Analog to Digital Converter (ADC) Selection Guide

To select the right ADC, you’ll first need to consider four critical factors:

- Resolution
- Speed
- Accuracy
- Noise

Once you’ve assessed your project’s needs in those areas, you can narrow your selection even more by considering less important variables, such as:

- Input voltage
- Interface
- Number of channels

ADC Selection Criteria to Consider

As you explore the world of ADCs, it may be helpful to refresh your memory on how resolution, speed, accuracy, and noise will affect your choice.

Resolution refers to the number of output bits that the ADC can generate per conversion. This figure determines the smallest input signal that the system can represent. Resolution also defines the smallest incremental change to the analog signal that the ADC can express.

Speed has to do with the device’s sampling rate – in other words, what is the highest number of conversions per second that the ADC can handle? Sampling rate is determined by how much time it takes to perform a single conversion. That figure determines how many samples per second are possible in a best-case scenario.

Accuracy is relatively straightforward. How closely does the output match the input? How much of the output is the desired signal? Generally, we assess accuracy in terms of the noise present in the output signal, using a figure called the signal-to-noise ratio (SNR), where higher is better (more signal per given amount of noise). Even in an ideal ADC, there will be some amount of noise present since rounding must necessarily occur in order to digitize an analog signal (quantization noise, explained below). Higher resolution generally leads to higher accuracy as well, since the smaller the rounding error, the more true the digital output is to the analog input.

Quantization noise is one of several noise types that contribute to the device’s accuracy. This type of noise deserves its own mention because quantizing noise is unavoidable in analog-to-digital conversion. Simply put, when a continuous set converts to a discrete set, we can expect to lose some information. We refer to that lost information as quantizing noise, and it manifests as a sawtooth noise signal. With high enough resolution, it is possible to overcome quantizing noise functionally, but it remains an inherent part of the ADC process.

Overview of Common ADC Architectures

The various ADC designs have their own strengths and weaknesses. As a result, your project and intended use case will largely dictate which type of ADC you choose to use. A clear idea of what you want your device to achieve will help you prioritize the four factors we described above and guide you toward the right type of ADC architecture.
The most common ADC architectures are:

- Flash
- Successive approximation (SAR)
- Delta-sigma
- Pipelined

For a given architecture type, the higher an ADC’s resolution is, the lower its speed will be (and vice versa), since higher resolution means more data to convert. Pipelined ADCs, however, are a somewhat unique method: they combine some of the best qualities of SARs and flash-type ADCs, managing to achieve both high speed and high resolution.

While flash ADCs are large and expensive, their speed makes them quality candidates for converting analog video recording to digital, a process that deals with huge amounts of data. SARs are very popular in data-acquisition and instrumentation applications, where ultra-high speed is less critical, and accuracy reigns supreme.

Delta-sigma architecture (one of the newest designs) can boast astounding accuracy, but it’s also the slowest of the popular designs, making it well-suited to high-fidelity audio applications. In these instances, capturing even small nuances is critical, but total amount of data isn’t extreme (compared to video, for example). As a result, you’ll see delta-sigma designs employed largely for digital audio and instrumentation.

Finally, pipelined ADCs are becoming more and more popular due to their ability to combine both reasonably high resolution and speeds. Pipeline designs are essentially a more refined version of the SAR, and they are well-suited to a wide range of applications, including:

- Ultrasonic medical imaging
- Digital video
- High-speed internet scenarios such as cable modems, xDSL, etc.

As for accuracy, it results from the combination of resolution and sampling speed. The resolution determines amplitude precision and rounding error (and thus the quantization noise and a baseline loss of accuracy). Meanwhile, sampling speed determines the accuracy and precision of timing (the more times a second the source is sampled, the more accurate the timing will be). An ADC stat sheet will capture overall accuracy and specific quantization noise in the “signal to noise ratio” line. You’ll want this value to be as high as possible (more signal power per given amount of noise).

Other Items to Consider

Don’t forget to consider the practicalities. Different ADC units vary widely regarding:

- Physical size
- Number of possible inputs
- Power requirements

Flash ADCs, for instance, are not well-suited to portable devices because their many components affect size and power draw. Anyone intending to build a portable device should probably look past flash ADCs and consider something more suited to a smaller device and power system. Similarly, many ADCs only allow two inputs, so if your application requires more than two possible inputs, you can save time by limiting your search to multichannel ADCs.

Remember, regardless of which architecture you choose, the minimum sampling frequency (called the Nyquist rate) should be at least twice the maximum frequency of the input. Based on the expected frequency of the incoming data, you’ll want to be sure your sampling rate is high enough.