This isolation is crucial for signal integrity, system protection, and user safety, but designers must choose from three main isolation technologies: magnetic, optical, and capacitive barriers.
Each of these options shares similar performance characteristics but has subtle differences that designers need to understand before making a selection. Therefore, this article will discuss the role of isolation in sensors, then introduce the various options, their different characteristics, and how to apply them.
It will also introduce digital isolation and provide more examples of digital isolators.
The basic elements of isolation
When a sensor or sensor subcircuit is “isolated,” there is no electrical path between it and the rest of the circuit, and no current flows between the two parts when performing basic measurements with an ohmmeter (Figure 1). Due to this barrier, the challenge lies in transmitting signals from the isolated subcircuit to the rest of the system. In many cases, there is an additional challenge: powering the isolated subcircuit without a “detour path” through the power subsystem, which negates any isolation.
Figure 1: In an isolated system, there is no current path between the ungrounded sensor and the system (which may be grounded), but some type of information carrying energy must still be transferred from one side to the other. (Image source: Digi-Key, based on original data from www.ee.co.za, 2010)
The reasons for quarantine include:
The sensor is "floating" and must not have any connection to the system "ground" (system "general" is a more accurate term, but "ground" is a common misnomer used here). * Even if the system is battery-powered and therefore not connected to "AC ground," the sensor output may still be above a high common-mode voltage (CMV). This CMV can damage the rest of the electronics. For example, the voltage of a single battery cell at the top of a series stack. * The sensor may be inadvertently connected to a high-voltage source or even an AC line. This will not only damage the electronics but also put the user at risk.
Fortunately, there are viable options for achieving analog sensor isolation, providing low-level and advanced isolation ranging from tens to thousands of volts. The latter is essential in mass-market applications such as EVs/HEVs and is often used for regulatory safety requirements. The three most common methods for providing isolation use magnetic, optical, and capacitive technologies, each rated for >1000 V or higher.
These technologies overlap significantly in their key performance attributes, but also have some differences. Deciding which one is best suited for the application is often a difficult decision.
Factors to consider include bandwidth, size and footprint, cost, isolation rating (volts), lifespan rating, and personal "comfort." Finding the right balance of performance parameters depends on the application. For example, battery monitoring does not require a fast response, while high-speed test sensors do.
Magnetic isolation: starting point
Magnetic isolation, using transformers, is the oldest technology; for many years, it was the only technology. Isolation transformers typically have a 1:1 turns ratio and can be quite small because they handle signals rather than power, and power consumption is very low.
Because transformers cannot carry direct current and do not handle very low frequencies well (unless they have a large magnetic core), the signal to be isolated cannot be applied directly to the primary (input) side. Instead, if needed, the sensor signal is amplified and then used to modulate the carrier at a higher frequency (amplitude modulation) or for pulse width modulation (PWM).
On the output (secondary) side, conventional techniques are used to demodulate the signal to extract and recover the original signal. An isolated power supply must be provided to the primary side; therefore, there is typically a separate dedicated isolated-side power supply, while the output side uses the system power rail.
An early example of an embedded isolated operational amplifier was Analog Devices' AD215 (Figure 2). It functioned similarly to a non-isolated operational amplifier but offered 1500 V (rms) isolation and a 120 kHz bandwidth. It included a signal modulator, transformer, and demodulator, as well as an isolated DC power supply. All of these components required current isolation while allowing analog signals to be passed from the input to the output.
Figure 2: The AD215 magnetic isolator contains a signal modulator, transformer, and demodulator, as well as an isolated DC power supply. (Image source: Analog Devices)
This device has a ±10 V input/output range and a rated gain range of 1 V/V to 10 V/V. The front end includes an internal isolated DC/DC power supply, so no separate power supply is required.
Although the AD215 is primarily used for feedback loops in switching power supplies, it can also be used in voltage monitors, motor current sensing, and large battery systems, all within its 400 kHz bandwidth (Figure 3).
Figure 3: While the AD215 is primarily used in the feedback loop of switching power supplies, it can also be used in voltage monitors, motor current sensing, and large battery systems. The AD210 shown here is a functionally equivalent version of the AD215, but with looser specifications; the AD620 is a precision instrumentation amplifier.
In such applications, when measuring the voltage across a motor sensing resistor to determine the current flowing through the motor, analog isolation is typically required. This is necessary because the sensing resistor is not ground-referenced but "floating" and may be at a very high potential relative to ground.
These early magnetic-based isolation devices used discrete transformers, making them relatively large and expensive. Newer designs use proprietary versions of planar coplanar transformers compatible with IC packages. For example, Analog Devices' ADuM3190 isolation error amplifier is packaged in a 16-pin QSOP package but provides a 2.5 kV isolation rating. Its planar transformers are placed parallel to each other for maximum power transfer (Figure 4).
Figure 4: The ADuM3190 isolation error amplifier looks and is handled similarly to the IC, but includes a silicon die and a pair of planar transformers placed parallel to each other for maximum power transfer. (Image source: Analog Devices)
Although it looks like a standard operational amplifier, it actually receives the input signal and uses it to generate a PWM signal through a planar transformer. This PWM signal is demodulated and filtered on the secondary side to produce an analog output signal. The datasheet includes a standard operational amplifier Bode plot for phase and gain margin (Figure 5). Designers using this device (or a similar device) can expect to model and simulate the loop stability and related aspects of a standard amplifier. Bode's conspiracy will be helpful.
Figure 5: Despite its complex isolated internal architecture, the ADuM3190 appears to designers as a conventional operational amplifier (Image credit: Analog Devices).
The ADuM3190 has a rated operating temperature range of −40°C to +125°C, which is realistic for some target applications. Note that because the isolation element is a rigid wire, there is no "wear" mechanism in the traditional sense unless the device exceeds its specifications.
However, due to voltage stress, all insulating materials will eventually decompose over a sufficiently long period of time, with the degradation rate being a function of the magnitude and type of voltage waveform applied to the voltage barrier. For the ADuM3190, the supplier guarantees a 50-year lifespan even under the maximum rated bipolar current waveform, which is greater than that under the same amplitude of unipolar current or DC stress.
Optical isolation: a newer option
An alternative to magnetic isolation is optical isolation, which is conceptually simple: the LED is driven on the input side, and the LED's output strikes a co-packaged phototransistor, resulting in a phototransistor current (Figure 6). The short optical path between the LED and the phototransistor within the package provides the necessary current isolation.
Figure 6: An optical isolator requires two active components: an LED for irradiating the light source (IR), and a phototransistor for converting the received photons into current. Current isolation is provided by the optical path within the package. (Image source: Sunpower UK)
Similar to transformer-based isolation, the input signal is used to modulate the LED current in digital mode using PWM or other methods. The Broadcom (Avago) ACPL-C87B/C87A/C870 series of opto-isolated amplifiers is a good example of a device that can be used for voltage sensing across a current-sensing resistor (Figure 7).
Figure 7: The opto-isolators in the Broadcom ACPL-C87B/C87A/C870 series are designed for lower voltage levels and use Σ-Δ modulation and chopper-stabilized amplifiers to achieve precision, accuracy, and consistency. (Image source: Broadcom)
This series of isolators features a 2 V input range and a high input impedance of 1 GΩ. These specifications make them ideal for isolated voltage sensing requirements in power converter applications, including motor drives and renewable energy systems. These devices combine optocoupler technology with Σ-Δ modulation, chopper-stabilized amplifiers, and differential outputs to provide high isolation-mode noise rejection, low offset, high gain accuracy, and stability. They are all packaged in a stretched SO-8 (SSO-8) package.
These devices, with a 100 kHz bandwidth (Figure 8) and high common-mode transient immunity (15 kV/μs), are ideal for power converter applications. Such transients are common in motor drivers.
Figure 8: Broadcom's ACPL-C87B/C87A/C870 series optical isolation amplifiers utilize their internal Σ-Δ analog-to-digital conversion technology to easily achieve a 100 kHz bandwidth and a flat response up to 100 kHz. (Image source: Broadcom)
Capacitive isolation: Newest option
Another isolation technique uses tiny gaps between capacitors and capacitor “plates” for isolation. This technique has become feasible and cost-effective in recent years due to advancements in IC and packaging technologies. A good example of this implementation is Texas Instruments. This is a precision isolation amplifier whose input and output sections are electrically isolated by matching 18 picofarad (pF) capacitors embedded in an SOIC-1 (or SOIC-16) surface-mount plastic package.
Like other analog isolation amplifiers, its high-level function diagram is simple (Figure 9).
Figure 9: A common symbol for analog isolation amplifiers is this "split" operational amplifier, as used in the ISO 124 datasheet; this clearly indicates that the input and output sections have their own independent "reasons" (although "general purpose" is a more accurate term). (Image credit: Texas Instruments)
Meanwhile, the detailed block diagram reveals the complexity of the internal content, which is not visible to the user (Figure 10).
Figure 10: Similar to magnetic and optical isolators, ISO 124 contains a large amount of analog and digital circuitry, enabling its unique capacitor-based isolation to function properly. (Image credit: Texas Instruments)
The ISO124 input is duty cycle modulated and transmitted digitally via a grid. The output receives the modulated signal, converts it back to analog voltage, and eliminates the ripple component inherent in demodulation. It features a maximum nonlinearity of 0.010%, a signal bandwidth of 50 kHz, and an operating system drift of 200 μV (μV)/°CV, requiring a single supply of ±4.5 V to ±18 V.
Similar to non-isolated operational amplifiers, the datasheets include both tabular data and graphical information on sine wave and step response performance under various conditions (Figure 11). Potential users of these isolated devices need to study the data and graphs to ensure that the device performance matches their application requirements.
Figure 11: Due to the operational amplifier nature of ISO 124 analog devices, designers need to pay close attention to many graphs, including standard sine and step response graphs. (Image credit: Texas Instruments)
ISO124 is ideal for low-speed applications, such as isolating the signal from the thermocouple temperature sensor on the resistance temperature detector (RTD) and receiver end 4-20 mA current loop and converting it to voltage (Figure 12).
Figure 12: ISO 124 is used to isolate RTDs connected through a standard 4-20 mA current loop and convert this current signal to a 0-5 V signal for control system compatibility. (Image source: Texas Instruments)
Temperature measurement applications typically require the sensor to be isolated from the rest of the system circuitry due to potentially high voltage points directly mounted to the sensor, such as on a motor housing. This voltage is then used by the system's analog-to-digital converter (ADC) for readout or closed-loop control, both common industrial scenarios.
Make a decision
All three analog isolation techniques—magnetic, optical, and capacitive—can provide excellent results under the right conditions. The designer's dilemma then becomes deciding which is "best" for a given situation.
Factors to consider include bandwidth, expected lifetime (failure or wear-out time), size, and power requirements. Each isolation technology balances these attributes and may offer specific products with different trade-offs within a range, complicating the decision-making process.
Regarding voltage isolation, all three types are certified for at least 1 kV (some reach 5 kV or even higher) and comply with relevant regulatory standards (IEEE, VDE, CIE, UL, CSA). Therefore, maximum isolation voltage is not an issue for most IoT applications. If this becomes an issue, dedicated isolators certified for higher voltages can be used.
For each isolation technology, some general statements can be made, but there are exceptions to every statement, and each technology vendor will present valid and legitimate arguments for why their approach is better. Generally:
Magnetic isolation offers a very long lifespan, and its passive barrier can withstand surges and spikes far exceeding the continuous voltage rating. However, due to its inductive coupling through the magnetic field, it is susceptible to interference from external magnetic fields. Some newer designs have successfully minimized this problem to such an extent that the unit has been certified for this interference sensitivity using industry-standard testing.
Optical isolation offers high immunity to electrical and magnetic noise, but its speed is moderate due to LED switching speeds, higher power consumption, and concerns about LED wear. The last issue is the most serious, as LEDs do experience degradation (dimming) during normal use, with a typical half-brightness cycle of about ten years. However, companies like Broadcom/Avago have pushed the boundaries of LED material technology, thus guaranteeing specifications for up to twenty years, which is usually sufficient for this situation.
Capacitive isolation offers high immunity to magnetic noise and, compared to optical isolation, can support a wider bandwidth because it eliminates the need for LED switching. In reality, most IoT sensor applications operate in low-bandwidth environments. Furthermore, capacitive coupling involves data transmission using electric fields, making it susceptible to interference from external electric fields.
Analog and digital isolation
So far, we have studied analog isolation techniques for IoT sensors (Figure 13).
Figure 13: In an analog isolation topology, sensor signals remain analog information (regardless of what happens inside the isolator itself) until they reach the non-isolated side, where they can be converted to digital format for further use. (Image credit: National Instruments)
However, there is a basic alternative architecture: digital isolation, where the output of the analog sensor is conditioned and digitized on the isolation side, and then the digital output passes through a digital isolation barrier (Figure 14**).
Figure 14: Another often attractive approach is to digitize the signal on the isolation side and then pass the converted result through a digital isolation barrier. This allows for very different isolation functionality compared to analog isolation designs. (Image credit: National Instruments)
Similar to analog isolation, this barrier can use any of these three technologies, but its internal design is specifically optimized for digital signals, typically supporting data rates of tens of Mbps. Furthermore, for digital isolation, there is a new fourth option that uses a modulated RF carrier instead of modulated (LED) light. Silicon Labs' Si863x series is a good example of such a device (Figure 15).
Figure 15: Silicon Labs' Si863x series digital isolators use modulated radio frequency carriers instead of light to transmit signals while providing isolation. (Image credit: Silicon Labs)
As ADC costs decrease, suppliers have found it increasingly attractive to use digital isolation in technologies such as I²C and LVDS. The downside is the need for more circuitry on the isolation side. This means more isolation power supplies are required, increasing cost and footprint.
However, advancements in low-power, high-performance ADCs have made this problem less of an issue. Standard interface digital isolators, such as the 1 MHz ADUM1250^2 for I/O and the ADN4651 for LVDS from Analog Devices, simplify this design alternative. Integrated isolation ADCs are also available in multi-chip IC packages, such as the 16-bit AD7401A from Analog Devices, which makes the entire conversion and isolation process transparent to the user.
Finally, there's the issue of multi-channel isolation. While many IoT applications have only one or two sensors that need isolation, others may have four, eight, or more. In these cases, a single analog isolator can be too large, too expensive, and too power-intensive overall.
Another approach is to use a multi-channel ADC or a single-channel device with a front-end multiplexer, all located on the isolation side, providing higher-speed digital isolation for transmitting results. This can be more power-efficient, space-efficient, and cost-effective than simple per-channel isolation.
in conclusion
Analog sensor isolation is a critical issue for signal integrity, system security, and user protection in many IoT applications. Three competing technologies can achieve isolation, each with many similarities in performance but some subtle differences. Digital isolation, which digitizes the sensor on the isolation side, is also an option to consider in many applications.
