The simple, ubiquitous, pervasive word "ground" is tossed around by electronic circuit and systems engineers casually and frequently. Phrases such as "that point is grounded," "connect that to ground," and "don’t worry, it's grounded" are among the many often heard. Connecting to "ground" is often invoked as some type of cure-all for whatever ails a circuit, whether related to noise, safety, or performance.
Yet the term "ground" is frequently misused, misunderstood, or outright misleading in the context used—all of which can lead to system issues, problems and even failure. The problem with using the term "ground" casually and incorrectly, even when the meaning is hopefully understood by all participants by implication and experience, is that sloppy verbiage leads to sloppy thinking and even design mistakes.
Start with a fact: there's only one real ground, Earth, and then there are many other signals that are called "ground" but are not Earth-ground at all. After all, a battery-powered smartphone or even an airplane has no connection or reference to the Earth "ground," or anything electrically close to resembling it. The use of the term "ground" was legitimate when AC-powered systems or those with exposed metal (such as a chassis or rack) had to be connected to Earth-ground for safety, in case there was an internal short circuit, high-voltage circuitry to a user-accessible panel.
By "grounding" the metal frame, the dangerous current would instead have a low-resistance path to the Earth (a near-infinite source and sink of electrons) and not go through the user. In most cases, the flow of excessive current would also cause the fuse or circuit breaker to trip, thus further protecting the user and the equipment. Similarly a lightning strike would have a directed path to Earth-ground and thus generally cause less damage.
An example with AC-powered hand tools illustrates the situation, when electric drills had metal bodies (cases), and these bodies were connected to an AC-line ground via a three-wire line cord (and the user had to hope the third wire for ground was actually connected to Earth-ground). If there was an internal short of the motor's wiring to the drill body, the current would flow to ground and the user would be safe (assuming the ground really was connected to ground).
A few decades ago, the metal bodies of AC-powered hand tools were replaced by high-strength plastic bodies, with no danger of the tool body becoming "electrified." These two-wire, ungrounded tools became the standard, and were called "double-insulated" due to the combination of insulation of the wires plus that of the tool body. Shocks and fatalities due to metal-body tools with false ground connections dropped dramatically, as "double insulated—so no need for ground" became a major selling point.
Along with the transition from line-powered electronics to battery-powered devices, the term "ground" morphed into a shorthand phrase for any set of points in the physical circuit which are (in theory) at the same potential with respect to each other, even if they have no relation to an AC power line. This equipotential zone is also often assumed to be at 0 V, but that, too, is a misconception.
The reason is that voltage does not exist at a single point; instead voltage is more properly known as a potential difference and only has meaning when it is defined between two points. In a battery-operated circuit, for example, one side of the battery—usually the negative—may be called the ground, but it is more accurate to say that any voltage should be measured with respect to that reference point.
Safety versus signal ground
While we have talked about ground in terms of high voltages and safety, the term is commonly used with respect to low-voltage signals, not power lines. It is in this context where it is a complete misnomer, as there is no Earth-ground anywhere in the application. Also, most designs that have (or require) a true Earth-ground also have one or more signal grounds—and that is where the confusion begins.
The more appropriate term for signal ground in most cases is signal "common." This is a single point or plane that is the reference against which other signals are measured or referred, and to which many circuit points are connected or switched. In many designs, there are many signals at different points that must have zero potential difference with respect to each other, and so are connected to this "common." Some designs even have multiple such commons, which may be eventually connected to each other, or they may be deliberately isolated from each other.
Using the term "ground" and its symbol, Figure 1, for what is really a signal common-point can lead to the design problems. A perfect ground or common would have zero potential difference anywhere across its physical implementation, and thus there would be no current flow between these points. Yet the reality is that no ground is perfect, and there is always some DC resistance and RF impedance. Therefore, there will be a potential difference—voltage—across two points of the common connection, and current will flow between these points.
How much voltage across a common? It depends on the situation, but it is not unusual to have tens or hundreds of millivolts, or even several volts, across two points of this common, especially if it exists across a large circuit board or between two boards. As a result, the many implicit assumptions about this common and its use as an equipotential reference plane are not quite true. This deviation, even if small, can affect sensitive analog circuits and cause errors in digital circuitry as well, due to what is called "ground bounce" and "ground noise." Just because various points are shown on the schematic diagram as being connected to the same "ground" does not mean they have zero voltage between them. The real issue is how small or large that difference is in the actual circuit layout and implementation.
"Ground rules" and when to break them
There are many regulatory mandates and best-practice guidelines related to the use of so-called grounds. The problem is that for each such rule or guideline, there are many legitimate and sanctioned exceptions, usually required to meet some sub-rule of the electrical code, or for proper circuit performance. Therefore, every statement about how these so-called grounds should be connected and used must be carefully reviewed. Among them:
• All "grounds" (actually, circuit "common points") should be connected together at one point, in a star configuration, Figure 2 (very easy to do in concept, but often very hard to implement in the actual PC-board layout, unfortunately). The idea is to minimize impedance between subsections or subcircuits, and establish a better equipotential point for all these commons. There are exceptions, such as when one or more of these sections must be electrically isolated from the others for system performance due to the presence of high voltage (even DC high voltages such as from a series string of batteries), or as a "floating" sensor (not referenced to any common point).
• AC-power grounds and signal commons should be connected: again, usually true, except when it is mandatory that the high voltage be absolutely prevented from reaching the user, such as in a medical device where an internal failure could put line voltage on the patient. In such cases, the AC-power supply and the electronic circuits must be isolated from each other, using techniques such as a transformer barrier, optical isolation, or captive coupling (Reference 1).
• There are many cases where the so-called ground has such a large potential difference or voltage drop from one end to the other that there is substantial current flow through this presumed equipotential path. For example, there is usually relatively high DC resistance and AC impedance between the two ends of a shield in a cable (especially at higher frequencies), Figure 3. As a result, the signal source and sink "grounds" are really at different potentials, which is noise-inducing and even potentially dangerous. The solution is to break this "ground loop" (as it is known in the design work), usually done with a transformer, although optical methods are sometimes used. The same problems can even occur on a localized circuit board.
Don’t be lulled into complacency by the indiscriminate and casual use of the term "ground" when the point is really a signal common. Clearly identify which points are connected to Earth-ground (if any), and what the various common points are. Also, realize that even though these commons are connected in the schematic, the real-world connection will have some impedance, and the situation becomes more challenging when analog and digital signals are in the same circuit, which is a normal situation.
Understand the standard approaches to actually constructing a topology for the common, and when there are exceptions to these guidelines. Look carefully at vendor component data sheets and application guides (Reference 2) for the "grounding" approach they use when testing and specifying devices, as any variation may degrade the performance you achieve yet be frustratingly inexplicable at first.
- The Why, Where, and Hows of Signal Isolation
- Staying Well Grounded (Analog Devices, Inc.)