Five Analog Front-End Mistakes That Undermine ADC Accuracy

Driving Higher-Resolution Analog to Digital Converters (ADCs): Five Analog Front-End Mistakes That Limit Measurement Accuracy

Modern high-resolution ADCs have made it possible to measure incredibly small changes in voltage, current and sensor outputs. Devices such as the MCP3561R enable precision measurements for industrial sensing, instrumentation and process-control applications. Yet in many systems, converter resolution alone does not determine overall performance.

As ADC resolution increases, errors that were once insignificant become increasingly visible. Offset voltage, settling behavior, noise, electromagnetic interference (EMI) and input bias current can all limit measurement accuracy long before the ADC itself reaches its limits. The challenge is no longer simply digitizing a signal. It is ensuring that the signal presented to the ADC accurately represents the physical quantity being measured.

Let's examine five common analog front-end mistakes that can prevent high-resolution measurement systems from achieving their full potential.

Mistake #1: Focusing on ADC Resolution Instead of the Entire Error Budget

When evaluating a data acquisition system, it is tempting to focus on ADC specifications such as resolution, signal-to-noise ratio, sampling rate and linearity. While these parameters are certainly important, they tell only part of the story.

The analog front end establishes the signal quality that ultimately reaches the converter. Any error introduced before the ADC input becomes part of the measurement. The front-end amplifier can be a significant contributor to the overall system error budget. In many precision measurement systems, input offset voltage is one of the dominant error sources because it is introduced before the ADC and is amplified along with the desired signal. This is especially true when conditioning low-level sensor outputs that require high gain. Consequently, designers who use a general-purpose op amp in a precision signal chain may find that amplifier offset error limits system accuracy long before the ADC reaches its theoretical resolution.

Example 1: Amplifier Offset Error Comparison

Consider a precision sensor interface using:

• MCP3561R ADC
  • 24-bit resolution
  • 2.4V reference
• Front-end amplifier configured for:
  • Gain = 20 V/V

Suppose we compare two amplifier options:

1. MCP6486, general-purpose op amp, Vos max of 1.6mV
   The output referred error due to the offset is 32mV, or over 223,000 ADC codes
2. MCP60711, precision op amp, Vos max of 180µV
   The output referred error due to the offset is 3.6mV, or close to 25,000 ADC codes, close to nine times less error

Although both amplifiers may be suitable for many signal-conditioning applications, the MCP60711's lower offset voltage reduces the output-referred offset error from 32 mV to 3.6 mV in this example, helping preserve more of the accuracy available from a high-resolution ADC.

Mistake #2: Ignoring Amplifier Settling Requirements

Modern data acquisition systems often monitor multiple sensors using an analog multiplexer. In these systems, the amplifier output must settle to its final value before the ADC begins the next conversion.

Unfortunately, settling time is frequently overlooked during the design process. Engineers may focus on gain-bandwidth product or sampling rate without considering how quickly the signal-conditioning circuit can respond to changing inputs.

This becomes particularly important when switching between channels that have significantly different signal levels. If the amplifier output has not settled before the ADC samples the signal, the resulting conversion error may be much larger than the converter's quantization error.

As channel count and sample rate increase, the amount of time available for signal acquisition continues to shrink. Under these conditions, amplifier dynamic performance can have a major impact on overall system accuracy.

Example 2: Comparing Settling Performance Between the MCP6071 and MCP60711

Consider a system in which eight channels are multiplexed, with each channel being sampled at 10kSPS. In this case, the available time per conversion is 12.5µs. Let’s also suppose that the sensor signal at the multiplexer switches from 0.5V to 2.0V. A front-end amplifier configured for a fixed gain of 15V/V must respond to this 1.5V step before the ADC converts.

Using the first order equation for op amp settling:

Let’s consider two op amps:

Although the MCP6071 1.2MHz op amp may seem sufficient for this application, utilizing the above settling time equation yields a settling time of 16.8us. This exceeds the available time per ADC conversion, resulting in measurement error. The MCP60711 10MHz op amp, on the other hand provides a settling time of only 1.75us, well within the available conversion time.

In high-throughput measurement systems, settling time can become a larger source of error than converter resolution itself. Selecting an amplifier with sufficient dynamic performance helps ensure the signal has reached its final value before conversion begins.

Mistake #3: Underestimating Op Amp Noise Contributions

A high-resolution ADC can only measure signals that rise above the analog noise floor. As converter resolution increases, understanding how much noise is contributed by the signal-conditioning amplifier becomes increasingly important.

One common misconception is that an amplifier's voltage noise density specification directly indicates how much noise will appear at the ADC input. In reality, total noise depends on several factors, including measurement bandwidth and circuit gain.

The amplifier's noise contribution must be integrated across the bandwidth of interest and then referred to the ADC input. Only after performing this calculation can a designer determine how much of the converter's available resolution is being consumed by amplifier noise.

This distinction becomes especially important in precision sensing applications such as bridge measurements, current sensing and industrial instrumentation, where small signal changes must be resolved with high accuracy.

Example 3: Estimating Op Amp Noise Using the MCP6006 and MCP60711

Let’s consider a bridge sensor circuit with an amplifier front-end configured for a fixed gain of 20V/V and a measurement bandwidth of 10kHz.

We have two op amps:

• MCP6006 general-purpose op amp
  • Input noise density: 22nV/√Hz
• MCP60711 precision op amp
  • Input noise density: 5.4nV/√Hz

Given the measurement bandwidth of 10kHz, the MCP6006 and MCP60711 op amps have an integrated input noise of 2.2µV RMS and 0.54µV RMS respectively. Considering the closed loop gain of 20V/V, the output-referred noise is 44µV RMS and 10.8µV RMS respectively. 

In this example, both amplifiers may be suitable choices depending on the required measurement accuracy. If the sensor, reference and other system components contribute significantly more noise than the amplifier, the difference between 44 µV RMS and 10.8 µV RMS may have little impact on overall performance. However, in systems where every microvolt matters, the lower noise contribution of the MCP60711 can help maximize the usable resolution of the ADC. This is one reason designers sometimes select a higher-bandwidth amplifier even when the required signal bandwidth is relatively modest.

Mistake #4: Forgetting That Modern Measurement Systems Live in an RF Environment

Today's measurement systems rarely operate in isolation. Wireless connectivity is now common in industrial automation, building management, process control and Industrial Internet of Things (IIoT) applications.

As a result, precision analog circuitry often shares PCB real estate with Bluetooth^®, Wi-Fi^® and cellular radios. While these technologies provide important connectivity benefits, they can also introduce new sources of measurement error.

RF energy can couple into high-impedance analog nodes and create apparent offset shifts that look like legitimate sensor signals. In many cases, engineers initially suspect calibration problems, software issues or sensor drift, when the actual culprit is electromagnetic interference (EMI).

As wireless connectivity continues to spread throughout industrial systems, EMI robustness is becoming an increasingly important consideration in signal-chain design. This is particularly true for low-level sensor interfaces where small disturbances can translate directly into measurement errors.

Example 4: RF-Induced Measurement Errors in a Connected Sensor Node

Consider a wireless industrial sensor node containing:

• Pressure sensor
• MCP60711 signal-conditioning amplifier
• MCP3561R ADC
• Bluetooth radio operating at 2.4 GHz

Suppose that through PCB coupling and nearby antennas, the amplifier input is exposed to 100mVp of RF energy. Without any EMI rejection, that RF signal could potentially create significant measurement errors.

The MCP60711 specifies an EMI Rejection Ratio (EMIRR) of 81dB at 2.4GHz. EMIRR is defined as:

Rearranging:

Due to the enhanced EMI rejection of the MCP60711 op amp, this 100mVp of RF energy only causes an input offset error of 8.9µV. In connected systems, RF interference can masquerade as a valid measurement. Robust EMI performance helps ensure that the ADC is measuring the intended signal rather than environmental interference.

Mistake #5: Ignoring Input Bias Current in High-Impedance Sensors

Input bias current is one of those specifications that often receives little attention because the numbers appear extremely small. However, when high source impedances are involved, even tiny input currents can introduce surprisingly large measurement errors.

This becomes particularly important when interfacing with sensors such as pH probes, electrochemical sensors, photodiodes and leakage-current measurement circuits. These sensors may exhibit source impedances ranging from several megaohms to hundreds of megaohms.

In these applications, the input bias current flowing into the amplifier creates an error voltage across the sensor's source impedance. The resulting voltage error becomes indistinguishable from the signal being measured.

As sensor currents become smaller and measurement sensitivity increases, input bias current can become one of the dominant contributors to overall system error.

Example 5: Input Bias Current Effects in a High-Impedance Sensor Interface

For this example, let’s use an electrochemical sensor with a source resistance of 10MΩ. We will compare two different front-end amplifiers: one with a bipolar input transistor and one with a CMOS transistor.

1. MCP616 precision, bipolar input op amp
   Input bias current: 15nA (typical)
2. MCP60711 precision, CMOS input op amp
   Input bias current: 0.1pA (typical)

As this bias current flows through the source impedance of the sensor, it creates an error voltage—in this case, 150mV for the MCP616 and 1µV for the MCP60711 based on the typical bias current specifications. When working with high-impedance sensors, amplifier input bias current should be evaluated just as carefully as offset voltage and noise. Otherwise, the measurement circuitry itself may significantly alter the signal being observed.

Conclusion

High-resolution ADCs have made precision measurements more accessible than ever, but converter resolution alone does not guarantee system accuracy. As ADC performance improves, analog front-end design becomes increasingly important.

Offset voltage can dominate low-level measurements. Settling behavior can limit multi-channel systems. Noise can reduce usable resolution. EMI can appear as false sensor signals. Input bias current can distort measurements from high-impedance sensors.

By looking beyond ADC specifications and evaluating the complete signal chain, designers can build measurement systems that are more accurate, repeatable and robust. Addressing these five common mistakes early in the design process helps ensure that the performance promised by modern converters translates into real-world measurement results.

The good news is that many of these challenges can be addressed through thoughtful signal-chain design and appropriate component selection. Whether an application requires a cost-effective general-purpose amplifier, a high-performance precision amplifier or a zero-drift amplifier for the lowest possible offset and drift errors, Microchip offers a broad portfolio of operational amplifiers to help designers optimize their analog front ends. When paired with high-resolution converters such as the MCP3561R, these devices can help maximize measurement accuracy and system robustness. Learn more about our operational amplifier portfolio here.