. it's not uncommon for the harmonic distortion and noise output of a UPS to interfere with the proper operation of some equipment. This will usually manifest itself with symptoms like displays that wiggle and dance, or other kinds of interaction with the frequency output of the PWM inverter. Does this mean we shouldn’t buy a UPS? Not in the least.

All products have various trade-offs to consider when balancing one technology against another. We will see this dramatically as this section progresses. On-line UPSs have a distinct place in the market where uninterruptible power can't be sacrificed for anything else. But they do have shortcomings other than common-mode attenuation and a noise output. They tend to be less efficient than other products because the inverter is on all the time. They also put out heat and are noisy in comparison to other products. These problems, like the negative points we listed for ferros, may or may not be significant, depending on the site-specific needs and tastes of the user.

Ferroresonant UPSs

A popular product design that has really come into its own, in the 1000-VA to 10-kVA area, uses a ferroresonant transformer in a unique way. We have not yet talked about a “quality” of ferros that makes them useful in many applications with UPSs. This quality is ride-through.

Because of the large amount of wire and steel in a ferro and the energy circulating in its tank circuit, a ferroresonant transformer has the curious property of being able to deliver stored energy to the load for many milliseconds after utility power has failed. We must point out that this stored energy is proportional to the size of the load. If the load and the capacity of the transformer are closely matched, very little ride-through exists. How ever, under normal circumstances, there will be sufficient ride-through to switch on an inverter without disturbing the flow of current to the load.

ill. 12-6 shows us the block diagram of this type of UPS. Under normal operation, power is delivered through the ferro to the load. This affords all the advantages we discussed in Section 10. The load receives regulated and conditioned power at all times. The ferro eliminates normal-mode noise, and since the neutral-to-ground lines are bonded on the secondary, no common-mode noise appears across its output.

ill. 12-6. The block diagram of a ferroresonant UPS.

When the utility power fails, the inverter is turned on and the ride through of the transformer provides continuous power to the load. No inverter noise or switching transients are passed on to the load, as we have seen with UPSs and SPSs. Since the output of a ferro, without additional windings, is a square wave, and since the inverter is only on during an outage, the manufacturer of these devices can get away with an inverter that is significantly less costly, lighter in weight, and less critical in terms of its waveshape output.

This design has a lot going for it. It also meets my personal prejudice in favor of transformers. Even though it uses a ferro, a transformer on the output of a UPS is a design criteria that is optimal. But let’s look a little deeper. Everything has its price in terms of trade-offs, both economically and technologically.

In comparing this design over an on-line UPS, several things must be pointed out. First of all, we are not going to get away from the problems of ferroresonant transformers. But there are problems with on-line UPSs too. Let’s assume these concerns balance out. More often than not, this unit will be too large in terms of its available sizes, as well as its physical size, for most PC installations. However, where there are multiple PCs, this unit can be a central power source. Even the 500-VA units that are available weigh about 90 pounds. This may hold back some purchasers.

The real controversy between manufacturers of on-line UPSs and the producers of the ferro product is a dispute over another issue which we have not touched on so far—reliability. From a load standpoint, this ferro design is truly an on-line unit. But, from an inverter point of view, it's a standby power system. The on-line people will tell us that their unit is more reliable because their inverter is operating all the time. They will tell you that a standby inverter is less reliable since it must cold-start every time the power fails.

This point has the ring of truth to it. But, on-line units are operating all the time, meaning they may have a greater chance of failing. it's not just the thermal stress caused by large inrush currents that causes component failure. Constant running causes component stress as well.

People who live in glass houses shouldn’t throw UPSs. The UPS manufacturers have problems of their own. and , new solid-state components, along with proper burn in and testing, may make this argument mute. All equipment fails. Who knows where it will happen or when? The plain truth is that batteries are the single most unreliable element of a UPS.

The standby people would not survive long in a market that placed the burden of reliability on them so competitively. Knowing this, they must go to extraordinary lengths to turn out a reliable product. However, this does not mean that an on-line manufacturer will or will not go to great lengths to turn out a reliable product. At 10 kVA, when powering a minicomputer system with 75 terminals attached, this discussion is of vital importance, and many DP professionals may feel more comfortable with one technology or the other. Although this comfortable feeling may be deceiving, there are strong pros and cons. But at 1000 VA, the whole discussion is somewhat misplaced. The economics of the marketplace force everything to take a back seat to cost considerations. Here an argument over reliability has a hollow echo.

Against the Ideal

In the last section, we held up ill. 11-10 as the ideal electrical environment for the PC. How does this ferroresonant UPS affect that ideal? Obviously, it was the addition of a transformer downstream from the SPS that was so attractive. But in the case of the ferro UPS, we have those same advantages. The question becomes one that is related to the site and the needs of the user. The ferro has advantages like voltage regulation and ride-through, but normal PCs don’t need either. However, if either becomes a concern, the ferro must become an option.

If we bring the on-line UPS into this framework, we loose all the benefits of a transformer. Possibly the mind is not built to balance all these variables. Add to this the fact that the ideal unit, as pictured in ill. 11-10, is contained in two boxes.

How does one decide? The site, the computer, the peripherals, and the needs of the users are all factored together, and this will suggest the proper product choice. All things being equal, there is the matter of money. Typically, the SPS, coupled with a line conditioner, is the least expensive. Ferroresonant UPSs are substantially more expensive, but still somewhat under the cost of most on-line UPSs.

Tri-ports

The tri-port design is relatively new to the UPS field. Currently, this design is not to be found in the PC market. However, it's beginning to make an impact on medium-size systems. Nevertheless, we feel that it warrants discussion here since its entry into the PC field in one form or another is inevitable.

ill. 12-7 shows a diagram of the tri-port design. This product is a combination of a UPS and a tap-switching regulator. The incoming utility power is fed through a low-impedance isolation transformer via a set of SCR switches that can be adjusted for surges and sags. The output of the transformer is noise free, just like the line conditioners we talked about earlier. It also has the neutral and ground bonded together.

ill. 12-7. Diagram of a tri-port UPS.

During normal operation, the inverter draws enough power to keep the batteries charged. Some additional circuitry has been added to allow the inverter to act like a battery charger. While it charges the batteries, power flows from the primary of the transformer through the inverter and then into the batteries. When utility power fails, the inverter changes the direction of current flow and pulls power from the batteries, thereby becoming the source of power for the load. The sensing circuit must do two important things. First, it must acknowledge the outage and turn on the inverter. Secondly, it must disconnect the input feeder so that the inverter does not try to operate and back feed the unit’s own power source.

The tri-port design has several advantages. First, the inverter is on all the time. Some think this makes the unit more reliable than an SPS, although still somewhat less reliable than an on-line unit since it's not carrying the entire load all the time. However, engineers will argue that being on-line all the time is less reliable. Again, for small systems, this point is not particularly important. It has the advantage of being a low-impedance power source that provides a good electrical environment, and , of course, all the noise-attenuation characteristics of a power conditioner.

The problem with the design in the form presented here is likely to be “cost.” This unit has all the advantages of the ferro UPS except ride-through, without the disadvantages of a ferro. Its inverter could be more reliable while its regulator may be less reliable. But to produce this device at a price that would compete favorably with a ferro would probably not be possible in sizes for PCs.

If this product design ever makes it to market, it will probably have to eliminate some features, like the tap-switching section, in order to be a price competitor. Even at that, it would still be attractive since it provides both conditioned and standby power. We have included it here since the design has important distinctions which fit well into our overall UPS discussion.

Line-Interactive UPSs

As the name implies, line interactive is a UPS designed to be on-line only to interact. Tri-ports are a variety of line-interactive units since they interact with the lines changing state. While tri-ports have yet to make an appearance as products for PCs, there is a line-interactive type of UPS that is making an appearance. it's called a load-sharing UPS. ill. 12-8 shows us this basic idea.

ill. 12-8. The block diagram of a load-sharing, line-interactive UPS.

This design looks strangely like an SPS. But there is an important difference. This is an on-line UPS. In other words, the UPS circuit from the rectifier charger to the battery, and through the inverter, is on all the time. The sensing circuit, which is very simple in this case, governs the amount of current that the load actually needs to draw from the inverter section.

During normal operation, the inverter might provide a small fraction of the current needs of the load. The bulk of current would be drawn from the line. Let’s say that the line voltage begins to sag. As this occurs, the inverter would be called on to provide more and more power to the load. Finally, if the power went out completely, the inverter would supply 100% of the power required by the load.

This design accomplished some marketable concepts. it's on line all the time, so any concern about cold starting an inverter is gone. There is no switching time since it's on line. Devices sensitive to the 4- to 10-milliseconds delay typical from an SPS would have nothing to fear from a load- sharing design.

On the other hand, there is no ongoing power conditioning with this approach. Of course, the inevitable answer to this is to throw a few MOVs into the unit and call it a surge suppressor as well. This design is beginning to crop up in those under-the-monitor-type control-center packages.

The RUPS

We can leave a section about UPS without touching on rotary technology. A more complete discussion of this topic is contained in my earlier guide, Computer Electrical Power Requirements. ill. 12-9 shows the basic block diagram of this type of device. Notice that the incoming utility power is used to drive a spinning motor which is mechanically linked to an AC generator. When utility power fails, there is enough inertial mass in the motor! generator set (MG Set) to ride through until the inverter section is able to provide power to the unit.

This seems like a strange technology to find in a guide about PCs. However, this unit is being built in sizes clear down to 1000 VA. This is almost unbelievable to many DP professionals who remember RUPS as being found only in large data centers. Presently, the military is the only real customer for this kind of technology in sizes that might fit in the PC environment. We include it here on the off chance that its use might become more widespread, although this is a remote possibility. Even so, the thoughtful reader may find it valuable to know such a technological approach does exist.

UPS Wrap-up

We have seen quite a plethora of power products in the last two sections. To be sure, the most popular UPS for PCs is the SPS, especially in the sizes under 1000 VA. Above 1000 VA, applications often call for the advantages of on-line UPSs. At two to three times the price, this can be a difficult decision.

The load-sharing design is an attempt to get at this price differential, while other technologies, like the ferro design, really compete head on with the on-line varieties.

In the next section, we will see two growing applications for power products—networks and desktop publishing. it's in considering these kinds of installations where we must put our knowledge to the test. Which will be the best, most cost-effective application of technology? This guide now serves us as a “handbook” for determining what makes the most sense in each situation.

NEXT: LANs, Desktop Publishing, and Other Final Thoughts

PREV: Standby Power Systems

All related articles   Top of Page   Home