Voltage Regulators

We will discuss two basic kinds of voltage regulators in this section. But, before we do, we must put the entire subject of voltage regulation in some kind of perspective where PCs are concerned. On the face of it, voltage regulation sounds like a very attractive proposition.

However, a glimpse back at Section 6 will confound this notion to a large degree. In that section, we discussed the historical transition of computer power supplies from linear supplies to switching power supplies.

As we will see in this section, this is approximately the same range as that of a good voltage regulator. To put it more directly, a voltage regulator has a range roughly approximating the acceptable limits of a switching power supply. If that’s true, who needs voltage regulation? Almost no one!

Like many things in life, there are no simple answers, and there are perfectly legitimate circumstances where voltage regulation is the correct application of technology. But strictly speaking, personal computers and their peripherals don't require voltage regulation.

Why Use Voltage Regulation?

Why use voltage regulation? If all of the preceding statements are true, why would a manufacturer build a voltage regulator in the first place? and , further more, what are the circumstances where voltage regulation is a requirement?

Let’s look at the first question first. As we will see, voltage regulation is a side benefit of some power-conditioning technologies. But it's also a costly add-on to others, which may significantly reduce the reliability of an other wise superior power conditioner. Why would a manufacturer do this? A lack of understanding of the PC marketplace and of technology is the only answer. The explanation for this leads us to the next question.

When is voltage regulation necessary? The obvious first answer to this question is in linear power supplies. There are a few applications where older or exotic equipment is used that contains linear power supplies. These pieces of equipment not only need power conditioning, but they need voltage regulation as well. Some of these applications might involve industrial, military, or medical environments. But even then, the power supply of choice is usually a switch that doesn’t care much about the voltage it sees.

Another area where voltage regulation may be required is in the mini- and mainframe environment. For many years, it was standard practice for power- supply engineers to design shutdown features into the power supplies of the larger computers. For a wide variety of reasons, they did not want their machines running on power that had surged or sagged beyond acceptable limits. When the power supply sensed the voltage wandering, it would shut down the equipment. It was for this reason that voltage regulation became so popular. No DP manager wants to go down every time the power fluctuates.

This practice has continued until today. Even after the rationale for such circuits has gone away because of design advancements, like switching power supplies and smaller machines, minicomputer manufacturers may include this feature in their units. This can bedevil a site for no good reason. But voltage regulation is stuck in the psyche of many in the industry. and many manufacturers of computer power equipment started by designing products for this industry. Now they are manufacturing equipment for the PC market and this prejudice toward voltage regulation has followed. This explains why they might include it as a design criteria.

While the problems of minicomputers don't affect PCs directly, they may affect some environments where PCs are installed and this will cause a need for voltage regulation. But beyond these general areas, the need for voltage regulation for PCs is questionable.

Tap-Switching Regulators

In an earlier section, we talked about a concept called “turns ratio.” We know that “power in” equals “power out.” So if the turns of the primary winding equal the turns of the secondary, the voltage on the output will equal the voltage on the input.

Let’s look back at ill. 9-7. We know that this power-line conditioner will give us a clean noise-free environment. But since “power in” equals “power out,” we will always see the same voltage at the output that appears on the input. The voltage may sag, and the voltage may surge—all slow-moving voltage changes will pass right through.

But what would happen if we could change the number of turns in the primary as the voltage fluctuated? Let’s say we could take a few turns off of the primary. Now the voltage at the secondary would be slightly higher since the turns ratio would have changed. More turns on the secondary in relation to the primary would mean a higher voltage.

For years, engineers have been trying to figure out how to change the turns ratio to achieve a stable voltage at the output of a device as the input changed. A number of unique approaches have been used. Most of these methods used some sort of mechanical motion to make and break new connections on the windings of a transformer, and thus change the turns ratio. But mechanical action proved far too slow for typical voltage changes. Then, with the revolution in semiconductor devices, engineers found solid- state switches that could handle the job of changing taps electronically. ill. 10-1 shows this basic concept at work.

Silicon-controlled rectifiers are usually used as solid-state switches to adjust the turns ratio for line-voltage fluctuations. A sensing circuit measures the voltage, and as it changes, the circuit sends the proper bias voltage to the correct SCR. The SCR begins conducting through the tap to which it's attached. Each tap is embedded in the windings of the primary or secondary, depending on the design. By properly placing these taps, the turns ratio can be adjusted in precise steps as the input voltage varies.

The design shown in ill. 10-1 measures the voltage on the secondary side of the transformer and feeds this information back to the SCRs on the primary side so action can be taken. This primary-to-secondary separation is somewhat slower than the next two designs we will see. The ability to switch fast is essential if the unit is to respond to sags and surges.

In order to improve this, a design like the one shown in ill. 10-2 can be used. Here we see the taps have been moved to the secondary side, where the association of the taps with the sensing circuit make for a few milliseconds faster response time. We might note that the advantage of having the taps on the secondary is a much finer voltage adjustment for the load. Dynamic changes in the load can affect voltage just as a sag might. Therefore, taps on the secondary can respond to both load changes and supply changes.

ill. 10.1. A tap-switching regulator changes the output voltage by changing the turns ratio of a transformer.

Another advantage of secondary regulation is the placement of the SCRs. As we stated at the outset of the section, the addition of voltage regulation to a power conditioner can sacrifice reliability. Increasing the parts count of any electronic system will inherently make the system less reliable. Tap switchers are no exception.

The line conditioner we discussed in the last section will always provide power to the load, unless the transformer burns up. That is extremely rare. Certainly, the capacitors and MOV may fail, but power will still flow to the computer. Forcing supply current to flow through a solid-state device, like an SCR, will make the unit less reliable. Putting the SCRs on the secondary side of the transformer will provide inductive cushioning that will protect them to a certain degree.

ill. 10-2. Changing taps on the secondary provides finer voltage adjustments.

It is no secret that switching creates noise. and placing SCRs on the load side of a transformer will transmit any noise created by the switching action directly toward the PC. In order to get away from this problem, a design like the one shown in ill. 10-3 is used. This combines the advantages of the noise-reduction capabilities of the transformer with a fast switching action. Obviously, it sacrifices SCR vulnerability and load-side regulation.

Let’s look at one more design consideration. In the early days of tap- switching technology, voltage was used as the switching trigger. In other words, the sensing circuit would wait until the voltage dropped to zero before switching. But, as can be seen in ill. 10-4, significant current is still flowing at the zero-voltage crossing point of the sine wave if the computer has a lagging power factor. This switching action created large impulses that often interfered with computer operations.

The answer to this was to sense for the zero-current crossing point and turn the SCRs on and off when no current was flowing through them. This change dramatically decreased the amount of switching noise generated. Superior tap switchers use the zero-current crossing design.

The typical range of regulation from tap switchers designed for the PC market is -15% to +10%. Or, to put it another way, from about 102 volts to 132 volts. There are some families of products that regulate -27% to +15%. Even this is still within the range of a switching power supply. Not only that, but as the voltage sags, the load draws more current to compensate. At some point, enough current should be drawn along a branch circuit that a circuit breaker should trip.

A note on SCR reliability. The reliability issue can't be ignored. SCRs have a high initial mortality. Studies have shown it to be as high as 14% in some cases. We are not saying that products fail. We are saying the SCRs fail. That is why proper burn-in is absolutely essential in order to weed out those semiconductors that are likely to fail during the first few days of operation. If care is taken in selecting the right SCR with the proper current-handling margin of safety, and if proper testing is done before the unit leaves the factory, tap switchers can be manufactured that are very reliable.

We will move on now to the other kind of voltage regulator that we find on the market. But, before the section is over, we will come back to the subject of tap switchers—once we have put the entire subject in perspective with a discussion of ferroresonant regulators.

ill. 10-3. Tap switching with a regulated input.

ill. 10-4. When voltage is zero, significant current flows.

Ferroresonant Voltage Regulators

Ferroresonant voltage regulators, or ferros, depend on a technology that is radically different from either tap switchers or the kind of power-line conditioners which we have been discussing. These products are also known as constant-voltage transformers, or CVTs, but we will refer to them as ferros. There are companies that like to call them other things which are meant to throw the buyer off the track. But most reputable manufacturers will state in their literature that ferroresonant technology is being used.

We mentioned that ferros are referred to as constant-voltage transformers. This indicates that there must be something about the way they operate that delivers a constant output for a varying input. This is true since ferroresonant transformers operate in a way that is different from the simple isolation transformers we normally find in power-line conditioners and tap- switching regulators.

We know from earlier discussions that the current flowing in the primary winding of a transformer induces current in the secondary winding, because magnetic lines of flux cut through the windings of the secondary. As current in the primary increases, the total amount of flux increases. More lines of flux means that more current is induced in the secondary. There is a point at which additional current flow in the primary will not create more flux in the core of the transformer. This point is called saturation. The core of the transformer becomes saturated with flux, and a further increase in primary current can't create more lines of flux, and , therefore, no additional current is induced in the secondary.

A ferro uses the principle of saturation to its advantage, in order to regulate voltage. If we could construct a transformer core so that it would easily achieve saturation, we would have a natural voltage regulator. We can do this with an isolation transformer since current at saturation is close to short-circuit levels. ill. 10-5 shows a simple diagram of a ferroresonant transformer.

ill. 10-5. A simple diagram of a ferroresonant transformer.

A ferroresonant transformer is specifically designed to go into saturation. ill. 10-5 shows that the core is constructed with an air gap between the primary and secondary windings of the transformer. Notice also that the output of the transformer is connected across a capacitor. In practice, this may be many capacitors connected together. The design of the air gap and the value of the capacitance are the key design elements affecting the operational characteristics of the transformer.

The air gap gives the transformer a high reluctance as current flow starts at a low level. This means that the flux is primarily through the outer regions of the core. As the current flow increases, secondary voltage is determined by the turns ratio. As the flux lines increase, the current in the secondary also increases, causing the reactance of the secondary winding to increase. When the inductive reactance of the secondary winding becomes equal to the capacitive reactance, the capacitor and the winding become resonant. This causes the reluctance of the center core to drop and the transformer to go into saturation. Subsequent increases in primary current have a small influence on the secondary.

To compensate for these small voltage variations, another winding is added to the transformer. Strangely enough, this is called the compensation winding (ill. 10-6). it's placed in the electrical circuit of the transformer, in series, opposing the secondary. This works to offset small voltage fluctuations that may come through the transformer even though it's in saturation.

ill. 10.6. The addition of a compensation winding will offset small voltage variations that appear on the secondary.

The design of the transformer shown in ill. 10-6 has been around for approximately 40 years. it's used as a regulator for regulating the voltage to lighting systems and other loads which were not particularly disturbed by harmonic distortion in the waveform. Computers are not one of those loads. Harmonics can be a source of noise that can disrupt data processing and can unnecessarily heat components. The transformer shown in ill. 10-6 puts out a square wave.

The combination of resonance and of the transformer being driven into saturation produces a square-wave output (ill. 10-7). For use with computers, we must have some way of eliminating the harmonics contained in the square wave. If we could see the equations that describe a sine wave, as opposed to a square wave, we would see that a sine wave is derived in a very straightforward manner using the sine of a function, along with values of time and frequency. The equation for a square wave has all the same elements, along with the numbers 3, 7, 9, 11, 13, 17, ..., etc. These numbers represent the derivation of odd harmonics that actually compose this type of waveshape. We will not reproduce these equations here since they are full of expressions that would take too long to explain and are not much fun to read—to most readers anyway.

To get around this problem, a neutralizing winding is added to the circuit. This winding is magnetically linked not only to the primary but to the secondary (ill. 10-8). The purpose of this neutralizing winding is to allow the harmonic voltage to appear on this winding, having been coupled from the primary. This harmonic voltage is then fed to the secondary in series opposing, thus cancelling out those nasty odd harmonics. This, along with some filtering, produces a sine-wave waveform that is suitable for use with computers (ill. 10-9). A note of caution: Not all manufacturers include a neutralizing winding.

ill. 10-7. A ferroresonant transformer produces a square-wave output.

A comment about ill. 10-9 needs to be made. This picture looks like a fairly decent sine wave. What is even more impressive is that ill. 10-10 shows the sine wave that was fed to the primary of the ferro. This is the same old, rumpled, flat-topped, distorted power waveform we saw in an earlier section. The ferro is literally putting out its own sine wave, with little regard for what is feeding it.

The transformer we have presented here is the ideal ferro, or CVT. The picture in ill. 10-9 was taken at the output of an actual product that is sold for PC users. It may very well not have the exact same winding design we have discussed, and it probably has no additional filtering, but the principles we have illustrated are common to all of these devices.

ill. 10-8. Adding a neutralizing winding, plus additional filtering, will produce a sinusoidal output.

ill. 10-9. This is the actual sine wave produced by a ferro.

Ferros have a wide regulation window—in some cases, from -40% to +20%. They are rugged. Other than the capacitors, which do fail with some regularity but which don't cause the load to loose power in most cases, there are no parts to fail—mechanical or electronic. They also have an inherent ability to suppress normal-mode noise. Because of the saturation of the transformer, impulses can't penetrate to the secondary windings. In other words, when an impulse appears across the primary windings, no additional flux lines will be created since the transformer is in saturation. This normal-mode immunity is in stark contrast to the isolation transformers we talked about earlier. They required additional filtering and surge suppression in the normal mode. On a ferro, neutral and ground are bonded together on the output for common-mode protection.

ill. 10.10. A picture of the sine wave feeding the ferro that produced the output shown in ill. 10-9.

So far, the ferro, or CVT, sounds like the absolute answer to any power problems. But before we compare the products we have discussed, let’s look into some of the drawbacks of ferros. Every technology has some inherent strengths and weaknesses. Usually the strengths are inextricably tied to those weaknesses. In the case of a ferro, this is once again true.

One of the ferro’s major shortcomings is its ability to transfer energy. A ferroresonant transformer is a high-impedance device. This means that its resistance to the flow of current is high. Now high is a relative term, but we are talking about the internal impedance of a ferro as opposed to an isolation transformer. The net effect of this high impedance is that a ferro reacts more slowly to instantaneous current demands placed on it by the computer system connected to it.

On the other hand, an isolation transformer instantly reflects the demand of the load back up the line to its power source. This difference in internal impedance is the source of much controversy between rival manufacturers. The argument centers around the current demands of the switching power supply.

It may have already occurred to you that a ferro is the ideal product for linear power supplies. Their need for a constant source of stable voltage makes a ferro one of the best choices for an ideal electrical environment, although the same could be said of a tap switcher. Where switching power supplies are concerned, the matter is entirely different. Remember the switch wants to pull all of its current during a very short portion of the sine wave. This pumping of short bursts of current places an unusual demand for current on the power source.

In the case of a ferro, all the available current must be circulating in the resonant tank circuit, a circuit made up of the secondary windings and a capacitor. If this circuit were to suddenly become drained of its current, the transformer would loose resonance and would no longer be in saturation. The job of the transformer is to pull enough current from its source to hold the transformer in resonance and , therefore, saturation, plus providing for the changing current demands of the load. If the transformer is not sized properly for the load, it's possible for the switching power supply to gradually become starved for current, resulting in computer malfunctions.

This current starvation results from the fact that the time taken to transfer all the needed energy (from primary to secondary) can be too slow to react to the dramatic demands of the switching power supply. (Again, this is the subject of much argument between manufacturers.) This kind of thing would never happen with a low-impedance transformer. This high-impedance factor leads to another related liability—inrush currents.

Since the startup-current demands of most computer devices are on the order of six to ten times their ordinary running levels, it's possible that a ferro might not be able to meet these inrush demands. Most ferros can't service current demands of more than 125% to 150% for more than a few milliseconds. The result of this is that the entire system might have to be brought on line, one device at a time. There have been instances where a properly sized ferro simply could not start up some system combinations at all, because of the inrush demands.

But let’s be reasonable. This almost never happens. The reason is simple. Designers of ferroresonant products are well aware of the problems we have been discussing. In order to avoid these situations, they often build any where from 20% to 30% more capacity into their products to avoid current demand problems. If the device is oversized to begin with, the likelihood that switching power supplies or inrush demands could cause problems becomes nearly insignificant.

Of course, nothing comes without a cost. A little more capacity means more steel in the core, more wire, more cost, and more weight. Watt for watt, a ferro has far more steel and copper in it than a low-impedance transformer. This is due to a core that is built to be driven into saturation, along with the addition of neutralizing windings and compensation windings. Also, adding hidden capacity to avoid field collapse and current starvation makes this difference even greater.

Efficiency is another concern. As we stated earlier, an isolation transformer of the type found in power-line conditioners and tap switchers is about 96% efficient. The efficiency of a ferro varies with the load placed on it. When a ferro is fully loaded, efficiencies approach 90%. But, as we have seen, running a ferro fully loaded runs the risk of overloading the capacity of the tank circuit, which will cause the field to collapse. As the load drops to 60% of rated capacity, efficiency can drop to as low as 65% to 70%. Remember that no matter how light the load, the transformer still must be driven into saturation. This uses energy even when nothing is plugged into the unit.

Inefficiency results in heat. This, coupled with the dynamics of the circulation of energy within the transformer, creates audible noise. Com pared to isolation transformers, ferros are heavy, hot, noisy, and big. But let’s not get too carried away. At 500 VA, for instance, the differences are minor. At 10 kVA, they are larger. and , as we stated in our other guide, Computer Electrical Power Requirements, three-phase devices display enormous variances. The differences are relative, according to the size of the requirement.

Let’s touch on two other areas of potential concern before we move on. Overload protection of the branch circuit supporting a ferro can be a problem. Since the current demands which a ferro places on its source of power are little different from no load to full load, even a short circuit across the output may not pull enough additional current to trip an “upstream” circuit breaker, in some cases.

The other problem is that in order to go into resonance, the ferro tank circuit must see a constant flow of 60-Hz power, plus or minus a fraction of a hertz. Since this is a tuned circuit at 60 Hz, minor variations in source power frequency may cause the transformer’s magnetic field to collapse. This is almost never a problem except when on-site generation of power is used—e.g., during prolonged outages. Diesel generators have long been notorious as a source of frequency drift.

Technologies Compared

So how is one to choose? Tap switchers are smaller, more efficient, more able to provide for large short-duration current demands. Ferros, on the other hand, have a greater regulation range, are probably more reliable, and have better normal-mode noise rejection. Let’s be honest. In the size required by PCs, the differences between these technologies shrink to insignificance. Arguments rage between rivals in the industry, but the customer will make his or her determination more on price than on any other point.

This is where ferros have a big edge. To date, it's far cheaper to build a ferro than to build a tap switcher for personal computer systems. Often the difference in price between two equal-sized units can be significant. When all the facts are considered, it's hard to see how tap switchers can beat ferros as the regulator of choice for systems under 2 kVA. However, technology changes, and different manufacturers do creative things that may nullify this assumption, and after all, the choice is not all that clear-cut in every circumstance. For readers of my earlier guide, Computer Electrical Power Requirements, this may seem to be an about-face, but, not at all. As power requirements increase, the disadvantages of ferros grow. This tilts the scales in favor of tap switchers for minicomputers and mainframes, especially as soon as we get into the three-phase power ranges.

But this is not the important conclusion to this discussion. The real decision for the PC user is not between tap switchers and ferros. The real decision is between power-line conditioners and ferros. This is clearly because of the lack of need for voltage regulation in small personal computers.

Let’s twist our perspective away from voltage regulation for a moment. We can do this because we have already learned that switching power supplies don't need voltage regulation except under rare circumstances. With this in mind, we can evaluate a ferro in terms of its qualities as a power conditioner, with voltage regulation as a mere side benefit.

Having stripped away regulation as an issue, we are left with three key areas of interest: noise rejection, physical characteristics, and the device as a power source. We have already shown that a ferro is a better conditioner for normal-mode noise, although a low-impedance transformer with some filtering is good. Both units have their neutral and ground lines bonded together, so let’s assume that they have comparable common-mode attenuation. We know that a simple power conditioner is smaller, lighter, more efficient, and less noisy. But these differences are tolerable in most cases, so let’s give the power conditioner a slight edge here. Now, we are left with the concept of the device as a source of power. We know that a power-line conditioner is a low-impedance power source that can deliver ample cur rent on demand. The ferro, however, is limited in this respect since it's a high-impedance device.

Being a good source of current for switching power supplies and short- term overloads is one aspect of power-source behavior. Another is, how does the device handle noise or harmonic distortion placed on the line from the load side? In order to test this, we fed our 900-volt spike generator power from both a ferro and a low-impedance transformer. Thus, we plugged the power conditioner into the wall and then connected the genera tor as the load into the power conditioner.

Figs. 10-11 and 10-12 show the results, which are fairly dramatic. ill. 10-11 is a photo of the power waveform available at one of the unused outlets of a low-impedance line conditioner. ill. 10-12 is the picture of the power sampled from a ferroresonant transformer. While the simple power conditioner shows clearly that the spike is gouging a hefty divot out of the peak of the sine wave, the ferro is showing us something radically different. Some sort of ringing oscillation has been set up by the windings and the resonant circuit caused by the pulsing of the generator. The generator is firing only once per cycle.

ill. 10.11. The effect of a spike generator on a simple power conditioner, when used as a power source.

A more realistic example of this kind of problem would be found with devices like printers or outboard tape backups. Large starting or cycling currents could reflect back enough of a transient to cause the kind of pattern shown in ill. 10-12. Let’s be careful not to exaggerate this. The chance of peripherals back feeding 900-volt impulses is rare. But we will see a device in the next section that generates impulses in excess of 300 volts every half cycle, and it could easily be connected to a ferro.

ill. 10-12. The effect of a spike generator on a ferroresonant regulator when it's used as a power source.

Summary

What does this really mean? Is this some horrible secret defect in ferroresonant power conditioners? No. What it means is that the interrelationship of elements within the device makes it more susceptible to generating harmonic distortion in certain unusual circumstances. If it sounds like we are building an enormous case against ferros, we are not. There are many interesting things to be said on the subject that the thoughtful buyer should be aware of.

In the final analysis, the decision of which unit to buy can be evaluated by answering a few direct questions.

1. Do I need voltage regulation?

If so, the choice between tap switchers and ferros is a matter of the needs of the specific site and the comfort of the buyer with the points we have discussed. Price will no doubt make the choice clearer.

2. Do I need power conditioning?

If so, the choice is between a ferro and a simple power-line conditioner. The relative strengths and weaknesses of both product types have been described. The potential buyer of multiple units should obtain evaluation units and “try before you buy” to measure actual on-site performance against the technological differences presented here. Price differences between the two are slight.

3. Do I need backup power?

Backup power will be the subject we will move to next. We include the question here only to lay the framework for further discussions of the subject of power conditioning. We will return to the subject of power conditioning and voltage regulation as we look at some new approaches to the issues. This includes the subject of providing power when the utility source fails.

NEXT: Standby Power Systems

PREV: Power-Line Conditioners

All related articles   Top of Page   Home