EPA-AA-IMS-/81-16
Technical Report
DRAFT EPA RECOMMENDED PRACTICE.
FOR NAMING I/M CALIBRATION GAS
September, 1981
Thomas L. Darlington
Inspection and Maintenance Staff
Emission Control Technology Division
Office of Mobile Source Air Pollution Control
Office of Air, Noise, and Radiation
U.S. Environmental Protection Agency
Ann Arbor, Michigan 48105
-------
Abstract
This report explains how calibration gases will be used in Inspection and
Maintenance (I/M) programs, and identifies the problems states may have in
obtaining accurate gases. A Recommended- Analysis Practice for gas
manufacturers to use when naming I/M calibration gases is presented. States
are encouraged to procure gases named according to this Recommended Practice
for their own use, and to require licensed inspection stations to procure them
to ensure that they are obtaining accurate calibration gases which meet the
terms of the Emission Performance Warranty [207(b)], and to improve the
general quality of their I/M programs.
-------
Table of Concents
1.0 BACKGROUND 4
1.1 EPA Regulations Concerning Calibration Gas Accuracy 5
1.2 Practical Problems With Warranty Accuracy Requirements 6
1.3 EPA Efforts to Help States Meet Warranty Requirements 7
for Gas Accuracy
2.0 INTRODUCTION 8
2.1 I/H Calibration Gas Users 8
2.1.1 Centralized Programs 8
2.1.2 Decentralized Programs 8
2.2 Components of I/M Calibration Gases 10
2.2.1 Carbon Monoxide and Propane 10
2.2.2 Carbon Dioxide 12
2.2.3 Hexane 15
2.2.4 Diluents (Balance Gases) 17
2.3 Concentrations of I/M Calibration Gases 18
2.3.1 Emission Performance Warranty Requirements 18
2.3.2 State Requirements 19
2.4 NBS and Other Standards 20
3.0 TRACEABILITY 24
3.1 Establishing Traceability: Overview of Procedures 24
3.2 Analysis of Pure Components 26
3.3 Cylinders 27
3.4 Instrument Preparation and Calibration 28
3.4.1 Definition of Linearity 28
3.4.2 C02 Interference Check 30
3.4.3 Calibration Curve for Linear Instruments 30
3.4.4 Calibration Curve for Non-Linear Instruments 32
3.5 Analysis of I/M Calibration Gas Cylinders 34
3.5.1 Re-Usable Cylinders 34
3.5.2 Disposable Cylinders 35
3.6 Calculating I/M Calibration Gas Concentrations 38
3.6.1 Equations Used' 38
3.6.2 Concentration Determination 38
3.7 Cylinder Labeling and Documentation 39
3.3 Audits of I/M Calibration Gases 40
Appendices
Appendix 1 Emission Performance Warranty Regulations
Pertaining to Calibration Gases 41
Appendix 2 Listing of Analysis Requirements for
Recommended Analysis Practice 42
•Appendix 3 . Accuracy Discussion 55
Appendix 4 Discussion of Statistical Sampling and Naming
-------
1.0 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 once per year. Vehicles that have tailpipe emissions
greater than state or locally established emission standards are required to
i
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. Twenty-nine states
are expected to have operating programs by Janurary 1, 1983, however, many of
these programs will commence before that date. At this time, 11 states are
expected to have decentralized programs, 16 states are expected to have
centralized programs, and two states are undecided.
The primary inspection and diagnostic tool of these I/M programs is a
Non-Dispersive Infrared (NDIS) 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, however, need
periodic maintenance to keep their accuracy, as operator misuse, or changes in
pressure, temperature, and other operating variables can render them
c*
inaccurate.
Many analyzer manufacturers recommend a periodic check of their analyzers with
an accurate calibration gas. State or local I/M program regulations will also
-------
require a periodic check. This check is performed by flowing a known
concentration of calibration gas into the analyzer and determining whether the
analyzer is reading this calibration gas correctly. If it is not, 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 also important in determining
analyzer accuracy with this maintenance check. If the labeled concentration
of the calibration gas is significantly different than the 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.
1.1 SPA Regulations Concerning Calibration Gas Accuracy
The EPA recently 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 of those conditions is that the
analyzer used to conduct the emission short test must be checked with gases*
"traceable to NBS standards +2Z" at least once per week (Section 85.2217, page
* Note: There are no C02 emission standards in the Emission Performance
Warranty, consequently, the Warranty accuracy specification of ^22 traceable
to NBS does not apply to C02« Also, since there are no NBS standards for
hexane, the Warranty accuracy.specification of traceable ^22 to NBS cannot be
met on an analyzer calibrated with hexane only.
-------
34808. The relevant portipn of this page is provided in Appendix 1 to this
document). There are other requirements for these gases in the Warranty;
however, these are beyond the scope of this section and will be addressed in
Section 2.3.1. The EPA thinks that many states implementing I/M programs will
want to make Warranty protection available to consumers participating in their
I/M programs. 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 ^22 to NBS standards. However, there are
some practical problems associated with this accuracy requirement which are
discussed in the following paragraphs.
1.2 Practical Problems With Warranty Accuracy Requirement
States might think that they have only to order calibration gases for their
analyzers that have been certified by the manufacturer to be "traceable to NBS
standards ±21" to be ensured of meeting the accuracy requirement in the
Warranty. However, the most conscientious gas manufacturers are very hesitant
to label cylinders of calibration gas that they have analyzed with the phrase
"^22 to NBS standards", as they point out that there is currently no clear-cut
definition of "traceable" to NBS stanards. They explain that traceability
implies a system of analytical procedures and documentation of data that is
used to express the concentration of a particular gas relative to a standard
gas. These analytical procedures and methods of handling data may vary
significantly between manufacturers, causing differences in the concentrations
of identically labeled calibration gases made by different manufacturers or
even by the same manufacturers. I/M calibration gas users would find it
-------
difficult to detect these differences and identify which gas cylinders were
correctly, labeled, with the possible exception of the I/M managers if they
maintain a sophisticated gas analysis program of their own.
1.3 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.
A special effort has been made to make these procedures workable from a
scientific gas manufacturer's viewpoint. A meeting was held between the gas
manufacturing industry, the National Bureau of Standards and the EPA to agree
on the original framework of the procedures. Manufacturers have also been
allowed to comment on the draft forms of the Recommended Practice: many of
their comments have been incorporated into the procedures. The end result is
a Recommended Practice which should contribute significantly to overall I/M
program accuracy and fairness.
-------
2.0 INTRODUCTION
2.1 I/M Calibration Gas Users
2.1.1 Centralized Programs
In a centralized I/M program where the official I/M short test is conducted in
state or contractor operated centralized test facilities only,, states are
likely to require that only those analyzers in the centralized test facilities
be checked at least once per week with an accurate calibration gas. Most
states operating centralized programs will not require garage owners who have
purchased analyzers to help them in the repairs of vehicles, failed from the
I/M programs to perform periodic calibration checks on their instruments.
Therefore, analyzer operators in the centralized facilities will be the
primary users of calibration gas in a centralized program. The- exception to
this situation is where a state running a centralized program licenses certain
garages for the reinspection function (New Jersey, for example). In this
case, the state will likely require reinspection garages to perform a periodic
calibration check of the garage analyzer with an accurate calibration gas,
also.
2.1.2 Decentralized Programs
In decentralized programs, there are likely to be two different categories of
I/M calibration gas users. The first are the garages licensed by the state to
perform emission inspections. States are likely to require that these garages
-------
perform a periodic check of the garage analyzer with an accurate calibration
gas. Also, decentralized programs will have official auditors who
periodically visit each facility to determine (among other things) if the
emission analyzers are accurate. Part of the auditor's job will be to check
the facility's analyzer with state audit gas, to provide a double check
against the facility's gas.
Many states are requiring their auditors to use calibration gases with
accuracies of ^12 to NBS standards rather than the accuracy requirement of ±22
which is specified for weekly calibrations in the Warranty. Since some states
have proposed regulations which would require auditors to use calibration gas
which is traceable HZ, included in this Recommended Practice are procedures
which will allow gas manufacturers to name certain audit gases to within ±1Z
of NBS standards.*
Decentralized programs are expected to use more calibration gas than
centralized programs. This is because decentralized programs consist of many
small neighborhood garages performing a few emission inspections each, where a
centralized program consists of a few high-throughput inspection facilities.
Consequently, there are more analyzers in a decentralized program, each
requiring the same periodic calibration checks as the fewer number of
analyzers used in a centralized program.' .
* The one exception is an I/M calibration cylinder which contains both C02
and CO. The best accuracy 'specification on this cylinder named according to
the Recommended Practice is H.5Z. This issue is discussed further in Section
2.2.2 and Appendix 3.
-------
10
2.2 Components of I/M Calibration Gases
A state has several options in determining how many and what components are in
the calibration cylinders which are bought for calibrating analyzers in that
state. The Recommended Practice generally does not limit the scope of these
options: within certain limits as discussed in this section, there should be
no difference in the quality or accuracy of gases in each option if the
components are named- in accordance with the procedures described in this
Recommended Practice.
2.2.1 Carbon Monoxide and Propane
States must first examine what components are needed in a calibration
cylinder(s) to ensure that inspected vehicles are accurately passed and
failed. Some I/M programs are being implemented with a carbon monoxide (CO)
cutpoint only.* The state may feel that analyzers therefore only have to be
accurate at reading a vehicle's CO emissions, and determine that a calibration
cylinder of CO is sufficient to ensure this accuracy. This determination is
consistent with the Emission Performance Warranty. However, the EPA
recommends that a state with a CO-only program require inspectors to calibrate
their analyzers with propane as well as CO, since a mechanic performing
repairs on a vehicle can use a vehicle's HC emissions to diagnose problems
* Some urban areas that are in non-attainment for CO but not for ozone will
implement an I/M program which uses CO cutpoints only. The state of North
Carolina is operating an I/M program which is in the voluntary inspection and
maintenance phase of implementation which has cutpoints for CO only.
-------
11
that may be causing a significant loss in fuel economy for that vehicle, and
the costs of calibrating an analyzer with propane as well as CO are not
significantly greater than the costs of calibrating an analyzer with CO only*
Most I/M programs will be implemented with CO and EC cutpoints. These states
therefore are likely to require all the analyzers used for inspections in that
state to be calibrated with both CO and propane** The most convenient
cylinder for them to order will be a three-blend cylinder of CO and propane in
nitrogen. Some states may wish to order separate cylinders of CO in nitrogen
and propane in nitrogen on the theory that there is "interference" between CO
and propane in- nitrogen such that both components appear to have slightly
higher concentrations than they would if they were in separate cylinders.
[However, the National Bureau of Standards (NBS) has studied the level of
interference between CO and propane in typical I/M calibration gas
concentrations, and determined that the interference effects are very
slight.**] Consequently, little accuracy is gained by using two binary, blend
calibration gases (i.e., CO in nitrogen and propane in nitrogen) than by using
a single tri-blend calibration gas (i.e., CO and propane in nitrogen).
* Hexane might be used sometimes instead of propane. Hexane is discussed in
Section 2.2.3.
** [Reference to not-yet-conducted NBS study.]
-------
12
2.2.2 Carbon Dioxide
Some I/M programs will have carbon dioxide (CO-) outpoints for vehicles as
well as HC and CO cutpoints. Checking for the amount of CO, in a vehicle's
exhaust helps the state to determine whether there are significant leaks in a
\
vehicle's exhaust system which are causing dilution of exhaust gases.
The Recommended Practice contains procedures which will allow gas
manufacturers to incorporate CO. into I/M calibration gas mixtures and
demonstrate traceability j^2Z to NBS of all components used. However,
traceability better than ±1.5Z cannot be attained for these mixtures because
of interference effects between CO and CO. These and other considerations
are discussed in the following paragraphs.
Inspection analyzers use infrared absorption principles to detect the
concentrations of propane, CO, and CO- in vehicle exhaust. These gases
absorb infrared radiation at different wavelengths, consequently the analyzer
can, for the most part, differentiate completely between each gas. However,
the wavelengths of absorption of CO and CO- are very close to each other.
This has the effect of making some infrared analyzers slightly inaccurate at
reading CO in the presence of CO , because they actually count a small
portion of the CO in the analysis of CO (but not vice versa).* This
phenomenon is known as interference.
, * The ability of infrared analyzers to read these components discreetly is a
function of the quality of filters used in the analyzers.
-------
13
Laboratory-type infrared analyzers which gas manufacturers will likely use to
name CO in an I/M calibration cylinder containing CO. and propane are
capable of keeping interference to minimum levels. The Recommended Practice
requires gas manufacturers to keep the level of CO. interference in CO to
less than .31 of the CO concentration when naming CO in an I/M calibration
cylinder containing CO^* Also, the Recommended Practice requires gas
manufacturers to use tighter tolerances in naming I/M calibration cylinders
containing CO.. The small amount- of additional interference error present
when CO is used requires these tighter tolerances so that all components
meet the ±2£ to NBS accuracy requirement. However, the tolerances within the
Recommended Practice have not been tightened to such a degree that
traceability _^1I for a mixture containing CO. and CO could be assured.
Consequently, M..5Z is the best specification for this mixture attainable with
this specification. A more complete discussion of how these accuracy
specifications are arrived at is presented in Appendix 3.
-------
14
Regardless of how accurately the calibration gases are named by
laboratory-type infrared analyzers used by gas manufacturers, inspection
analyzers are still likely to exhibit small interference effects.* A recent
study by EPA determined the CO interference levels in several common
inspection analyzers. The results of the study are presented in Table 1.**
* Another question raised by interference effects of C02 on CO is: What is
the effect of calibrating an analyzer with /a calibration cylinder containing
CO and C02 on a vehicle's apparent CO/- 'amiss ions? If an analyzer's CO
response is calibrated with a CO cylinder containing C02» the analyzer will
be set at slightly lower than the actual CO concentration in the calibration
cylinder because the analyzer will be responding to a small amount of CO2
also. Since the analyzer's CO response will be set slightly lower, it will
read a vehicle's CO emissions at a slightly / lower level, -thereby making a
given short test slightly less stringent than it would have been had a CO
cylinder that did not. contain C02 been used for the calibration.
** "Operational Evaluation of Vehicle Exhaust Emission Inspection Analyzers",
Volume 1, Contract # 63-03-2747, EPA 460/3-80-019.
-------
15
Table 1
Combined Z Full Scale*
. EC and CO Responses of
Analyzers to 15Z C02
Analyzer Z Combined F.S. Response
A O.OZ
B 1.74Z
C 2.04Z
* The combined interference levels reflect the percentage full scale responses
of the EC and CO meters to CO2 gas. The percentages were added to get a
"combined" response.
Based on the data in Table 1, it is probably better for an I/M inspector to
check the CO response of an analyzer without using a CO cylinder that contains
CO.. If a state determines that CO must also be used, then the CO
should be in a separate cylinder from the CO (i.e., one cylinder of CO and
propane in nitrogen, the other cylinder CO. in nitrogen). If the state
determines that the CO. must be in the same cylinder as the CO, the state
should select the lowest CO. concentration acceptable to maintain, the
analyzer's CO. accuracy. This will minimize the interference effects of
CO. on CO.
2.2.3 Besane
Idle emission inspection analyzers, when determining the amount of
hydrocarbons (EC) in vehicle exhaust, are really measuring the amount of
hexane (actually n-hexane) in vehicle exhaust, plus some fraction of the
-------
16
remaining hydrocarbons which appear to be partially hexane because of
interference effects. The assumption inherent in the measurement technique is
that the quantity of actual hexane plus other hydrocarbons which appear to be
partially hexane due to interference is roughly proportional to the total
quantity of all of the various types of hydrocarbons.
Because analyzer readouts are based on apparent hexane concentrations in
vehicle exhaust, hexane is in theory the best gas with which to calibrate the
HC portion of the analyzer. However, there are several problems associated
with hexane which make it impractical as an inspection analyzer calibration
gas, except under carefully controlled conditions.
First, there are no NBS standards for hexane. This means that the accuracy
specification of "traceable ±ZZ to NBS" is by definition impossible for
hexane. However, that does not necessarily mean that one cannot obtain"highly
accurate hexane calibration gases. There are scientific gas manufacturers who
have the resources to blend and name hexane accurately.
Second, hexane is a "sticky" molecule, and therefore has a tendency to cling
to the sidewalls and interior fittings of a cylinder if the cylinder is not
periodically agitated and/or heated. If sticking occurs in a cylinder of
hexane, the concentration of hexane in the cylinder changes as the cylinder is
used, since the diluent (balance gas) will flow from the cylinder first,
leaving the hexane behind. Agitation of the cylinder by rolling it,., and
heating of the cylinder can significantly reduce this problem.
-------
17
Lastly, hexane is much more expensive than propane to use as a calibration gas
because of the problems in handling and blending.it. Nonetheless, it still
has potential uses in an I/M program. For example, because propane is usually
used instead of hexane as a calibration gas, analyzer manufacturers must
develop correction factors for their analyzers to convert a propane reading on
the analyzer to a hexane reading. This correction factor can change slightly
if the analyzer is roughly handled. A hexane calibration cylinder is needed
to check or re-establish this correction factor periodically. Arizona and
California I/M programs are currently using hexane to check analyzer
correction factors.
2.2.4 Diluents (Balance Gases)
Throughout this report we have assumed that calibration gases would be blended
with nitrogen as a diluent or balance gas. Nitrogen is the easiest gas to use
as a diluent from a production standpoint since it is readily available in
very pure forms.
Air could also be used as a diluent, however, gas manufacturers cannot use
room air or ambient air because of the impurities present in these sources of
air which could cause errors in the blending and naming processes. In order
to use air as a diluent, gas manufacturers must buy pure nitrogen and oxygen
and blend it together prior to filling I/M calibration cylinders. This
process adds to the production costs for I/M calibration gases, therefore,
most states will order I/M calibration gases which use nitrogen as a diluent.
-------
18
2.3 Concentrations of I/M Calibration Gases
2.3.1 Emission Performance Warranty Requirements
In addition to the accuracy requirement for gases used to check emission
analyzers (i.e. , ±2Z to NBS standards) the Warranty also requires the
calibration gas to be vithin certain concentration ranges. The limitations
are that the calibration gas oust have concentrations of between -50Z and
+100% of the state emission standards if the state adopts the numerically
minimum emission standards specified in the Federal Warranty Regulations, or
have concentrations between these numerically minimum standards and the state
selected emission standards if the state selected emission standards are
numerically higher than the minimum emission standards. The minimum emission
standards specified in the warranty are, depending on the short test procedure
used, either l.OZ CO and 200 ppm hexane HC or 1.22 CO and 220 ppm hexane for
HC. If a state selects emission standards of l.OZ CO and 200 ppm hexane, the
concentration of calibration gas used to periodically check the analyzers must
be between 0.5Z and 2.0Z for CO, and between 100 ppm and 400 ppm hexane (or
200 ppm and 800 ppm propane) for the emission short test to meet the terms of
the Emission Performance Warranty. Alternately, if a state selects emission
standards of 4.0Z CO and 400 ppm HC (the highest standards any state is likely
to adopt for 1981 and later vehicles), the concentration of calibration gas
used to periodically check the analyzer must be between l.OZ and 4.0Z for CO,
and between 200 ppm and 400 ppm hexane for HC (or 400 and 800 ppm propane).
See Table 2. The lower limits of l.OZ CO and 200 ppm hexarie could be 1.2Z CO
and 220 ppm hexane if a different test procedure is used.
-------
State Adopts
Warranty Emission
Standards of
Warranty Emission
Standards of
State Emission
Standards of
19
Table 2
Example of Calibration Gas Requirements
for 207(b) Warranty Coverage
Emission
Standard
200 ppm £6. HC
l.OZ CO
220 ppm C6. HC
1.2Z CO
400 ppm C£ HC
4.0Z CO
Calibration Gas Concentration.*
200 ppm C3- 800 ppm C3_
.52 CO - 2.0Z CO
220 ppm C3- 880 ppm £3
.62 CO - 2.4Z CO
200 ppm C3- 800 ppm C3_
l.OZ CO - 4.0Z CO
Note: Standard.'is expressed in ppm hexane (C6). Span gases are expressed as
ppm propane (C3) assuming an approximate correction factor of 0.5.
2.3.2 State/Requirements
In addition to meeting the Warranty requirements for propane and CO
calibration gas concentrations, some states may wish to use other
concentrations of propane 'and CO in performing multipoint calibrations on
analyzers used in the I/M programs. Also, some states may wish to use CO
in multipoint calibrations also. In these cases the state would need several
cylinders of propane, CO and/or CO, calibration gas at varying
concentrations, all with accuracies of _^2Z to NBS standards for best results.
For these reasons, we have designed the Recommended Practice in such a way
that a gas manufacturer can demonstrate traceability for any ordered gas
concentration which falls between the range of NBS standard reference material
concentrations discussed in Section 2.4 (see Table 3). I/M calibration gas
-------
20
concentrations ordered either above the highest standard or below the lowest
standard listed in Table 3 are not assured traceability ^22 to NBS standards
within the scope of this Recommended Practice*
2.4 NBS and Other Standards
The National Bureau of Standards tries to make available standard gases in
commonly requested concentrations. Up until this year, the only NBS standards
available were called Standard Reference Materials (SRM's). The
concentrations of various NBS CO, propane, and CO- SRM's which are in the
range of interest as far as I/M calibration gas is concerned are listed in
Table 3. SRM's have an analytical uncertainty of ^1% to the true value of the
concentration in the cylinder.
-------
21
Table 3
NBS Propane, CO, and C02 Standards in Nitrogen
in the Concentration Range
of Typical I/M Calibration Gases
Propane
NBS Standard
2643
2644
2645
2646
2647
2648
Propane
Concentrations
100 ppm
250 ppm
500 ppm
1000 ppm
2500 ppm
5000 ppm
Carbon
Monoxide
NBS Standard
2638
2639
2640
2641
2642
Carbon
Monoxide
Concentration
5000 ppm C.5Z)
l.OZ
2.0Z
4.0Z
8.0Z
Carbon
Dioxide*
NBS Standard
2620a
2622a
2624a
2626a
1674b
167 5b
Carbon
Dioxide
Concentration
l.OZ
2.0Z
3.0Z
7.0Z
14. OZ
* Other standards are available for C02- A list of available standards may
be obtained by writing the Office of Standard Reference Materials, Chemistry
Building, Room B311, National Bureau of Standards, Washington, B.C. 20234.
-------
22
Historically, SRM's have been the only NBS standards. The EPA has therefore
required SRM's as the standard to which calibration gases used in many mobile
and stationary source programs had to be traced. This created a large demand
for SBM's since the tracing process consume the SSM's. Also, the analysis
\
procedures for SBM's are very detailed and lengthy, so SRM's have been in
short supply. In an attempt to alleviate these problems associated with the
lack of supply of NBS standards, the EPA Emission Monitoring and Support
Laboratory at Research Triangle Park (EMSL-RTP) and NBS developed procedures
that would allow gas manufacturers to make a new kind of standard, called a
Certified Reference Material (CRM), which if made properly has an analytical
uncertainty of about ±1.3Z to true value. These CRM'3 can be made in batch
quantities, but a limitation of these CSM's is that they can be made only at
SRM concentrations.
The analytical procedures described in this report rely on there being an
adequate supply of NBS standards, either SRM's or CRM's. The CRM program
should ensure that availability.*
Because of the difficulty in obtaining NBS standards, scientific gas
manufacturers have, over the years, developed their own in-house standards
which they have used in the analysis of various scientific gases. These
in-house standards, called primary standards, are for the most part very
* At the time of publication of this report, one gas manufacturing company was
offering CSM's for sale.
-------
23
accurately named. However, there are no uniform procedures throughout the
scientific gas industry for naming these standards. Therefore, differences in
accuracies of primary standards between manufacturers probably exists. We
mention these standards in this section because they are used by manufacturers
along with NBS standards (if a gas divider is not used) to demonstrate
traceability within the context of this Recommmended Practice. The UBS
standards serve the function of providing a check on the accuracy of the
in-house primary standards* The primary standards serve the function of
providing more data points by which to construct an analyzer calibration curve
than would be obtained if NBS standards were used alone.
-------
24
3.0 TRACEABILIT7
People familiar with the blending and naming (the affixing of a concentration
label to a cylinder) of scientific gases use the word traceability to describe
an analytical link between the concentration of two separate cylinders of the
same kind (not necessarily the same concentration) of gas, where one of the
cylinders is usually some kind of standard gas. The standards most often
referred to in any traceability definition are NBS standards (these standards
were discussed in section 2.4). For the purpose of this report, "traceable"
will mean the use of the standards and analytical procedures discussed in this
report. These procedures, if used properly, should yield an I/M calibration
gas which is accurately named relative to NBS standards (either SRM's or
CRM's) within ^22, _+1.5Z, or ±1%, depending on the criteria selected with the
Recommended Practice. A brief discussion on how this accuracy is attained in
this Recommended Practice is presented in Appendix 3.
3.1 Establishing Traceability; Overview of Procedures
The analytical procedures described in this report for establishing
traceability between an I/M calibration gas and NBS standards are designed so
that traceability can be demonstrated for practically any concentration of I/M
calibration gas a state or I/M customer may desire to order. The general
procedures are described in the body of this report. A more concise listing
of the procedures is provided in Appendix 2.
-------
25
Generally, the procedures for the analysis of propane, carbon monoxide, or
carbon dioxide in the I/M calibration gas are dependent on the characteristics
of the analysis instrumentation used to determine the concentrations of each
gas* A linearity check is first conducted to determine the extent, of
linearity betveen input gas concentration and response in the instrument.* If
a four-blend CO., CO, and propane in . N cylinder is to be named, an
interference check is first conducted betveen the CO. and CO. If only a
three-blend CO, propane and N. cylinder is to be made, the interference
check is omitted. If instrument response during the linearity check to a
given component in the subject gas is linear, a first or second order equation
is developed for the instrument from analysis of six standard gases which have
the same kind of component, but at different concentrations than the subject
gas. This equation is used in naming the subject component in each of the I/M
calibration gas cylinders* If the instrument has a non-linear response to a
given component in the subject gas, a third or fourth order equation is
developed for the instrument from analysis of eight standard gases which have
the same kind of component, but again at different concentrations than the
subject gas. The equation is then used in determining the concentration of
the subject component in each of the I/M calibration gas cylinders.
* Although there are exceptions to the rule, the instrumentation most commonly
used to analyze carbon monoxide and carbon dioxide (usually Non-Dispersive
Infrared-NDIR-analyzer) is non-linear, while instrumentation commonly used to
analyze propane (usually a Flame lonization Detector-FID) is linear.
-------
26
Before presenting details of the analysis procedures, discussions of the
analysis of pure components used in the blending process and the selection of
cylinders used to hold I/M calibration gas are presented.
3.2 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
nitrogen used as diluent in the calibration gas must be 99.99% pure nitrogen
to comply with this Recommended Practice. This grade of nitrogen is readily
available at a reasonable cost and is frequently used in blending scientific
gases. An analysis does not have to be performed on the nitrogen. Propane
used must be "instrument grade" propane which is certified to be 99.5Z
propane, and the balance will be other hydrocarbons (e.g., methane,
iso-butane, butane and ethane). An analysis must be performed of this bulk
propane to determine in fact that the propane gas used is at least 99.52
propane. This analysis can be easily performed with a Gas Chromatograph (GC)
which is appropriately fitted to read propane. Lastly, the carbon monoxide
used must have a methane count of less than 100 ppm, and the carbon dioxide
used must have a total hydrocarbon (THC) count of less than 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
-------
27
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.
3.3 Cylinders
Any size disposable or reusable 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. Also, reusable
cylinders must be fitted with CGA-350 valves (CGA stands for the Compressed
Gas Association). Disposable cylinders must be fitted with a CGA .1/4-inch
flare-fitting valve. The use of these valves only will not be a burden to gas
manufacturers, and they will allow I/M calibration gas users to switch
suppliers (gas manufacturers) without having to purchase new regulators.
I/M calibration gas users should be aware that the analysis of each cylinder
of I/M calibration gas contributes a significant portion to the ultimate price
of the calibration .gas. It probably costs no more for a manufacturer to
perform an analysis o.f a large cylinder than to perform an analysis on a small
cylinder (except in the case of calibration gas sold in disposable containers,
where special abbreviated analysis procedures are allowed). Therefore, the
cost per cubic foot of calibration gas sold in a large cylinder will likely be
significantly less than the cost per cubic foot of calibration gas in. a small
cylinder.
-------
28
3..4 Instrument Preparation and Calibration
3.4.1 Definition of Linearity
The NBS standard reference materials (either SHM's or CRH's) listed in Table 4-
or a gas divider which can be used to dilute the highest concentration
standard to at least three lover concentrations which match within ,+SZ the
standard concentrations, and a instrument grade 99.99Z pure nitrogen cylinder,
must be used to determine 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 l.OZ CO to 8.0Z CO, 250 ppm
propane to 2500 ppm propane and/or l.OZ CO. to 7.0Z or 14. OZ CO. which
deviate by ±2Z of point* or less from 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 ±2Z of the
concentration values of the intermediate points. The range of allowable
deviations from the straight line are illustrated .in Table 4.
The reader will note that linearity ia defined only between endpoints which
coincide with available NBS standards which are in the range of interest for
I/M calibration gases (see Table 4). This range in which linearity is defined
is wide enough to accomodate practically any concentration of I/M calibration
gas ordered by a state or I/M area.
* ±2% of point is j^2Z of .reading. For example, if a reading of a certain gas
is 6.0Z CO, +2Z of 6.0Z CO is +.122 CO.
-------
29
Table 4-
NBS Standards 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
CO in N?
1.02 + .022 CO
2.02 +.042 CO
4.02 ±.082 CO
8.02 +.162 CO
C02 in
N-,
1.02 * .022 C02
3.02 +. .062 C02
4.02 +• .082 C02
7.02 • T .142 C02
14.02** *_ .282 C02
* An instrument-grade 99.992 pure nitrogen cylinder must also be used to
establish the instrument zero.
** This C02 standard must be used in the linearity check if the. C02
concentration to,, be named is over 7.02. The 1.02 C02 standard can be
dropped from the check in chis situation.
-------
30
Linearity need only be determined once on a particular instrument, unless
service is performed on the instrument, in which case the linearity check must
be conducted after service is completed.
3.4.2 C0» Interference Check
The following check is to be made only if CO is to be named in a cylinder that
also contains CO.. Otherwise, this step shall be omitted.
Carbon dioxide gas (with nitrogen as a diluent) at the same concentration
which is to be used in the I/M calibration cylinder must be analyzed with the
CO analysis instrumentation. The response of the CO analyzer to- CO. must be
less than .5Z of the concentration of CO which is to be used in the I/M
calibration cylinder.
3.4.3 Calibration Curve for Linear Instruments
Prior to the analysis of each batch of I/M calibration gas, a calibration
curve must be generated on the range that is to be used to analyze the I/M
calibration gas. No more than one hour must elapse between the time at which
the calibration curve is generated and the start of the analysis of the batch
of I/M calibration gas begins. If analysis of a portion of the batch is
postponed for two hours or more, a completely new calibration curve must be
generated.
-------
31
A minimum of six cylinders must be used to generate the calibration curve.
One cylinder must be instrument grade 99.99% pure nitrogen, tvo other
cylinders must be undiluted NBS standards which are above and below the
concentration of the I/M calibration gas to be named, and the other three
cylinders can be primary standards or gases obtained by using a gas divider to
dilute either NBS 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 resonses 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 standards, with the others being either primary standards
or diluted NBS standards.
A first or second order equation must be calculated from the instrument
response for the six gases. If the difference between any point and the first
or second order curve is less than ^1.0%' for gases intended to be traceable
_+2% to NBS (_^. 5% if gases traceable _£!% to NBS are to be named or if CO is to
be named in a cylinder containing CO.), the first or second order equation
thereby derived may be used to name I/M calibration gases in this Recommended
Practice. Inability to satisfy this criteria with a first or second order
equation is an indicator 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.
-------
32
It is suggested that if a first-order 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 (i.e., there is a cluster of nearby points
above the line and another cluster below), then a second-order equation should
be generated from the calibration points. This will increase the accuracy of
the I/M calibration gas naming process. The second order curve must meet the
^l.OZ of point 0^.5Z if gases traceable ±IZ to NBS are to be named or if CO is
to be named in a cylinder containing CO ) criterion stated in the previous
paragraph.
3.4.4 Calibration Curve for Non-Linear Instruments
Prior to the analysis of each batch of I/M calibration gas with a non-linear
instrument, a calibration curve must be constructed on the range that is to be
used in analyzing the I/M calibration gas. No more than one hour must elapse
between the time at which the calibration curve is generated and the start of
the analysis of the batch of I/M calibration gas begins. If analysis of a
portion of the batch is postponed for two hours or more, a completely new
calibration curve must be generated.
A minimum of eight cylinders must be used to construct the calibration curve.
One cylinder must be instrument-grade 99.99% pure nitrogen, two other
cylinders must be undiluted NBS 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 NBS standard. Instrument responses for the seven points other
-------
33
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 third or fourth-order polynomial equation must be calculated from the
instrument responses to the eight gases obtained during the calibration step.
No inflection points are allowed in the equation of the curve generated from
analysis of the eight gases.* If an inflection point occurs in a fourth-order
equation, a third-order equation should be tried. If the differences between
any point and the curve is less than ^l.OZ for gases intended to be traceable
_+2% to NBS (^.5Z if gases traceable ^11 to NBS are to be named or if CO is to
be named in a cylinder containing CO.), the equation thereby developed may
be used to name I/M calibration gases in this Recommended Practice. Inability
to satisfy these criteria with a third or fourth-order equation is an
\
.indicator of improperly named primary standards, a malfunctioning gas divider,
and/or malfunctioning analysis instrumentation.
* 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.
-------
34
It is suggested that if a third-order 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
(i.e., there is a cluster of nearby points above the curve and another below
the curve), then a fourth-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
fourth-order curve. The fourth-order curve must meet the ±IZ of point (^.51
if gases traceable ±1! to NBS are to be named or if CO is to be named in a
cylinder containing CO.) criterion stated in the previous paragraph.
3.5 Analysis of I/M Calibration Gas Cylinders
3.5.1 Re-Usable Cylinders
After valid calibration curves are obtained for all instruments that are to be
used to analyze the I/M calibration gas, the analysis of I/M calibration gas
can begin. Each component (with the exception of the diluent) of every
re-usable I/M calibration gas cylinder must be discreetly analyzed. I/M
calibration gas cylinders can be analyzed -in any sequence, however, the
highest concentration NBS standard used1 to generate the calibration curve and
the zero gas cylinder must be repeated after -every ten. analyses of I/M
calibration gas cylinders to ensure that substantial instrument zero or span
drift has not occurred which could detrimentally affect the I/M calibration
gas naming process. No analyzer adjustments aye permitted during the naming
process. Zero and span drifts are calculated as a percentage of full scale.
-------
35
They are added together to arrive at a total drift error. If the total drift
error changes by more than^l.OZ (_+.5Z if gases traceable'^1Z to NBS are to be
named or if CO is to be named in a cylinder containing CO to _+1.5Z to NBS)
throughout the analysis process, a new calibration curve must be constructed
for the analyzer in accordance with the procedures described in section 3.4.2
or 3.4.3, and the analyses of the previous ten cylinders of I/M calibration
gas must be repeated.
3.5.2 Disposable Cylinders
It is likely that much of the I/M calibration gas sold to I/M users will be in
the form of disposable cylinders. Many analyzer manufacturers are designing
the stands that hold their analyzers to include a holder for a disposable
cylinder that can be used to periodically calibration-check the analyzer.
These disposable cylinders hold a small amount of calibration gas, usually 6
to 10 cubic feet. The disposable cylinders are likely to be used
predominantly in licensed facilities of decentralized programs, where (l)
there may be no room for a large cylinder of I/M calibration gas, or (2) the
facility owner does not want to incur the higher initial cost of a larger
cylinder. Fewer vehicles will be inspected on the basis of each disposable
cylinder.
As stated previously, for re-usable cylinders the Recommended Practice
requires that an analysis be performed according to the prescribed techniques
on each component (with the exception of the diluent) in every cylinder.
However, the result of insisting on this requirement for disposable cylinders
-------
36
also might be to make them prohibitively expensive for small facilities in
decentralized programs. .
Disposable cylinders are usually filled in one of two ways: (1) they can be
filled one-by-one from a re-usable cylinder or bulk homogeneous mixture, or
(2) separate cylinders of pure components can be piped together and connected
to a common manifold to which the empty disposable cylinders are also
connected, thereby filling the disposable cylinders simultaneously from a
single stream of "dynamically" blended gases. In this latter filling process,
the blend is usually continuously analyzed while the disposable cylinders are
being filled. In the first filling process mentioned, all components of the
mixture have usually been analyzed already. In both filling processes
mentioned, analysis is usually performed on the gas blend somewhere along its
path to the disposable cylinder, which makes the probability of the mixture in
the disposable cylinder being accurate much greater than if an analysis of the
blend entering a disposable cylinder had not been performed. However,
blunders (cylinders with contents significantly different than the labeled
concentrations) can still occur with the disposable cylinders, particularly
where a disposable cylinder is not properly evacuated prior to filling, or a
disposable cylinder is not properly filled because of some malfunction in the
ganging hardware.
For these reasons, this Recommended Practice requires only that an analysis be
performed on (1) one component of each filled disposable cylinder, and on (2)
the other components (with the exception of the diluent) in 10% and no. less
-------
37
than ten* total cylinders of the filled disposable cylinders that have been
blended with a continuous analysis system or filled from a previously blended
and analyzed bulk mixture. The mean concentration of the first component
which is analyzed in all disposable cylinders in the batch is calculated. If
the concentration of the first component is any cylinder in the batch is more
than ±2% different from the mean concentration, that cylinder must have all
components discreetly analyzed. Any cylinder whose first component
concentration is within ±2% of the mean concentration of the batch may apply
the criterion of the following paragraph in the determination of the
concentration of other components in that cylinder.
The mean concentration of the other component in the 10Z sample(s) are
calculated. The standard deviation of the concentration values of the other
components" in the 10Z samples must be less than or equal to i»5Z of each
sample's mean.** If they are-, all cylinders in the batch shall be labeled
with the mean concentrations of the other components in the 10Z samples. If
they are not, an analysis of all cylinders for "the other components is
required.
* If less than 10 disposable cylinders are to be named, all components (except
the diluent) must be analyzed in each cylinder.
** This will assure that 95Z of the sample will be within ±1% of the mean
concentration (^2 standard deviations), and ?9Z of the sample will be within
^2% of the mean concentration (±4 standard deviations). For an explanation of
this statistical criteria see Appendix 4.
-------
38
If a continuous analysis system for all components (propane, CO and CO ) has
not been used in the case where disposable cylinders are filled from pure
components flowing through a common manifold, or where an analysis has not
been performed on each component in the bulk mixture in the case where
disposable cylinders are filled from the bulk mixture, an analysis must be
done on each component of every disposable cylinder.
3.6 Calculating I/M Calibration Gas Concentrations
3.6.1 Equations Used
Linear Instruments - The concentration of each component of an I/M calibration
gas (propane and CO) that has been analyzed with a linear instrument shall be
determined by the first or second order equation generated by the calibration
curve as discussed in Section 3.4.2.
Non-Linear Instruments - The concentration of components in I/M calibration
gas analyzed with a non-linear instrument shall be calculated by the third or
fourth-order calibration curve equation discussed in section 3.4.3.
3.6.2 Concentration Determination
Re-Usable Cylinders - Propane, carbon monoxide, and CO. if used are analyzed
in each re-usable cylinder of I/M calibration gas blended. The labeled
concentration of each component in each cylinder must be calculated from the
meter responses for each component using the equations specified in Section
3.6.1.
-------
39
Disposable Cylinders - For disposable cylinders in which each component has
been analyzed in all cylinders of a batch (see Section 3.5.2), the same rules
apply for determining- the concentration label of each component as apply for
re-usable cylinders (see previous paragraph). For disposable cylinders in
which one component has been analyzed in every cylinder, while the other
components have bean analyzed in 10Z (and not less than 10) of the cylinders
of the batch, the concentration label of the first component shall be
calculated separately from the meter response of each cylinder using the
equations specified in Section 3.£.1. The concentration label for the other
components on all cylinders shall be the mean concentration values of the LOZ
samples, provided all other conditions in Section 3.5.2 have been met.
3.7 Cylinder Labeling and Documentation
All I/M calibration gas cylinders 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 disposables, where the batch
number is required
(ii) Concentration of propane (in ppm), CO (in molZ),. and CO. (in molZ)
in cylinder gas (determined from Section 3.6.2), and accuracy
specification (i.e., ^22, +1.5Z or ^l.OZ)
(iii) Balance gas
(iv) Analysis date
(v) Cylinder numbers of NBS standards used in determining instrument
calibration curves
-------
40
(vi) Vendor name
' /
(vii) The statement that "This gas has been named in accordance with the
EPA Recommended Practice for Naming I/M Calibration Gas".
The gas manufacturer must retain calibration curve data on each batch analysis
of I/M calibration gas for a minimum period of two years.
3.8 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 in
decentralized I/M. programs will likely be checking on a monthly basis the
accuracies of calibration gases used with inspection analyzers. Facilities
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 cylinders named according to the Recommended Practice on the open
market and analyzing these cylinders. The results of such audits are likely
to be published by the EPA for the benefit of I/M programs.
-------
41
APPENDICES
-------
41
Appendix 1
Emission Performance Warranty
Regulations Pertaining to I/M.
Calibration Gases
§ 8&2217 Calibrations, adjustments.
(a) Equipment shall be calibrated in
accordance with the manufacturers'
instructions.
• (b) WHhin 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 ±
236 within one week of the test. These
span gases shall have concentrations
either
(i) Between the standards specified in
this subpart and the jurisdictions
inspection standards for 1981 model
year light duty vehicles, or
pi) Be within -50% to +100% of the
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 33S shall
require repair of leaks. No analyzer
adjustments shall be permitted during
this check.
40 CPS. Part 85
Page 34308
May 22, 1980
-------
42
Appendix 2
Listing of Analysis Requirements
for Recommended Practice
Step 1 : Analyze pure propane used to blend I/M calibration gas for
Analysis other hydrocarbons such as methane, iso-butane, butane and
of Pure ethane. Analyze pure carbon monoxide for methane. Analyze
Components the pure CO. (if used) for total hydrocarbons. These
analyses are not necessary prior to the blending of each batch
of I/M calibration gas, but should be performed sometime prior
to the blending of the first I/M calibration gas batch made
from newly received pure components.
Acceptance Criteria: The total "other" hydrocarbon count of pure propane
(i.e., excluding propane) must be less than .5%. The methane
count of pure carbon monoxide must be less then 100 ppm. The
total hydrocarbon count of the CO- must be less than 100 ppm.
Step 2 :
Determination
of Insrument
Linearity
Determine whether instrument(s) to be used in analysis of
I/M calibration gas is (are) linear or non-linear. The
following NBS standards (either CRMs or SRMs) must be used in
determining linearity, or a gas divider may be used with the
highest concentration standards (i.e., 2500 ppm propane in
N2, 8.0% CO in NZ or 14.0% C02 in Nj) listed below to
-------
43
obtain three points below the highest concentration standard.
If a gas divider is used, the lower points should be targeted
to within ±5Z of the concentration of the standards listed in
the table. An instrument grade 99.99Z pure nitrogen must be
used to establish the instrument zero. The 14.01 CO.
standard must be used in the linearity check if the CO.
concentration in the I/M calibration gas to be named is over
7.0Z CO.. The l.OZ CO. standard can be dropped from the
check in this situation.
-------
44
NBS Standards Used to Determine
Instrument Linearity
Allowable
Deviations
For Linear
Propane in N? Instruments
250 pptn ^ 5 ppm
500 ppm ^10 ppm
1000 ppm ^20 ppm
2500 ppm ^50 ppm
CO in N-?
l.OZ + .02Z CO
2.02 ,+.04Z CO
4.0Z. "+.08Z CO
8.0Z ™.16Z CO
l.OZ ^.02Z C02
3.0Z ^.06Z C02
4.0Z +.08Z C02
7.0Z ^.14Z C02
14.OZ + .28Z C02
•Acceptance Criteria: The criteria for determining instrument linearity
is as follows. A linear analytical instrument is defined as
one which yields three intermediate points in the range of
l.OZ CO to 8.0Z CO, 250 ppm propane to 2500 ppm propane,
and/or l.OZ CO. to 7.0Z CO. or 14. OZ C02J (depending on
the CO. concentration to be named) which deviate by ±2% of
point or less from a straight line drawn from the point
-------
45
determined by zero gas to the highest calibration point. To
be considered linear, the difference between Che
concentrations indicated by Che intermediate points and the
straight line must not exceed' _±2Z of the concentration values
of the intermediate points* The range of allowable
deviations from the straight line is illustrated in Che Cable
above. Instruments not meeting this criteria are classified
as non-linear instruments for the purpose of Chis Recommended
Practice.
Seep 3; : The following check is Co be made only if CO is to be named
CO- in an I/M calibration cylinder Chat also contains CO.,
Interference otherwise Chis step shall be omitted.
Check
CO. gas (with nitrogen as a diluent) at Che same
concentration which is to be used in Che I/M calibration
cylinder must be analyzed with Che CO analysis
instrumentation.
Acceptance CriCeria; The response of Che CO analyzer Co CO^ must be
less Chan .52 of Che concencration of CO which is Co be uaed
in Che I/M calibration cylinder.
-------
46
Step 4;
Generation
of
Calibration
Curve for
Linear
Instruments
Prior to the analysis of each batch of I/M calibration gas, a
calibration curve must be generated on the range that is to
be used to analyze the I/M calibration gas. No more than one
hour must elapse between the time at which the calibration
curve is generated and the start of the analysis of the batch
of calibration gas begins. If analysis of a portion of the
batch is postponed for two hours or more, a completely new
calibration curve must be generated.
A minimum of six cylinders must be used to generate the
calibration curve. One cylinder must be instrument grade
99.99% pure nitrogen, two other cylinders must be undiluted
NBS standards which are above and below the concentration of
the I/M calibration gas to be named, and the other three
cylinders can be primary standards or gases obtained by using
a gas divider to dilute either NBS 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 resonses 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 standards, with the others being either primary
standards or diluted NBS standards.
-------
47
A first or second order equation must be calculated from the
instrument response for the six gases.
Acceptance Criteria; If the difference between any point and the first or
second order curve is less than ±1.0Z for gases intended to
be traceable ^22 to NBS (+.5Z if gases traceable +IZ to NBS
are to be named or if CO is to be named in. a cylinder-
containing CO.), the first or second order equation thereby
derived may be used to name I/M calibration gases in this
Recommended Practice* Inability to satisfy this criterion
with a first or second order equation is an indicator of
improperly named primary standards, a malfunctioning gas
divider, and/or malfunctioning analysis instrumentation.
Rectifying these problems will likely result in the criteria
being- satisfied.
It is suggested that if a first-order 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 (i.e., there is a cluster of nearby
points above the line and another cluster below), then a
second-order equation should be generated from the
calibration points. This will increase the accuracy of the
I/M calibration gas naming process. The second order curve
must meet the _^1.02 of point C^.5Z if gases traceable ^12 to
NBS are to be named or if CO is to be named in a cylinder
containing CO.) criterion stated in the previous paragraph.
-------
Step 5;
Generation
of
Calibration
Curve for
Non-Linear
Instruments
Prior to the analysis of each batch of I/M'calibration gas
with a non-Linear instrument, a calibration curve must be
constructed on the range that is to be used in analyzing the
I/M span gas. No more than one hour must elapse between the
time at which the calibration curve is generated and the
start of the analysis of the batch of I/M calibration gas
begins. If analysis of a portion of the batch is postponed
for two hours or more, a completely new calibration curve
must be generated.
A minimum of eight cylinders must be used to construct the
calibration curve. One cylinder must be instrument-grade
99.99Z pure nitrogen, two other cylinders must be undiluted
HBS 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 NBS 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 third or fourth-order polynomial equation must be
calculated from the instrument responses to the eight gases
obtained during the calibration step.
-------
49
Acceptance Criteria; No inflection points are allowed in the equation of
the curve generated from analysis of the eight gases.* If an
inflection point occurs in a fourth-order equation, a
third-order equation should be tried. If the differences
between any point and the curve is less than ±1.0Z for gases
intended to be traceable ±2Z to NBS (+.5Z if gases traceable
±IZ to NBS are to be named or if CO is to be named in a
cylinder containing CO.), the equation thereby developed
may be used to name I/M calibration gases in this Recommended
Practice. Inability to satisfy these criterion with a third
or fourth-order equation is an indicator of improperly named
primary standards, a malfunctioning gas divider, and/or
malfunctioning analysis instrumentation.
It is suggested that if a third-order 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 (i.e.,
there is a cluster of nearby points above the curve and
another below the curve), then a fourth-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
fourth-order curve. The fourth-order curve must meet the
+IZ of point (*.5Z if gases traceable j*lZ to NBS are to be
named or if CO is to be named in a cylinder containing CO )
criterion stated in the previous paragraph.
-------
50
Step 6A
Analysis of
Re-Usable
I/M
Calibration
Gas Cylinders
Steps 5A and SB differ in that Procedure A is for analysis of
re-usable cylinders, and Procedure B is for analysis of
disposable cylinders.
After valid calibration curves are obtained for all
instruments that are to be used to analyze calibration gas,
the analysis of I/M calibration gas can begin. Each
component (with the exception of the diluent) of every
re-usable I/M calibration gas cylinder must be discreetly
analyzed. I/M calibration gas cylinders can be analyzed in
any sequence, however, the highest concentration NBS standard
used to generate the calibration curve and a zero gas
cylinder must be repeated after every ten analyses of I/M
calibration gas cylinders to ensure that substantial
instrument zero or span drift has not occurred which • could
detrimentally affect the I/M calibration gas naming process.
No analyzer- adjustments are permitted during the naming
process. Zero and span drifts are calculated as a percentage
of full scale. They are added together to arrive at a total
drift error.
Acceptance Criteria; If the total drift error changes by more than ^1.0%
(_*. 5Z if gases traceable ,+lZ to NBS are to be named or if CO
is to be named in a 'cylinder containing CO. to ^1.52 to
NBS) throughout the analysis process, a new calibration curve
must be constructed for the analyzer and the analyses of the
\i
previous ten cylinders of I/M calibration gas must be
\
repeated.
-------
51
Step 6B
Analysis of
Disposable
I fli
Calibration
Gas Cylinders
After valid calibration curves are obtained for all
instruments that are to be used to analyze the I/M
calibration gas, the analysis of I/M calibration gas can
begin. For disposable I/M calibration gas cylinders, the
requirement is that an analysis be performed on (l) one
component of each filled disposable cylinder, and on (2) the
other components, (with the exception of the diluent) on only
10Z and no less than ten cylinders total of the filled
disposable cylinders that have been blended with a continuous
analysis system or filled from a previously blended and
analyzed bulk mixture*
The mean concentration of the first component in all
cylinders in the batch is calculated.
The highest concentration NBS standard used to generate the
calibration curve and a zero gas cylinder must be repeated
after every ten analyses of Che component which is being
analyzed in 10OZ of the disposable cylinders. No analyzer
adjustments are permitted during these analysis processes.
Acceptance Criteria; If the concentration of the first component in any
cylinder in the batch is more than _£2Z different from the
mean concentration, that cylinder must have all components
discreetly analyzed.
-------
52
If the total drift error does not change by more than ^1.0%
(±.5Z if gases traceable ^12 to NBS are to be named or if CO
is to be named in a. cylinder containing CO. to ^1.52 to
»
NBS) throughout the analysis process, the process is valid.
Changes which exceed the criteria above require that a new
calibration curve must be constructed and the previous ten
cylinders be re-analyzed.
For the other components in which only 10Z of the batch has
been analyzed, the standard deviation of the concentration
values of the other components in the 10Z samples must be
less than or equal to ^.52 of each sample's mean. If they
are, all cylinders in the batch shall be labeled with the
mean concentrations of the other component in the respective
101 samples. If it is not, an analysis of all cylinders for
the other components is required. If less than 10 disposable
cylinders are to be named, all components except the diluent
must be analyzed in each cylinder. Also, if a continuous
analysis system for each component -has not been used in the
case where disposables are blended from pure components
flowing through a common manifold, or where an analysis has
not been performed on each component in the bulk mixture in
the case where disposables are filled from the bulk mixture,
an analysis must be done on each component of every
disposable cylinder.
-------
53
Step 7A
Determination
of
Concentration
Values for
Re-Usable
Cylinders
Steps 6A and 6B differ in their applicability for either
re-usable or disposable cylinder.
The labeled concentration of each component in each re-usable
cylinder must be calculated from the meter responses for each
component using the equations developed for the calibration
curves.
Step 7B
Determination
of
Concentration
Values for
Disposable
Cylinders
For disposable cylinders in which each component has been
analyzed in all cylinders of a batch, the same rules apply
for determining the concentration label of each component as
apply for re-usable cylinders. For disposable cylinders in
which one component has been analyzed in every cylinder,
while the other components have been analyzed in only 102 of
the cylinders of the batch, the concentration label of the
first component shall be calculated separately for each
cylinder using the equations developed for the calibration
curves. The concentration label for the other components in
all cylinders shall be the mean concentration values of the
10Z samples, provided all other conditions pertaining to the
10% samples have been met.
-------
54
Step 8 : All I/M calibration gas cylinders named according to these
Cylinder procedures must be labeled with a tag which contains at a
Labeling minimum the following information:
and
Documentation
(i) Cylinder number, except in the use of disposables, where
the batch number is required
(ii) Concentration of propane (in ppm) CO (in molZ), CO (in
mol Z) in cylinder gas (determined from Section 3.6.2),
and accuracy specification (i.e., ±2Z, H.5Z or^l.OZ)
(iii) Balance gas
(iv) Analysis date
(v) Cylinder numbers of NBS standards used in determining
instrument calibration curves
(vi) Vendor name
(vii) The statement that "This gas has been named in accordance
with EPA Recommended Practice for Naming I/M Calibration
Gas".
Fhe gas manufacturer must retain calibration curve data on
each batch analysis of I/M calibration gas for a minimum
period of two years.
-------
55
Appendix 3
Accuracy Discussion
The purpose of this discussion is to explain how the accuracy specifications
for I/M calibration gases to NBS are derived within the Recommended Practice
for I/M calibration gas filled in a re-usable cylinder. The discussion for
disposable cylinders would be very similar although it would include a small
statistical sampling error for components which are not analyzed in every
disposable cylinder.
,•
At certain places in the Recommended Practice small working tolerances are
allowed which help make the process of naming I/M calibration gases ' a
relatively smooth process without sacrificing a great deal of accuracy. These
tolerances are limited to certain amounts so that the overall accuracy
specifications can be maintained.
The maximum total error or uncertainty in a given I/M calibration cylinder can
be estimated by adding together the maximum tolerances allowed from each
source of error.* Of course, gas manufacturers in many cases may not use—up
the maximum tolerances allowed in this Recommended Practice. But if even the
worst case stack-up of tolerances meets the required accuracy specifications,
all cylinders in a batch should be ^22 to NBS or better.
* A certain maximum percentage error allowable in the naming process does not
always translate to the same percentage error in the labeled concentration of
a cylinder. In most cases the error of the labeled concentration value would
be less than the error allowed in the naming, process. However, for the
purpose of this analysis, we have assumed that a certain percentage error
allowed in the naming process translates to the same percentage error in the
labeled concentration value of a cylinder.
-------
56
Table .5 presents an accuracy analysis of two different types of calibration
gas mixtures theoretically named by the Recommended Practice. These gases are
representative of what most I/M programs will be using. The error tolerances
for each source of -error in the Recommended Practice are added to determine an
overall accuracy (bottom of the table), which can be compared to the desired
accuracy specifications at the top of the table.
.Table 5
Accuracy Analysis of Two
Common I/M Calibration Gas Mixtures
Named to Different Accuracy Specifications
With the Recommended Practice
Accuracy Spec Desired in Accuracy Spec Desired in
I/M .Calibration Gas I/M Calibration Gas
of Propane, & CO in N2 of Propane, CO, & C02 in
Error Sources
(1) CO2 Interference
(2) Calibration Curves
(3) Total (Zero and
Span) Drift
+22 to
NBS
N/A*
I1*
+12
+12 to
NBS
N/A*
+.52
+.52
+1.52 to
~~ NBS
+.52
+ .52
+.52
+22 to
NBS
+.52
+..52
+ 12
Overall Accuracy
(Total of 1, 2, and 3)
+22
+ 12
+ 1.52
+22
* Not appliable because this gas mixture does not contain C02
-------
• 57
The reader will note that an accuracy specification of ±IZ to NBS is not
available in the Recommended Practice for an I/H calibration gas containing
CO and CO. This is because the added error of the interference between
CO and CO cannot be reasonably compensated for by lowering the acceptable
tolerances of the other sources of error (i.e., calibration curve and zero and
span drift error). To do so would make the process of naming I/M calibration
gases unduly restrictive and expensive.
Many states have specified that their auditors shall use calibration gases
which are traceable ±1Z to NBS, while the inspectors shall use gases which are
traceable ±2Z to NBS. The reasoning was that audit gas being used to pass and
fail the inspector's gas should be more accurate then the inspector's gas.
From Table 5 it is apparent that auditors are not assured with this
Recommended Practice of obtaining a four blend CO , CO, and propane in N
audit gas that is ±1Z to NBS. Some gas manufacturers, however, may have the
capability to name all components of a four—component mixture to within ^IZ of
NBS standards. The best specification within the scope of the Recommended
Practice for the four blend mixture containing CO and CO- mixture is H.5Z.
Auditors either can use this gas or, if they desire £LZ gases, may (1) carry
two cylinders which are ^12 to NBS, one being CO. in N_, the other being
CO and propane in N,, or (2) contact a scientific gas manufacturer to
' r •
discuss whether or not a four component, gas mixture named ^1% to NBS standards
can be supplied.
-------
58
Appendix 4
Discussion of Statistical Sampling
and Naming Procedure for I/M Calibration
Gases Filled in Disposable Cylinders
The Recommended Practice uses statistical sampling criteria to determine
whether or not an abbreviated procedure can be used in analyzing and naming
I/M calibration gases filled in disposable cylinders. An example is presented
below.
A batch of one-hundred disposable cylinders filled with 2.0% CO, 5.02 CO.,
and 600 ppm propane in N_ is to be analyzed and named. The batch has been
filled simultaneously from a previously analyzed bulk mixture. The following
procedures illustrate the requirements of the Recommended Practice.
All one-hundred cylinders are analyzed for CO concentration. The CO labeled
values are determined by the analyzer meter response to each disposable
cylinder and the third or fourth order equation developed for CO. A 10%
sample (with no less than 10 cylinders) is randomly selected from the batch of
100 cylinders for further analysis of CO. and propane. If the batch had
consisted of less than one-hundred cylinders, ten cylinders would still have
to be selected. For batches larger than one-hundred cylinder, 10% of the
batch is required. The ten cylinders are analyzed for CO. and propane
-------
59
concentrations. These concentrations are calculated from the analyzer meter
responses to the cylinders and the equations developed for each analysis
system.
The mean CO and propane concentration of the sample is determined. If the
mean concentration values for C0_ and propane in the subject sample are to
be used for the entire batch, the sample must display a certain amount of
homogeneity. This is established by setting limits or tolerances around the
meac concentration, outside of which the standard deviation of concentration
values must not vary if homogeneity is assumed.
The tolerance limits set by the Recommended Practice are ^.52 of the mean
concentration. If the standard deviation is within these limits, one can be
a-ssured that 66Z of the batch is within i«5Z of the mean concentration (±1
standard deviation), 952 of the batch is within £LZ of the mean concentration
0^2 standard deviations) and somewhat over 99% of the batch is within ^22 (+4
standard deviations) of the mean concentration. If the mean concentrations of
CO- and propane are 4.982 CO and 603 ppm, the tolerances around these
values are 4.982 CO. ^.0252 CO., and 602 ppm .+3 ppm. If the standard
deviation of CO- and propane concentration values are • within these limits,
' ** * A
all cylinders in the batch can be labeled with these mean concentrations. If
the standard deviations lie outside of these tolerances, all cylinders in the
batch must be analyzed for CO and propane, the concentrations of each
cylinder being calculated from the analyzer meter responses to each cylinder
and the equation developed for each analysis system.
------- |