Understanding the analog aspect of data converters
Smart devices make it easy to believe we live in a digital world. Instead, the world is entirely analog. IoT makes use of nature’s data, like time, temperature, light, sound and pressure. Transducers convert these phenomena into electrical signals. Analog-to-digital (ADC) and digital-to-analog (DAC) converters allow us to access the information hidden inside these signals.
A simple converter works by comparing an unknown value to a known reference. A comparator is the simplest type of 1-bit ADC. Most data converter architectures are more useful but also more complex. Understanding these architectures makes it simpler to choose the right converter for your application.
The most common ADC architectures in use today include:
- Sigma-delta ADCS with 12- to 24-bit resolution. Sigma-delta converters have sample speeds from single-digit up to thousands of samples per second (ksps). The applications include low-frequency precision industrial measurement and instrumentation as well as high-frequency audio.
- Successive approximation register (SAR) ADCs with resolutions ranging from 8- to 18-bits. SAR ADCs are the most popular architecture used in data acquisition applications. SAR is a good option when multiple input channels need to be multiplexed at tens to hundreds of ksps.
- Pipelined 8- to-16-bit ADCs are used in applications like radar, high-speed instrumentation (spectrum analyzers, medical imaging), and ultra-high-definition TV, which require sampling rates of more than 5 Msps.
ADC architectures are characterized by samples per second. Settling time is the key differentiator for DAC architectures. Commonly used DAC architectures include:
- Sigma-delta DACs with resolutions up to 24-bits. SAR DACs are used to generate analog signals for the calibration of instrumentation equipment. These have settling times in the order of milliseconds down to tens of microseconds.
- Voltage output resistor string, R-2R, and multiplying DACs (mDAC) with resolutions ranging from 8- to 16-bits. These DACs can have settling times as low as tens of nanoseconds. They are commonly used in industrial applications.
- Current steering DACs with ultra-fast settling times (down to the picosecond range). These types of DACs are used in high-speed video and communication systems.
What are the main data converter specifications?
After choosing the best architecture, you may want to compare individual devices. It is important to understand the main design parameters and specifications for data converters. These are found on the device datasheet. Use them to evaluate performance.
Resolution is a design parameter common to both DACs and ADCs. For a DAC, it refers to the smallest possible change in output voltage (or current) in response to the single bit change in digital input code. The smallest digital change is defined by the least significant bit (LSB). The resolution of an ADC is the smallest possible variation in binary output code in response to a change in the analog input voltage. Resolution is usually specified in bits or percent of the full-scale range (%FSR). The LSB is calculated by dividing the reference voltage by 2n (where n is the number of bits in the data converter).
For example, an ADC with 10-bit resolution can resolve one part in 210 (1 in 1024) or 0.0976% of FSR. A 10-bit ADC with a full scale of 10V could resolve a 9.76mV input change. Likewise, a 10-bit DAC would exhibit the same change in output voltage when the input increases by one bit (1 LSB). A converter with the highest number of bits will resolve the smallest input increments.
Resolution is not the same as accuracy. The level of accuracy required defines the resolution (number of bits) needed. The important static (DC) specifications for accuracy are differential nonlinearity (DNL), integral nonlinearity (INL), offset error, and gain error. These describe how much the behavior of a real device deviates from its ideal transfer function.
DNL measures the width of a digital code relative to its ideal value. Each code transition should occur at an interval equal to 1 LSB, meaning if the measured code width is equal to its ideal value, then the DNL is zero. If it is wider than the ideal the DNL is positive. If it is shorter, the DNL is negative. For example, in a 3-bit ADC, if the first code transition occurs at one-eighth of full scale (0.125 FSR), the second transition should occur at 0.250 FSR. The deviation from that ideal figure is the differential linearity error for that code.
The DNL specification for a converter is the worst-case DNL across all code transitions. Sometimes, DNL can be large enough to cause a missing code transition. Converters with missing codes are non-monotonic and this is undesirable in applications where an electronic system must be settable to a particular state corresponding to the missing code. A converter with a DNL less than +/-1 LSBs is guaranteed to have no missing codes. Many manufacturers test their data converters so they can guarantee no missing codes to a defined number of bits.
The linearity of a data converter will define how accurately it represents real-world values in the digital domain.
INL describes the overall shape of the transfer function of a data converter. This error is sometimes also referred to as static (or absolute) linearity and is calculated by summing (integrating) the individual DNL errors. INL is a measure of the maximum deviation of the actual transition points in the transfer function of a data converter from a straight line (ideal, endpoint or best fit) and is usually expressed in LSBs.
Gain error is the deviation of a data converter from the ideal gain slope of +1 and is usually expressed as %FSR but may also be specified in LSBs or volts. The reference voltage is the main contributor to gain error as it sets the full-scale range of the data converter.
Offset (or zero-input) error describes the amount by which a transfer function of a converter differs from its zero-input value. For example, an input voltage of 0V should produce 000 (binary) at the output of a 3-bit ADC. If this code appears in response to an input of 10mV, then this is the level of offset error, and it will appear on all other codes in the transfer function. The converse is true for a DAC.
It is possible to use calibration techniques to remove gain and offset errors from the transfer function of a data converter, but linearity errors cannot be removed because they are a feature of the internal design used in the construction of the device.
Apart from DC specifications, datasheets also commonly include some AC performance metrics. Most data converters operate differentially i.e., they only convert the difference between the signal voltage levels present on the input terminal pair, while rejecting common-mode signals coupling onto them from external sources of electrical noise. The ability of a converter to reject these common-mode signals is called the common mode rejection ratio (CMRR) and is specified in decibels (dB).
Drift (usually long-term) is primarily caused by the aging of a data converter’s internal components. This can make linearity, gain and offset errors worse over time. Scale change due to aging of the voltage reference is usually the biggest contributor to these errors.
What are the auxiliary functions of an ADC?
It takes a finite amount of time for an ADC to convert an analog signal to a digital value. ADCs can only convert at intervals when it holds a sample of the signal. The sample is stored using a “sample-and-hold” circuit. The circuit holds the latched value steady until the conversion is complete. The process is then repeated. The sample-and-hold circuits are usually integrated into the converter.
The stability of the power supply and the voltage reference are key contributors to the overall performance of a data converter. Power to the converter can be stabilized using either a low-voltage dropout regulator (LDO) or a DC-DC converter. Voltage references are now also commonly integrated into the same package as the data converter to help minimize the effect of temperature differences. Finally, converters now also typically include an I2C or serial peripheral interface (SPI) port. This port allows them to communicate with other integrated devices, such as a microcontroller.
Transducers, data converters and microcontrollers are at the heart of IoT. Converters process physical quantities from the analog world in an actionable way. This article provides an overview of data converter architectures. It identifies suitable architectures for a variety of applications. It explains the key performance criteria that can be used when comparing the performance of ADCs and DACs.
With a wide selection of analog solutions from the industry’s leading suppliers, Avnet provides the technology that brings the analog world into the digital domain.