Data integrity series: Input accuracy and precision issues

This is the third in a series of discussions with Mike Glass, owner of Orion Technical Solutions, on ensuring data integrity for process control and industrial automation systems. In previous articles of this series, we examined signal filtering and digital degradation factors. Now we'll explore how measurement errors originate at the source—in the sensor itself or its installation.

Greg: In our previous discussions, we covered how filtering shapes signals and how digital systems can degrade data quality. But even the best filtering and digital architecture can't save you if the original measurement is wrong. Let's talk about input accuracy and precision issues—where do they typically start?

Mike: The reduction in training investment has had a big effect on instrumentation and controls in many ways – one of which is that instruments are increasingly installed, configured, and setup incorrectly – and then are also often tested improperly, resulting in problems and mistakes going undiscovered.

According to studies by the ARC Advisory Group and major instrument vendors, primary element issues and installation problems now account for over 60% of measurement system errors, while actual transmitter calibration issues contribute only 5-15%. Yet at most facilities, the vast majority of instrument maintenance time is still spent on (possibly unnecessary) transmitter calibrations.

Greg: What are some of the major mistakes and problems that are occurring?

Mike: The biggest issues now come from three main areas: improper installation, incorrect sensor selection, and poor sensor configuration. Let's break these down.

Improper installation issues include impulse line problems like pockets, plugging, or leaks; incorrect mounting orientation; insufficient thermowell/probe installation; poor signal loop or loop-power design, and improper electrical connections or grounding (among others). These physical installation issues can create significant measurement errors.

Incorrect sensor selection involves choosing technologies ill-suited to the application, like using a capacitance probe in a coating application or an uncompensated DP level transmitter in a variable-density fluid. These types of gross errors are often induced after initial design as products, systems, and operations change.

Configuration problems include incorrect settings like wrong RTD coefficients, thermocouple types, damping values, or pressure compensation parameters. A surprising number of transmitters operate with mismatched or incorrect configurations that create persistent errors.

Greg: What are some of the most common installation issues you see affecting instrumentation data quality?

Mike: Pressure and DP measurements suffer most frequently from impulse line issues. A while back I was on a deepwater oil production platform where beautifully installed tubing runs had zero slope. This created dozens of potential gas or liquid pockets throughout the impulse lines that caused erratic and inconsistent readings. Worse yet, there were numerous large traps and pockets formed in some parts of that picturesque tubing where they simply ran tubing over or under obstructions. If the installers aren't trained (or at least closely checked or guided by personnel who are) these things will happen – even on critical systems.

For temperature measurements, inadequate thermowell immersion or poor thermal conductivity is a common problem. Sometimes a probe is replaced with one that is slightly shorter in a pinch and then left in service. Adequate immersion typically requires 8-10 times the diameter of the sensing element and should typically have solid contact with the bottom of the thermowell. Probes using thermal conductive paste can also have problems if the paste dries out or becomes 'crusty'.

Level measurement problems often stem from incorrect reference leg configurations and errors in the calculations or assumptions. This is extremely common source of errors.

Diaphragm and capillary lines should be equal lengths and at same temperatures and should have the same exposure to heat/sunlight.

Greg: I'm intrigued by what you're calling "primary element issues." Can you elaborate on those?

Mike: Primary elements are the mechanical components that interact directly with the process—orifice plates, flow tubes, displacers, thermowells, diaphragms, and similar devices. These components physically transform the process variable into something the transmitter can measure.

Unlike the electronics, these components experience wear, corrosion, coating, and mechanical damage. An orifice plate might develop a buildup that changes its effective diameter – or may erode over time. A thermowell might become coated with process material, slowing its thermal response.

One lesson I learned about twenty years ago was how tiny changes in the sharpness of the edge of an orifice in a DP flow system can impact the readings. I saw an orifice that looked fine to my untrained eye – but it failed the micrometer tests. And sure enough, when replaced with a new orifice the readings were spot on. I have to admit I was surprised by that one – but have now seen that many times since.

What makes these issues particularly challenging is that they're often gradual and variable. The errors they introduce might depend on flow rate, temperature, or other operating conditions, creating nonlinear error patterns that can be mistaken for legitimate process behavior. This makes them especially problematic for data analytics.

Greg: What about sensor configuration issues? I imagine those can be just as problematic.

Mike: Absolutely. Configuration errors create persistent, systematic bias in measurements. Some common examples include:

1. RTD Alpha Value Errors: Using an incorrect alpha value for an RTD creates temperature-proportional errors. These errors become more pronounced at temperature extremes, potentially causing significant measurement discrepancies in high-temperature applications. The blind assumption of the 0.00385 alpha causes many problems – especially in plants where certain skids or systems may have come from places that aren't as standardized on a 0.00385 alpha for RTD's. A tip I use to confirm the alpha of an RTD is to place it in a dry block calibrator set to 100C. a PT100 with a 0.00385 alpha will read (approximately) 138.5 ohms. If a PT100 with 0.00392 (or 0.003916) alpha is being used it will read approximately 139.2 ohms. That may seem minor – but it results in large errors as temperature gets further from 0C.
2. Thermocouple Type Mismatches: A Type K thermocouple configured as a Type J creates substantial measurement distortion. This is surprisingly common in facilities where multiple thermocouple types are used and color-coding standards aren't consistently followed. Many people are unaware that different countries have different color code standards for thermocouples.
3. DP Level Calculation Errors: Incorrect height values, specific gravity settings, or reference leg configurations in DP level measurements create systematic errors that vary with level. These configuration errors can significantly impact inventory calculations and process control.
4. Compensation Parameter Errors: Modern multi-variable transmitters apply numerous compensation factors for ambient temperature, static pressure, fluid properties, and other variables. Errors in these parameters compound, especially at operating extremes.

What makes configuration errors particularly insidious is their persistence. Unlike mechanical issues that might show symptoms like noise or instability, configuration errors simply produce wrong values that look completely normal. The data is smoothly, consistently, precisely wrong.

Greg: I'm guessing the older "calibrate everything annually" approach doesn't catch many of these issues?

Mike: Exactly. Traditional calibration procedures often completely miss primary element and installation issues because they focus solely on the transmitter. A typical technician simulates inputs to the transmitter and verifies outputs, but never checks whether the primary element is actually measuring the process correctly.

ISA-RP105.00.01 (Management of Instrumentation Calibration and Maintenance Programs) provides excellent guidance for modernizing maintenance approaches. It recommends comprehensive measurement system verification rather than simple transmitter calibration, and it provides statistical methods for determining appropriate intervals based on actual performance data.

One particularly critical aspect of this standard is its emphasis on "as-found" data collection before any adjustments are made. This practice allows organizations to analyze actual drift rates and patterns, which typically reveals that modern transmitters require far fewer adjustments than traditional maintenance programs assume.

Another related suggestion is to 'leave it alone unless it is out of spec'. A huge portion of the problems that modern transmitters experience are actually caused by well-intending (but inadequately trained) personnel attempting to 'perfect' a reading. Most plants would benefit by setting up realistic tolerance envelopes and using them properly rather than making those frequent adjustments that likely cause more problems than they resolve.

Greg: How can organizations shift their focus to address these more significant issues?

Mike: The most effective approach involves redirecting some of the time currently spent on routine calibrations toward comprehensive inspections and targeted verification procedures.

For example, a simple 5-minute "walk-by" inspection can detect many developing problems before they affect process control. This includes visual inspection of the installation, checking for transmitter alerts or warnings, and comparing the local display value with the control system reading. For additional validation, compare readings between different instruments measuring the same variable (if not already done in control system logic).

Another high-value activity is checking impulse lines for proper slope and absence of pockets, inspecting orifice plates for wear or buildup, verifying thermowell installation depth, and examining installation compliance with best practices.

Configuration verification is equally important. Review the parameters in smart transmitters against design specifications, especially after any maintenance or replacement activities. Parameters to verify include sensor types, alpha values, compensation settings, and physical dimension values used in calculations. It is also wise to have a system to save instrument configurations so they can be loaded from an 'official' master source when needed – vs field personnel needing to guess on details like damping or other settings.

Greg: You mentioned an additional voltage check during calibrations in one of our previous discussions. Expound on that.

Mike: When performing an instrument calibration on a typical 4-20mA loop powered transmitter it is helpful to record the voltages across the transmitter at 4mA and 20mA (or at low fault and high fault) values to help ascertain and track any developing issues in the signal loop.

Get your subscription to Control's tri-weekly newsletter.

Moisture in j-boxes or associated leakage current, excess impedance, power supply changes, will each cause changes in the overall loop electrical characteristics, and will all impact the voltage across transmitter. If the voltage changes, you know something is happening to the loop and can investigate – before it causes problems.

Many transmitters have this functionality built into their diagnostics capabilities as well, and could even report these issues in real time if on a HART or similar network.

In my opinion, this functionality is grossly underutilized considering how well it would identify developing problems and how much of a percentage the signal loop is in modern instrumentation applications.

Greg: Those all sound like valuable alternatives to the traditional "calibrate everything annually and assume it's all fine" approach. Any final thoughts on maintaining measurement accuracy?

Mike: I'd emphasize that measurement accuracy is a system property, not just a device specification. A highly accurate transmitter connected to an improperly installed primary element might deliver significantly worse actual accuracy. Organizations need to think holistically about their measurement systems.

I also can't stress enough the importance of training. Many technicians haven't received adequate education on modern smart instruments or measurement system verification. Without this knowledge, they often apply (and pass on) practices developed for 1980s technology to modern equipment, sometimes introducing more errors than they fix.

According to research from the ISA, technicians actually introduce errors approximately 17% of the time when making calibration adjustments. This statistic alone should cause organizations to reconsider their approach to instrument calibrations.

Based on where the errors are nowadays (per ARC and other studies) I'd also suggest doing more frequent visual checks of transmitter displays (when available) to check for any alerts or warnings and to cross check that transmitter PV matches the controller or HMI data – this quick check can be a huge boost to overall reliability and one tech can typically check 50 transmitter loops in the time it would take to complete a single calibration check. Or better yet – gather the digital data and setup to do this all automatically, via one of the many ways available today (at least for most critical loops).

Greg: Your insights highlight how maintaining data integrity requires focusing on the right problems, not just following traditional practices. As we wrap up this part of our series, what would you say are the key takeaways for improving input accuracy and precision?

Mike: I'd summarize with these points:

1. Modern transmitters drift far less than older models, so recalibration frequency can typically be reduced significantly using the methodology in ISA-RP105.00.01.
2. Primary element issues and installation problems now cause far more measurement errors than transmitter calibration issues in modern systems.
3. Redirect maintenance resources toward regular inspections of the entire measurement system rather than just transmitter calibration.
4. Document configuration parameters carefully and verify them during maintenance activities.
5. Train technicians on modern verification procedures that address the actual sources of measurement error.
6. Use the "as-found" data from calibration checks to analyze drift patterns and optimize maintenance intervals.
7. DO NOT make calibration adjustments unless they are outside of the allowable tolerance (unless demanded by regulations or other requirements).

By addressing these aspects of input accuracy and precision, organizations can dramatically improve their overall measurement system performance—and consequently, their control system performance and data quality.

Greg: Mike, thanks for these insights. I offer the following advice as to how to maximize measurement accuracy and minimize maintenance considering device selection and installation based on my personal experience. Lower maintenance costs and financial benefits from increases in process efficiency and capacity often outweigh increased initial hardware costs. 

1. For temperature measurements preferably use RTDs for applications less than 600 degrees F sheathed and spring loaded in tapered thermowell whose tip for vessels is past baffles near agitator and for pipelines is near centerline with an insertion length of 8 to 10 thermowell diameters.
2. For pressure measurements preferably use direct mounted transmitters to eliminate need for impulse lines.
3. For level measurements preferably use radar (for which type of radar is best to use see TABLE 2.2.4.1 Selection and Best Practices for Typical Radar Applications in Process/Industrial Instruments and Controls Handbook, Sixth Edition, McGraw-Hill 2019)
4. For liquid flow measurements preferably use Coriolis mass flow meters that have a measurement accuracy and rangeability that is an order of magnitude better than differential head and vortex meters with an inherent accurate correction for changes in density that could be caused by composition changes. The meter can in many cases provide an inferential measurement of composition. Second choice is magnetic flow meters if fluid conductivity is good enough since these meters like Coriolis meters have a great rangeability due to an accuracy in percent reading instead of percent span and are minimal piping installation requirements reducing noise. 
5. For pH measurements preferably use glass bulb or dome electrodes with glass designed to handle acids or bases and temperature and ideally a flowing liquid junction reference electrode with second choice being a double junction with replaceable outer junction and electrolyte selected to be compatible with process fluid. Use shroud designs to minimize coating or abrasion. For much more, see Advanced pH Measurement and Control – Digital Twin Synergy and Advances in Technology, Fourth Edition, Wiley, 2024.
6. For pH measurements ensure electrode tip for vessels is past baffles near agitator and for pipelines is near centerline. Electrodes in pipelines should be more than 10 pipe diameters downstream of pump or static mixer and apart. Velocity should be greater than 1 fps to minimize electrode response time and around 5 fps to minimize coatings.
7. For pH measurements and any other measurements prone to resolution, response time and reliability issues, use middle signal selection to inherently ignore a single failure of any type, an in-progress calibration, and to reduce noise, drift, and slow response. 
8. Minimize measurement calibration span to maximize accuracy since many instruments and analog input cards have accuracies in percent span and to better recognize trends, disturbances, and noise by reducing data compression. 

Next time, let us discuss how final control elements —valves, actuators, and positioners—can create further data integrity issues. Even with perfect measurements, valve problems can introduce their own forms of data corruption.

Mike: I look forward to it, Greg. Output device issues are indeed a major source of process data corruption—and often the most overlooked. While most organizations focus heavily on measurement quality, the mechanical components that actually control the process and inject variability, process noise, and other errors often receive far less attention than they deserve.

Mike Glass is the founder of Orion Technical Solutions. His company specializes in hands-on training and assessments of I&E personnel.