Do Standards Matter In Evaluating ADCs?

Our flight had just started its descent when the gentleman sitting next to me turned to me and we started talking engineering. (He had seen me reading an engineering periodical.) My seatmate mentioned that he was a member of the Institute of Electrical and Electronics Engineers (IEEE) and that he had made a personal goal to serve on a standards committee.

I asked about what standard he was working on. It had something to do with power-station security. We continued our conversation until we arrived at the airport terminal and went our separate ways. During our discussion, I remarked that standards were very important and told him how, in my corner of the industry, it was amazing to me that we didn’t have an IEEE standard that addressed definitions of specifications and how to test them for analog-to-digital converters (ADCs) until the year 2000.

I find this remarkable because analog-to-digital conversion has been known since at least the 1920s, and commercially available ADCs began to emerge in the 1960s.1 For decades, manufacturers of ADCs could define specifications for these devices and tests for those specifications purely on their own. Naturally, there became some de facto standardization of how tests were done, but guidelines from an actual standards body didn’t exist.

The first real standard work on ADCs began in the 1980s, culminating in IEEE1057,2 and later, IEEE1241.3 IEEE1241 specifically targets the ADC device itself, as opposed to an entire data acquisition or recording system. IEEE1241-2000 was the first real standard for manufacturers of ADC components, and it was updated in 2010.

The Evaluation Begins

The primary task in evaluating an ADC is determining its transfer function. Ideally, a converter would have a transfer function similar to that in Figure 1, which shows a transfer function for a 3-bit converter. In an ideal converter, the width of each code is exactly the same and a straight line can be drawn through the midpoints of each code “plateau.” In practice, this is not always the case. Determining the transition points and code widths is critical to ADC testing and characterization, as it lets us see where the actual transfer function deviates from the ideal.\

1. The ideal ADC transfer function has evenly spaced transition points that are exactly 1 least significant bit (LSB) wide.

To find the real transfer function, the IEEE standards suggest several possible test setups and methods. One method is a complicated servo-loop system that requires a digital-to-analog converter (DAC) of a higher resolution than the ADC under test. Another method uses a sine-wave oscillator, which must have total harmonic distortion and noise (THD +N) at least 20 dB better than the expected signal-to-noise-and-distortion (SINAD) of the ADC under test.

For example, an ideal 16-bit ADC has a signal-to-noise ratio (SNR) of 98 dB and no distortion (it’s ideal, after all), so the SINAD would be 98 dB. Testing this ADC would require an oscillator with a THD+N better than –118 dB. When you start looking at high-resolution ADCs, the sine-wave generator may require filtering to achieve a spectrally pure signal, if the generator alone is not adequate.

Finding such high-resolution DACs or spectrally pure oscillators, and building the complicated test setups needed, is something that ADC manufacturers are willing and generally equipped to do. Some of these instruments can be quite costly. If your business is making ADCs, though, the investment is worth it. But what does the person who is actually designing a system with the ADC do about evaluating and testing the ADC?

The Right Tools

Many designers turn to the manufacturer’s evaluation boards and kits to perform their tests (Fig. 2). These systems make it easy to connect the ADC under test to a computer over a USB connection and provide software to capture the data and analyze it.

2. ADC manufacturer’s evaluation boards and kits, like this ADS1281EVM-PDK from Texas Instruments⁴, often provide the complete data capture system, but not the signal sources. Software provided often implements tests similar to the IEEE standards.

Some people attempt to use evaluation kits to duplicate the results shown in the ADC datasheet specifications. This is likely not completely possible, especially with high-resolution converters, because the required sine-wave generators probably aren’t available. With some care, though, meaningful results often can be seen using the evaluation board and its software.

The evaluation hardware and analysis software generally operate in what one could term a block-mode, meaning that it collects a block of a fixed number of samples and sends it to the software, and the software analyzes that block or record of data. Most of the tests in the IEEE standards are defined so they operate on such blocks of data.

The question is if the tests outlined in IEEE1241 really help you evaluate an ADC for its suitability for your system. If you were the trusting type, what would it tell you beyond what you can see in the device datasheet? The answer for many people is that beyond seeing the device actually in action, the evaluation board provides a reference design and layout that can be helpful in guiding its usage in the actual system.

For some, though, the IEEE1241 tests aren’t what are required. Depending upon the type of ADC, some people would like to use the evaluation board and software as a digital oscilloscope or chart recorder, streaming data constantly as opposed to doing block-by-block data transfers.

Some customers I’ve dealt with who were trying to determine long-term stability or drift performance have asked to record the data to the computer’s hard disk, for hours or even days on end. While IEEE1057 more typically addresses these sorts of uses, neither standard discusses long-term drift or stability tests.

Most manufacturers’ evaluation boards and software won’t support these types of usages or tests. Should ADC component manufacturers enable their evaluation boards and software to implement streaming as well as block capture, in addition to other features needed to perform these types of tests not specified in the ADC-related standards?

For ADC manufacturers, standards absolutely matter. But as circuit designers who use those ADCs, those same standards used by the ADC manufacturers may not really help you. What standards are you thinking about (or not) when evaluating ADCs? I’d like to know. Send me an e-mail and tell me, and I’ll share the results in a subsequent article.