PC Power Supplies

It may come as somewhat of a shock to some that electronic circuitry like that found in a personal computer uses DC power. The role of the power supply is to convert the incoming AC power to DC power at the logic voltages used by the internal circuitry of the PC. it's the power supply that is tied directly across the 120-volt AC line. it's the front end of the computer’s power structure.

Over the last few years, there have been major changes in the technology behind the design of power supplies. Interestingly enough, the technology of some power products have not kept up with these changes and , as we shall see in a later section, some of those products are no longer appropriate for the newer PCs that are on the market.

There are basically two different kinds of power supplies found in modern computer equipment: the linear power supply and the switching power supply. The linear-type supply was popular for many years after World War II. We will discuss its design characteristics shortly. More recently, the switching power supply, often referred to as the switch-mode power supply, appeared in computer equipment. These two different types of power supplies take radically divergent approaches to converting AC power into DC power. Their unique behaviors and needs are the focus of this section.

Principles of DC Power Supplies

Most microprocessors and other associated electronic circuitry need one level of voltage from the power supply, + 5-volts DC. Other common voltages that may be required are +12 volts, -5 volts, -3 volts, and +10 volts. For simplicity, we will use +5 volts in our examples. Clearly, 120 volts is larger than 5 volts so a transformer will be needed to step the voltage down to the 5-volt level. Next, we must worry about how to convert alternating current to direct current so our +5 volts will be usable by the logic circuits connected to it.

Rectifiers

Alternating current, as we have seen, moves in two directions. During a portion of the cycle, it moves in a positive direction. Then it turns around and moves in a negative direction. This change of direction is not consistent with what we want to feed to logic circuits. To correct this changing of direction, a device called a rectifier was invented. It consists of one or more diodes arranged in a circuit so that current can flow in only one direction.

ill. 6-1 shows a diode rectifier in action. Notice that when the anode is positive with respect to the cathode (ill. 6-1A), the diode is forward biased and current will flow. But when the voltage polarity reverses, the anode is negatively biased with respect to the cathode (reverse biased), and no current flows. The result is a pulsating output that is unidirectional. This is not pure DC, but at least the direction of current flow does not change.

ill. 6-1. Current flow through a diode rectifier. (A) Forward biased, current flows. (B) Reversed biased, no current flow.

There is a small problem with this arrangement. ill. 6-2 shows the output waveform of the rectifier. Notice that every negative-going half cycle is missing. Because of these missing half cycles, our rectifier is called a half wave rectifier. Unfortunately, the output voltage of this type of rectifier is only 0.45 times the input RMS voltage. Also, the half-cycle effect makes it necessary to have a transformer with a VA rating that is 40% higher than the VA of the load. Obviously, these characteristics make the half-wave rectifier extremely inefficient.

ill. 6-2. Output waveform of a half-wave rectifier.

Intuitively, we might expect to discover another design that would overcome these inefficiencies, and which would rectify the full sine-wave waveform of the input. In this case, our expectations are correct. We simply modify the circuit slightly, add another diode rectifier for the missing half cycle, and we have a full-wave rectifier. ill. 6-3 shows this design.

ill. 6-3. A full wave-rectifier circuit that uses a center-tapped transformer. For one half cycle, current flows through diode D In the next half cycle, current flows through diode D2.

During one half cycle, diode D1 will be forward biased and diode D2 will be reverse biased. This will cause current to flow from the center tap, through the load, and back through Dl. During the alternate half cycle, diode D2 will be forward biased and diode Dl will be reverse biased. This time, current will flow from the center tap, through the load, and through diode D2. The current flows in the same direction during the full cycle. The output waveform looks like the one shown in ill. 6-4. Since the current flows in the same direction through the load twice during one full cycle, there are two humps per cycle in the output waveform.

ill. 6-4. The output waveform of a full-wave rectifier.

The average output voltage from this type of full-wave rectifier is 0.90 times the applied RMS input voltage. By comparison, this makes the full-wave rectifier twice as efficient as the half-wave variety. You might ask yourself, why would any design engineer ever call for a half-wave rectifier? The half-wave rectifier allows the use of certain designs that call for rectification without the use of a transformer. Often times, consumer- electronic devices are made using this kind of design—for the sake of cost, weight, and size.

There is yet another design that can be used without a transformer. That is the full-wave bridge rectifier. For now, we will discuss the use of the bridge rectifier with a transformer. ill. 6-5 shows a bridge rectifier. We can compare this with the diagram of a full-wave rectifier and see that the center-tapped transformer has been swapped for two additional diodes. During one half cycle, the current flows through diodes D1 and D2. During the next half cycle, the current flows through diodes D3 and D4. Another thing to notice is that although the output waveform of the bridge is basically the same as that of the full-wave rectifier, we are now using the entire secondary winding of the transformer to produce the desired voltage. Therefore, the design considerations used with the transformer is different due to the differing turns ratios and power ratings.

ill. 6.5. A full-wave bridge rectifier.

The Sine Wave

You may have wondered by now about the relationship between various voltage measurements of the sine wave. We’ve used the term RMS or root- mean-square. Later in the guide, we will talk about peak voltages or peak- to-peak voltages. ill. 6-6 shows how these different terms relate.

ill. 6.6. Sine-wave relationships.

Filters and Regulators

So, now we have an output waveform that looks like a bunch of humps (ill. 6-4). This is called pulsating DC, and as such is still unusable by our personal computer. Filtering is the next step toward producing a clean, ripple-free DC. ill. 6-7 shows the bridge rectifier with a filter section and voltage regulator added. ill. 6-8 shows the output waveform after filtering.

ill. 6-7. A bridge rectifier with a filter section and voltage regulator added.

The output of a full-wave rectifier (ill. 6-7) has a 48% ripple component. As we might expect, the frequency of this ripple is twice the line frequency, or 120 Hz. By selecting the proper value for capacitor C1, we can fill in this ripple. Capacitor C1 will charge to a value equal to the peak of the voltage magnitude. After the peak has passed, C1 will begin to discharge back into the circuit. The net affect of this is to fill in valleys between voltage humps (ill. 6-8). Resistor R1 is what’s known as a bleeder resistor. Its purpose in the circuit of ill. 6-7 is to drain the charge of the capacitor when the power is turned off—to eliminate possible shock hazards.

Inductance L1 is a coil of wire wound around an iron core. This is known as a choke. If we select the proper value of inductive reactance, at the ripple frequency, the choke will oppose the flow of ripple current.

A voltage regulation section is then added. Most often, this will be a series-pass regulator. A series-pass regulator is a series-connected transistor (Q1) that varies its voltage drop to keep the output voltage at a constant level. This is known as linear regulation and is where that type of power supply gets its name.

ill. 6-8. Output waveform of ill. 6-7 rectifier circuit, after filtering. The discharging capacitor fills in the area between voltage humps and smooths out the waveform.

Linear power supplies have the disadvantage of being both large and heavy, because of the 60-Hz transformer. They are also inefficient, operating at around 50% for the typical 5-volt supply. This means they give off significant heat. Also, linear supplies are voltage sensitive. The linear volt age regulator can compensate for only small voltage excursions. Beyond that, it may cause system failures or even shut the system down. Its internal ride-through time is about 10 milliseconds on the average.

Switching Power Supplies

ill. 6-9 shows the schematic for a switching power supply, sometimes called a switch-mode power supply. A switching power supply is 70% efficient, is smaller, lighter, and cooler, and has a ride-through on the order of 20 to 40 milliseconds. Let’s take some time and see how this circuit works. We must point out that there are a variety of designs and variations to this concept. We are only concerned with the design principles involved.

In ill. 6-9, we see a number of circuit elements that we have seen before. The first of these is the bridge rectifier. Notice that it's connected directly across the power line. During alternate half cycles, current flows and charges capacitor C1. Next, we see an inverter circuit made up of transistors Q1 and Q2. These two transistors conduct when they are turned on by the pulse-width modulation (PWM) control circuit. For the sake of simplicity, we have drawn this circuit as a box.

When either transistor Q1 or Q2 conducts, current flows through the high-frequency transformer and through the full-wave rectifier. The choke and capacitor filter the ripple current and we have 5-volt DC output for our logic circuits.

ill. 6-9. The circuit of a switching power supply.

At the output of the power supply, the circuit takes a sample of the voltage and compares it to a reference voltage. If the output voltage is too high or too low, an error amplifier tells the PWM control circuit to vary the amount of time that transistors Q1 and Q2 are turned on. In other words, what is happening is that the timing of current flow through the inverter controls the voltage across the final stage of the power supply.

We have shown a 20- to 100-kHz oscillator. The oscillator frequency is fixed for any given switching power supply. But designs are in use that use differing rates—from tens of thousands to hundreds of thousands of cycles per second. The trend more recently has been toward faster switching rates.

The switching power supply is current hungry. The amount of current drawn from the line during any given half cycle is largely dependent on the charged state of capacitor C1. ill. 6-10A shows a photo of the voltage and current sine wave of a switching power supply. Clearly, the bulk of the current is drawn at the peak of the voltage sine wave. This is, from a power standpoint, the most unique feature of this design. The capacitor gulps large amounts of current in short pulses twice every cycle. Meanwhile, the switch is chopping the conduction of current flowing from the rectifier and capacitor into thousands of chunks every second.

ill. 6.10. Power supply waveforms. (A) Voltage and current waveforms of a switching power supply. (B) Voltage and current waveforms of a linear device.

The net effect of this action is the extremely fine correction of voltage variations at the output. The input voltage, since the supply is mostly current sensitive, can swing from 85—90 volts to 130—140 volts. The previous discussion shows why a switching power supply is so popular among PC power-supply design engineers.

Noise and Power Supplies

Since this guide is about computer power problems, it would seem logical to question the ability of a given power supply to handle both normal-mode and common-mode noise. Let’s review what we mean by noise. Noise is any signal in a circuit other than the desired signal. Noise has many sources and many magnitudes. When we say noise in the context of power supplies, we are talking about repetitious or transient impulses that appear across the terminals of the supply. These may be large enough to do actual damage or can interfere with the data processing.

ill. 6-11 reminds us of what common-mode and normal-mode noise is. Normal-mode noise can be measured between the line and neutral. Common- mode noise can be measured from line to ground or neutral to ground. From this, we see that common-mode noise must appear on the line and neutral at the same time. Often we hear the term ground noise. This is common-mode noise that appears on the ground side as measured from the line and /or from neutral. Usually this voltage is measured simply between neutral and ground.

ill. 6.11. Normal-mode noise appears between the line and neutral. Common-mode noise appears between ground and the line and neutral.

Noise in Linear Power Supplies

Let’s consider the linear power supply first. ill. 6-12 shows the same diagram that we used earlier. Notice that across the input, we have put both normal-mode noise (the little impulse in the circle) and common-mode noise (the impulses on both line and neutral).

ill. 6-12. Noise across the transformer windings of a linear power supply.

In a later section, we will take up the subject of transformers in greater depth. For now, let’s take a simplified look at the common-mode noise that is simultaneously appearing across the primary of the transformer. These impulses are 1800 out of phase and will cancel each other out. For reasons we will explain later, this is an imperfect process and there isn’t a complete cancellation. The resultant of this is a normal-mode impulse that appears on the secondary of the transformer.

In other words, the common-mode impulses that move toward the center of the transformer primary winding cancel each other to a large extent. What is not cancelled is transferred to the secondary and appears as a single normal-mode impulse. From this, we can see that linear power supplies are nearly immune to common-mode noise by virtue of the transformer input to the power supply.

What about normal-mode noise? Clearly, if an impulse of sufficient magnitude enters the power supply on the live side, damage will be done. However, we must remember the components that would be subjected to the impulse first. The first line of defense is the transformer. Transformers of the type used in linear power supplies have almost no ability to attenuate normal-mode impulses. The inductance of the windings can change the waveform of the impulse, but the overall magnitude will be unchanged.

The next line of defense is the bridge rectifier. If the polarity of the impulse is not in phase with the conduction of the rectifier, the impulse will not be able to pass through; it will be dissipated as heat if this voltage finds a leakage path through the components. If the magnitude of the impulse is large enough, it could exceed the junction breakdown voltage of the power diodes. This would take a lot of joules of energy.

The filter section and voltage regulator are rugged components designed to handle the stress of both high currents and heat. They have large energy-absorption capacities. In short, the linear power supply is well able to handle all but the most destructive normal-mode impulses.

Noise in Switching Power Supplies

What about switching power supplies? ill. 6-13 shows the diagram we used earlier along with normal-mode and common-mode impulses. Notice that we have added a block in the circuit called “Logic.”

ill. 6-13. Noise across a switching power supply.

Let’s start with normal-mode noise. Normal-mode noise is trying to complete a path from line to neutral. As it tries to find this path, it appears across the terminals of the power supply just as it did in the previous case. and , again, we see that those components which are well able to absorb this energy will block the path of normal-mode noise. Unless the magnitude of the noise is high enough to break down power diodes and capacitors, normal-mode noise will have little effect on the PC.

Common-mode noise is another matter. While normal-mode noise is trying to complete a path from line to neutral, common-mode noise is trying to complete a path from neutral to ground or from line to ground. While the power supply would be the first casualty from normal-mode noise, this is not necessarily the case with common-mode noise.

A PC’s susceptibility to common-mode noise is a highly controversial subject. In order for damage to be done, a voltage must appear across a component that is high enough to break down whatever insulation there is, so that a current path is established. As we stated in our section on “grounding,” unless this path exists, we would expect the PC to float with respect to any noise appearing between line or neutral and ground. This current path would almost certainly be through the power supply.

The debate that rages concerns the probability that noise could flow through the high-frequency transformer, or through other paths which are made up of stray capacitance. If it could, then, obviously, common-mode voltages would appear between logic ground and the supply-voltage pins. In some systems, there are decoupling capacitors and other methods which attempt to isolate logic ground from power ground. Most PCs have these two ground systems tied directly together.

Electricity behaves in strange ways, and high-frequency noise can find coupled paths that designers might never have dreamed of. The high- frequency transformer is designed to pass a signal that, by its nature, is like a noise signal. The probability that unwanted noise would see this transformer as a coupling capacitor is high.

All components have some inherent leakage current. The output electrolytic capacitor changes characteristics dramatically as frequency increases. A switching power supply is a tightly packed circuit. In fact, space is at a premium, which is one main reason why the industry moved in the direction of the switch. But, ever since the switching power supply’s inception, stray coupling has been a problem. and , stray coupling is not confined to a neat band of interference. Low-frequency as well as high-frequency noise can cause problems. No one can design around such a broad range of potential noise.

Two things are clear regarding switching power supplies and noise, however. They are much more vulnerable to common-mode noise than they are to normal-mode noise, and a damaging common-mode impulse will be magnitudes smaller than a damaging normal-mode impulse.

Switching Noise

It may have occurred to you that all this switching between 20 kHz to 100 kHz just might generate some noise. The fact is that noise from the power supply, as well as the clocks inside the logic, is a problem for PCs. That is why PCs are now subject to strict requirements from the FCC, due to their tendency to emit EMI. EMI stands for electromagnetic interference. In recent years, it has replaced another term which had essentially the same meaning—radio-frequency interference or RFI.

In ill. 6-10, we saw the rather strange current sine wave of the switching power supply. This waveform is made up of odd harmonics which are reflected back into the line. This harmonic content can appear as noise on the input of other devices.

A PC, then, is a source of EMI from both its clocks, and from the oscillator and inverter in its power supply. and it’s a source of noise to other devices connected along the same line. Let’s take this a step further. In Section 2, we made the point that in a three-phase system, the current of each phase adds algebraically to a vector sum of 0. Odd harmonics, how ever, actually add mathematically. But the difficulty does not just stop there. The current needs of switching power supplies, along with a lagging power factor, can further unbalance a transformer. it's not uncommon to see neutral currents higher than the sum of the three individual phase currents, when the load is that of many switching power supplies.

This type of circumstance is not unusual. In fact, it's typical. Now, power lines become like antennas. The switches are sources of multiple kinds of noise. and the neutral line carries large amounts of current. If we add a simple ground loop, we have a receipt for constant soft errors, excessive power-supply heating, and the eventual breakdown of solid-state components.

Beyond Power Supplies

What is beyond power supplies? Why, logic, of course—the little computer chips scattered all over the motherboard. Those integrated circuits are the ones we see pictures of all the time. You know, the little black speck sitting on the end of someone’s finger?

We’ve made a big deal out of the ability of an IC to withstand noise voltages or induced currents from ground loops. The fact is that the PN junctions of an IC can't hold back more than a few dozen volts, and usually substantially less. Current flows that are in excess of a few dozen milliamperes (thousandths of an ampere) can't be tolerated.

The reason for this is not surprising when we consider the size of the circuits. A typical gate inside an IC is on the order of the width of a human hair. A half dozen or so gates might easily fit into the period at the end of this sentence. The distances between current-carrying paths are measured in microns (one millionth of a meter, or 1 x 10^-6). A distance of a few microns between the logic voltage supply and ground is typical. it's these tiny distances that make IC chips so vulnerable to the high-energy impulses coming through the power supply or discharging through the chassis, especially the common-mode potentials.

Noise immunity is another factor. Noise immunity is the chip’s ability to distinguish data from noise. For TTL-type ICs, this is only 400 to 500 millivolts, or roughly 1/2 volt. For the more rugged logic families, this can be as high as 3.2 to 6.5 volts, but this better immunity often comes at the expense of operating speed. Of course, a chip is most vulnerable to noise when it's in transition from one logic state to the next. This is why soft errors can seem so bedeviling. The noise impulse must occur at exactly this transition time of the chip and in such a way that the noise event actually interferes with the chip.

NEXT: Transient Suppression Devices

PREV: Grounding

All related articles   Top of Page   Home