Dave KressIn industrial, or data-acquisition projects, designers are likely to confront one or more of these issues: 
• Digitizing an input signal that extends over a very wide dynamic range, such as delivered by an environmental sound pressure meter able to detect signals over a 60-dB to 80-dB signal range; 
• Accommodating signals from different sources that exhibit quite different signal ranges; 
• Resolving small changes around a certain value, so that the objective would be to expand the range around that point.

If a relatively low-resolution ADC is used — say with 10-bits of effective resolution — for the high level signal resolution would be close to 10 bits. However, for the low level signals, if they are less than 10 percent of full scale, effective resolution might be no more than six or seven bits. So in many cases, 10 bits — equivalent to 0.1 percent — would suffice for a sensor with an accuracy of only 1 percent. However, for lower level signals the effective resolution might be less than 1 percent.

Approaches to these Design Issues
There are a number of ways to approach these design issues but the three primary ones would be: 
• Connect a programmable-gain amplifier (PGA) in front of a relatively lower resolution ADC. 
• Apply the input signals to buffer amplifiers connected ahead of the ADC 
• Use a high-resolution ADC.
Let's evaluate these approaches, one at a time:

The PGA Approach. Historically, the PGA approach has been popular largely because, when paired with lower cost ADCs, it winds up being less costly than higher-resolution ADCs. This approach is particularly useful if all of the input signals are near zero Volts, but cover wide dynamic ranges. Shown in Figure 1 is a simplified schematic of an ADC with a built-in PGA. 

Figure 1. An ADC with built-in PGA (simplified schematic).

This might be the case in a process control system that is monitoring signals from a variety of sensors that have different ranges, such as sound pressure meters. If gain-ranging is used on a wide-dynamic range signal, the key error that can show up is 'cross-over mismatch'.

This means that when the PGA switches to a different gain value, the digitized output may jump a little up or down at that point. This therefore requires careful matching of the gains at each level to reduce that effect. This issue is of less importance when multiplexing signals from different signal sources. However, it depends on whether the system is designed with a fixed gain for each signal, as depicted in Figure 2, or has dynamic gain-switching as would be the case with wide-ranging single inputs.

The issues with the gain ranging approach include: 

• When driving a 12-bit ADC and putting an amplifier ahead of it with a gain of 27 = 128, the effective input noise and offset voltage of the amplifier must be essentially 18-bit accurate. This would be a problem using a fixed-gain op amp, and a more severe problem with a PGA that must be switched. So, the accuracy requirement is moved from the ADC to the PGA, without any benefit. 

Figure 2. An ADC with independent buffer amplifiers built-in PGA (simplified schematic).

• You must know something about the signal in order to switch the gain. The over-range output of the ADC can be used for this, in conjunction with software, or you can do it with comparators. The process is messy, and the switching time can be a problem. (Remember the old gain-ranging DVMs and how slow they were when they changed ranges?) 
• A simple analysis could be performed on a precision low-noise op amp at a gain of 128: Calculate the effective output noise and offset voltage and compare it to an LSB of a low resolution ADC. However, the linearity of the op amp in a high gain mode could be a problem.

Information on designing PGAs with discrete components can be found in Reference 1.

The Multiple Buffer Amplifier Approach. The multiple buffer amplifier approach may be used if the sensors or signal sources are at some distance from the data acquisition unit containing the ADC. See Figure 2.

A Single High-Resolution ADC. What is attractive about a single high-resolution ADC is its simplicity. See Figure 3. If a 16-bit ADC is used, losing three, four or five bits to a smaller dynamic range signal reduces the effective resolution of that signal to 11 to 14 bits. However, this is still sufficient for the accuracy of most transducers, because its accuracy is on the order of 0.05 percent or better.

Since the prices of these devices have recently dropped to $5 or less, cost will seldom be a factor. If higher effective resolution is needed, or a wider dynamic range needs to be accommodated, ADCs in the 18- to 24-bit range can still provide cost-effective performance as well as a much simpler system.

Figure 3. A single high-resolution ADC (simplified schematic)

Using a high-resolution ADC is clearly the design choice for resolving a small change in signal around a certain point off zero. This is an alternative to using a digital-to-analog converter (DAC) to offset most of the signal. This is still a viable choice for some situations.

Application Concept – Multiple Thermocouples. In many process control situations, the process temperature must be known at several different places within the system. These temperatures can differ can a few degrees to possibly a hundred degrees or more. While it is important to know the absolute temperature at each location, it is often just as important to know the temperature variation around a certain point for fine control, and also to know the temperature difference between certain points. The cold-junction temperature must then also be known from a temperature sensor near the ADC, which will be much different from the process temperature, and possibly measured with a different type of sensor. A single high-resolution ADC tracking all of these sensors will give the best accuracy and differential accuracy for each measurement point. Any small drift in the full-scale of the ADC will affect all of the inputs by the same amount, so that the fine measurements will be preserved, this would be more difficult to achieve with different gain networks around multiple amplifiers.

Application Concept – Photodiode measurement. Photodiodes are used in instrumentation in many areas from astronomy to gas chromatography. Photodiodes are especially useful because of their wide dynamic range, approaching six orders of magnitude. When used in a wide-dynamic range application, the output of the photodiode can be sent through a logarithmic amplifier, which compresses the signal, or it can be sent directly to a high resolution ADC. The log-amplifier approach tends to give the same signal accuracy of 0.1 to 0.5 percent throughout the signal range, and is suitable for many applications. The high resolution ADC approach bypasses the precision resistors and tuning required for the log stage, while providing very high accuracy, and predictable accuracy for all signal levels.

1."Basic Linear Design", Chapter 2, Secction2.13
2."The AD8250 Programmable Gain Amplifier"

Dave Kress is director of technical marketing for Analog Devices, Inc. He has BSEE and MSEE from MIT in electrical engineering. . He can be reached via email at