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.

-------