Grounding

Every time we flush the toilet at my house, the waste water flows into a cesspool. The cesspool is constructed of cleverly aligned brick which gives the support of a hard structure while allowing the water to gradually leach into the soil around it. Someone, somewhere along the line, must have picked up on this as an analogy for electrical grounding. Unfortunately, the analogy is totally false.

And yet, what a concept! Imagine being able to drive a copper rod into the ground, and having the moist spongy soil greedily gobbling up unwanted noise and voltage spikes that have invaded the wiring of our home or office. The ground becomes the repository for unwanted electricity—-a path of no return for the gremlins looking to garble our granules and deceive our data.

Unfortunately, like many fantasies, this scenario has no—repeat—no basis in fact. The opposite is sadly the case. Misbegotten beliefs about “ground” are a common source of the problems that bedevil computer operations. Let’s throw a grenade and explode this myth once and for all. From an electronic standpoint, earth is not ground.

Since the early days of electrical power distribution, the earth has been used as a ground return path for lightning, fault current, and other utility related occurrences. Proper earth grounding is important in the construction of buildings, swimming pools, and overhead power lines. But what these techniques do is use the earth as a grounding path. This is logical when we remember that lightning uses the earth as part of a current path.

But earth grounding does not have the magical capacity to absorb electrical energy. A path for electrical energy? Yes. A sponge that soaks it up? No. Computers run fine in outer space or aboard planes. The Space Shuttle does not drag a ground wire behind it that is connected to the earth. So just what the heck is this grounding?

Safety

ill. 5-1A shows a safety hazard. The ground wire has been eliminated from the circuit, and fault current could flow from the equipment cabinet to ground through the first person who touches it. In ill. 5-1B, we see that fault current will flow through the extremely low-impedance path of the “safety ground,” thereby tripping the circuit breaker. If we think about Ohm’s law, we will see why this happens. Remember “E,” or voltage, will equal resistance (R) times current (I). If the resistance is very low, maximum current will flow. The impedance of the ground wire is such that enormous current flows through in an instant, thereby tripping the breaker. A human body, on the other hand, has a high resistance, and through a high resistance, less current will flow. But there will still be enough current to cause death, while possibly not being sufficient enough current to trip the circuit breaker. That’s why it's so important not to use radios and hair dryers in the bath tub. Since these devices don't have a ground wire, and the water itself provides an alternate path, the only current path is the most dangerous path.

Two potentially dangerous circuits. (A) A potential safety hazard exists should someone close the fault path by touching the cabinet. (B) Fault current will flow through the safety ground, tripping the breaker immediately. No touch hazard exists.

ill. 5-1 clearly shows the advantage of proper safety grounding. Safety must not be sacrificed in any wiring scheme. That is what proper installation of the green wire (in electrical wiring) is all about. But let us leave safety aside for a moment so we can talk a little theory. As the discussion continues, however, safety grounds (those green wires) will pop up again.

Floating Grounds

Let’s imagine a power system that is not referenced to an earth ground. In fact, let’s imagine that it's referenced to a 20,000-volt power line. What? How can we do that? ill. 5-2 shows what this might look like. For many, this might appear to be the first step into the twilight zone. But this deceptively simple drawing shows a very important concept.

This computer is fed by a system that floats with respect to the earth. As long as the voltage relationships don't change with respect to ground, the potential between ground and earth might be most any value.

The computer has no idea that it's at 20,000 volts with respect to the earth. It only cares about what its relationship is to ground. In this case, we have assumed that the computer is using a standard type of power supply that converts 120-volt AC power into usable DC power for the logic circuits. As long as the power supply sees this voltage, everything will operate normally. The logic circuits will be referenced to ground the same way. No problem.

Think for a moment. Let’s take a mental journey. Take the diagram off the page and put it inside an aircraft. We will assume that our working computer is flying from New York to Los Angeles. On the way, we fly through a variety of storm systems and clouds. Because of this triboelectrical effect, the structure of the plane picks up a charge of 20,000 volts with respect to the earth. Even though we are aware of this, we find the computer working just fine.

When we land, our plane keeps its charge because it's insulated from earth by rubber tires. Unfortunately, the ground crew forgets to equalize this buildup of electrical charge through the use of a metal strap or chain, and the first passenger off the plane dies a horrible death as the charge passes from the off ramp to the tarmac through his body. But that’s another story.

The point is that the PC worked fine no matter what its grounding system was referenced to. There are safety issues here that we are obviously ignoring. But the point applies to many other types of equipment that are around us and in use every day. Consumer electronics, like stereos, TV sets, and VCRs, are not connected to even a safety ground. They are what’s called double insulated. We know this because the round ground-wire prong on the power plug is missing. This means that the chassis of the device can float with respect to the building’s power ground. The touch-voltage danger does not exist because of the double insulation.

We might ask, why don’t we have a floating ground for our PC if all of this is true? This is a good question and we will answer it as the section progresses. For now, let’s frame another question while we are on the topic of other electronic devices. Why do manufacturers recommend that the chassis ground of a record player be tied to the chassis of the stereo amplifier? The answer to both of these questions has something to do with a technique called single- point grounding or signal referencing for equalization. Let’s see what this is all about.

Single-Point Grounding

Every building has a built-in floating ground. ill. 5-3 shows this in simplified form. This is a safety precaution that equalizes any potential differences between neutral and ground. If we look at Ohm’s law, we can determine that no voltage can exist between ground and neutral because of this bonding. If there is no resistance, there can be no voltage drop. So now, even if the power ground varies with respect to some other grounding point somewhere else, the power system for the building is properly referenced to a single point.

ill. 5-3. The service transformer has the neutral-to-ground bonded on the secondary. This point allows everything connected to it to be referenced to this single-point ground.

This is a point of vital importance. We can reestablish this neutral-to- ground bonding point anywhere we desire by simply placing a transformer at the desired location. This is consistent with the National Electric Code and proper safety practices. A good way to visualize this is to think of the PC as a boat (ill. 5-4). If the neutral and ground are tied together, the computer’s ground potential floats with the nearby power-ground single-point reference. If, however, we tie the boat to the ocean floor by referencing our PC ground to some other point with respect to the power ground, we may find ourselves bailing water.

ill. 5.4. An analogy. (A) Our boat is referenced to the top of the water and therefore floats on the waves. (B) Our boat is referenced to the ocean floor and a wave can swamp our boat.

Let’s expand this concept a bit further. ill. 5-5 shows the electrical distribution of a typical personal computer system, with CPU, printer, modem, etc. Each of these devices might have its own power cable. At 60 Hz, each device’s chassis is at zero potential with respect to the others because they are all tied to the neutral-to-ground bonding point of the building service trans former. Because of this, the entire group of devices float together with respect to ground. This means that voltages are equalized between cabinets. The happy consequence of this is that we don’t get a severe shock when we touch two cabinets at once.

ill. 5.5. Since each device is ultimately tied to the single-point ground, they float together.

Grounding Noise

Single-point grounding is a valuable tool in the fight to keep unwanted signals or noise from getting into our PC system. But it must be applied properly or it will do us absolutely no good. We may have a very safe system, but it may be subject to chronic malfunctions. Relying on the single- point ground at the building service will not be sufficient to accomplish what we want. The goal of any critical computer installation, whether it be a PC or a mainframe, should be a noise-free environment.

You will notice we use the term, “Power Frequencies.” Of course, it's no mystery that the frequency of power is 60 Hz, or 60 cycles per second. At noise frequencies, something much different happens. We touched on this in an earlier section, but in the context of grounding, we ought to look at it again.

At the frequency of power, the current path is only through the resistor, but at 10,000,000 cycles per second, the circuit acts totally different. The resistance has stayed the same, but the actual wire in the circuit, or what are called the leads of the resistor and capacitor, have become inductors of some value. That’s right. Straight wire will behave at noise frequencies just like a coil of wire does at power frequencies. The capacitor which appeared to be an open circuit at power frequencies has become a short circuit at noise frequencies. If it's truly a short circuit to the 10-MHz signal, nearly all the current will flow through that path. An experienced electronic student would have seen that the lead inductance in the circuit of ill. 5-6B has added a large amount of impedance to the circuit. This is illustrated in ill. 5-6.

ill. 5.6. At 10 MHz, the circuit elements appear totally different from what they are at 60 Hz. (A) Circuit at 60Hz. (B) Circuit at 10MHz.

Let’s concentrate on the concept of lead inductance. Look back at ill. 5-5. There is one electrical path that we did not show. That path is the path that data must take between each item of the system. ill. 5-7 shows us this concept both in block form and as a circuit. We still have our single- point ground. At power frequencies, the green wires are at a much lower impedance path and the cabinets still float. The little tiny ground wire that runs through the printer cable is a high-impedance path and very little ground current will flow in it.

However, at noise frequencies, the path through Li can become a higher impedance path if the ground wire is unusually long. In our diagram (ill. 5-7A), this noise has two paths to choose from. All that wire running clear back to the neutral-to-ground bonding point of building service trans former looks like a high value of inductive reactance. The shorter path through the printer cable to the CPU may be of nearly equal or lower impedance and , therefore, will carry some of the noise. The fact is that even if the power ground path were a lower impedance path, some current would still flow in the printer cable. This unwanted current can couple signals into nearby wires, which might be interpreted as data—causing print errors. To prevent this, the power ground system must be held to an extremely low impedance.

ill. 5-7. If the path through the power ground is at a higher impedance than the data line, then more ground-noise current will flow through the data line. (A) Block diagram. (B) Circuit schematic.

Let’s consider the voltage drop across the inductive reactance. At power frequencies, the low impedance of the wire will have a low voltage drop across it. As frequency goes up, the value of “Z” goes up, and the “E,” or voltage drop, across the inductance increases. This voltage offset appears as a noise voltage between the two cabinets.

How do we equalize the voltage and provide for an alternate current path? There are a couple of things we can do to keep noise from using data lines as a return path. One thing we could do is tie the chassis together with an additional ground bonding strap. This ground strap should be short and wide. Because of the skin effect at higher frequencies, a braided copper strap has enough cross-sectional area to present a much lower impedance path than a data cable might.

This alternative might be somewhat cumbersome. Who wants to drill holes in modems that have no grounding lugs? Another alternative is to buy a power strip. Most of us already use power strips just because we need the additional receptacles to plug things into. But we also gain the benefit of using a single-point ground that is very near the cabinets. This serves to equalize voltages between all of the devices plugged into the power strip. The short lengths of the individual power cords, themselves, are impedances too small to worry about. and at noise frequencies, everything plugged into the power strip is floating without respect to the single-point ground at the building service.

Two points must be made here. First, we are not talking about those power strips that have surge suppressors built into them. We will cover those later. We are talking about a simple power strip with multiple outlets, that is all. Secondly, while a power strip will equalize noise voltages between cabinets through their respective power grounds, this does not always make them immune to any noise that might appear between ground and neutral or between ground and the line. But it's a desirable goal to equalize any potential differences between the PC and any peripheral so data communication cabling remains free of interference.

Remember our floating boat. Computer chips that process signals are expecting to see a logic level with respect to ground. If we present the chip with an offset voltage to look at, in addition to the logic signal, the intelligence is no longer floating above ground but may be swamped out as the ground-noise impulses rise and fall. If this offset voltage is large enough, it may exceed the breakdown voltage of the data communication chip itself and will bridge the insulating barriers within the chip causing its destruction.

Shielding

Whenever we think of keeping data communication wiring free of unwanted interference, we are likely to think of shielding. The common wisdom is that a shield will prevent noise from being coupled, either capacitively or inductively, to the wire or wires carrying the intelligent signal. By shielding, we mean the practice of wrapping a mesh of wire around an inner conductor, or, in some fashion, the enclosing of conductors with some other conductor. Coaxial cable and wires run inside a metal conduit are examples of this.

The unfortunate fact is that as frequency increases the “shielded” effect starts to disappear. ill. 5-8 illustrates this effect. We see that a noise impulse can be capacitively coupled between a water pipe and the shield of a conductor, and the shield and the conductor become a low-impedance path at some noise frequency.

Curiously enough, current flowing in our water-pipe illustration induces an opposite current flow in both the shield and the conductor. However, the induced flow in the shield sets up an induced flow of its own in the conductor (ill. 5-8B). These two currents are 1800 out of phase and will tend to cancel or “buck” each other. But, some residual noise current will be left from this imperfect cancellation. This “bucking” process does not help with capacitively coupled noise, however.

ill. 5-8. Noise coupling versus shielding. (A) Noise is capacitively coupled into the conductor. (B) Two induced currents are bucking’ each other and cancel in the conductor.

We might be tempted to solve both of these problems by grounding one end of the shield. Of course, in normal practice, the shield would be terminated at the chassis of the computer on one or both ends of the cable. Grounding only one end of the cable adds an additional problem to the situation, as we can encounter standing waves, unwanted resonances, and hot spots that are caused by the actual amplification of signals (at multiples of the signal’s quarter wavelength). A detailed discussion of why this hap pens is beyond the scope of this guide. There is a simple answer, however. Ground the shield at both ends. If the cable is very long, don’t hesitate to also ground it at multiple points along its length.

What we are attempting to do by using this method is reduce the ground-lead lengths, so as not to present a high-impedance path to a noise signal. it's much more desirable to give the noise signal multiple short paths which complete a loop circuit, than forcing it to appear along the entire length of the cable. In actual practice, this is seldom done because of misunderstandings about ground-path impedances, and it's sometimes a difficult solution to implement. The typical response to a data-cable corruption problem is usually to lift one or both grounds so that the ground noise can't reach the chassis of the affected equipment. Of course, this often makes the problem worse since the shield has no path to ground and acts as nothing more than a capacitive coupling device. The use of multiple bonding points to a common ground plane, like building steel, is usually the answer. One creative solution, using this idea, is to ensure that every equipment chassis connected by a data cable is also connected to the same column or building steel member. This is a more practical way of equalizing the potential differences between PCs or terminals that are scattered throughout a building. Then there are baluns, traps, and filters available as well, that can be used to help solve these problems. and there is always the option of finding the noise source and eliminating it, but this may be easier said than done.

As we stated earlier, multiple short grounds connected from the shield to the ground plane can help reduce the problem of noise being capacitively coupled into a shield. This technique is the formation of a ground loop. The use of ground loops in grounding shields is one of the few instances where ground loops are beneficial. Usually they cause problems, as we shall see.

Ground Loops

There are a number of ways that we might define ground loops. Basically, a ground loop is formed when a potential difference exists between any two places that are, by definition, at ground potential. Let’s start with a simple diagram of this. ill. 5-9 illustrates a simple ground loop. In the drawing, we are using a ground rod to make a point, but the loop could be formed by any two grounds of differing potential. ill. 5-10 shows this as a schematic.

ill. 5-9. A potential difference appearing across the two separate grounds will force current to flow through the data line.

ill. 5.10. This is a schematic of the sketch shown in ill. 5-9. Here we see R1 as the resistance between the two grounds. R2, then, is the resistance through the chassis of the two PCs and the data line.

There are a number of important points to be made about these diagrams. First, if we had one ground connected to each PC, there would have been no loop current flowing since voltage can't develop across a near-zero impedance. The obvious solution to this problem would be to get rid of one of the ground connections.

This tells us something about ground loops. A ground loop may be formed when there are two or more different ground connections in a circuit. This circuit might be also formed between boards inside a PC or in a local area network.

Another point to be made is that driven-rod earth grounds are never appropriate as a solution to ground noise. They are a source of ground noise! However, there are situations where a driven-rod ground is appropriate, like on the secondary of a building transformer, e.g.,. But, to arbitrarily bond a ground plane to a driven rod just for the sake of the cesspool effect hazards the potential of degrading system operations by injecting unwanted ground currents.

Notice that, consistent with Ohm’s law, the bulk of the current is developed through the path of least resistance, or lowest impedance. We are assuming that whatever is generating the noise voltage between the grounds is capable of delivering the full flow of current we find from the calculations. (This energy is called a joule, which is a unit of energy of work that is related to watt-seconds.) But, the noise source might not have enough kinetic energy to release a current flow of the kind we might expect, or, possibly, the current flow might only last for nanoseconds or picoseconds. Also, we are using Ohm’s law here in its simplest form. For the sake of simplicity, we are not taking a number of complicating factors into consideration. Nevertheless, the flow of anything like 10 amperes of current through the chassis of a PC and a data line is surely a disaster. It will clearly override data signals, break down insulation, or induce sufficient energy into nearby circuits to cause hardware failure.

We have presupposed the values of voltage, current, and impedance. The fact is that the voltage and current are not the only concerns here. The voltage drop caused by the impedance of a ground loop can cause serious problems. A noise voltage offset of several hundred or several thousand volts is likely in the real world. Impulse values that large are injected into signal circuits only as they appear across high-impedance differences in the circuit. This may hold the current to small fractions of an ampere. But any current bridging a microcomputer chip will ruin it, whether it causes a puncture hole that is a micron in diameter or the total destruction of a circuit board.

We might attempt a solution to this problem by adding a grounding strap to equalize the potential of the two cabinets. By doing so, we divert some energy away from the data line by providing a lower impedance path for current to flow. We could calculate this addition by adding another parallel branch in ill. 5-10 to see what current would flow through this bonding strap. This approach will not reduce the voltage appearing across the data line, however. Our calculation, using Ohm’s law, would have to assume that the noise source was capable of delivering additional current. The reason for this is that in a parallel circuit, the voltage appears simultaneously across each parallel element. Our bonding-strap solution, how ever, might be able to divert enough current through its low impedance to significantly reduce the exposure of the data line to unwanted current flow.

A better way to think of this is to think in percentages. In ill. 5-10, a much greater percentage of noise current will flow through the path where impedance is expressed as 1 ohm than through the 10-ohm path. If we add a low-impedance bonding strap, the percentage of energy flowing in the data- line path will be significantly reduced. This is nothing more than a substitute type of single-point grounding.

To summarize this, remember that whenever there is some impedance existing between two points, voltage can appear across these points and current can flow. When it comes to grounding, it should be a goal to hold all grounds to the same potential by not allowing this impedance to develop. Any impedance between two differing points along the ground plane should have a significantly lower impedance than the alternative paths that could divert current through sensitive circuits.

Maybe now we can see why computers don't have a two-pronged plug on their power cords. If our PC is to communicate with the outside world, we must have some way to reference the signals between devices. Like the example of a stereo amplifier and a turntable, we can't allow the grounds of two devices to move independently of one another, or we will develop noise voltages that interfere with the desired signals.

Loops Due to Wire Length

Ground loops can also develop because of the sheer length of a circuit wire. As frequency goes up, the inductive reactance of a piece of straight wire begins to increase. The flow of high-frequency current sets up a magnetic field that induces an opposing current in the wire. This has the effect of raising the overall impedance of the line to high-frequency noise. At power frequencies, we have no problem plugging a PC into a power outlet that terminates at the power source 50 to 200 feet away. This distance, however, looks like a resistor of high value to a noise signal of, say, 10 MHz. Some of the things that may contribute to this are loose conduit fittings, ground terminations that are not tight, or small-sized ground conductors. Typically, we would not expect normal “green-wire” grounds to show much impedance at any frequency, as long as the length(s) of the power cord(s) are reasonable. But common-mode noise problems are usually caused by a combination of things that work together to develop excessive voltage drops.

ill. 5-11 shows how this might work. Because of the long run of wire from the grounding point of the PC and the entrance of the neutral wire into the PC, a voltage drop occurs across this impedance. This voltage simultaneously appears from neutral-to-chassis ground and may find current paths through the stray capacitances that exist in connectors and along phenolic-board traces, or worse yet, across ICs. If no stray paths are found, obviously the PC will float on the common-mode noise and no harm will be done. The detrimental effect of common-mode noise to a single PC, where no ground loop through an external device exists, is the subject of some controversy in the industry. Will it float or will it be affected? There is evidence to support both notions. and some equipment designs are more immune than others.

ill. 5-11. Because of the high impedance of the connecting wire, a ground loop is formed with the neutral wire. Thus, loop current may find a path through the stray capacitive paths that exist inside the PC.

There is disagreement about the ability of long neutral-to-ground loops being able to develop significant noise voltage. Many experts contend that the sheer size of the wires is enough to present low impedances, even at high noise frequencies. and yet common-mode noise voltages do cause processing errors and equipment damage. There are always those locations where strange unexplainable things are happening. The solution almost always is found through proper bonding, single-point grounding, or the elimination of ground impedances.

Naturally, had the lead length been shorter, with good integrity, no ground loop would have existed. Just because the neutral and ground wires are bonded at power frequencies does not guarantee they are bonded at noise frequencies.

Any time that ground wires have excessively long lengths, we can expect ground-loop currents to flow or high offset voltages to be present. This is the case with one of the most commonly misunderstood devices, the IG (isolated ground) receptacle. They are those receptacles having the funny orange plastic faceplates. They are misused more often than driven ground rods. ill. 5-12 shows the construction and application of an IG receptacle. When the IG receptacle is installed according to the National Electrical Code (NEC), the ground pin must terminate at the appropriate power-source transformer. On its way there, it must not bond to any enclosures, including the enclosure for the receptacle itself. Enclosures are bonded with a separate wire, or sometimes with the conduit. In effect, this method uses the conduit as the safety return path for fault current. This is a poor idea because of its reliance on the tightness of mechanical fittings and the continuity of the metal, which might be broken if used with plastic fittings or PVC.

If the power-source transformer were placed close to the computer devices, IG receptacles might be a reasonable idea. This is usually not the case, especially where PCs are concerned. Often the IG wire does nothing more than act as an antenna for noise, which is impressed upon its lead inductance. While the enclosure ground must go the same distance, its multiple bondings, with enclosures, can act as a noise return path for loop currents which might otherwise appear as common-mode noise at the PC. In addition, the IG receptacle method forces equalization to occur along the entire distance of the IG path. Safety grounds will come together either at the nearest receptacle enclosure or at the nearest panel board.

ill. 5-12. The isolated ground receptacle, when installed according to the NEC, must terminate the ground pin at the power source and must not touch enclosures.

Another typical problem is the proper installation of the IG receptacle itself. There is much misinformation about how the receptacles are to be installed, and , in addition, many electricians have no understanding of signal referencing to begin with. Often times, signal lines are mistaken for dedicated lines, a branch circuit dedicated to the computer. But, this doesn’t necessarily mean it has an isolated ground. The manufacturers are to blame for much of this. Some of the printed material that accompanies the receptacles shows the isolated ground wire going off to a driven-rod ground or to a water pipe. This is a violation of the NEC and a safety hazard. IG receptacles are not a substitute for proper grounding of computer equipment.

While we are on the subject of the National Electrical Code (NEC), it might serve to point out that the code allows a smaller wire to be used for the safety ground than for the phase and neutral wires. This is a mistake from the computer’s point of view. As noise frequencies increase, the inductive reactance of wire increases. All things being equal, a larger wire has considerably less inductive reactance than one of smaller cross-sectional area. it's recommended that green-wire grounds be selected that are at least the same size as the phase wire to ensure a low-impedance path to ground.

Some Final Comments

Earlier, we made the point that common-mode noise is noise that appears between neutral and ground and between the line and ground. We can see from the previous discussion that ground loops create common-mode volt ages. As it's assumed that we want to keep noise out of signal circuits, the elimination of ground loops as a source of noise would seem to be desirable. The main principle behind this goal is equalization of voltages along the ground plane. If logic signals are tied to the ocean bottom, as in our earlier example, they will be swamped as the unreferenced ground level sweeps above the signal level. Circuits that are referenced to a ground at zero potential, with respect to one another, float on the ground plane no matter what its relative voltage might be in relation to some other ground.

This gets tricky as the frequency of noise increases. The electrical behavior of various components change, causing a divergence from safety grounding practices. While not sacrificing safety, we must see that impedance differences don't exist at noise frequencies.

Let’s look at this another way. ill. 5-11 shows us a ground loop formed by the long loop of wire that runs from neutral to ground. Let’s imagine for a moment that this bonding point is infinitely short; that right—short. If this distance is infinitely short, it has zero impedance at any frequency. If E = IR and R = 0, then E must equal 0.

Now, let’s imagine what happens as the loop gets longer. As the loop gets longer, two things begin to happen. The likelihood that noise will be impressed, either by direct physical contact or by capacitive or inductive coupling, increases. Also, the impedance of the wire to that noise signal increases, thereby creating a larger voltage drop across this impedance. As this happens, this signal voltage appears as a potential difference between the PC’S chassis and the neutral wire.

We might not be too concerned at this point since our ground loop might still have basically a short-circuit value of impedance. If current were to flow, it would surely flow through the neutral-to-ground circuit rather than through a data-communications cable or telephone line. At some point of increasing length, we reach values of loop impedance that cause damaging amounts of current to flow through the wrong paths. Meanwhile, the voltage appearing across the ground loop increases. This might exceed the breakdown voltage of the microcomputer chips or power transistors that are in parallel to the voltage drop, thus creating some new current path and destroying components in the process.

Imagine that the lines get even longer, so that, to the noise signal, the neutral-to-ground circuit has a very high value of impedance. Now other alternative paths that might exist inside the computer look to have a much lower impedance than does the power ground path. If a path were ever found, such as through a data line or through stray capacitance, then damaging current will flow through this unwanted path. As the neutral-to- ground connection moves further away from the computer, these problems grow worse. The value of the “length” impedance is enough to create a disruptive ground loop all by itself.

Hopefully, this mental journey has shown why low-impedance ground paths are so important, and why a short single-point ground is the only way to connect power to a computer system. The goals are to equalize chassis, prevent ground loops from developing, and keep the noise current away from the signal circuits by reducing the coupled voltage that appears across them. Low-impedance grounds should be short and thick. Even lengths over four or five feet can be a problem at high frequencies since these represent quarter wavelengths.

The basic themes that are laid out in this section will be touched on many times in the sections to come. We will find that sometimes, as in local area networks, there is little choice but to live with ground loops. But there are products and strategies that deal with this problem.

Grounding is like religion. Everybody believes ardently in it, but every one has their own doctrine that support those beliefs. The subject of grounding can be confusing and controversial. We have tried to lay out our discussion in a way that is consistent with the advice of experts in the field.

Let’s summarize by restating a few tips:

1. Whenever possible, connect devices to a single-point ground.

2. Attempt to equalize potential differences between cabinets through single-point grounding, using a power strip or supplemental ground straps between the cabinets.

3. Keep ground connections as short as possible in order to make the ground path a low impedance path.

4. Keep the neutral-to-ground bonding point as close to the computer system as is practical.

5. Use large conductors for green-wire grounds to ensure a low lead impedance, and avoid reliance on conduit as a ground path because of its tendency toward loosening, rust, and corrosion.

NEXT: PC Power Supplies

PREV: Voltage Transients

All related articles   Top of Page   Home