Analytical requirements for Hydrocarbon Dewpoint calculation.

Chris Cowper, EffecTech Ltd., UK.

1. Introduction.
Hydrocarbon dewpoint is an important characteristic of natural gas in respect of pipeline operation
and critical for fuelling of gas turbines. It is usually specified as a limiting temperature above which
no liquids can separate from the gas. Sometimes this is replaced with a maximum quantity of
condensed liquids specification. In Northern European countries a dewpoint temperature would
typically be not greater than -2
0
C (28
0
F) at any pressure. A quantity specification could be not
more than 5 mg of condensate per cubic metre (0.04 gallons/MMSCF) at -2
0
C (28
0
F) and any
pressure. The at any pressure phrase recognises that the maximum dewpoint temperature (the
cricondentherm) is found at a pressure intermediate between those used in transmission and those
used in distribution.
The de facto method for dewpoint temperature is the cooled mirror device, in which the temperature
of a mirror surface in contact with the gas at a selected pressure is recorded as the first appearance of
liquid condensate is observed. This may be manual (1) or automatic (2). Hydrocarbon content is
measured by treating a flow of gas to the specified temperature at a selected pressure and measuring
the rate of liquid condensate formation (3). Measurement of gas composition (4) and calculation of
phase properties using an appropriate equation of state (5, 6) gives dewpoint temperature or amount of
condensate at specified conditions or both. This paper reviews the physical and analytical approaches
and gives guidelines for the latter.
2. Physical measurement.
Of the physical methods, it may be assumed that dewpoint temperature is a more rigorous
specification than quantity of condensate. It assumes, after all, zero condensate at the specified
temperature rather than not more than a defined amount. Is this a correct assessment?
Figure 1 shows a cooled mirror device diagrammatically. The cell, of volume V contains gas at
pressure P. The film of condensate is of diameter D and thickness T.
Figure 1. Cooled mirror measurement cell.
condensate film
gas in
gas out
cooled surface
diameter D
thickness T
observation
window
light
cell volume V
EffecTech
Specialists in Gas Measurement
Assuming that the observed film is 10 mm in diameter, and that its thickness is 500 nm (the
wavelength of light), and that the liquid density is 0.6, then the mass of liquid is
024 0
10
6 0 500 5
6
2
.
.
=
t
mg
Assuming a physical cell volume of 10 ml, containing gas at 28 bar, with a compression factor of
0.9, then the volume of gas is
00031 0
10 9 0
28 10
6
.
.
=

m
3
which gives a condensate concentration of 76 mg/m
3
or 0.6 gallons/MMSCF. This is the amount of
condensate necessary to register a dewpoint temperature and is a considerably greater amount than the
quantity alternative of 5 mg/m
3
.
To consider the implications of this necessary amount of condensate, we have studied ten natural
gases sampled from different delivery points at inputs to the U.K. National Transmission System.
These gases had dewpoint temperatures ranging from -38.4 to +0.9
0
C (-37.1 to +33.6
0
F). The
amounts of condensate which would separate from these gases at different temperatures below their
dewpoints were calculated and are shown in figure 2.
0
30
60
90
120
150
-30 -25 -20 -15 -10 -5 0
temp below dewpoint
0
F
c
o
n
d
e
n
s
a
t
e
g
a
l
l
s
/
M
M
S
C
F
gas 1 gas 2 gas 3 gas 4 gas 5
gas 6 gas 7 gas 8 gas 9 gas 10
Figure 2. Amount of condensate at different temperatures below dewpoint.
The amount of condensate varies widely. Of greater interest is the region closer to the dewpoint
which is shown with expanded scales in figure 3. Also shown in the figure is the line equivalent to
0.6 gallons/MMSCF, or 76 mg/m
3
. Depending on the gas composition, this value is not achieved until
the mirror reaches a temperature below the theoretical dewpoint varying between 0.3 and 2.5
0
F.
Exceptionally, gas 9 does not deposit the appropriate amount of condensate until 10
0
F below the
theoretical dewpoint. This gas has a dewpoint temperature in the region of -40
0
, and so for normal
operational needs so large an error may not be significant. It may, however, become important where
dewpoint control is being planned, particularly if the treatment involves a supercooled sidestream
rather than the entire process flow.
If dewpoint temperature is taken to mean the value which is measured with a cooled mirror device,
then these variations are, by definition, both invisible and unimportant. They do, however, represent a
difference between the measured and theoretical values which varies in magnitude between different
gases.
0
1
2
3
4
-15 -12 -9 -6 -3 0
temp below dewpoint
0
F
c
o
n
d
e
n
s
a
t
e
g
a
l
l
s
/
M
M
S
C
F
gas 1 gas 2 gas 3 gas 4 gas 5
gas 6 gas 7 gas 8 gas 9 gas 10
Figure 3. Amount of condensate at different temperatures below dewpoint.
Another consideration is the effect of measurement pressure. A cooled mirror device is used at a
selected, usually specified, pressure which is assumed to be close to the cricondentherm. If different
gas compositions lead to cricondentherms at different pressures, then an error will be introduced.
Figure 4 shows the dewpoint lines of two gases used in the study.
Figure 4. Effect of dewpoint measurement pressure.
0
200
400
600
800
1000
1200
10 15 20 25 30
Dewpoint temp
0
F
P
r
e
s
s
u
r
e
p
s
i
Measurement pressure
0.6
gas 1 gas 3
Both gases are measured at 400 psi, and give dewpoint temperatures of 26
0
F (gas 1) and 22.5
0
F (gas
3). When the full dewlines are included, however, a different picture emerges. The cricondentherm
for gas 1 is close to 400 psi, and the error from the maximum dewpoint temperature is small. For gas
3, the cricondentherm is at a considerably higher pressure, and the maximum dewpoint temperature is
26.5
0
F. This dewpoint is underestimated by 4
0
F. If the measurement pressure were adjusted to be
suitable for gas 3, then the use of that value would give an underestimate for gas 1 of a similar size.
The comprehensive GPA study (7) quotes repeatability and accuracy of the vast majority of
individual Dewscope observers as less than 1.3
0
F and 1.9
0
F respectively. There is no reason to doubt
the repeatability claim, but it is difficult to reconcile the accuracy claim with the foregoing.
3. Analysis and calculation.
Gas chromatographic analysis of natural gas for the purpose of hydrocarbon dewpoint calculation
needs to be comprehensive and competent. For the purpose of calculating heating value, density and
Wobbe Index, the analysis can be simplified by, for example, recombining all C
6
and heavier
components and measuring then as a single entity (C
6
+) (8). Their contributions to such properties are
relatively small, and so the group can have properties attributed to it which are sufficiently
representative of the sums of the individual parts. Where more detail is available, for example a
breakdown of higher components by carbon number, there is no need to distinguish between different
isomers at a particular carbon number as their properties are quite uniform.
This type of approach is not suitable when hydrocarbon dewpoint is the aim. Higher hydrocarbons,
although present in diminishing quantities with increasing carbon number, have a disproportional
influence on hydrocarbon dewpoint. Components or groups of components which can be lumped
together or even ignored for the purpose of heating value determination must be measured with low
uncertainty for dewpoint calculation. Assumptions about similar behaviour of isomers cannot be
used. Whereas the heating values of individual C
9
isomers are very similar, boiling points and vapour
pressures are not.
To measure higher hydrocarbons in detail, a high efficiency column with a sensitive detector is
needed, capable of handling wide boiling range samples. An analyser capable of temperature
programming, fitted with a wide bore non-polar capillary column and a flame ionisation detector
(FID) represents one of a number of approaches. However, rather than offer prescriptive detail of a
particular configuration, it is preferable to propose performance requirements, which may be
demonstrated (or not) to be satisfied by a number of analytical options.
In this regard, questions to be answered include:
- How far in carbon number should the analysis proceed? It would be theoretically possible,
with concentration sampling techniques, to analyse natural gas to C
20
, but would this
significantly add to our knowledge?
- What limit of detection is needed? Again, parts per trillion may be possible, but do such trace
quantities influence the hydrocarbon dewpoint temperature?
- How much detail is necessary or indeed possible? How do we account for unidentified
components?
- How do we calibrate the system?
Figure 5 shows a chromatogram produced on a laboratory analyser fitted with a wide bore capillary
column and a flame ionisation detector and used with temperature programming. It is a useful
reference point for the above questions.
C
5
C
7
C
6
C
9
C
8
C
10
benzene
cyclo-C
6
me-cyclo-C
6
toluene
C
11
C
12
1 2 4 16
19 ppm
6 ppm
0.8 ppm
0.1 ppm
Figure 5. Chromatogram.
The chromatogram has been recorded with changes in detector sensitivity, as indicated by the scale
at the top of the figure. Component separation is not optimised, since the need for low limits of
detection requires a relatively large sample size. In response to the above questions:
- How far to go? Previous work (9) shows that for gases with hydrocarbon dewpoints not
higher than 0
0
C (32
0
F), analysis to C
12
is sufficient. Analysis to higher carbon numbers will
give more information, but the difference in calculated dewpoint temperature is
insignificantly small. If calculation of higher dewpoints is required, then the analysis needs to
be extended, but such gases are likely to be considerably outside specification limits. With
lower dewpoint temperatures, it may not be necessary to analyse beyond C
9
, but if a method is
set up to handle gases near the dewpoint limit, then the normal procedure will be to go to C
12
.
- Limit of detection? Again, previous work suggests 0.1 ppm for individual peaks. As figure 1
shows, this is achievable without the need for concentration sampling. Again, lower limits of
detection give more information, but the difference in calculated property is small.
- How much detail? Figure 1 shows benzene, toluene, cyclohexane and methyl cyclohexane in
addition to the n-alkanes. C
5
and C
6
alkane isomers, of which there are few, are identified and
handled as individual components. While it is possible to give component names to some of
the alkane peaks in the C
7
and C
8
range, these should not be regarded as unambiguous
identifications. The presence of C
6
and C
7
aromatics and cycloalkanes strongly suggests that
higher homologues will also be present, but although the lighter members are identified with
some confidence, the same cannot be said of the heavier ones.
Thus, while it is an approximation, we treat all measured components emerging after n-C
8
up
to and including n-C
9
as C
9
alkane isomers. An equivalent judgement is applied to C
10
, C
11
and C
12
components. We therefore know what these components are as a category C
9
hydrocarbons, C
10
hydrocarbons, etc., but not as individuals which we could include in an
equation of state calculation. A strategy for coping with this is described in the next section.
- How do we calibrate? Generally the type of analysis shown in figure 5 is quantified by using
the carbon counting property of the flame ionisation detector. A component such as butane or
pentane, which has been measured as part of the major component analysis, is used as a
bridge transferring quantitative information to the higher hydrocarbons. Components are
measured using the ratio of their responses to that of butane or pentane and the ratio of carbon
numbers.
This use of predictable detector responses is an assumption, albeit one which is more often
correct than incorrect. It should be shown to be true for each application and on a regular
basis. This requires either the use of blends of liquid hydrocarbons, which is a difficult and
time consuming procedure, or higher hydrocarbon calibration gases. Traceable calibration
mixtures containing n-alkanes up to C
10
, benzene, toluene, cyclohexane and methyl
cyclohexane are now available in the UK. These are referenced to blends prepared and
validated by the National Physical Laboratory. It is planned to extend this to cover C
11
and
C
12
.
3.1 Data handling.
Figure 6 expands part of the figure 5 chromatogram to show more detail of the C
9
components.
n-C
8
n-C
9
126.0
0
C
151.3
0
C
Figure 6. C
9
portion of chromatogram.
Approximately 20 peaks are visible between n-C
8
and n-C
9
. While some could be tentatively
identified, all of them cannot. If we assume them to be C
9
hydrocarbons, then we can treat them as a
single group or fraction for input to the equation of state. To handle a user-defined fraction of this
type, we need information on the quantity and on the average boiling point and specific gravity of the
fraction (10). These last two items allow the critical properties to be inferred, which are needed for
the equation of state calculation.
If all the components which constitute the fraction are C
9
hydrocarbons, then we may take their
relative responses as being identical to that of n-C
9
, which means that we can quantify them either by
reference to n-C
9
in a calibration gas, or, using relative response factors, to some lighter component in
the sample (n-C
5
or n-C
6
) which has been measured independently.
The n-alkanes are clearly identified in most natural gases, and have well known boiling points.
Temperature programmed analysis using a non-polar column gives component retention times which
have a linear relationship to their boiling points. By linear interpolation between adjacent n-alkanes,
we can infer the boiling points of the individual isomers, even though they are not identified, as
indicated in figure 6.
We therefore know the quantity and boiling point of each peak, and can calculate the average
boiling point of this group of components according to equation 1.
Eq. 1
where x
i
is the amount of each peak and B.Pt.
i
is its calculated boiling point. Specific gravities vary
little between isomers of a particular carbon number, and so the well known specific gravity of the n-
alkane can be applied to the fraction.
Fractions are thus defined for each carbon number group, so that all peaks are accounted for. The
C
7
and C
8
fractions do not include those components which have been separately identified, i.e.
benzene and cyclohexane are not summed with the C
7
fraction, and toluene and methyl cyclohexane
are left out of the C
8
fraction.
The input data for the equation of state thus consists of a mixture of individual components
(nitrogen, carbon dioxide, alkanes from C
1
to C
6
, benzene, toluene, cyclohexane and methyl
cyclohexane), and carbon number fractions (FR7, FR8, FR9, FR10, FR11 and FR12).
3.2 Quantity calculation and uncertainty.
If the molar concentration of, for example, n-C
5
has been established by a separate major component
analysis, and it is also measured without interference in the minor component analysis, as in figure 5,
it can be used as an internal standard for measurement of other minor components, using equation 2.
5
5 5
nC
C n i C n i
i
A
RRF x A
x


=
,
Eq. 2
Where x
i
is the amount of unknown component i,
x
n-C5
is the known amount of n-C
5
A
i
and A
n-C5
are the areas of component i and of n-C
5
respectively, and
RRF
i,n-C5
is the response factor for component i relative to n-C
5
.
( )

=
i
i i
x
Pt. B. x
Pt. B. Fraction
The uncertainty of the measurement, Ux
i
, can then be derived from the relative uncertainties of the
contributing terms, summed in quadrature, as in equation 3.
( ) ( ) ( )
2
2
2
2
2
2
2
2
5
5
5
5
5
5
C n i
C n i
C n
C n
C n
i
C n
i
i
i
RRF
RRF u
x
x u
A
A
A
A
u
x
x u

+ +
|
|
.
|

\
|
|
|
.
|

\
|
=
,
,
Eq. 3
These individual uncertainty contributions are described below.
3.2.1 Quantity of n-C
5
.
This is derived from a major component analysis of the gas. In our case, this is a UKAS accredited
method (11), following ISO 6143 Gas analysis Comparison method (12). The best measurement
uncertainty for n-C
5
is1.4% relative, with a coverage factor (k) =2.
3.2.2 Precision of peak area ratio measurement.
Gases with individual minor component concentrations ranging from 0.03 to 230 parts per million
were analysed repeatedly, and the ratios of peak areas relative to that of n-C5 were measured and the
standard deviations of these ratios derived. These are plotted against concentration, using logarithmic
axes, in figure 7.
-14.0
-12.0
-10.0
-8.0
-6.0
-4.0
-9.0 -7.0 -5.0 -3.0 -1.0 1.0
ln(area ratio)
l
n
(
s
t
d
d
e
v
r
a
t
i
o
)
Individuals Groups Regression
Upper 95% Lower 95% Limit
Figure 7. Precision of peak area ratios.
In figure 7 both individual peaks and component groups (FR7, FR8, etc.) are plotted. Regression
analysis shows that the relationship is described by
ln(standard deviation) = -7.6189 + 0.4619 ln(peak area ratio) (Eq. 4)
The upper and lower 95% confidence limits for the regression line are calculated and also shown in
figure 7. These limits show little curvature, and so the upper limit, which is the criterion which
should be used, can be closely approximated by a straight line which has the equation
ln(standard deviation) = -5.9 + 0.47 ln(peak area ratio) (Eq. 5)
3.2.3 Relative response factor uncertainty.
As mentioned above, the flame ionisation detector is deemed to be a carbon counter: without such a
belief a huge number of analyses would not be possible. This observation and the development of the
effective carbon number concept followed hard on the heels of the introduction of the FID, which
means that many of the references which are still used (13-15) are forty years old. The most recent
(16) is from 1985. Without tempting fate by challenging these authorities, we thought it worth
checking this assumption because by doing so we could make an estimate of the uncertainty of such
factors. None of the earlier references mention uncertainty, which is, of course, a relatively recent
consideration in analytical chemistry. Without an estimate of uncertainty on relative response factors,
we cannot complete the uncertainty budget shown in equation 3.
Liquid hydrocarbon mixtures were prepared gravimetrically, containing n-alkanes from C
5
to C
12
, i-
alkanes 2-me-C
5
, 2,3-di-me-C
5
, 2,5-di-me-C
6
and 2,3,4-tri-me-C
5
, aromatics benzene and toluene, and
cycloalkanes cyclohexane and me-cyclohexane. These were analysed on the same column using the
same conditions as for the gas analyses. The mixtures were injected directly on column, using a
splitless temperature programmed vaporiser so that there was minimum chance of discrimination.
The liquid mixtures were diluted successively so as to cover three orders of magnitude. All analyses
were repeated at least three times.
Response factors were calculated relative to n-C
6
, on the basis of moles of carbon. The results are
shown in figures 8 to 11. Among the alkanes, the greater variability for n-C
5
is almost certainly due
to its volatility. It is very difficult to keep stable mixtures containing n-C
5
for more than a short
period of time. Other than that, alkanes show RRF values which are close to 1 over three orders of
magnitude.
0.90
0.95
1.00
1.05
1.10
0.01 0.1 1 10 100
micrograms
a
r
e
a
/
C
r
e
l
t
o
n
-
C
6
n-C5 n-C7 n-C8 n-C9
Figure 8. Relative response factors for n-alkanes
0.90
0.95
1.00
1.05
1.10
0.01 0.1 1 10 100
micrograms
a
r
e
a
/
C
r
e
l
t
o
n
-
C
6
n-C10 n-C11 n-C12
Figure 9. Relative response factors for n-alkanes
0.90
0.95
1.00
1.05
1.10
0.01 0.1 1 10 100
micrograms
a
r
e
a
/
C
r
e
l
t
o
n
-
C
6
i-C6 i-C7 i-C8(1) i-C8(2)
Figure 10. Relative response factors for i-alkanes
0.95
1.00
1.05
1.10
1.15
0.01 0.1 1 10 100
micrograms
a
r
e
a
/
C
r
e
l
t
o
n
-
C
6
Benzene Toluene cyclo-C6 me-cy-C6
Figure 11. Relative response factors for aromatics/cycloalkanes
. Figure 11 is more difficult to explain. Analysis of the original mixture gave a RRF for benzene
which was 15% higher than the theoretical carbon-based value, and for toluene 10% higher. This
anomaly was visible at the first stage of dilution, though to a reduced degree. Further dilution gave
values close to the theoretical value of 1. The presence of this anomaly emphasises the need to check
assumptions about detector response and not just to take it on trust.
We have no explanation for this behaviour. The entire experiment was repeated, with new mixtures
prepared and analysed, with the same result. Fortunately, when analysing these components in gases,
we are measuring fractions of micrograms, which is far below the levels at which this anomalous
behaviour is seen. There appear to be no implications for gas analysis, but there may be when
analysing liquids such as condensates.
For alkanes over the full range tested, and for aromatics and cycloalkanes over the range 28 ng to
1.7 g, the response factor per mole of carbon relative to n-C
6
may be taken as 1.000, with a standard
uncertainty of 0.012 or 1.2% relative. Combining this with the gravimetric uncertainty of mixture
preparation gives a combined uncertainty of 1.3% relative, or an expanded uncertainty (k=2) of 2.6%
relative. Note that while these values may offer a useful indication to other users, they are only
strictly valid for the instrument on which these measurements were made, under the conditions in use
at that time.
4. Dewpoint temperature uncertainty.
From the foregoing, we can derive an overall uncertainty for the concentration of each component
or fraction in a gas sample following equation 3. These uncertainties then need to be converted to an
uncertainty of dewpoint temperature. The equation of state calculation which derives phase data from
composition uses iterative numerical methods. It is not based on a relatively simple algorithm which
would be amenable to an analytical solution, such as is used to create equation 3 from equation 2.
The approach therefore is to use a Monte Carlo technique, creating a large number of randomised
compositions, and solving the dewpoint temperature for each. This methodology is consistent with
the GUM (17).
Typically, 1000 compositions may be postulated. For each component, a random number generator
creates 1000 values which are normally distributed, with a mean of the analysed value and a standard
deviation of the overall uncertainty. A dewpoint temperature is calculated for each composition, and
the mean and standard deviation of these 1000 values are taken. The standard deviation then
represents the uncertainty of calculated dewpoint temperature.
For the configuration described here, the composition contribution to the uncertainty of dewpoint
temperature for a range of natural gases is around 0.3
0
C (0.6
0
F). There is another source of
uncertainty related to the properties which are attributed to the fraction data, and to the interpolation
whereby the boiling points of unidentified peaks are calculated (clause 3.1). These uncertainty
sources do not influence composition, and so cannot be included with the previously described
calculations. These latter influences are estimated to create an uncertainty in dewpoint of similar
magnitude. Conservative combination of these sources means that the uncertainty of calculated
dewpoint due to the analytical processes should be better than 1
0
C (1.8
0
F).
There are unquantified uncertainties associated with the choice of equation of state. Use of the
Peng-Robinson equation gives lower dewpoint temperatures than the Redlich-Kwong-Soave
calculation by 2.5 to 3
0
C (4.5 to 5.5
0
F). Where comparative analysis and calculation approaches are
required, it is essential that the same equation and the same interaction parameters are used.
5. Conclusions.
- Direct measurement of hydrocarbon dewpoint underestimates the true value because of two
factors; the need to form a film of liquid thick enough to be visible and the fact that different
gases have cricondentherms at different pressures, not necessarily the chosen measurement
pressure. The size of the error is not related to the actual dewpoint temperature.
- Detailed analysis and calculation can give, among other properties, dewpoint temperature at
any pressure, the dewline, the amount and composition of condensate formed at any chosen
conditions, and the cricondentherm. There is also the advantage that when a change in
property is observed, the analytical data shows why it has occurred.
- A carefully selected and properly controlled analytical procedure can generate data which can
contribute an uncertainty of not more than 1
0
C (1.8
0
F) to a calculated dewpoint temperature.
- The analytical method must be capable of analysing components to C
12
, and have a limit of
detection of not less than 0.1 ppm. Sample handling must be paid careful attention, to ensure
that the analysed sample is not changed by component loss or enrichment. A competent
calibration procedure is necessary for accurate quantitative data.
6. References.
1. Deaton, W.M. and E.M. Frost, Jr., Bureau of Mines Apparatus for Determining the Dew
Point under Pressure, U.S.B.M. R.I. 3399, May 1938.
2. Bannell, J.L.K., A.G. Dixon and T.P. Davies, The Monitoring of Hydrocarbon Dewpoint,
Gas Quality, edited by G.J. van Rossum, Elsevier, April 1986, pp. 263-271.
3. ISO 6570:2001 Natural gas Determination of potential hydrocarbon liquid content
Gravimetric methods. International Standards Organisation, Geneva.
4. ISO/TC 193/SC 1/WG 14 CD 23874 - Natural gas Gas chromatographic requirements for
hydrocarbon dewpoint calculation. International Standards Organisation, Geneva.
5. Soave, G., Equilibrium Constants for a Modified Redlich-Kwong Equation of State, Chem.
Eng. Sci., 27, (1972), pp.1197-1203.
6. Peng, D.Y. and D.B. Robinson, A New Two-constant Equation of State, Ind. Eng. Chem.
Fundam., 15, (1976), pp.59-64.
7. Warner, H.R. Jr. et al., Hydrocarbon Dewpoint Determination of Lean Natural Gases, Gas
Processors Association.
8. ISO 6974:2001 Natural gas Determination of composition with defined uncertainty by gas
chromatography. International Standards Organisation, Geneva.
9. Cowper, C.J., Natural gas hydrocarbon dewpoint; comparison of measurement and
calculation methods, 2
nd
Gas Analysis Symposium and Exhibition, Maastricht, (2002).
10. Kesler, M.G. & B.I. Lee, Improve Prediction of Enthalpy of Fractions, Hydrocarbon
Processing, (March 197, pp.153-158.
11. Squire, G.D., Natural gas calibration mixtures Getting it right in an accredited calibration
laboratory, Natural Gas Quality Conference, Loughborough, (2002).
12. ISO 6143:2001 Gas analysis Determination of composition and checking of calibration gas
mixtures Comparison methods. International Standards Organisation, Geneva.
13. Sternberg, J.C., W.S. Gallaway and D.T.L. Jones, The mechanism of response of flame
ionisation detectors, in Gas Chromatography, edited by Brenner, N., J.E. Callen and M.D.
Weiss, Academic Press, New York, (1962), pp. 231-267.
14. Ettre, L.S., Relative molar response of hydrocarbons on the ionisation detectors, in Gas
Chromatography, edited by Brenner, N., J.E. Callen and M.D. Weiss, Academic Press, New
York, (1962), pp. 307-327.
15. Kaiser, R., Gas Phase Chromatography, Volume 3. Butterworths, Washington, (1963), pp.
99-103.
16. Scanlon, J.T. and D.E. Willis, Calculation of flame ionisation detector relative response
factors using the effective carbon number concept, J. Chromatogr. Sci., 23, (1985), pp.333-
340.
17. Guide to the Expression of Uncertainty in Measurement, BIPM/ IEC/ IFCC/ ISO/ IUPAC/
IUPAP/ OIML, International Standards Organisation, Geneva, (1995).