EPA-AA-TSS-83-8-B
Technical Report
EPA RECOMMENDED PRACTICE
FOR NAMING I/M CALIBRATION GAS
September 1983
Technical Support Staff
Emission Control Technology Division
Office of Mobile Sources
Office of Air and Radiation
U. S. Environmental Protection Agency
Ann Arbor, Michigan 48105
-------
Table of Contents
1.0 INTRODUCTION
2 . 0 BACKGROUND
2.1 EPA Regulations Concerning Calibration Gas
Accuracy
2.2 EPA Efforts to Help States Meet Warranty
Requirement for Gas Accuracy
3.0 DISCUSSION OF RECOMMENDED PRACTICE
3.1 Analysis of Pure Components
3.2 Cylinders
3.3 Instrument Preparation and Calibration
3.3.1 Using Hexane
3.3.2 Definition of Linearity
3.3.3 Generation of Calibration Curve for Linear
Instruments
3.3.4 Generation of Monthly Calibration Curve for
Non-Linear Instruments
3.4 Pre-Analysis Calibration Check - All Instruments
3.5 Analysis of I/M Calibration Gas Cylinders
3.5.1 Dependence on Previous Analysis of Bulk Mixture
or Mixture Stream
3.5.2 Homogeneity Check
3.5.3 Determination of Component Concentrations
3.6 Analysis of Five or Fewer Cylinders
3.7 Calculating the Accuracy of I/M Calibration Gases
3.8 Cylinder Labeling and Documentation
3.9 Audits of I/M Calibration Gases
3.10 Age of I/M Calibration Gas
3
4
5
5
6
6
6
8
9
10
11
11
11
12
13
14
14
15
15
Appendix 1
Appendix 2
Appendix 3
Appendix 4
Section 85.2217 of Emission Control
System Performance Warranty Regulations
Overview of Analytical
Procedures
Alternative Procedures for Compensating
for Instrument Drift
Accuracy Discussion
-------
1.0 INTRODUCTION
This report presents a set of procedures scientific gas
manufacturers can use to blend, analyze, and label
calibration gases for infrared inspection analyzers used in
vehicle inspection and maintenance (I/M) programs. The
procedures are intended to be used on gases a manufacturer
labels with the phrase "made in accordance with the EPA
Recommended Practice For Naming I/M Calibration Gas." EPA
has established no legal requirement that gas manufacturers
follow this set of procedures. However, claims by a gas
manufacturer that a gas has been made in accordance with this
set of procedures may create rights and obligations under
existing State or Federal law, particularly those related to
fair marketing practices and product warranties.
EPA encourages States and other I/M authorities to procure
gases named according to this practice for their own use, and
to require inspection stations to procure them to ensure they
are obtaining accurate calibration gases which meet the terms
of the Emission Control System Performance Warranty, and to
improve the general quality of their I/M programs. EPA
recognizes that there may be many other techniques and
analytical procedures for blending and naming I/M gases which
could achieve the same results as this recommended practice.
However, these are the only techniques that may be followed
if a gas manufacturer is to label a gas with the phrase "made
in accordance with the EPA Recommended Practice for Naming
I/M Calibration Gas".
Propane, hexane, carbon monoxide (CO), and carbon dioxide
(C02) can be named with these procedures as two component
mixtures of one of the components in nitrogen (N2> , or as
multi-component mixtures such as CO and propane in N2 or
CO, hexane, and C02 in N2- The possible combinations are
shown in Table 1.
Table 1
Possible Mixtures of I/M Gases
Covered by the Recommended Practice
Number of Components
Two Three Four
Propane in N2 Propane, CO in N2 Propane, CO, C02 in N2
Hexane in N2 Hexane, CO in N2 Hexane, CO, C02 in N2
CO in N2 Propane, C02 in N2
C02 in N2 Hexane, C02 in N2
CO, C02 in N2
-------
-2-
A separate report entitled "EPA Recommended Practice for
Naming I/M Calibration Gas: A Discussion For I/M Programs"
(EPA-AA-TSS-83-8-A) discusses issues concerning calibration
gas which are of interest to States and other I/M
authorities. The body of this report discusses the actual
procedures gas manufacturers must use. The Appendix contains
the relevant part of the Emission Performance Warranty
Regulations which require traceability of +2.0% to NBS for
gases used to calibrate I/M analyzers, an overview of the
analytical procedures, an explanation of an alternative
procedure for accounting for instrument drift, and a
discussion of techniques used to estimate the accuracy of
gases named according to this practice.
-------
-3-
2.0 BACKGROUND
An inspection and maintenance (I/M) program is a State or
locally run program in which registered vehicles are required
to obtain and pass a tailpipe emission inspection on a
periodic basis. Vehicles that have tailpipe emissions
greater than State or locally established emission standards
are required to obtain maintenance to pass that standard.
The emission test (called a "short test") can be conducted by
the State (or locality) or a contractor to the State, and
this is referred to as a "centralized program." The emission
test can also be conducted by private garages which are
licensed by the State. This latter system is called a
"decentralized" program.
The primary inspection and diagnostic tool of these I/M
programs is a non-dispersive infrared (NDIR) analyzer which
is capable of determining the concentrations of hydrocarbons
(as hexane), carbon monoxide, and, in some cases, carbon
dioxide in raw vehicle exhaust. These analyzers need
periodic calibration and maintenance to keep their accuracy,
as age, operator misuse, or changes in pressure, temperature,
and other operating variables can render them inaccurate.
Many analyzer manufacturers recommend a periodic check of
their analyzers with a calibration gas. State or local I/M
program regulations also require a periodic calibration
check. This check is performed by flowing a calibration gas
of known concentration into the analyzer and determining
whether the analyzer is reading this calibration gas
correctly. If it is not reading this gas correctly, then a
simple adjustment (i.e., a calibration) of the analyzer can
usually be performed which will result in the analyzer being
accurate again. However, the accuracy of the calibration gas
used is very important in determining analyzer accuracy with
this maintenance check. If the labeled concentration of the
calibration gas is significantly different from the true
concentration in the cylinder, the analyzer could become
significantly misadjusted. An operator who had performed the
calibration would be unaware that he/she had actually
misadjusted the analyzer, since he/she trusted the label on
the calibration cylinder to be correct.
2.1 EPA Regulations Concerning Calibration Gas Accuracy
EPA has promulgated Emission System Performance Warranty
Regulations for 1981 and later vehicles which entitle a
vehicle owner to emission-related repairs at the
manufacturer's expense if, among other things, the vehicle
fails an "approved" emission short test. One condition is
that the analyzer used to conduct the emission short test
must be checked with gases "traceable to NBS standards +2.0%"
-------
-4-
within a week of the test (40 CFR 85.2217). The relevant
portion of this page is provided in Appendix 1 to this
document. There are other requirements for these gases in
the Warranty regulations; however, these are of interest more
to States ordering gases than to manufacturers producing
gases and are discussed in the companion report,
EPA-AA-TSS-83-8-A. EPA thinks that many States implementing
I/M programs will want to make Warranty protection available
to consumers participating in their I/M programs.
Consequently, these States will try to make sure that the
gases used to periodically calibrate analyzers used in the
I/M programs meet the Warranty requirement of having an
accuracy of +2.0% to NBS standards.
2.2 EPA Efforts to Help States Meet Warranty Requirement
for Gas Accuracy
In order to improve the general quality of I/M programs, and
to help States in meeting the accuracy requirement of the
Warranty with respect to calibration gas used to check
emission analyzers, the EPA has published this Recommended
Practice which States can require gas manufacturers to follow
in naming and labeling calibration gas for the I/M programs
in those States. A State requirement would be imposed
directly by the State in its own purchases of gas and/or
indirectly by establishing rules and regulations which
require other I/M gas users (contractors or licensed
inspection stations) to buy only gas that the manufacturer
certifies was named and labeled according to this Recommended
Practice.
-------
-5-
3.0 DISCUSSION OF RECOMMENDED PRACTICE
3.1 Analysis of Pure Components
Scientific gas manufacturers must follow certain procedures
in ordering and analyzing pure components used to make I/M
calibration gas. First, the impurity concentrations in the
nitrogen used as a diluent or a gas to establish instrument
zero shall not exceed 1 ppm equivalent carbon response, 1 ppm
CO, 0.04% C02, and 0.1 ppm nitric oxide. These conform to
the existing Federal regulation specifications for zero gases
(40 CFR 86.114-78) . Propane or hexane used must be
"instrument grade" which is certified to be 99.5% propane or
hexane, and the balance will be primarily other hydrocarbons
(e.g., methane, iso-butane, butane and ethane). An analysis
must be performed of this bulk propane or hexane to
determine, in fact, that the gas used is at least 99.5%
pure. Lastly, the carbon monoxide and carbon dioxide used
must have a total hydrocarbon (THC) count which must not
exceed 100 ppm.
Although these specifications for the pure components might
seem moderately stringent when one considers that the
dilution of the propane, carbon monoxide and carbon dioxide
with nitrogen will result in these trace components
(particularly methane) having very little effect on the
resultant accuracy of the I/M span gas, they nonetheless
contribute to error in naming the I/M calibration gas and
should be limited. Also, the small extra cost of the
components ordered and analyzed to these specifications will
have virtually no effect on the overall cost of making the
I/M calibration gas. The bulk of the cost in making an I/M
calibration gas is incurred in filling and analyzing the
final blend.
3.2 Cylinders and Waiting Time
Any size low pressure or high pressure cylinder may be used
as a container for I/M calibration gas named according to
this Recommended Practice; however, the cylinders must be
constructed of either steel or aluminum. Aluminum must be
used for mixtures containing CO with concentrations less than
1.0% CO unless the gas manufacturer can provide statistically
valid data to show that the stability of the gas mixture is
not degraded by using other types of cylinders. High
pressure cylinders must be fitted with CGA-350 valves. Low
pressure cylinders must be fitted with a CGA 1/4-inch
flare-fitting valve. The use of these valves will allow I/M
calibration gas users to switch suppliers (gas manufacturers)
without having to purchase new regulators.
-------
-6-
In some cases, an incubation period is necessary to order to
ensure the accuracy of the mixture. For CO-C02-propane
mixtures in dry nitrogen, the only potential problem is
nonhomogeneous mixtures. In the case of high pressure
cylinders, once the cylinders are filled, they must be rolled
and/or convection heated for a period of 1 hour to ensure
homogeneity. In the case of low pressure containers, the gas
mixtures must originate from one homogeneous, well
characterized bulk mixture; therefore, no incubation period
is needed.
3.3 Instrument Preparation and Calibration
3.3.1 Using Hexane
If hexane is to be blended and named, the manufacturer must
use in-house hexane standards which were gravimetrically
blended using NBS weights in the place of NBS gas standards
(which are unavailable) for the remainder of this Recommended
Practice. The hexane must have been weighed-in to an
accuracy of +0.1% of the weight being added to a cylinder.
The hexane in-house gravimetric standards must also be used
in the monthly curve generation process (Sections 3.3.3 and
3.3.4) and the pre-analysis curve check (Section 3.4).
3.3.2 Definition of Linearity
The NBS Standard Reference Materials (SRM's) or gas
manufacturers' Certified Reference Materials (gm-CRM's)*
listed in Table 2 and an instrument grade 99.9% pure nitrogen
must be used to determine instrument linearity. A gas
divider which can be used to dilute the highest concentration
standard to at least three equally spaced lower
concentrations is also acceptable for determining instrument
linearity. For the purposes of these analytical procedures,
a linear analytical instrument is defined as one which yields
three intermediate points in the range of 1.0% CO to 8.0% CO,
250 ppm propane to 5000 ppm propane and/or 1.0% CO? to 7.0%
or 14.0% C02 which deviate by +2.0% of point** or less from
*CRM's refer to gas standards prepared and analyzed by a gas
manufacturer according to EPA-600/6-81-010 ("A Procedure for
Establishing Traceability of Gas Mixtures to Certain National
Bureau of Standards SRM's"). CRM's are recognized by EPA as
equivalent to SRM's for establishing traceability.
**+2.0% of point is +2.0% of reading. For example, if a
reading of a certain gas is 6.0% CO, +2.0% of 6.0% CO is
+0.12.% CO.
-------
-7-
Table 2
SRM's or Gas Manufacturers' CRM's
Used to Determine Instrument Linearity*
Allowable
Deviations
For Linear
Propane in N? Instruments
250 ppm +5 ppm
500 ppm +10 ppm
1000 ppm +20 ppm
2500 ppm +_50 ppm
5000 ppm +100 ppm
CO in N?
1.0% +0.02% CO
2.0% +0.04% CO
4.0% +0.08% CO
8.0% +0.16% CO
C02 in
N?
1.0% +0.02% C02
3.0% +0.06% C02
4.0% +0.08% C02
7.0% +0.14% C02
14.0%** +0.28% C02
*Instrument-grade 99.9% pure nitrogen must also be used to
establish the instrument zero. The gas manufacturers'
gravimetric hexane standards should be used to determine
instrument linearity when hexane is to be analyzed.
**This C02 standard must be used in the linearity check if
the C02 concentration to be named is over 7.0%. The 1.0%
C02 standard can be dropped from the check in this
situation.
-------
-8-
a straight line drawn from the point determined by the zero
gas to the highest calibration point. To be considered
linear, the difference between the concentrations indicated
by the intermediate points and the straight line must not
exceed +2.0% of the concentration values of the intermediate
points. The range of allowable deviations from the straight
line are also shown in Table 2.
Linearity need only be determined once per year on a
particular instrument, unless service is performed on the
instrument, in which case the linearity check must be
conducted immediately after service is completed. Also, in
the case of THC analyzers, if either the FID fuel or support
air is changed, a linearity check must be conducted
immediately thereafter.
3.3.3 Generation of Monthly Calibration Curve for Linear
Instruments
If a linear instrument is to be used in the analysis of I/M
calibration gases, at least once per month a calibration
curve must be generated on the instrument in the range that
is to be used to analyze the I/M calibration gas.
A minimum of six cylinders must be used to generate the
calibration curve. One cylinder must be instrument grade
99.9% pure nitrogen, two other cylinders must be undiluted
NBS-SRM's or gm-CRM's which are above and below the
concentration of the I/M calibration gas to be named, and the
other three cylinders can be gas manufacturer primary
standards which have a minimum accuracy of +_2.0% to NBS -
SRM's or gm-CRM's or other gas concentrations obtained by
using a gas divider to dilute either an NBS SRM or gm-CRM or
primary standard. Instrument responses for the five points
other than the zero point should be approximately equally
spaced on the range of the instrument which is to be used to
name the I/M calibration gas. For example, if a 3.0% CO I/M
calibration gas is to named, instrument responses could be
obtained at 1.0% CO, 2.5% CO, 4.0% CO, 6.5% CO, and 8.0% CO.
The 1.0% and 8.0% CO could be NBS-SRM's or gm-CRM's, with the
others being either primary standards or gas concentrations
obtained by diluting the NBS-SRM, gm-CRM's, or primary
standards.
An equation must be calculated from the instrument response
for the six gases. The labeled values of the cylinders
analyzed must be the independent variable, with their
responses being the dependent variable. If all points are
within +_0.5% of the equation, the equation may be used to
name I/M calibration gases in this Recommended Practice.
Inability to satisfy this criterion with an. equation is an
-------
-9-
indication of improperly named primary standards, a
malfunctioning gas divider, and/or malfunctioning analysis
instrumentation. Rectifying these problems will likely
result in the criterion being satisfied.
It is suggested that if an equation yields a calibration line
which meets the above criteria, but calibration points do not
appear to be randomly distributed above or below the line
(e.g., there is a cluster of nearby points on one side of the
line and another cluster or clusters on the other side) then
a higher-order equation should be generated from the
calibration points. This will increase the accuracy of the
I/M calibration gas naming process. However, no inflection
points are allowed in this higher-order equation in the range
of data (see discussion of inflection points in the next
section) . The higher order curve must also meet the _+0.5% of
point criterion stated above.
3.3.4 Generation of Monthly Calibration Curve for Non-Linear
Instruments
If a non-linear instrument is to be used to name I/M
calibration gas, at least once per month a calibration curve
must be generated on the instrument in the range that is to
be used to analyze the I/M calibration gas.
A minimum of eight cylinders must be used to construct the
calibration curve. One cylinder must be instrument-grade
99.9% pure nitrogen, two other cylinders must be undiluted
NBS-SRM's or gm-CRM standards which are above and below the
concentration of the I/M calibration gas to be named, and the
other five cylinders can be primary standards or gases
obtained by using a gas divider to dilute either an SRM,
gm-CRM, or a primary standard. Instrument responses for the
seven points other than the zero point should be
approximately equally spaced on the range which is to be used
to name the I/M calibration gas.
A polynomial equation must be calculated from the instrument
responses to the eight gases obtained during this step. The
labeled values of the cylinders must be the independent
variable, with their responses being the dependent variable.
No inflection points* are allowed in the equation of the
curve generated from analysis of the eight gases over the
range of the data. If an inflection point occurs in an
*Inflection points can be determined by taking the second
derivative of the resultant calibration curve equation,
setting it equal to zero, and evaluating over the range of 0
to 100% full scale on the given range.
-------
-10-
equation in the range of the data, a lesser order equation
should be tried. If all points are within +0.5% of the
equation, the equation may be used to name I/M calibration
gases. Inability to satisfy these criteria with a particular
equation is an indicator of improperly named primary
standards, a malfunctioning gas divider, and/or
malfunctioning analysis instrumentation.
It is suggested that if a particular equation yields a
calibration curve which meets the above criteria for
non-linear instruments, but calibration points do not appear
to be randomly distributed above or below the curve (e.g.,
there is a cluster of nearby points on one side of the curve
and another cluster or clusters on the other side of the
curve), then a higher-order equation should be generated from
the calibration points. This will increase the accuracy of
the I/M calibration gas naming process. However, no
inflection points are allowed in this higher-order curve in
the range of the data. The higher-order curve must meet the
+0.5% of point criterion stated in the previous paragraph.
3.4 Pre-Analysis Calibration Curve Check - All Instruments
Prior to the analysis of each batch* of I/M calibration gas
cylinders,** the instruments being used to analyze the
concentrations of the cylinders must be zeroed with
instrument grade 99.9% pure nitrogen, spanned with an NBS-SRM
or gm-CRM with a concentration at least 80% of full scale of
the range being used to analyze the I/M gas (and adjusted if
necessary) , and then checked with two intermediate NBS or
primary standards whose concentrations must be higher and
lower than the I/M gas to be analyzed. These intermediate
standards must be from the same group of standards that were
used to develop the monthly curve. The equations generated
by the monthly calibration curve for each instrument must be
used in calculating concentrations for the intermediate
standards from their responses. If the calculated
concentrations do not match the labeled concentrations within
+0.5%, new calibration curves must be generated according to
the procedures discussed in Sections 3.3.3 and 3.3.4.
*A batch is defined as more than five cylinders. The naming
process for five or fewer cylinders is discussed in Section
3.6.
**As stated earlier, manufacturers may need to allow an
incubation period between the time at which filling of the
cylinder is complete and the time at which the analysis of
the cylinder begins.
-------
-11-
3.5 Analysis of I/M Calibration Gas Cylinders
3.5.1 Dependence on Previous Analysis of Bulk Mixture or
Mixture Stream
Manufacturers have many different techniques for filling
cylinders. In some of these techniques, an analysis is
performed on each component in the mixture prior to filling a
batch of cylinders. Examples of these techniques are where
analyses are performed on a bulk homogeneous mixture, or
performed continuously on a mixture of gas from the raw
components that flow to the individual cylinders. Other
techniques use no analysis other than that of the pure
components. For example, a manufacturer may mix components
together by partial pressures into a cylinder without
analyzing the mixture.
The first step in the analysis process is the determination
of homogeneity in the I/M gas batch. For filling techniques
in which a previous analysis on every component except the
diluent has been performed on the bulk homogenous mixture or
on the mixture stream (as the cylinders are being filled),
the gas manufacturer need only perform the homogeneity check
on one component in every filled I/M cylinder. Where
previous analysis of every component except the diluent of
the bulk mixture or mixture stream has not been performed,
the homogeneity check must be performed on every component in
every I/M cylinder.
3.5.2 Homogeneity Check
No more than 1 hour may elapse between the time at which the
calibration curve is checked and the time at which the
homogeneity check of I/M cylinders begins. The homogeneity
check is conducted by performing one analysis of the same
component in every cylinder. Instrument drift must be
determined and compensated for by reintroducing a primary
standard with concentration between 80% and 100% of full
scale after analysis of every five I/M cylinders.
Adjustments must be made to the analyzer if the span drift
exceeds +0.5%. If an adjustment is made to the analyzer to
compensate for drift after the analyses of 5 I/M cylinders,
those 5 cylinders must be reanalyzed after the adjustment is
made. One other procedure which can be used in compensating
for drift is discussed in Appendix 3.
A mean response is calculated from all of the analyses of
that component in the I/M cylinders. Any cylinder with a
response which exceeds +0.5% of the mean response can be
reanalyzed. If reanalyzed and still found to exceed +0.5% of
the mean, it must be rejected. if any one of the rejected
-------
-12-
cylinders is reanalyzed and found to be within +0.5% of the
mean response, the cylinder can be reaccepted as a part of
the batch.
3.5.3 Determination of Component Concentrations
A random sample of 10% of the batch and no less than six
cylinders must be selected from the batch for the purpose of
determining the concentrations of each of the components in
the batch. In some cases, such as with high pressure
cylinders, a gas manufacturer may blend gas mixtures directly
into the cylinders by partial pressure. In such cases, the
size of the batch is defined by the number of cylinders being
blended at one time.
In other cases, such as with low pressure cylinders, a gas
manufacturer must first blend a bulk homogeneous mixture into
a large storage vessel and then transfer portions of the
mixture into the individual cylinders that will utlimately go
to the gas user. In this case, the size of the batch is
defined by the amount of gas blended into the bulk storage
vessel. For instance, assume that a gas manufacturer uses a
bulk container to blend enough gas to fill 200 low pressure
cylinders. Once the bulk mixture is blended and homogenized,
the manufacturer then transfers the gas mixture from the bulk
container to individual low pressure cylinders using a
manifold which accommodates 20 cylinders at a time until all
200 cylinders have been filled. In this example, the batch
size would be 200 cylinders. The 10% sample for analysis
must be selected randomly from the entire batch.
If the mixtures are two-component mixtures (such as CO in
N2) , any appropriate instrument may be used to analyze the
components in the representative sample. However, if the
mixtures are three-component mixtures of CO and propane (or
hexane) in N2, a gas chromatograph (GC) which is
appropriately fitted to completely remove all other
components from the analysis of the subject component must be
used to analyze all components in the representative sample,
whenever the gas concentrations are greater than 4% CO and/or
500 ppm propane (or 250 ppm hexane) . In cases where the
mixtures are three-component or four-component mixtures
containing C02, a gas chromatographic analysis is also
required.
Each component in each cylinder of the representative sample
must then be analyzed. For mixtures of three or four
components, each component must be analyzed simultaneously in
each bottle (if the manufacturer has several GC's), or one
component must be analyzed in every cylinder before the next
component is analyzed.
-------
-13-
Instrument drift must be determined and compensated for by
reintroducing an NBS-SRM or gm-CRM or primary standard with
concentration between 80% and 100% of full scale after the
analysis of every 5 or fewer cylinders in the representative
sample. Adjustments must be made to the GC if the span drift
between any two analyses of the standard is +0.5% or
greater. If an adjustment is made to the GC to compensate
for drift after the analyses of 5 I/M cylinders, those 5 I/M
cylinders must be reanalyzed after the adjustment is made.
One other procedure which can be used for compensating for
drift is discussed in Appendix 3.
After analyses of all components is complete, a mean
concentration is calculated for each component using the
response for each component in the monthly equations as
developed in Sections 3.3.3 and 3.3.4.
If all cylinders in the representative sample have component
concentrations within +0.5% for each component, the
calculated mean concentration for each component in the
representative sample can be used as the concentration for
each component in the entire batch.
If any cylinder in the representative sample has a component
with concentration that exceeds +0.5% of the mean
concentration for that component, all of the cylinders in the
batch must be analyzed for that component. If the mixture
being analyzed is a three or four component mixture, GC
analysis must be performed on the entire batch for that
component. For a two component mixture, any process
instrument may be used. However, drift must be determined
and compensated for during these analyses in the same manner
as for the representative sample. An alternative procedure
for compensating for drift is discussed in Appendix 3. A
mean concentration is calculated for the entire batch from
this analysis. Cylinders with concentrations exceeding +0.5%
of the mean concentration must be rejected. The mean
concentration of the remaining cylinders is then calculated
for that component. All cylinders remaining in the batch can
then be labeled with the mean concentration for that
component.
3.6 Analysis of Five or Fewer Cylinders
For a batch of five or fewer cylinders, each component in
each cylinder must be analyzed twice. The response for each
cylinder is then the average of the two measurements. The
component concentrations should then be calculated in the
same manner as for the representative sample described in
Section 3.5.3.
-------
-14-
3.7 Calculating the Accuracy of I/M Calibration Gas
A summary of the maximum allowed working tolerances for each
step of the analysis process is shown in Table 3.
Table 3
Summary of Working Tolerances
Error Source Allowed Tolerance
Monthly curve generation +0.5%
Pre-analysis curve check +0.5%
Homogeneity check - batch +0.5%
Determination of component +0.5%
concentrations in
representative sample
If the I/M gas has been blended and named according to this
procedure and with the maximum allowable error tolerances
listed in Table 3, the gas may be labeled with an accuracy of
_+2.0% to NBS. The methods used in arriving at this accuracy
are discussed in Appendix 4.
Some states may desire to order gases to better than a +2.0%
accuracy. In these casesr manufacturers will have to use
more stringent working tolerances than those in Table 3.
Manufacturers must use the methods outlined in Appendix 4 to
calculate accuracies for gases ordered to be named according
to this Recommended Practice but with better than +2.0% to
NBS accuracies.
3.8 Cylinder Labeling and Documentation
All I/M calibration gas cylinders ordered and named according
to these procedures must be labeled with a tag which contains
at a minimum the following information:
(i) Cylinder number, except in the use of low-pressure
cylinders, where a batch number is required.
(ii) Concentration of hydrocarbons (in ppm - either
hexane or propane), CO (in mol %) , and CC>2 (in
mol%) in cylinder gas (determined from Section
3.5.3), and accuracy specification (e.g., +2.0%,
or other as determined with methods in Appendix 4.
-------
-15-
(iii) Balance gas (diluent) .
(iv) Analysis date.
(v) Cylinder numbers of NBS and primary standards used
in determining instrument calibration curves.
(vi) Vendor name.
(vii) Expiration date of cylinder (see below).
(viii) The statement that "This gas has been named in
accordance with the EPA Recommended Practice for
Naming I/M Calibration Gas."
The expiration date is the date beyond which the manufacturer
does not suggest using the gas mixture to calibrate an
analyzer. The length of time of use may vary with different
mixtures.
The gas manufacturer must retain calibration curve data on
each batch analysis of I/M calibration gas for a minimum
period of 2 years. The data should include cylinder numbers
of all standards used in the practice, and should allow
someone to determine if the naming process has been performed
correctly.
3.9 Audits of I/M Calibration Gases
A formal audit of I/M calibration gas which has been named
according to this practice is not necessary prior to the sale
and delivery of the I/M calibration gases. However, it
should be remembered that State auditors will likely be
checking the accuracies of calibration gases used with
inspection analyzers. Inspection stations with improperly
labeled calibration gases will likely be required by auditors
to suspend the conducting of inspections until the problem
with their calibration gas is resolved. The labeling
requirements of the Recommended Practice should assist all
concerned parties in determining why a cylinder or group of
cylinders was improperly labeled.
The EPA may, from time to time, conduct audits of I/M
calibration gas by acquiring on the open market cylinders
named according to the Recommended Practice and analyzing
these cylinders. The results of such audits are likely to be
published by the EPA for the benefit of I/M programs.
3.10 Age of I/M Calibration Gas
Once a cylinder of I/M calibration gas has been properly
named, it is assumed to have the same concentration until it
is nearly empty,* regardless of how slowly it is used or how
*A cylinder is normally considered empty when the pressure in
the cylinder drops to one-tenth of its original pressure.
-------
-16-
much time expires between the time it is named and the time
it is sold. EPA believes that this is a good assumption,
since I/M calibration gases are high enough in concentration
so that interior cylinder wall absorption does not materially
affect the cylinder's concentration. This is a problem which
is limited more to much lower concentration cylinders.
-------
Appendices
-------
Appendix 1
EXCERPT FROM: 45 FR 34808, MAY 22, 1980,
EMISSION PERFORMANCE WARRANTY REGULATIONS
jaS.2217 Calibrations, adjustments.
(a) Equipment shall be calibrated in
accordance with the manufacturers'
. instructions...- • .
(b) Within one hour prior to a test the
analyzers shall be zeroed and spanned.
: Ambient air is acceptable as a zero gas; .
• an electrical span check is acceptable. .
Zero and span checks shall be made on
.. the lowest range capable of reading the.
•. short test standard. ••: •
(c) Within eight hours prior to a •
loaded test the dynamometer shall be •-
checked for proper power absorber
. settings.
(d){l) The analyzers shall have been
• spanned and adjusted, if necessary;- ~:
using gas traceable to NBS standards ±
2% within one week of the test These
span gases shall have concentrations' . '• •
. either - - ..' ' v .
(i) Between the standards specified in" '
this subpart and the jurisdictions
inspection standards for 1981 model
year light duty vehicles, or
(ii) Be within -50% to -i-100%ofthe •
standards in this subpart
(2) For analyzers with a separate
calibration or span port. CO readings
using calibration gas through the probe
and through the calibration port shall be
made: discrepancies of over 3% shall
require repair of leaks. No analyzer
adjustments shall be permitted during
this check,
-------
Appendix 2
Step Procedure
1 Check for impurities in pure components.
2 Determine whether instrument response is linear or
non-linear.
3 Generate a monthly equation for each instrument to be
used with NBS and other standards.
4 Fill I/M cylinders. Hold for incubation period, as
necessary.
5 Check monthly equation with several gases used to
generate monthly equation.
6 Perform homogeneity check on one component in all
cylinders. Check to see that all cylinders are within
+_0.5 of mean response.
7 Randomly select a representative sample of 10% of the
batch and no less than six cylinders.
8 Analyze each component in mixture. For two component
mixtures any process instrument. For three and four
component mixtures GC must be used, except for low
concentration CO-propane (or hexane)-N2 mixtures.
9 Calculate mean concentrations of components in
representative sample. Check to see that all
cylinders in representative sample are within 4-0.5% of
mean concentrations for each component.
10 Calculate accuracy of process (or use +2.0%).
11 Label cylinders.
-------
Appendix 3
Alternative Drift Procedure
This procedure can be used to compensate for drift in either
the homogeneity check of the the batch or the concentration
determination in the representative sample. In this
procedure, a sample is selected at random to serve as a
reference gas. The reference gas should be analyzed before
the rest of the cylinders are analyzed and after the analysis
of every subgroup of five or fewer cylinder in the batch. No
analyzer adjustments are permitted during the analysis
process.
The response of each I/M cylinder is then adjusted for drift
with the following equation:
Ri/mc = Ri/muc x _, Risl
Risl - N(Risl-Ris2)
Q
where
Ri/mc = tne corrected response of the I/M cylinder
Ri/muc = tne uncorrected response of the I/M cylinder
Risl = the response of the reference gas before
analysis of the subgroup of I/M cylinders
Ris2 = the response of the reference gas after the
analysis of the subgroup of the I/M cylinders
Q = number of segments between analyses of the
reference gas.*
N = the Nth cylinder in the analysis of the
subgroup.
This equation develops a ratio with which to correct the
response of the I/M cylinder. The ratio is essentially a
factor which measures the drift of the reference gas,
corrects it to the time at which the I/M cylinder was
analyzed, and applies it to the response of the I/M
cylinder. Drift is assumed to be linear between analyses of
the reference gas. For example, if an instrument drifted
-0.5% between analyses of the reference gas and four I/M
cylinders were analyzed between analyses of the reference
gas, use of the ratio would say that the magnitude of drift
at the time of analysis of the first cylinder was -0.1%. The
ratio would then multiply the uncorrected response of the I/M
*For example, if 5 cylinders are analyzed between analyses of
the reference gas, there are six potential segments of drift,
.one between analysis of each cylinder (RG-1-2-3-4-5-RG).
-------
cylinder by 1.001 to obtain the corrected I/M cylinder
response. This method of compensating for drift is identical
to the method which is suggested in the naming of gm-CRM's.
The drift from one analysis of the reference gas to the next
analysis of the reference gas must not exceed +0.5%. If it
does, the homogeneity check or concentration determination
must be restarted. Once all I/M cylinders have been analyzed
and corrected for a drift, a mean response is calculated.
Cylinders whose responses differ from the mean response by
more than +0.5% of the mean response must be rejected as
outliers. The remaining cylinders can then be labeled with
the mean concentration.
-------
Appendix 4
Accuracy Discussion
The model used to calculate the accuracies of I/M calibration
gas is one that involves estimating the total uncertainty, U,
with the following equation. This method is recognized by
the National Bureau of Standards and has been widely used
throughout the industry.*
U = (B + t90.0S)
where
B is the bias limit
S is the precision error index
t is the 90th percentile point for the two-tailed
Students "t" distribution. The t value is a function
of the number of degrees of freedom used in
calculating S. For small samples, t will be large and
for larger samples t will be smaller, approaching 1.65
as a lower limit. The use of the t arbitrarily
inflates the limit U to reduce the risk of
underestimating S when a small sample is used to
calculate S.
For the Recommended Practice, we are assuming that the sum of
the bias errors is zero. One example of a bias error is the
interference of CO on propane measurements with a THC
analyzer. However, for three component mixtures (such as CO
and propane in N2) we have required that the representative
sample be analyzed with GC techniques so that the bias errors
would be close to zero. The value of S is calculated by the
root mean square of the remaining precision errors, namely,
Monthly Calibration Curve +0.5%
Pre-Analysis Curve Check +0.5%
Homogeneity Check - Batch +0.5%
Concentration Determination -
Representative Sample +0.5%
*The 95th percentile point has also been used as a t-factor.
-------
The root mean square of these sources of errors at +0.5% is
+1.0%.* The model now takes on the form U=_+tgo (1.0%).
The tgo factor is taken from the two-tailed Students "t"
distribution with 10% of the area under the distribution
falling in the tails. These factors are shown in Table 4.
The number of degrees of freedom is the sample size. For
example, if a manufacturer makes a batch of 30 cylinders, the
degrees of freedom is 30, and the tgg value is 1.697. The
uncertainty then becomes (+1.0%)x(1.697)=+!.697%. The
manufacturer should round this off to the nearest 0.1%, so
this becomes +1.7%.
In Section 3.7 "Calculating the Accuracy of I/M Calibration
Gas" of the Recommended Practice, we stated that
manufacturers following the practice and using the maximum
allowable working tolerances for the four sources of error of
+0.5% could label the gas as having an accuracy of +_2.0%.
Although the actual accuracy depends on the batch size (and
the number of degrees of freedom), for batches of six
cylinders or more the Student's t-factor is less than 2.0.
This means that for any batch of six or more, the overall
accuracy should be within +2.0% (+I%x2) . For batches of five
or fewer, the recommended practice requires multiple analyses
of each component in each cylinder (the number of analyses
depending on the batch size) such that the Student's t-factor
is less than 2.0.
Some States or I/M users may desire to order gas to better
than +2.0% accuracy. In these cases if the State or I/M user
orders the gas blended according to the Recommended Practice,
the manufacturer will have to reduce the working tolerances
below +0.5%. For instance, if the manufacturer is capable of
reducing all of them to +0.3%, a batch of 30 cylinders could
be labeled with an accuracy of +1.0%**. Manufacturers must
use this technique to calculate the accuracy of a batch which
is requested to be blended and named with the Recommended
Practice with an accuracy different than +2.0%.
* (0.52 + 0.52 + 0.52 + 0.52)1/2
**(0.32 + 0.32 + 0.32 + 0.32)1/2 (1.697) = +1.0%
-------
Table 4
Student's t - Distribution for 10%
of Area in Both t Calculations Combined*
Degrees of Freedom
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Factor
6.314
2.920
2.353
2.132
2.015
1.943
1.895
1.860
1.833
1.812
1.796
1.782
1.771
1.761
1.753
1.746
1.740
1.734
1.729
1.725
Degrees of Freedom
21
22
23
24
25
26
27
28
29
30
40
60
120
Normal Distribution
Factor
1.721
1.717
1.714
1.711
1.708
1.706
1.703
1.701
1.699
1.697
1.684
1.671
1.658
1.645
*Hamburg, Morris, Statistical Analysis for Decision Making,
2nd edition, Harcourt Brace Jovanovich, Chicago, 1977, page
704.
------- |