TERMINOLOGY USED IN INSTRUMENT ACCURACY

Rick Williams
Rawson & Co., Inc.
P.O. Box 924288, Houston, TX 77292

The purpose of this paper is to offer a brief explanation

and discussion of many key terms used in describing
instrument accuracy. The terms included within this
discussion are most commonly used for defining
performance standards with primary sensing elements
typically used in the measurement of flow, level pressure
and temperature instruments. Many of the terms used
may apply to controllers, recorders and final control
elements. However, the focus provided herein is the
primary element device. The specific devices include
transmitters (differential pressure and temperature) and
flow meters (e.g., magnetic, vortex, turbine, variable area
and positive displacement).
This paper is written for the benefit of the typical user of
instrumentation products to include instrument engineers
and technicians. A comprehensive discussion of
precision measurement must address calibration and
traceability issues. The scope of this paper will be limited
to discussing the terms associated with the application
of instruments rather than addressing the issues of
maintaining accuracy. Calibration is a key issue in
determining the continuous performance of an
instrument and is worthy of discussion. However, it is
complex enough to warrant a separate discussion.
A fundamental understanding of gas laws and the effects
of compressibility are necessary for selecting and
applying instruments used for gas measurement. This is
obvious since the majority of the flow measurement
devices used in industry are volumetric, and volume
changes under actual conditions. We will begin our
discussion with the assumption that this base
understanding exists. There will be no debate regarding
the merits of direct or indirect mass measurement
devices nor a comparison the advantages of mass
measurement over volumetric. The decision to use any
specific technology for a solution must be weighed upon
the need for mass measurement vs. volume, the
accuracy requirements from the measurement, the
overall installed cost of the solution while taking into
consideration user and industry standards and the
additional ownership costs affected by maintenance and
life expectancy of the product.
Many of the concepts discussed herein will be painfully
obvious to the experienced instrument user. The goal is
to provide a logical discussion of most key terms and
thereby offer a condensed reference guide for any user
of instrumentation products. These terms cross the
boundaries of ISA, ASME, AGA, ANSI, API and various
manufacturers terminology.

ACCURACY TERMS
Accuracy is a composite statement of performance that
defines the quality of the instrument measurement. It is
the difference between the reading of an instrument and
the true value of what is actually being measured.
Accuracy is normally expressed as plus or minus a
percentage of either reading, calibrated span or the full
scale of the instrument. Accuracy is one of the most
critical factors to consider when applying an instrument
for a given application. Accuracy is also a term that is
most misunderstood by the typical user. The reason for
this misunderstanding is due to the myriad of effects on
instrument accuracy and the ability of manufacturers to
offer varying interpretations for the expression of
instrument accuracy. In the world of marketing this is
commonly known as creative specmanship.
ACCURACY STATEMENTS

% Reading (% Rate)
% Calibrated Span
% Full Scale

% Input
FIGURE 1.

As stated in the discussion on accuracy, the specification

may be expressed in terms of a percentage of reading,
calibrated span or full scale. As illustrated in Figure 1,
the comparison is enlightening. An instrument accuracy
stated as a percent of reading maintains a constant
window of error throughout the measurement range. A
device with an accuracy expressed as a percent of span
or full scale possesses an ever-widening envelope as
2002 PROCEEDINGS

PAGE 271

AMERICAN SCHOOL OF GAS MEASUREMENT TECHNOLOGY

the measurement reading moves farther downscale. This

makes it imperative that the user operates the instrument
as high as practical in the span in order to maintain
accuracy.
Absolute accuracy is a term that defines how the
performance of an instrument relates to a traceable
standard.
Composite Accuracy includes the total effect of accuracy,
linearity and repeatability on an instrument at reference
conditions.
Reference Accuracy is the baseline accuracy for many
instruments. This specification does not include some
of the effects that cause instrument error such as
temperature and static pressure effects. Reference
accuracy is the percentage of error associated with the
instrument operating within designed constraints under
reference conditions. This is the most liberal of accuracy
statements and is commonly misinterpreted as a
benchmark for evaluating one instrument against another.
To further confuse the user, this accuracy statement may
not take into consideration errors induced by the output
mode, such as the error associated with digital to analog
conversion necessary to obtain a current output.
Reproducibility is an ISA term that is the combination of
linearity, repeatability, hysteresis and drift.
System Accuracy is a term used to define the overall
accuracy of a process measurement involving more than
one component used in series or parallel. This is an
excellent means to determine the total error induced on
a process variable that will be used for recording or
controlling purposes. Individual instrument errors must
be calculated as an error of reading (not a percentage
error) for a given measurement point. The combination
of errors then is averaged using a statistical approach
such as taking the square root of the sum of the squared
errors. For example, a flow instrument has an error of
plus or minus one percent of calibrated full scale. The
scale is 0-100 SCFH. The reading is 50 SCFH. The error
is plus or minus 1 SCFH. The recording or controlling
device has an error of plus or minus one half percent of
range. The range is 0-100 SCFH. The reading is 50 SCFH.
The error is plus or minus .5 SCFH. Using the formula:
(12 + .5 2)1/2 = a system accuracy of plus or minus 1.1
SCFH.
TERMS ASSOCIATED WITH ACCURACY
Damping does not affect instrument accuracy but does
have an effect on the quality of the control loop. Damping
is used to slow the response of the instrument to process
changes. It is necessary to use damping when there is
process noise and/or input fluctuations that are
undesirable for control. Damping may be implemented
by mechanical or electrical conditioning. Damping is a
means to average the process variable over time in order
to stabilize the output to the controller. For that reason

PAGE 272

2002 PROCEEDINGS

AMERICAN SCHOOL OF GAS MEASUREMENT TECHNOLOGY

(particularly with electronic means) the adjustment is

expressed in terms of seconds of damping. Damping
may be applied to an instrument with poor installation
technique or inappropriate application of the instrument
for the process. For this reason, it is important to
recognize when applying damping that the need is valid.
Deadband: All instruments exhibit a point where some
change in process data cannot be measured. This error
is commonly expressed as a percent of reading.
Deadband is simply the ability of the instrument to react
to small process variable input changes. Deadband may
be associated with resolution because it does affect the
ability of an instrument to measure continuous data.
However, resolution addresses the ability to continuously
measure the process variable. Deadband only addresses
the ability to monitor minor changes from a given process
variable reading. An instrument with a deadband of plus
or minus one tenth of a percent and a span of 100 psi
has the ability to not read small changes of plus or minus
1 psi at any reading point. If a resolution of plus or minus
1 psi is necessary, a tighter span or an alternate
technology is advised. Deadband has an effect on
instrument accuracy when very small changes are
expected.
Filtering is sometimes confused with damping. Filtering
is actually the ability of an instrument to accept or reject
an input signal. There is more than one interpretation for
filtering an instrument signal. Filtering is commonly used
to ignore intermittent spike input signals to the instrument
that may not be actual changes to the process input.
For example, noise may be measured by vortex flow
meters or magnetic flow meters from hard solids in a
liquid flow application. The solids interfere with the
continuous measurement of the flow signal by
introducing an error normally interpreted as a flow spike.
Damping of the signal may help to alleviate the problem.
However, this type of signal spike is irrelevant to the
measurement of the process. Therefore, the ability to
ignore the spike is advantageous. Another common type
of filtering is the use of capacitors on a DC power supply
to minimize the effects of AC ripple voltage.
Hysteresis: A true and accurate measurement of a
process variable shall yield a beginning measurement
and an ending measurement cycle that is equal. That is,
the upward curve from zero to one hundred percent and
the downward curve from one hundred percent back to
zero are identical. Any deviation from these two curves
is defined as hysteresis. Hysteresis is also linked to
deadband. Since deadband affects the ability of the
instrument to react to small instrument input signals, this
effect amplifies the hysteretic error.
Linearity is described in ISA Standard S51.1 as the
deviation from the calibration curve of an instrument from
a straight-line relationship between zero input and 100%
input. Ideally this is a forty-five degree slope. Most
instruments do not possess a linear output with respect
to input under reference or actual conditions.

Depending upon the instrument, there are many variables

that can affect linearity. For example, most differential
pressure transmitters use a sensor technology that is
inherently nonlinear under reference conditions. As
process and ambient temperature and static pressure
change during actual conditions, there are effects on
linearity. Pressure transmitters vary by class. A baseline
for a pressure measurement is the common pressure
transducer. The transducer converts a pressure signal
input into a voltage signal output. A pressure transmitter
extends the pressure input range and typically provides
for a current output in order to extend the transmitted
distance. There is the added value of compensation for
these transmitters to stabilize the output for the negative
effect of temperature on the linearity of the output. This
transmitter class is referred to as conventional. The
pressure effect is compensated by zero adjustment while
under static pressure conditions. The top of the
transmitter class is referred to as smart due to the
added capability of higher performance standards and
the ability to offer intelligent communications from the
transmitter for configuration and diagnostics information.
Smart transmitters have multivariable sensors for
temperature and pressure compensation. The sensors
feed information to microprocessors that manipulate the
output by comparing the input process variable, ambient
temperature and static pressure (flow transmitters) with
a characterization algorithm stored in an EPROM within
the transmitter. The characterization feature of smart
transmitters offers the ability to accurately produce an
output with respect to input change as the actual
environmental conditions change.
Piping effects (flow profile effects): The installation of the
instrument has the potential for greatly affecting
accuracy. Whether the application is liquid, gas or steam,
the flow profile entering the instrument typically dictates
the ultimate performance. Reynolds number constraints,
turbulence, grounding, buildup on pipe walls,
degradation of orifice plates, positioning of elbows,
valves, reducers, cavitation, multiphase flow, internal pipe
imperfections and other variables all play into the ability
of a specific instrument technology to measure
accurately. It is beyond the scope of this paper to address
these effects. Many instruments are blamed for
incompetent measurement when actually those
instruments have suffered from incompetent installation
guidelines. The guidelines for standard installation
applications are difficult to keep up with. To complicate
the issue, manufacturers vary on the recommended
installation practices for a given technology. This adds
to the confusion for users. The point is that it is
reasonable to adhere to the manufacturers
recommended installation practices. This should be
compared with industry standards such as those
published by ISA and API.
Pressure effects on instrument accuracy typically apply
to the static pressure effect on zero or span of a
differential pressure transmitter. For conventional
transmitters, this error may be partially resolved by

adjusting the zero while the transmitter is under operating

line pressure. If there are swings in operating pressure,
then this error cannot be properly compensated for with
conventional transmitters. New generation smart
transmitters offer characterization with the ability to sense
pressure changes and automatically correct for the error
thereby minimizing the effect.
Rangeability is closely linked with turndown. Most
instrument technologies offer the ability to adjust the
measurement range. It is important to discuss
rangeability because the use of instruments is limited to
certain ranges in order to maintain accuracy. Instruments
may be over-ranged and operate reliably within design
limits. However this is often at the sacrifice of some
accuracy. The degree by which an instrument may be
over-ranged depends upon the technology and the
manufacturer of that technology. Under-ranging an
instrument often has negative effects on accuracy. Some
accuracy statements qualify the degree of accuracy
depending upon where the instrument is operated within
the span capabilities. A case can be made that liberal
use of range application can save the cost of ownership
by reducing the number of spares. After all, a transmitter
with a maximum range of 100 IWC and a minimum range
of 5 IWC can be used for the 70 IWC and the 10 IWC
applications. However, the combined possible errors as
stated in typical specifications will yield lesser
performance on the lower span. Prudent users will apply
the appropriate range as they would use the proper tool
for the job and carefully evaluate the desired results.
Repeatability is the ability of an instrument to precisely
duplicate a measurement while operating under the same
conditions while the input signal is made in the same
direction. The input signal may be consecutive
movements from zero to one hundred percent or vice
versa. Many individuals believe that repeatability is as
valuable as a high degree of accuracy. In actuality, good
accuracy cannot be achieved without good repeatability.
What causes poor repeatability? The instrument has a
poor repeatability specification, the instrument has a
defect, or there are one or more influences causing the
problem with a good instrument. Some of these
influences are piping installation problems, cavitation
from valves and pumps, hammering from vapor pressure
problems, temperature changes (ambient and process),
noisy power or grounding and vibration.
Resolution is the ability of the instrument to continuously
measure and transmit all process variable data. In digital
systems the smallest interval that exists between two
measurement samples defines resolution. Most current
technologies offer microprocessors for the benefit of
enhanced performance. The tradeoff from the use of
microprocessors is speed and resolution. Digital
measurement involves sampling of the process variable.
While the process variable is being sampled, there may
be small dead spots where process data is lost. On many
smart instruments, there may be multivariable data
sampled. Microprocessor instruments typically process

2002 PROCEEDINGS

PAGE 273

AMERICAN SCHOOL OF GAS MEASUREMENT TECHNOLOGY

input information via analog to digital converters. The

resolution of these types of instruments is dependent
upon the resolution of the converters. For example, a
twelve bit resolution A/D converter has the resolution to
within 1/2 of 4,096 counts or in electrical terms 1 mV for a
2.048 volt input. Converters include signal conditioning,
sampling, multiplexing and conversion processes and
there are accuracy and linearity specifications tied to
converter. This is may or may not be incorporated into
the accuracy specification of the instrument. In addition,
there are roundoff errors to consider with converters. New
technology is enhancing the ability to increase the
resolution and accuracy of converters. The fact that this
resolution error exists is not all bad. It must be considered
however in the application of the instrument.
Other sources of resolution errors involve the specific
technology used. DC Pulsed magnetic flowmeters
sample a portion of a process variable at a fixed
frequency typically between 7.5 Hz to 30 Hz. The lower
the frequency equals lower resolution and therefore
inadequate response time for noisy slurry applications
or fast acting positive displacement pumps. Vortex
flowmeters interpret vortex swirls commonly using piezoelectric sensors that appear to the electronics as a
square-wave pulse input. The frequency is directly
proportional with the velocity of the fluid. As line size
increase each pulse represents larger volumes of flow
thereby reducing resolution. Fieldbus networks offer an
advantage of transmitting multivariable data as well as
diagnostics from instruments. The data obtained from
fieldbus instruments is valuable but because the data is
intermixed with multivariable and diagnostic data. Since
this data is shared with data from other instruments over
the same wires, Fieldbus strategies must be weighed
with the need for resolution and response time. Smart
instruments with built-in PID algorithms help to address
this issue by truly distributing the control to the field
where the primary and final elements are co-located.
Many current technologies involve a means of sampling
data in order to provide an output that implies real
process conditions. Near instantaneous measurement
is possible with some analog instruments but is not
possible with instruments with digital electronics. Oddly
enough, there is an explosion of accuracy statements
that imply enhanced performance from smart (by default
digital) instruments. The accuracy statements are correct
but the response time and resolution of the instrument
must be taken into consideration to qualify performance.
The ability of an instrument to respond to and transmit
process data quickly is important for some critical flow
or pressure applications associated with fast-acting
positive displacement pumps, compressor control and
safety systems.
Response Time is the ability of an instrument to react to
process variable changes. This is closely related to the
terms resolution and damping. Most electronic
instruments have a delay from the moment the process
variable changes and the moment the change is indicated

PAGE 274

2002 PROCEEDINGS

AMERICAN SCHOOL OF GAS MEASUREMENT TECHNOLOGY

by the output signal. This delay will depend upon the

type of instrument, the circuitry design (analog vs. digital),
the deadband for the measurement reading, and the
setting of damping adjustment that may exist on the
instrument. The time delay in producing an output signal
is typically expressed in terms of milliseconds. A very
fast response time will be less than ten milliseconds.
This type of device is appropriate for fast acting pressure
measurement such as compressor applications.
Microprocessor instruments may have a response time
expressed in hundreds of milliseconds. These
instruments may be very accurate and appropriate for
many flow, level and pressure applications. A common
sense approach is to compare the measuring instrument
with the controller. Comparing the response time of the
instrument with the desired sampling and execution rate
of the controller provides data that is helpful in making a
qualified decision about the use of that instrument for
an application. A controller with sampling and execution
rates in excess of 500 milliseconds does not necessarily
need an input device with a response time of ten
milliseconds.
Reynolds Number Effects: Reynolds number (Re) is
defined by formula as the velocity of the flow times
density times pipe diameter divided by viscosity. This
specification is important for flow meter applications. A
great deal of time may be used to discuss Re. To keep it
simple there are fundamentals that help to keep this
concept into perspective. First, there is the difference
between laminar and turbulent flow. Re numbers below
2,000 are laminar. Between 2,000-4,000 the Re is
transitional. Above 4,000 the Re is turbulent. Most flow
meters need a Re in the turbulent region to operate
accurately. The Re calculation is relatively
straightforward. If this calculation is strictly applied and
other installation requirements are adhered to, then the
instrument should be expected to perform within
specified tolerances.
Span Shift may have a negative effect on the accuracy
of an instrument. If an instrument span is changed, there
needs to be a calibration verification of the change. The
instrument should be expected to perform within
published specifications providing that a calibration
device is used to confirm the zero and span of the
instrument. Again, on pressure transmitters, temperature
and pressure effects may cause span shift. Also,
adjustment of zero alone may affect the span.
Temperature effects must be broken into two categories:
process temperature effects and ambient temperature
effects. Process temperature affects instrument
performance in two ways. For liquid processes
temperature variation has an affect on viscosity. For those
flow instruments that are viscosity sensitive, the variation
in viscosity must be understood and compensated for
or there will be an error associated with the viscosity
shift. Secondly, many sensors are typically located in
close proximity to the process, there is the potential for
an effect on the sensor with process temperature

variation. In addition, for gases there is an effect on the

mass calculation with respect to temperature changes.
Ambient temperature has an effect on many sensors and
electrical components. Those instruments with an analog
current output typically have a higher error associated
with ambient temperature changes than the instruments
that use either a pulse output or a digital communications
output. An instrument specification must be scrutinized
to determine if the accuracy statement is qualified to
include the temperature effect for the type of output
intended. Depending upon the technology, ambient
temperature may have an effect that is equal to or greater
than the reference accuracy of the instrument. Ambient
temperature may affect both zero and span of an
instrument.
Turndown may be expressed in more than one way
depending upon the instrument technology It may be
expressed as either the ratio of the maximum
measurement capability to the minimum measurement
capability while maintaining accuracy, or as the ratio of
the maximum span capability to the minimum span
capability while maintaining accuracy. The latter definition
normally applies to pressure transmitters and must be
carefully evaluated for a given application. Instruments
universally have a point on the low end of the scale where
accuracy degrades. Specifications must be qualified to
determine where accuracy falls off and to what degree
the degradation occurs. Turndown is a term where much
creativity is used with specmanship.
Uncertainty and accuracy are closely related. However,
uncertainty is more definitive when used in the context
of instrument and system errors. This is due to the strict
guidelines and equations associated with flow instrument
standards. Uncertainty must take into consideration the
actual operating conditions. Formulas used to calculate
are common in flow measurement. An example of a
standard formula can be obtained from ANSI/ASME
MFC-2M-1983 titled Measurement Uncertainty for Fluid
Flow in Closed Conduits. The uncertainty revealed in
these types of calculations are important in that they
take away the focus on individual instrument errors by
assigning a relative importance for each variable that
impacts the flow calculation. These formulas introduce
two very interesting concepts. First is the concept of
two types of errors: Bias vs. Precision. Bias error may
be obtained from the manufacturer specifications.
Precision error is calculated using independent
comparison tests. This concept is worthy of discussion
for two reasons. First, bias error is commonly referred to
as reference accuracy. Since instruments are not used
under reference conditions, this accuracy statement is
not valid. The errors associated with having an effect on
reference accuracy should be taken into consideration.
Second, not all manufacturers use the same standard to
publish specifications. Some are more conservative than
others in order to mass produce instruments. For a given
product, there may be variance in absolute accuracy from
one instrument to the next, although all the instruments
in a particular class would meet the published

specification. This is obviously done with the intent to

reduce manufacturing costs. Precision error calculation
may allow for evaluation of a population of like-products
in order to determine worst case and average accuracy.
Taking into consideration the effects of temperature and
pressure on accuracy requires some sophisticated test
equipment that is normally associated with independent
laboratories. Of course independent evaluation may
reveal the conservative nature of some manufacturers
to understate product specifications.
The second important concept revealed from the use of
standard flow calculations is the use of square root of
the sum of the squares for all errors after a weight has
been applied to the individual errors. Because of the fact
that errors are expressed in terms of plus or minus, it
can fairly be argued that no instrument is likely to indicate
a worst-case scenario whereby all of the combined errors
will sum to a total error in one direction or the other. The
use of the square root of the sum of the squares offers a
fair averaging of these errors. A simple example of this
analysis is given under system accuracy.
Zero Shift is caused by numerous factors. What causes
this error? The answer should be evident within the
published product specifications. Zero shift that is
measured outside the specifications might be caused
by such variables as power or ground induced noise or
a defect within the product. Process and ambient
temperature effects and process pressure changes on
some devices might inhibit the instrument to reproduce
a zero input. Some technologies offer a feature to perform
auto-zeroing to minimize zero shift. Smart transmitters
offer real-time temperature and pressure compensation
to help minimize the effect. The knowledge that zero shift
can affect your measurement will help to define a
calibration program to correct for this problem.
Other sources of error include mounting position effect
of the instrument, RFI interference, power supply
influenced error and gross calculation errors in the
controller. The terminology used to describe instrument
accuracy is broad with some overlap. There is casual
misuse of some of these terms by users and sales people
whereby confusion is created in the marketplace. Care
must be exercised to learn the terminology and apply it
appropriately.
The ability of an instrument to perform accurately is
dependent upon many variables. It is the human factor
in the equation that ultimately determines how well a
technology performs. Process knowledge must be
coupled with the knowledge of instrument technology
and proper installation techniques to ensure success.
Process operations and support personnel are living in
a world of ever tightening constraints with higher
performance expectations. The tools available to help
guide the way are imbedded in the knowledge base of
those who have been there before. A wealth of data is
available in the form of text books and technical papers
and bulletins. This author is grateful for the availability of

2002 PROCEEDINGS

PAGE 275

AMERICAN SCHOOL OF GAS MEASUREMENT TECHNOLOGY

these tools and acknowledges authors of those sources

by offering a list of valuable references.
REFERENCES:
1. ABB Instrumentation Flow Products Bulletin 10E-12
Predicting Flow Rate System Accuracy as reprinted
from the proceedings of the 1979 ISA Symposium,
William S. Buzzard.
2. How Accurate is Accurate?, Control Magazine,
William L. Mostia Jr. PE, June 1996.
3. Flow Measurement, Practical Guides for
Measurement and Control, ISA Text, David W.
Spitzer, Editor, 1996.
4. Industrial Flow Measurement, ISA Text, David W.
Spitzer, 1995
5. Digital Control Devices, Equipment and Applications,
ISA Text, J.A. Moore, 1986.

PAGE 276

2002 PROCEEDINGS

AMERICAN SCHOOL OF GAS MEASUREMENT TECHNOLOGY