You use galvanic isolation to separate functional sections of electric systems so that charge-carrying particles cannot move from one section to another—that is, no electric current can flow directly from one section to the next. The sections of the system can still exchange energy and information, however, using capacitance, induction, or electromagnetic waves, as well as optical, acoustic, or mechanical means. The technique finds use in situations in which two or more electric circuits must communicate, but their grounds may be at different potentials. It is an effective method of breaking ground loops by preventing unwanted current from traveling between two units sharing a ground conductor. You also use galvanic isolation for safety considerations, preventing accidental current from reaching ground—a building’s floor, for example, through a person’s body. The word “galvanic,” in fact, means having the effect of an electric shock. Knowledge of isolation and its uses may galvanize you into designing it into all your future systems.

The technique should be an integral part of any design because it prevents ground loops, minimizes noise, and, most important, keeps users safe. A lack of isolation can have disastrous results. For example, Sandy Templeton, director of isolator development and applications at NVE, relates that, on two electrical supplies lacking isolation between them, one supply’s ground currents can raise the potential of ground to 50V, melt the cable between two computers, and cause a fire (see sidebar “What is ground, anyway?”).

“You don’t know isolation is there, but it’s everywhere,” says Ahsan Javed, product-marketing manager for isolated products at Silicon Labs. He lists a variety of applications requiring isolation, including power supplies, lighting, defibrillators, and hybrid-electric vehicles. Yet, Javed believes that car buyers, for example, are more interested in miles-per-gallon figures or fancy chassis than whether the electrical components integrate isolation. “The end user is ambivalent to it because it is a safety component,” he says. “You don’t care about your air bag until you need it.”

In another application, medical electronics, it is imperative that you design your systems in such a way that no high voltage from the wall socket or power system can reach—and perhaps kill—the patient. Fortunately, US products must meet the strict certification guidelines of the FDA (Food and Drug Administration). UL (Underwriters Laboratories) also reviews product designs in conferring UL listings (see sidebar “Isolation glossary”).

In a less dramatic application, isolation also filters out noise in electrical systems due to analog and digital grounds and circulating currents (Reference 1). “Galvanic isolation can allow data to pass across two completely isolative ground references,” says David Krakauer, product-line manager of iCoupler products at Analog Devices. Consultant Henry Ott advises careful part placement and a disciplined routing strategy on the traces to prevent noise from one part of your circuitry from polluting signals in another part. Sometimes, however, you simply can’t obtain the placement you want. In these cases, Ott advises, use circuit-isolation techniques to ensure that noise from the outside world or other parts of the circuit cannot ruin the signals in your design. Poor ground-plane design can also cause noise. If your isolator allows you to cut up the ground plane in your system and you then run fast digital signals across that cut, the return currents for those signals must now seek the long way around the cut (Figure 1). This scenario is sure to cause EMI (electromagnetic-interference) problems (Reference 2).

When considering techniques for isolating signals, examine the difference between using an isolator as a linear part and as a digital component. When using an isolator in linear mode, you bring an analog signal level across an isolation boundary. When using it in digital mode, you simply bring a high or a low signal across the isolation boundary. You can bring an analog signal across the isolation boundary with discrete parts or by using isolation amplifiers. Alternatively, if you want to bring signals across the boundary as digital representations, you can use one of many digital isolators. A hybrid approach to the analog/digital-architecture decision employs the use of a delta-sigma modulator to turn your analog signal into a digital PWM (pulse-width-modulated) signal. The part sends that signal across the isolation boundary. Once the signal is over the boundary, you can either use the digital signal as is or send it to a lowpass filter and turn it back into an analog signal.

A discrete transformer approach is a traditional way of providing isolation (Reference 3). Transformers can transmit pulse trains to control an H bridge across a 10,000V boundary (Figure 2). Click here for a detailed schematic. You make the pulse transformers by passing three single-turn, 18-kV, UL3239-certified, FEP (fluorinated-ethylene-polypropylene)-insulated, isolated wire loops through a toroid. It is difficult to find off-the-shelf pulse transformers with 10-kV isolation and UL certification, so you may have to wind your own. A discrete design is also complex, continuously pulsing the FETs with on or off pulses. This approach is preferable, however, to designs that use gate capacitance to maintain the on- or off-state (reference 4 and  reference 5). In those designs, voltage swings on the FET drain push charge from the gate-to-source capacitance by virtue of the FETs’ Miller capacitance—an increase in the equivalent input capacitance of an inverting voltage amplifier due to amplification of capacitance between the input and output terminals. The resulting reduction in gate drive may cause the FET to enter the linear mode and burn up.

Along with discrete designs using transformers, analog optocouplers represent a time-honored—yet tricky—way to bring an accurate analog signal back across an isolation boundary. In optocouplers, the clear plastic inside the part can degrade and get cloudy, and the IR (infrared) LED inside the part can age and produce a lower output. One clever way of surmounting this difficulty is to use a dual optocoupler in your circuit. You use one optocoupler as a reference; the second transmits the dc level across the isolation boundary. Avago’s HCNR200 optocoupler has one LED and two photodiodes that you can use for a referenced servo system; Solid State Optronics offers a similar device (Figure 3 and Reference 6). Like all other isolators, optocouplers also have a phase delay, and you may have to compensate for it (Reference 7).

Texas Instruments’ ISO124 and Analog Devices’ AD204 op amps, both with built-in isolation, also provide approaches for bringing analog signals across isolation. The ISO124 uses capacitors formed by metal plates on the lead frame, and the package’s molding compound acts as a dielectric. The AD204 uses transformers rather than capacitors to provide the same results. The Analog Devices product also has a power section that sends power across the boundary for the other side of the device and for any other ancillary functions.

If you can possibly build your system to pass digital representations of your analog signals across the isolation boundary, then you can take advantage of a slew of newer digital-isolator parts, which represent a new trend in system design and data acquisition. “I still get people who want to talk about analog isolation,” says Tim Lafferty, a product-marketing manager at Texas Instruments. “I show them the ISO124, but I explain that [digital isolation] is really the way that the world is going.”

The idea is to put a serial ADC on the isolated side of your design and feed it isolated power that can also supply the op amps or signal-conditioning circuitry. You then use digital isolators to bring the ADC data back across your isolation boundary. “You might use these isolated amplifiers in a feedback system, such as a motor drive,” says Analog Devices’ Krakauer. “But, more and more, those feedback systems are going digital.”

“The ADC has worked its way closer and closer to the sensor bridge in almost every application,” says NVE’s Templeton, who notes that you can use analog isolation in voltage-controlled systems, such as those for thermal and pressure control, but delta-sigma modulators are replacing isolation op amps in even those applications. Keep in mind, though, that you must build an isolated power supply for the front end of a digital system, whereas the AD204/210 parts have built-in isolated power.

The oldest digital isolators, optocouplers, work at 50 Mbps; include analog optocouplers that drive integrated digital gates; and are available from dozens of companies, including Vishay and Toshiba. Several companies, including Fairchild and International Rectifier, make isolators for power-supply feedback. Clare and Crydom offer another class of isolation products, solid-state relays, for ac-line control. Several companies also offer digital isolators using capacitive, inductive, and other isolation methods. Vendors of these systems claim that they consume less power and fit into smaller packages than do optical isolators.

You should pay attention to the method by which digital isolators encode the input signal and carry it across the boundary. TI integrates two differential channels in its capacitive isolators because it is impossible to send a dc level across a capacitor (Figure 4). The company’s new isolator line includes well-matched die capacitors to provide common-mode rejection of 50-kV transient spikes. The ac channel takes the edges of the data stream across the capacitive boundary with no encoding, making the chips speedy. The second channel encodes the dc level of the input signal and sends it across two more capacitors as a differential signal. Decoding of this signal takes place in the receiver chip and provides the dc information if the signal lingers at 0 or 1V.

Also consider whether an asynchronous clock performs the encoding that brings the signal across the boundary. Several vendors warn that these “level-triggered” systems can change the shape and duration of fast pulses. In “edge-triggered” systems, the logic is not simply gating a free-running clock. Instead, the gates act as oscillators that emit a pulse within a gate delay of the incoming data and then continue to pulse until the input signal goes away. Simply gating a fast-enough pulse train across the boundary can be an effective approach. For example, Silicon Labs uses an internal asynchronous, 700-MHz RF signal to encode input data, and the pulse-width errors are in the nanosecond range (Figure 5).

Linear Technology has leveraged its module-building expertise to incorporate power and signal isolation in a module that transmits RS-485-bus signals across a 2500V boundary (Figure 6). This approach may appeal to many engineers because it is a repackaging of time-tested technology. The module does not represent the company’s only foray into isolation, however; years ago, it introduced the LTC1535 RS-485 isolator. The device uses capacitors on the lead frame to bring the signal across the isolation boundary. The new LTM2881 μModule uses transformers built into a PCB (printed-circuit board).

Some digital parts comply with high-level protocols, such as I²C (inter-integrated circuit), RS-485, CAN (controller-area network), and USB (Universal Serial Bus), and several vendors make parts that comply with these standards. For example, Analog Devices’ ADuM4160 iCoupler provides an isolated USB system, and other products provide isolated I²C interfaces. Texas Instruments recently introduced the ISO1050 for CANbus applications in cars or factories. A hybrid digital/analog isolation technique uses an isolated delta-sigma modulator followed by a lowpass filter. In this case, you convert an analog signal into a PWM digital pulse train, send it across the boundary, and then use a filter to turn the PWM signal back into the analog domain. Be aware that you must create an isolated power supply to feed the front end of the modulator. For example, Avago’s ACPL-785J optically isolated delta-sigma modulator (Figure 7) lacks a lowpass filter, but you can use a digital filter to get a representation of the analog signal, or you can put your own lowpass filter on the output.

Similarly, Texas Instruments’ AMC-1203 delta-sigma modulator has no built-in filter. The 16-bit, 10-MHz AMC1203 also feeds the AMC1210 digital filter to allow you to create an isolated resolver-interface circuit. The device provides 4000V isolation and comes with agency approvals. Analog Devices’ 16-bit-resolution, 20-MHz AD7401 sigma-delta modulator also can bring an analog signal across an isolation boundary and, in compliance with UL1577, can stand off 3750V for one minute.

Vendors use a number of techniques, including RF, optical, capacitive, transformer, and GMR (giant-magnetoresistive) sensing, to provide isolation in their products. The insulators the companies use are as varied as the number of dielectric compounds. According to Texas Instruments’ Lafferty, the properties of the insulation in a part may translate to its performance and long-term reliability. “One of TI’s objectives [for] digital isolators was to make isolation completely out of semiconductor materials in a semiconductor-[manufacturing] flow,” he says. Lafferty claims that this approach makes the company’s new digital-isolator line more repeatable, more dependable, and more reliable.

Analog Devices’ iCoupler line instead uses transformer coupling with polyimide film as the dielectric (Figure 8), whereas Texas Instruments uses capacitive coupling with glass as the insulator in the ISO72xx. Linear Technology uses discrete capacitors on the lead frame of its LT1535, and its new LT2881 uses PCB material as the isolation for spiral transformers. In contrast, NVE uses electron spin across a proprietary polymer film to convey the signal (Figure 9). NVE’s approach uses spin valves employing GMR. A huge change in resistance occurs when you expose the devices to a magnetic field. The company builds parts with a small coil to generate a magnetic field and GMR sensors on the other side of the boundary. Analog Devices has also introduced a transformer-coupled line that uses die glass as the insulator. “If safety isn’t important [and] if the only thing customers care about is breaking a ground loop or reducing noise, then they don’t need the high-isolation capability of our standard products,” says Krakauer. The parts use the same spiral-transformer technology as the high-voltage parts. With these new parts, 5 microns of silicon dioxide separate the transformer windings instead of 20 microns of polyimide in the earlier parts. Designers can also use the company’s transformer technology to send power across the boundary, resulting in higher system integration.

Silicon Labs creates an RF signal inside the chip and beams it over to a receiver antenna. The dielectric is glass that grows on the die. The company also uses on/off keying for modulation. “The benefit of this [approach] is that the output always unconditionally follows the input,” says Javed, noting that this approach makes the system more immune to noise interference than parts that use latch-based or pulse topologies. Acoustic coupling is yet another approach to isolation (Reference 7).

The overriding consideration in selecting any isolator is the standoff voltage. The need for UL or international-standard certification may—more than any other factor—determine the part selection. You can get UL approval for designs that use non-UL-listed parts, but it is a much more time-consuming process. You must prove to UL that the isolation boundary is adequate. It is far easier to select isolator parts that have the safety approvals.

The first decision to make with isolators is to consider whether you must bring an analog signal back across the isolation boundary. If the signals are slow enough and you have the budget, using the legacy ISO124 or AD204 is an acceptable approach. You should always consider using optocouplers because they are ubiquitous, but you must either live with their inherent aging problems or provide a reference servo design that compensates for those problems. Delta-sigma modulators from Avago, Analog Devices, and Texas Instruments are available for designs requiring higher bandwidths. These units bring digital signals across a boundary, but you can then make it analog using a simple lowpass filter. TI uses capacitive coupling, Analog Devices uses transformers, and Avago uses optocouplers. Because the signal across the barrier is digital in Avago’s products, optocoupler aging is not a problem.

Most digital-isolation applications operate at speeds of 1 to 50 Mbps. If speed is not your first concern, then power may be. Optocouplers in a logic-low state are not also driving the LED, in which case the drive current is 0A. You may be able to design your system to take advantage of that feature. NVE offers a device that provides the magnet-coil inputs with no interposed driver chip. “With these coil inputs, the signal provides the power itself,” says Templeton. “You don’t need to supply any power on the input side.” In this way, you can drive the coil in any way that you choose—perhaps in a way that saves a significant amount of power. Most of the alternative methods to optocoupling claim lower power consumption. If your application operates at 1 to 50 Mbps, you have many technologies from which to choose. One important spec may be transient immunity. “If there is one parametric Achilles’ heel that you must watch out for, it is the dV/dt [change in voltage over time] of the common-mode voltage,” warns TI’s Jerry Steele, a strategic-development engineer at Texas Instruments. This test slews one side of an isolator across a large voltage difference while you determine whether the data is still valid. Vendors routinely specify parts at 25 kV/µsec. Texas Instruments points out that its isolators work properly with a 50-kV/µsec transient across the boundary.

Another consideration is electromagnetic susceptibility. Texas Instruments claims that its capacitive isolation has greater immunity to magnetic fields than do other technologies. Analog Devices, however, says that the level of magnetic fields that would cause an error in its part would have to be so huge that it would ruin the signal integrity of every other part and trace in your design. NVE bases its isolators on a magnetic field but says that they are close-coupled and shielded and that strong magnetic fields impinge on the isolator in only a few applications. “[Magnetic fields] just don’t come up [with our customers],” says Dan Baker, president and chief executive officer of NVE. “Our device has more practical aspects of an EMC [electromagnetic-compatibility] footprint, immunity to external fields, and [reduced] transmitted external fields.”

Whereas you might think an isolator from Silicon Labs would be too sensitive to external fields because it is an RF system, you might want to reconsider. “We tried several technologies and found the Silicon Labs parts to be best for immunity and radiation in our application,” says John James, a lead engineer at Crossbow Technology. You must characterize how the data flowing across your isolator affects the ability of your product to pass FCC (Federal Communication Commission) and CE (Conformité Européenne) radiated-EMI tests.

As with a lot of other complex decisions, when choosing an isolator, you must divide what you need from what you want. The requirement for a certain voltage standoff or UL listing narrows your search. If a compact footprint is a requirement, shy away from optoisolator circuits and instead look at integrated products from Texas Instruments, Silicon Labs, and others. For designs requiring a USB interface, Analog Devices offers a system that can use two chips to give you isolated data and isolated USB 5V power. When cost is a concern, consider legacy optoisolators from Avago, Vishay, and NEC. The fact that these devices are legacy products means that you won’t have to worry about their obsolescence. All the companies making digital isolators pledge to keep them in their portfolios for as long as a decade, but using optocouplers in standard package pinouts is a safe bet for ensuring that obsolescence does not ruin your design.

When you have listed necessities and desires, you will reach a decision on a product that is specific to your design. A part’s EMI/RFI (radio-frequency-interference) immunity, magnetic fields, pulse-width fidelity, and long-term reliability all play into your decision. Set up test scenarios specific to your application and test the part in an environment that proves that the part will work for you.

Remember that it is sometimes better to keep things in the analog domain and use legacy parts. No matter whether you use analog or digital isolators, you should understand how the parts work. Capacitor, transformer, RF, optical, acoustic, and GMR techniques are all available, and they all behave differently. One exciting development is the expansion of an isolator’s operating-temperature range into the military realm. All of these advances build on a solid foundation of isolation techniques that manufacturers have perfected over the decades.