Isolation is a means of preventing current from flowing between two communicating points. Typically, isolation is used in two situations. The first is where there's the potential for current surges that may damage equipment or harm humans. The second is where interconnections involve different ground potentials and disruptive ground loops must be avoided. In both cases, isolation is used to prevent current flow, yet allow for data or power flow between the two points.
Recent changes in legislation with regard to both the design and use of machinery and equipment require the isolation of almost any type of data-acquisition system in harsh environments. In addition, the trend from historic single-channel isolated systems to applications utilizing multichannel isolation led to the introduction of new isolation strategies. These applications involve high-voltage, high-speed/high-precision communications, or communication over large distances. Common examples include industrial I/O systems, sensor interfaces, power-supply/regulation stems, motor-control/drive systems, and instrumentation.
This article gives an overview of previous isolation methods and their components, then continues with the principles of operation of digital isolators and their applications in multichannel data-acquisition systems.
Early Isolation Techniques
In addition to the use of transformers, early designs utilized analog isolation amplifiers (iso amps) to isolate the sensor circuitry on the factory floor from the signal-processing system in the control room. Some of these amplifiers are still in use today in applications with limited channel count and small signal bandwidth. Figure 1 illustrates this kind of isolation in a single-channel temperature measurement.
These isolation amplifiers were precision amplifiers incorporating a novel duty-cycle modulation-demodulation technique to digitally transmit the input signal across a differential capacitive barrier (Fig. 2). With digital modulation, the barrier characteristics don't affect signal integrity, resulting in excellent reliability and good high-frequency transient-immunity across the barrier.
Referring to Figure 2, the input amplifier A1 integrates the difference between the input current, (VIN/RIN), and a switched current source. The integrator ramps in one direction until it exceeds the comparator threshold. The comparator and sense amplifier AS1 force the current source to switch. The resulting signal is a triangular waveform with a 50% duty cycle. The internal oscillator forces the current source to switch at high frequency (i.e., 500 kHz). The resultant capacitor drive is a complementary, duty-cycle-modulated square wave.
At the same time, sense amplifier AS2 detects the signal transitions across the capacitive barrier and drives a switched current source into integrator A2. The output stage balances the duty-cycle-modulated current against the current through the feedback resistor RF, resulting in an average value at the VOUT pin equal to VIN. The sample-and-hold amplifiers in the output feedback loop remove undesired voltage ripples inherent to the demodulation process.
Despite their high reliability and precision, isolation amplifiers were limited in signal bandwidth to 50 kHz. Their older technology requiring a minimum supply of Â±4 V doesn't support today's low-voltage applications of 3 V and below. Also, their manufacturing process - which includes the separate fabrication of the input and output sections, laser trimming for exceptional circuitry matching, and mounting both sections with isolating capacitors in between - made these devices rather expensive.
Many data-acquisition systems in industrial automation use multi-input channel analog-to-digital converters (ADC) to capture the input data (measurands) of multiple analog inputs (Fig. 3).Â Most delta-sigma ADCs feature serial interfaces to reduce package size and board space. The complexity of serial interfaces varies in the number of slow-speed control signals required, such as chip-select, power-down, gain- and speed settings, and multiplexer addressing. Common to all serial interfaces, however, are the high-speed transmission lines for the clock signal and the output data (conversion results).
Because signal capture and conditioning occur within the ADC, the best-suited location to isolate the sensor circuit from the signal-processing circuitry is at the digital interface using digital isolators. As mentioned before, due to interface complexity, the isolators must be able to transmit high-speed ADC conversion results, as well as low-speed control data. The next section explains the internal operation of a digital isolator showing how these devices are capable of high- and low-speed data transmission.
The isolator in Figure 4 is based on a capacitive isolation barrier technique. The device consists of two data channels, a high-frequency channel (HF) with a bandwidth from 100 kHz up to 150 MHz, and a low-frequency channel (LF) covering the range from 100 kHz down to dc.
In principle, a single-ended input signal entering the HF-channel is split into a differential signal via the inverter gate at the input. The following capacitor-resistor networks differentiate the signal into transients, which are then converted into differential pulses by two comparators. The comparator outputs drive a NOR-gate flip-flop whose output feeds an output multiplexer. A decision logic (DCL) at the driving output of the flip-flop measures the durations between signal transients. If the duration between two consecutive transients exceeds a certain time limit (as in the case of a low-frequency signal), the DCL forces the output multiplexer to switch from the high- to the low-frequency channel.
Because low-frequency input signals require the internal capacitors to assume prohibitively large values, these signals are pulse-width modulated with the carrier frequency of an internal oscillator, thus creating a high-frequency signal that can pass through the capacitive barrier. As the input is modulated, a low-pass filter (LPF) is needed to remove the high-frequency carrier from the actual data before passing it on to the output multiplexer. Figures 5 and 6 present the high- and low-frequency channels and representative waveforms.
The single-ended input signal is split into the differential signal components A and /A. Each signal component is then differentiated into the transients B and /B. The following comparators compare the differential transients to another. As long as the positive input of a comparator is on higher potential than its negative input, the comparator output will present a logical high, thus converting an input transient into a short output pulse.
The output pulses set and reset a NOR-gate flip-flop. From the truth table, we see that the NOR-gate configuration presents an inverting flip-flop, meaning that a high at input C sets output /D to high, and a high at /C sets D to high. Because the comparator output pulses are of short duration, there will be times where both outputs are low. During this time, the flip-flop stores its previous output condition. Since the signal at /D is identical in shape and phase with the input signal, /D becomes the output of the high-speed channel and is connected to the output multiplexer.
Slow input signals are pulse-width modulated with a high-frequency carrier so that a signal high yields a 90:10 duty cycle and a low yields a 10:90 duty cycle at location A. From there on, signal processing is identical with asymmetrical signal processing in the high-speed channel. The only exception is that the high-frequency content of the low-speed channel (/D) is filtered by an R-C low-pass before being passed on to the output multiplexer (E).
The successful proof of concept through the single isolator's capability of transmitting wideband data (from dc to above 100 MHz) inspired isolator manufacturers to fabricate unidirectional and bidirectional devices in dual-, triple- and quad versions. These accommodate the most common digital interfaces encountered in industrial applications.
When isolating industrial interfaces, we need to distinguish between process control and factory automation applications. That's because their differences will impact the isolation efforts of the digital interface design.
Process control typically involves the detection of various physical quantities, (i.e., pressure AND temperature) of some equipment, system, or process. Each physical quantity uses a specific type of sensor or transducer whose output signal requires specific signal conditioning. Consequently, a variety of different sensors requires different parametric settings, such as in-gain, sampling rate, measurement repetition, or impedance buffering. ADCs supporting a wide range of settings provide multiple interface control lines, all of which require isolation in addition to the standard serial interface lines.
In Figure 7, a number of sensors with different sensitivities (mV/K) measure different process parameters, i.e., temperature, pressure, and current. Various gain settings are required to maximize the input dynamic range of the ADC for each sensor. A possible switch between sampling rates (speed) might be required if one or more channels are expected to show faster input variations than others. The power-down features, used to save power consumption after measuring, allow the controller to perform other system functions. This high versatility requires many control channels to be isolated via two quad isolators.
In contrast to process control, factory automation is usually concerned with the monitoring of a single physical quantity (i.e., temperature OR pressure) of multiple devices or equipment. These systems, therefore, employ multiple sensors of the same type, exhibiting uniform characteristics in sensitivity and response time.
Figure 8 presents such a circuit using four thermocouples of the same type for temperature measurements of different equipment. This application uses the same ADC as the circuit in Figure 7. Due to the uniform sensor characteristics, however, the settings for gain and sample rate are fixed by connecting the associated control pins, (Gain1, Gain2, and Speed), to the appropriate supply rails (VDD or GND). Many autonomous systems in factory automation measure their inputs continuously, which necessitates the connection of the /PWDN-pin to the positive supply rail.
This system configuration drastically simplifies the interface down to the isolation of the data, clock, and address lines. Thus, it requires only one 3:1 quad isolator.
In the previous examples, interface isolation occurred between the ADC and the system controller. This approach works well for input modules with channel counts requiring only one or a maximum of two ADCs per module. Above that, isolating each data converter becomes uneconomical. Hence, implementation of a local controller is recommended. In such a case, each ADC communicates with the local controller via a GPIO bus interface. The actual isolation, however, occurs at the local-to-system controller interface.
In conclusion, it's safe to say that isolation amplifiers are out - and digital isolators are in. Understand your system requirements before deciding what type of isolator to use, and where to place it in the system.
For more information on Texas Instruments interface solutions, visit www.ti.com/interface.