Have you ever studied the A / D converter data sheet in depth? This is an adventure, even for people with analog engineering backgrounds, and it can be daunting for many manufacturers from digital and software backgrounds. The parameter table is helpful, but for manufacturers, understanding the work at hand is more important than filtering the specification page when selecting an A / D converter.

Source: Georgia Institute of technology, through Youda learning city

Usually on the manufacturer’s website, we will see many categories of a / D converters, which shows how professional and optimized the products have become. In addition to some unique cases, I think manufacturers will encounter one of the four basic cases that require an A / D converter, so optimization will help because many tradeoffs have been designed in these components.

Trapped sampling

Over the years, microcontroller based design has proved that a tested principle is “sleep” behavior. When the sampling rate is very slow (in seconds or even minutes), our idea is to wake up, sample and perform calculation and communication, and then return to sleep as soon as possible. This behavior can minimize power consumption and prolong battery life.

In this case, please select a converter with as few bits as possible to represent the range of the input signal. For example, when designing a weather station, the range provided by the 8-bit converter is sufficient to read the closest temperature. Fewer bits means shorter conversion time and less power consumption. There may be an on-chip A / D converter on the accessible microcontroller pin, so there is no need for an external converter at all.

Low rate sampling

Faster sensors with output data rate (ODR) up to 200 Hz need to consider sampling. Rule of thumb: the sampling speed is at least 2.5 times faster than the required data. Strictly speaking, Nyquist says 2x is the minimum, but add some margin. For example, if the rocket altimeter needs an accurate reading every 1 / 10 second in flight, set the sampling rate to 25 Hz or faster.

The idea here is to control the trade-off between sampling rate and resolution bit and power consumption. Have the idea of significant digits (ENOB). In a noisy environment, without stronger filtering or oversampling, it may be difficult to obtain a useful resolution of more than 16 bits, resulting in the development of conversion time and power consumption in the wrong direction.

Low noise, precision conversion

One of the areas where professional components come in handy is low noise, precision sampling. These are not horticultural varieties; Noise reduction steps are taken at each stage of the conversion, and care needs to be taken in the circuit board layout to avoid eliminating these reductions. Here, numbers such as signal-to-noise ratio (SNR), dynamic range (usually abbreviated as DNR) and total harmonic distortion (THD) come into play. Typically, these components have a medium sampling rate of up to 1 million samples per second (MSPs).

Faster signal analysis

The field of radar design and 5g signal processing may be beyond the scope encountered by most manufacturers, but there are more common applications to meet faster component requirements. The sampling rate of hundreds of MSPs makes applications such as software defined radio (SDR) and DIY oscilloscope possible.

Faster sampling increases the importance of low-pass filtering to strongly reject things higher than twice the Nyquist frequency, which may really mess up fast Fourier transform (FFT) analysis. It also puts forward higher requirements for memory throughput, and may need to use FIFO to buffer the results for analysis on a slower memory bus, which will increase the system cost.

finish the work

Note that we did not compare the various A / D architectures — flash memory, pipelining, SAR, and delta sigma. These are interesting things. If you are really interested in the way they work, you can read a lot of articles about them elsewhere in planet analog.

But I will take a risk here: for all applications except cutting-edge applications, chip manufacturers have completed most of the work of selecting and optimizing the architecture for their intended use. For example, they did not attempt to adapt the successive approximation register (SAR) design to higher sampling rate applications that encounter constraints; They are optimizing the SAR architecture for medium sampling rates.

In most cases, if you understand the resolution and sampling rate, filtering and oversampling, and how noise affects the results in some applications, manufacturers can quickly narrow the selection range of a / D converters according to the application.

Browsing the data table may provide information, and careful study of programming models such as polling operations and free running sampling can help you understand how to make full use of parts. You can also guide the selection of which a / D converters are provided on the distribution board through the manufacturer’s site, which also makes prototyping and software development easier.

After ten years of working on missile guidance system in general dynamics, don dingee became a preacher of Motorola VMEbus and single board computer technology.

Editing: hfy

Leave a Reply

Your email address will not be published. Required fields are marked *