Control Engineering

Charlie Masi

What is the difference between digital and analog measurement systems?
May 7, 2007

This is a very concise overview of a very involved subject. For an in-depth treatment, see Understanding New Developments in Data Acquisition, Measurement, and Control, published by Keithley Instruments.

Digital measurements are fundamental to the use of computerized measurement and control systems, but, as my friend Steve Scheiber once said, “The real world is analog.”

By that he meant that the measurable parameters nearly all engineers run into are continuous in value and in time. That is, they can take on any conceivable value within their dynamic range, and they always have some value. That is not true, however, of digital representations of those analog values. Digital measurements can take on only a limited number of pre-defined values and are valid only at certain pre-defined times.

The process of extracting digital measurements from analog signals is called digitization and occurs in two steps. The first step, called sampling uses an analog circuit called a sample and hold (SH) to determine the average value in a predefined time slot called a sampling interval, and holds it (usually in the form an amount of charge on a capacitor) until it’s time to take the next sample. The second step, called digitization, uses an analog to digital converter (ADC) to produce a digital bit pattern four or more bits long that represents an approximation to the (analog) value in the SH.

Digital measurements represent an analog signal (red line) as a series of bit patterns representing values (blue diamonds) of samples taken during small time intervals (green vertical bars).

Do not confuse the sampling interval with the sampling rate. The sampling interval tells you how long the SH looks at the analog signal to capture its sample. The sampling rate tells how often the SH goes back to get another sample. Generally, the sampling interval is shorter (sometimes very much shorter) than the time between samples.

Most engineers use the term resolution to indicate the number of bits the ADC uses to represent the value in each sample. It would be better, however, to call that precision of the measurement because the output bits serve the same function as the fiducials on a ruler or the face of an analog meter. On an analog meter, the result you quote is the value indicated by the nearest fiducial you can see on the meter. It’s the closest value you can “read” off the meter.

The ADC precision is not, however, always the precision of the measurement. Most ADCs determine the most significant bit first, then move on to successively less significant bits. (Don’t confuse this effect with the term “successive approximation” ADC. That term indicates a particular method of doing the analog-to-digital conversion, but other circuit types also determine the most significant bit first, then move on to finer distinctions.) Any particular ADC takes a certain amount of time to work out the whole bit pattern representing the sample value. Most computer-based measuring systems allow you to sample faster than would be required to get full ADC precision. When sampling faster than the ADC can follow, you get garbage in the least significant bits. I’ve seen 12-bit measuring systems whose real measurement precision drops to 4 bits when operating at maximum sampling rate.

Analog measurements, on the other hand, use a continuous phenomenon, such as the turning of a dial needle, to represent the signal value. The venerable D’Arsonval meter movement balances the restoring force of a spring with the magnetic force created by an electric current coursing through a small coil of wire. The more current, the more magnetic force, which requires more restoring force to balance, which (by Hooke’s Law) requires more stretch in the spring, which, finally, calls for more pointer deflection. Everything is continuous in both current-value and time. By the way, the term analog comes from the idea that the meter deflection behaves in a way that is analogous to the current's behavior.

Posted by Charlie Masi on May 7, 2007 | Comments (0)