Author: Mark Patrick, Mouser Electronics
We live in an analog world, but our experience of it is increasingly mediated by digital technology. If you look at the sky, you’ll see it is one of an infinite number of colours. If you look at a digital image of the sky, it can only be shown as one of a relatively limited number of fixed shades. And if you look at a compressed digital image of the sky, for example in an online video, you will often see the colour banding that reveals the gap between analog reality and digital representation.
We can usually live with such shortcomings in digital media. But when an analog signal forms the key input to a digital control system, we need to take more care of it. After all, no-one would drive a car whose brake sensors could only report that the brakes are ‘fully on’ or ‘fully off’. It just isn’t safe.
And so we can see that it is increasingly important to maintain the integrity of the analog signals that provide the raw materials for digital systems. Without robust analog input signals, the most sophisticated control systems in the world are just guessing.
Even so-called ‘digital sensors’ rely on analog sub-systems that contain the core analog sensor, conditioning circuitry and an analog-to-digital converter (ADC) that turns the input signal into a digital number that can be easily used by the rest of the system. Designers therefore still need to understand how to capture and process a low-frequency analog signal correctly, so that it can be presented to an ADC for conversion into the digital domain for onward processing. This is especially true for those working on sensor-based systems such as the Internet of Things (IoT) devices. Such sensor nodes are quite likely to suffer from low signal levels, reduced supply voltages that limit signal-to-noise ratios, and electromagnetic interference (EMI) in their operating environments.
Optimizing analog signal capture
A typical analog signal chain usually includes a signal source (often a sensor), buffering to protect the sensor’s operation from the influence of subsequent circuitry, filtering to remove noise and other unwanted signal components, amplification to make the signal easier to measure, and then the ADC. Each of the active elements, and their associated passive components, plays an important role in ensuring that the signal’s integrity is preserved so it can be faithfully converted into the digital domain.
The design process begins with understanding the source signal’s amplitude and its susceptibility to EMI. Most sensors output a DC voltage proportional to the parameter being measured, which could be light, temperature, pressure or many other physical parameters. Sensors often produce relatively weak signals, which makes them susceptible to EMI that can ruin the integrity of the ADC’s measurement. To counter this, it makes sense to place a follower circuit, such as a non-inverting amplifier based on an operational amplifier (op-amp), as close to the sensor as possible. Putting the follower circuit close to the sensor stops long signal traces acting as antennae that pick up ambient electrical noise. If it’s not possible to protect the signal in this way, it may make sense to shield long traces between the sensor and its follower circuit.
The gain of the follower circuit is determined by the ratio of the resistors R2 and RF shown in Figure 1 below. These resistors’ values affect the amplitude of the signal passed down the signal chain, with a follow-on effect on factors such as signal to noise ratio and the accuracy of the analog to digital conversion. If the exact level of gain is critical to the circuit design, it is important to specify resistors with tight tolerances. If a circuit needs to achieve a very precise amount of gain then it’s possible to specify matched resistors, such as the Maxim MAX5491LC parts. These mount two resistors, matched in value to within 0.035% of each other, in a SOT23 package. Mounted this way, the resistors also exhibit a very low thermal drift of 2ppm/°C over the operating temperature range of –40°C to +85°C.
The single op-amp circuit shown in Figure 1 will act as an ideal buffer, especially for driving a signal from remotely located sensors down long signal lines. It is simple to use but needs careful configuration to achieve the levels of precision needed to amplify small signals accurately.
For example, theory tells us that the op-amp’s output should be zero if the potentials at the two inputs are the same. In practice, a small error voltage, known as the input offset voltage, is usually present. If the op-amp draws a lot of power to operate this can also affect very weak sensor signals. Designers should, therefore, consider the input impedance of the op-amp (which often reaches MW), and the input bias current required by the op-amp to operate as part of their sensor-signal buffering strategy.
One way to overcome the limitations of a single op-amp is to use a matched pair of op-amps mounted in one package. The circuit in Figure 2 below is often used with dual op-amps, such as the Analog Devices’ ADA4522 series.
The circuit offers low noise and zero drift because the two op-amps are co-packaged. Its low input bias current of 50pA and input offset voltage of 0.7mV means it has little effect on high-impedance signals, and so it can accurately amplify signals from any sensor. The common-mode input impedance of the circuit is around 100GW.
Texas Instruments also offers a dual op-amp part, the high-precision OPA2156. It has a very low maximum offset voltage of ±200mV, a low bias current of ±5pA and low thermal drift of ±3mV/°C> this makes it a very low-noise, high-precision buffer amplifier. The device can also swing both its input and output between its supply rails, has a bandwidth of 25MHz, and a high slew rate of 40V/ms. This makes it useful for high-speed sensing applications. It is supplied in an eight-pin SOIC package and operates over the industrial temperature range of –40°C to +125°C.
Some sensors have outputs that can be set up to work with a variety of op-amp configurations to achieve the correct amplification of their signals. Figure 3 shows a photodiode sensor that delivers an output current of 0mA to 90mA, and an output voltage ranging from –5V to +5V. Adding a capacitor to the feedback loop has created a filter, which reduces the amount of additional discrete filtering needed. For example, the 2.7pF capacitor in this example creates a filter cut-off frequency of 1MHz.
Choosing the right data converter
The circuitry described so far is meant to ensure that a relatively fragile analog signal, subject to corruption by external noise sources and poor circuit design, can be presented for measurement in the best condition possible. The next step is to ensure that its translation from the analog to the digital domain is as accurate as possible. Specifying the right ADC is critical to this process. Making this choice requires a good understanding of the various parameters that affect an ADC’s ability to transform an analog input signal into the kind of digital output that enables overall system design goals to be met.
The four primary parameters of any ADC are resolution, speed, accuracy, and noise.
Resolution is often confused with accuracy but is entirely different. The resolution of an ADC defines the number of different values it can output, which is usually encoded as a set of binary digits. For example, a 1bit ADC can output two values, while a 4bit ADC can output 16 different values. This resolution figure also, therefore, defines the smallest incremental change that the ADC can output, as well as its smallest output value. In the case of the 4bit ADC, therefore, the smallest value and the smallest increment that could be recorded would represent one-sixteenth of the total range of the converter.
Accuracy, on the other hand, defines how faithfully the digital output represents the analog input and includes any offsets or nonlinearities within the ADC. In general, increasing resolution helps raise the accuracy of the conversion.
Speed affects applications such as video processing since an ADC’s conversion rate should be at least twice as fast as the highest frequency that needs to be captured in the source signal. (If you think of an ADC’s input as a basic sine wave, you can see that you need to sample it at least twice per cycle to accurately represent it). In many sensing applications, though, especially if you’re measuring physical parameters such as temperature, the conversion rate of the ADC is rarely a concern.
All ADCs generate some form of quantization noise as a by-product of the conversion. Going back to our sine wave input, the digital output will be a series of numbers that, given a fast enough ADC, would show a sine-wave shaped curve defined by a series of discrete steps. The difference between the true analog representation of the signal and the digital ‘model’ of it is the quantization noise. It can be countered by increasing the resolution of the ADC.
It’s important to choose an ADC that strikes the right balance between resolution, accuracy, conversion speed, and quantization noise, as well as enough channels to sense all the necessary signals. But ADCs are the gateway between the analog and digital world, and so it is equally important to consider how the ADC will interact with its digital host system.
For example, the Texas Instruments ADS1219 is a precision, 24-bit ADC whose inputs can be configured as four multiplexed single-ended inputs or as a pair of differential inputs. It has an onboard buffer stage, which means that high-impedance signals can be directly connected to it. The ADC also has a programmable gain stage for further signal conditioning. The digital interface is a two-wire, 1Mbps I2C-compatible bus, and the part is housed in either a 16-pin WQFN or TSSOP package.
As our analog world becomes increasingly mediated by digital technology, it is vitally important to do a good job of capturing, buffering, filtering and conditioning real-world analog signals so that they can be accurately represented digitally. This demands a strong understanding of how sensors work, the characteristics of their outputs, and how those outputs need to be conditioned for conversion. This analog signal chain will invariably include op-amps, discrete components, and data converters, each of which has to be thoughtfully specified to achieve the correct point and systemic performance of the target device.