ISO/TR 18231:2016

TECHNICAL ISO/TR
REPORT 18231
First edition
2016-05-01
Iron ores — Wavelength dispersive
X-ray fluorescence spectrometers —
Determination of precision
Minerais de fer — Spectromètres à fluorescence à rayons X à longueur
d’onde dispersive — Détermination de la précision
Reference number
©
ISO 2016
© ISO 2016, Published in Switzerland
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized otherwise in any form
or by any means, electronic or mechanical, including photocopying, or posting on the internet or an intranet, without prior
written permission. Permission can be requested from either ISO at the address below or ISO’s member body in the country of
the requester.
ISO copyright office
Ch. de Blandonnet 8 • CP 401
CH-1214 Vernier, Geneva, Switzerland
Tel. +41 22 749 01 11
Fax +41 22 749 09 47
copyright@iso.org
www.iso.org
ii © ISO 2016 – All rights reserved

Contents Page
Foreword .iv
Introduction .v
1 Scope . 1
2 Frequency of testing . 1
3 Counter tests . 2
3.1 Counter resolution . 2
3.1.1 General. 2
3.1.2 Procedure . 4
3.1.3 Assessment of results . 6
3.2 Conductivity of the gas flow proportional counter window . 6
3.2.1 General. 6
3.2.2 Procedure . 7
3.2.3 Assessment of results . 7
3.3 Pulse shift corrector . 7
3.3.1 General. 7
3.3.2 Procedure . 8
4 Spectrometer tests . 8
4.1 General . 8
4.2 Precision . 9
4.2.1 General. 9
4.2.2 Calculation of counting statistical error .10
4.3 Test specimen .11
4.3.1 General.11
4.3.2 Sequential spectrometers .11
4.3.3 Simultaneous spectrometers .11
4.4 Instrumental conditions .11
4.4.1 General.11
4.4.2 Sequential spectrometers .12
4.4.3 Simultaneous spectrometers .12
4.5 Stability test .12
4.6 Specimen rotation test .13
4.7 Carousel reproducibility test .13
4.8 Mounting and loading reproducibility test .13
4.9 Comparison of sample holders .13
4.10 Comparison of carousel positions .14
4.11 Angular reproducibility .14
4.12 Collimator reproducibility (for sequential spectrometers fitted with an
interchangeable collimator) .14
4.13 Detector changing reproducibility (for sequential spectrometers fitted with more
than one detector).14
4.14 Crystal changing reproducibility .14
4.15 Other tests .15
4.16 Note on glass bead curvature .15
5 Determination of the dead time and the maximum usable count rate of the equipment .15
5.1 General .15
5.2 Methods of determination of dead time .16
5.2.1 General.16
5.2.2 Recommended method for determining dead time .17
Annex A (informative) Calculation of the coefficient of variation of duplicates .24
Bibliography .26
Foreword
ISO (the International Organization for Standardization) is a worldwide federation of national standards
bodies (ISO member bodies). The work of preparing International Standards is normally carried out
through ISO technical committees. Each member body interested in a subject for which a technical
committee has been established has the right to be represented on that committee. International
organizations, governmental and non-governmental, in liaison with ISO, also take part in the work.
ISO collaborates closely with the International Electrotechnical Commission (IEC) on all matters of
electrotechnical standardization.
The procedures used to develop this document and those intended for its further maintenance are
described in the ISO/IEC Directives, Part 1. In particular the different approval criteria needed for the
different types of ISO documents should be noted. This document was drafted in accordance with the
editorial rules of the ISO/IEC Directives, Part 2 (see www.iso.org/directives).
Attention is drawn to the possibility that some of the elements of this document may be the subject of
patent rights. ISO shall not be held responsible for identifying any or all such patent rights. Details of
any patent rights identified during the development of the document will be in the Introduction and/or
on the ISO list of patent declarations received (see www.iso.org/patents).
Any trade name used in this document is information given for the convenience of users and does not
constitute an endorsement.
For an explanation on the meaning of ISO specific terms and expressions related to conformity
assessment, as well as information about ISO’s adherence to the WTO principles in the Technical
Barriers to Trade (TBT) see the following URL: Foreword - Supplementary information
The committee responsible for this document is ISO/TC 102, Iron ore and direct reduced iron,
Subcommittee SC 2, Chemical analysis.
iv © ISO 2016 – All rights reserved

Introduction
If an X-ray fluorescence spectrometer is to be used for precise analyses, it needs to be functioning
correctly to specification, that is, the errors associated with the various functions of the instrument
have to be very small. It is important therefore that the spectrometer be tested to ensure that it is
indeed functioning to deliver the required precision. The objective of this Technical Report is to set
out tests that can be used to ascertain the extent of the errors and to suggest procedures for their
rectification. These tests are not used to ascertain whether the instrument is operating optimally but
to determine whether the instrument is capable of giving a preselected precision.
TECHNICAL REPORT ISO/TR 18231:2016(E)
Iron ores — Wavelength dispersive X-ray fluorescence
spectrometers — Determination of precision
1 Scope
This Technical Report describes methods of test that can be applied to wavelength dispersive X-ray
fluorescence (WD-XRF) spectrometers to ensure that the spectrometers are functioning in a manner
that allows precise analyses to be made.
The tests outlined are designed to measure the errors associated with the operation of certain parts of
the spectrometer. They are not designed to check every part of the spectrometer but only those parts
that may be the common sources of error.
It is assumed that the performance of the instrument has been optimized according to the
manufacturer’s instructions. For all tests, the two-theta angle should be carefully set for the line being
measured. The pulse height window should be set according to the manufacturer’s instructions and
should have a broad setting which may also include the escape peak for gas proportional counters. The
instrument and detector gas environment should be as specified by the manufacturer, as should the
power supply to the instrument.
NOTE Where no distinction has been made, it is assumed that a test is applicable to both sequential and
simultaneous spectrometers.
2 Frequency of testing
Testing is not required to be carried out with each batch of analyses. The frequency of testing varies
depending on the test involved. Table 1 lists the suggested frequency with which each test should be
carried out. Where specific problems are encountered, more frequent testing may be required and
remediation work performed.
Table 1 — Suggested frequency of precision tests
Frequency Test
Monthly Resolution of the gas-flow proportional counter
Resolution of the scintillation and sealed gas counters
a
Operation of the pulse height shift corrector
Half yearly Conductivity of gas-flow proportional counter window
General stability
Collimator reproducibility
Detector changing reproducibility
Crystal changing reproducibility
Angular reproducibility
Yearly Carousel reproducibility
Comparison of carousel positions
Comparison of sample holders
Sample loading and unloading
a
The position of the pulse height peak should also be checked after changing
a bottle of detector gas since a variation in the methane content of the gas will
change the position of the peak.
The frequencies with which the tests listed in Table 1 are carried out are suggested on the basis that
there have been no changes to the spectrometer. If mechanical or electronic maintenance of a major
nature is carried out, the appropriate tests should be made before the spectrometer is taken back into
routine service.
3 Counter tests
3.1 Counter resolution
3.1.1 General
3.1.1.1 Theoretical resolution
Impurities in the flow gas and contamination of the anode wire may cause gas flow proportional
counters to gradually deteriorate, which will result in both a shift and a broadening of the energy
distribution (pulse height) curve. Similarly, scintillation counters and sealed gas counters may, for
various reasons, exhibit the same gradual deterioration. This can, ultimately, adversely affect the
measurements. Impurities in detector gas can be minimized by the use of gas filters.
The resolution (RES) of a counter is related to its energy distribution curve, and is given by the measured
peak width at half height (W) expressed as a percentage of the maximum of the pulse amplitude
distribution (V), using Formula (1) where the values of W and V are in terms of arbitrary units (which
vary between instrument manufacturers) obtained from the X-axis (see Figure 1):
W
RES=×100 (1)
V
The theoretical resolution (RES ), using the full width at half height of a Gaussian distribution, can be
th
calculated using the following formulae:
RES =23, 6σ (2)
th
2 © ISO 2016 – All rights reserved

σ= n (3)
Expressed as a percentage relative to n, Formula (3) becomes:
σ = in% (4)
()
n
where
n is the number of primary electrons per incident photon (gas counters) or number of
photoelectrons collected by the first dynode of the photomultiplier tube (scintillation coun-
ters), calculated using Formula (5):
E
x
n= (5)
V
i
E is the energy of the incident radiation, in kilo electron volts (keV);
x
V is the effective ionization potential of Argon for a flow counter, in kilo electron volts (keV)
i
= 0,026 4.
Substituting Formula (5) into Formula (4), and Formula (4) into Formula (2) gives:
236× 0,026 4
38,4
RES = = (6)
th
EE
x x
Hence, for Cu Kα (E = 8,04 keV), the theoretical resolution of an Ar gas counter is 13,5 %.
3.1.1.2 Scintillation counter
For a scintillation counter:
RES = (7)
th
E
x
[1]
and for Cu Kα, the resolution should be approximately 45 %.
3.1.1.3 Practical resolution
In practice, however, the measured resolution achieved (RES ) is given in Formula (8):
m
RESk= R (8)
m
where
k is a factor that varies with the design of the counter, phosphor efficiency (scintillation
counters), diameter, cleanliness and composition of the anode wire (gas counters).
For a well-designed and clean gas-flow proportional counter, k should be less than 1,15. Hence, for such
a counter, RES should be less than 15,6 % for Cu Kα radiation. For the scintillation counter, this value
m
should be less than 52 %.
3.1.2 Procedure
This test should be carried out on all counters used in the spectrometer. Most modern instruments
provide the facility to measure pulse height distributions and to print out the counter resolution and
this facility should be used if available.
For sequential spectrometers, it is recommended that the test be carried out using either Cu Kα or Fe Kα
radiation for both detectors. However, if these lines are measured using only the scintillation counter
in actual analysis, measure an X-ray line of a major element analysed with the gas proportional counter
for testing.
If the spectrometer does not provide automatic functions to determine RES then the following
m
procedure should be used.
a) Select a sample containing the appropriate analyte and, using a lower level setting and the pulse
height analyser (PHA) window set to “threshold” (no upper level), adjust the X-ray tube power to
give a count rate of about 2 × 10 cps (counts per second).
b) Select a narrow pulse height window (2 % to 4 % of the peak voltage V of Figure 1) and decrease
the lower level setting until the count rate drops to essentially zero.
c) Increase the lower level stepwise, noting the count rate at each step, until the peak has been passed
and the count rate drops again to a very low value. Each step should be of the same width as the
pulse height window width, i.e. if the pulse height window width corresponds to 0,2 units, then
each step of the lower level should be 0,2 units.
d) Plot the count rate obtained at each step against the lower level values. An example is shown in
Figure 1.
4 © ISO 2016 – All rights reserved

W
AV B
Key
1 half peak height
2 peak width at half peak height
Figure 1 — Intensity as a function of lower level setting (arbitrary energy units) displayed for
Fe Kα radiation measured on a flow counter
The measured counter resolution RES (in %) is obtained from the plot as follows:
m
BA−
RES = ×100 (9)
m
V
where
B and A are the lower level settings at the half height positions on either side of the peak;
V is the lower level setting at the maximum of the pulse height distribution.
The pulse height distribution should be determined at a count rate indicated by the instrument
manufacturer, or just below 2 × 10 cps (see Figure 1).
When determining the pulse height distribution automatically, A, B, V and W are not usually displayed,
but the graph, peak position and detector resolution are shown together with instrument specific data.
3.1.3 Assessment of results
3.1.3.1 Gas flow proportional counter
Since the resolution of a proportional counter depends on design, an absolute resolution value cannot be
given but the value should be near to that given in Formula (6). The optimum resolution of a particular
counter may be specified by the manufacturer or it may be found with use. When RES increases by a
m
factor of 1,2 times that value, the counter should be serviced.
3.1.3.2 Scintillation and sealed gas counters
These counters usually do not exhibit significant changes in resolution unless the counter is beginning
to fail, which can be quite early for some sealed gas counters for light elements and where they have
been used at very high flux for extended periods. Their resolution can be calculated according to
Formula (9) but the k factor is instrument specific. Hence, it is sufficient to measure their resolution
experimentally on a regular basis. If a significant increase in resolution is noted or the pulse height
distribution extends outside the energy window when it did not on installation, then the counter should
be replaced.
Scintillation counters can fail due to leaks in the beryllium window, thereby admitting moisture to the
very hygroscopic NaI crystal. The effect of such moisture reduces count rates for the longer wavelengths,
e.g. ~2Å but may have relatively little effect on short wavelengths, ~0.6Å. As more moisture reacts with
the crystal, its sensitivity will be reduced for all wavelengths.
Similarly, the sensitivity of sealed counters will reduce due to leaks in the detector window. A leaking
sealed gas counter window will result in a loss of gas density and thus gas amplification.
Although a reduction in sensitivity of the scintillation and sealed gas counters does not affect counting
precision, the signal-to-noise ratio will be reduced and longer counting times will be required to achieve
a given precision. Furthermore, if the reduction in sensitivity is due to failure of the window, then this
reduction will be continuous and total detector failure will be imminent.
The best method of detecting deterioration of such counters is to set aside a stable specimen (such as an
instrument monitor) and to periodically measure the intensity of a reasonably long wavelength line (for
example, Fe Kα for the scintillation counter). Any pulse height shift or decrease of intensity over time
should be investigated to ascertain whether the scintillation (or sealed gas) counter requires replacing.
3.2 Conductivity of the gas flow proportional counter window
3.2.1 General
The gas-flow proportional counter window consists of a thin layer of plastic material (such as Mylar or
polypropylene) coated with a very thin electrically conductive layer (typically aluminium). For practical
considerations, the window material should be able to support atmospheric pressure without breakage.
Where air or helium paths are frequently used over a period of time, this material may stretch and cause
the aluminium coating to suffer from micro-cracks, resulting in the window becoming non-conducting.
A gas-flow proportional counter having a poorly conducting or non-conducting window appears to
function normally in most respects. However, measurements made with such a counter will be in error,
as the counter will give reduced count rates for samples which would normally give a high count rate.
This is especially noticeable when low and high count rates are measured consecutively.
Where 6 μm windows are used, loss of conductivity is seldom encountered, as the window usually has
a life of some years, requiring replacement only when the counter is serviced for other reasons. The
useful life of thinner windows (≤1 μm polypropylene windows are commonly used in most modern
spectrometers) varies from several months to more than a year. Where the spectrometer chamber
6 © ISO 2016 – All rights reserved

is continuously under vacuum, the window life is longer than for those where the spectrometer is let
down to normal air or Helium pressure frequently.
NOTE If the gas-flow proportional counter has a beryllium window, this test may be necessary to occasionally
check for the formation of hairline cracks.
3.2.2 Procedure
3.2.2.1 Sequential spectrometers
The procedure for sequential spectrometers is as follows.
a) Carry out the test using K Kα radiation.
b) Using low X-ray tube power (low kV and mA), select a sample that gives a count rate of between
1 000 cps and 5 000 cps.
c) Set the 2θ angle and pulse height analyser using this sample and then, using these settings, measure
the count rate for 10 s.
d) Replace the sample with one having a high potassium concentration (a briquette of potassium
chloride or potassium hydrogen phthalate is particularly suitable or a sample advised by the
instrument manufacturer); set the X-ray tube to such a power that the maximum count rate allowed
by the manufacturer is achieved. Maintain these conditions for 2 min.
e) Reduce the X-ray tube power to its starting value and measure the count rate for the original
sample again.
3.2.2.2 Simultaneous spectrometers
For simultaneous spectrometers, the test is made on those channels having gas-flow proportional
counters fitted with thin plastic windows. The test is carried out in a similar manner to the procedure
specified in 3.2.2.1, except that the radiation used is that specific to the selected channel.
NOTE Tests for simultaneous spectrometers are also applicable to fixed channels in a sequential
spectrometer.
3.2.3 Assessment of results
If the window is conducting properly, the initial count rate will be within 5 % of the final count rate
when the X-ray tube power is returned to the original settings. If the window is poorly conducting, a
reduced count rate will be observed. In such cases, the count rate should slowly return to the original
rate but for non-conducting windows, the time involved may be in excess of 12 h. The window should be
replaced if there is any evidence of poor conduction.
3.3 Pulse shift corrector
3.3.1 General
At high count rates, high detector gains will result in a shift to lower pulse amplitudes. In modern
spectrometers, the effects of pulse shift have been minimized by designing counters with a lower
dependence on gas density and composition and by the use of automatic pulse shift correction
electronics. Nevertheless, the effect of pulse shift errors on analytical accuracy can be significant.
All modern spectrometers are fitted with automatic pulse shift correction but the operation of
detectors should also be checked when the automatic shift correction is not operating at low count
rates. The threshold count rate at which the automatic shift correction operates can vary between each
detector type, instrument model and spectrometer manufacturer. This may even be programmable
with automatic pulse shift disabled or enabled in software or hardware.
For gas-flow detectors, since the pulse shift is dependent on the gas composition (proportion of argon
to methane and the level of contaminants) and gas density (temperature or pressure of the gas within
the detector), pulse shift should be checked whenever the flow gas supply cylinder has been replaced or
whenever the gas flow detector or gas density stabilizer has been repaired.
NOTE Leaks in the gas-flow circuitry may result in the erroneous entry of air, or helium in the case of liquid
measurements, which will alter the flow gas composition.
As modern spectrometers are fitted with gas density stabilizers of significant volume, detector gas lines
should be allowed to purge prior to checking pulse heights after any of the above work has been carried
out. As the volume of flow gas circuits may be up to 3 l, the purging of lines may take up to 3 h unless
the possibility exists for a higher purge rate for a short period of time. Advice from the manufacturer
should be taken here.
3.3.2 Procedure
a) For all XRF instruments, use a sample whose radiation wavelength is suitable for the detector and
crystal combination. Below are the recommended lines for each crystal:
1) TlAP – Al Kα or Mg Kα;
2) 2,5 nm to 5 nm 2d crystals (synthetic multilayer) — Al Kα, Na Kα or Mg Kα;
3) 0,18 nm to 0,41 nm 2d crystals (LiF , LiF & LiF ) — Fe Kα or Cu Kα;
200 220 420
4) Ge — S Kα;
5) PE — Al Kα, Si Kα or P Kα;
6) InSb — Si Kα.
b) Using high X-ray tube power (with high mA), select a sample, or samples, that give(s) a count rate
near to that specified by the manufacturer as the maximum for the counter.
c) Measure the pulse height using a small step size (1 % to 2 % of the peak voltage) and a step time of
not less than 1 s. Record the peak maximum position.
d) Check manufacturer instructions for the count rate at which the automatic pulse shift correction is
enabled. Reduce the X-ray tube mA setting so that a count rate below this level is obtained or use a
second sample.
e) Repeat the procedure c) and record the peak maximum position.
f) Repeat the whole procedure for each detector.
If the peak maximum positions at the low count rate and the high count rate differ by more than 5 %
then the high voltage gain of the detector requires adjusting. This adjustment may require the skills of
a qualified technician.
If the instrument uses pulse shift correction at all count rates, the test will check that the correction is
working.
4 Spectrometer tests
4.1 General
X-ray spectrometers are mainly used for quantitative analyses. The degree of precision required
in various applications varies considerably and the following tests are to determine whether the
spectrometer can deliver the required precision.
8 © ISO 2016 – All rights reserved

Determining the measurement error component of the total analytical error can be achieved by
measuring one sample 20 times (for example) and then processing the data to obtain the mean, standard
deviation and % coefficient of variation (%CoV) of the concentration or intensity values. The %CoV of
the concentration values will be that of the intensity values multiplied by the matrix correction term
for the element of interest. The standard deviation of the intensities should be close to the counting
statistical error.
If the measurement error is outside of the statistical error limits given in Table 2 and also outside of
the acceptable error for the analyses being carried out, then the following tests can be used to find the
instrumental variables which are causing the problem and they can then have their influence on the
total analytical error minimized.
For typical routine analyses, a %CoV of 0,1 % is generally satisfactory for analytical requirements.
If higher precision is required, then the instrument should be carefully maintained and the testing
frequency of parameters listed in Table 1 might need to be increased.
Testing of the spectrometer is carried out by making repeated measurements to determine the basic
instrument stability and then the repeatability as various instrumental parameters are changed, one
variable at a time, so that any error source can be found.
The statistical bounds of tests conducted present a probability distribution about the counting
statistical error (%CSE). For tests conducted at low levels of precision, it is frequently possible to obtain
results that are less than the %CSE. The probability of this occurring decreases when increasing the
number of counts accumulated per measurement and when increasing the number of measurements.
At higher levels of precision, residual instrumental errors will have an increased impact on results and
the %CoV will generally be greater than the %CSE.
Precision measurements are made on the major element of the material to be analysed using a sensitive
line. In the case below, the Fe Kα line is used.
Dead-time corrected intensities should be used for measurements since these intensity values are used
to convert intensity data into concentration values during the analytical procedure. Such intensities
are also used in calculating measuring times required to achieve the desired counting statistical error
when setting up an analytical method.
The %CSE should theoretically be based on the actual counts a detector registers. At high count rates
above about 1 000 000 cps, dead-time effects have an increasing effect on the calculated %CSE and this
value is better established using the non-dead-time corrected count rate.
At high count rates, where dead time and pulse pile up effects are high, the time interval distribution of
counts also varies from a Poisson distribution. However, such differences need not be considered when
simply trying to determine if a spectrometer is capable of obtaining a required degree of analytical
precision or if some component of the spectrometer needs to be fixed.
4.2 Precision
4.2.1 General
The precision of WD-XRF analytical methods is dependent on the error of sub-sampling the material
brought to the laboratory for analysis, the sample preparation error and the sample measurement error.
The total analytical error (T) is determined from contributing error components using Formula (10):
22 2
T= (subsampling error) +(preparationerror) +(measurementerrorr) (10)
Factors contributing to measurement precision errors are counting statistical errors, instrumental
stability errors, errors associated with moving instrumental parts and errors due to measurement
parameter settings. These error sources will be addressed in the subclauses below.
The above errors are based on 1 standard deviation (sd). Generally, the precision of an analytical
method is described using 2 sd, which corresponds to a 95 % probability that a result will be within
those bounds. The maximum counting statistical error required for an analytical procedure can be
calculated by rearranging Formula (10) and having minimized and knowing the sub sampling and
sample preparation errors. These can be obtained experimentally.
4.2.2 Calculation of counting statistical error
For an infinite number of measurements and assuming no instrumental errors:
100 N Standard deviation
%%CSEC==oV =×100 =×100 (11)
N Mean
N
where
N is the total counts accumulated per measurement.
For a finite number of measurements and assuming Poisson statistics apply:
Standard deviation
%CoV =×100 (12)
Mean
And the standard deviation = Mean , as shown in Formula (11).
If the number of measurements (n) is <20, then the sample standard deviation (based on n−1) is used
instead of the population standard deviation.
[2]
Where 20 measurements are made, the observed %CoV should not exceed 1,4 times (1 % probability )
the %CSE, as shown in Table 2.
Table 2 — %CSE and upper statistical limit for %CoV
Total counts
a a b b
%CSE 1,4 × %CSE %CSE 1,4 × %CSE
(N)
1 000 000 0,100 0,140 0,116 0,160
10 000 000 0,032 0,044 0,037 0,051
20 000 000 0,022 0,031 0,026 0,036
25 000 000 0,020 0,028 0,023 0,033
40 000 000 0,016 0,022 0,018 0,026
100 000 000 0,010 0,014 0,012 0,016
a
Using dead-time corrected counts at a count rate of 2 000 000 cps.
b
Using non-dead-time corrected counts at a count rate of 1 481 636 cps
Counting statistical errors should be calculated on the counts actually detected by the detectors and
not the dead-time corrected counts. At low count rates <1 000 000 cps, there will be little difference
in the dead-time corrected and non-dead-time corrected %CSE. At high count rates, the difference
becomes noticeable.
The non-dead-time corrected count rate can be calculated using the formula for extendable dead time
[Formula (14)]. A dead time of 0,15 μs was used in the calculation, with the value entered in seconds, i.e.
0,000 000 15 s and the count rate in counts per second.
b
The total counts (N) in Table 2 are the dead-time corrected counts accumulated. Calculation of %CSE
is based on the same counting time used to accumulate the dead-time corrected counts with a non-
dead-time corrected count rate of 1 481 636 cps.
10 © ISO 2016 – All rights reserved

a b
The difference between %CSE and %CSE will decrease with decreasing count rate and increase with
increasing count rate. There is little difference between the values below count rates of 1 million cps.
b
The %CSE values of Table 2 should be close approximations on the statistical limits obtained on flow
counters at 2 000 000 cps for most modern instruments.
Test measurements are always carried out using the dead time corrected counts. Only the counting
statistical limits are adjusted for dead-time effects.
If the %CoV for a precision test is larger than the values given in Table 2, which are limits based on
statistics alone, then the measurements should be repeated. If the measurements are outside the limits
again, then corrective measures should be taken to minimize the problem if the results do not satisfy
analytical requirements. If the spectrometer cannot meet the required precision for an analysis, tests
can be carried out at a lower count rate or its use should be limited to less precise work.
When testing spectrometers at the above levels of precision, it should be remembered that all
spectrometers have some residual instrument errors and, therefore, will not pass precision tests when
tested at some very high level of precision such as 100 million accumulated counts per measurement.
However, most spectrometers are expected to be within the statistical limits or plus 20 % of the
maximum limit when accumulating between 20 million to 40 million counts per measurement under
standard laboratory conditions since measurement errors should be and are generally small.
4.3 Test specimen
4.3.1 General
The test specimen should be robust and stable and should have a flat analytical surface. A metal alloy
or a glass disc into which the analyte(s) has/have been incorporated by fusion with a borate flux can
be used. The use of compacted powder pellets is to be actively discouraged as these are not stable over
long periods of time. The sample should be firmly fixed in the sample holder so that no movement is
possible during measurement.
4.3.2 Sequential spectrometers
Any analytical line that gives a high count rate can be used for the precision test and it is appropriate
to use the line of a major component of a typical sample. In the examples given below, Fe Kα is used
and the amount of iron in the test specimen should be such that, under normal conditions of X-ray tube
power, the maximum count rate stipulated by the manufacturer is not exceeded.
4.3.3 Simultaneous spectrometers
For simultaneous spectrometers, a test specimen that allows testing of several channels simultaneously
is desirable. Several samples may be required to cover all element channels. The test specimen should
be such that, under normal operating conditions of X-ray tube power, a count rate of around 10 cps
or higher is obtained in each of the channels under test. For channels measuring very light elements,
where such intensity may not be possible, use as high a count rate as is achievable. Very high count
rates should be avoided as there will then be significant errors in the recorded counting times used
unless three decimal places are available. Alternatively, use the fixed time method to collect sets of
counts which are very close to the target value (N) for all power settings.
4.4 Instrumental conditions
4.4.1 General
The X-ray tube should be operated at the normal working power. The sample spinner should be used, if
possible, unless otherwise stated in the specific test.
Where possible, measurements should be made using the pre-set count mode. The time, in seconds,
required to accumulate the counts should be recorded to at least three decimal places. If a pre-set count
facility is not available, or if the timer does not read to the required precision, the pre-set time mode
should be used and the measuring time should be adjusted so that approximately the same number
of counts is accumulated at each power setting. Alternatively, a pre-set counting statistical error
mode can be used for the desired level of precision. The actual number of counts accumulated for each
measurement should then be recorded for these alternate methods.
4.4.2 Sequential spectrometers
For sequential spectrometers, the collimator normally used for the major element being measured
should be used if there is more than one available. The detector should also be that normally used for
the element being measured.
A broad energy window should be used for the pulse height distribution (PHD) which may also include
the escape peak.
Setting an upper energy level at a position where the intensity is still significantly higher than zero will
lead to instability in a detector. This can be seen when comparing stability determinations between
Fe Kα to Zn Kα radiation on a scintillation detector at similar count rates, with the Zn Kα peak being
less broad, or when carrying out stability tests at high or lower intensity levels for a specific line.
4.4.3 Simultaneous spectrometers
A broad PHD window should be used with each channel under test. A trial measurement should be
made to ascertain the correct PHD window width.
4.5 Stability test
When an X-ray tube is turned on, or when total power settings are changed, it may take some time
for the output of the X-ray tube to stabilize. The period of instability, and its magnitude, varies even
between instruments of the same manufacture and it should be determined for each spectrometer.
Instability can be observed by taking repetitive counts and plotting the count rate against time.
Modern spectrometers often have this facility built into their software. The tests given below should
be carried out after the spectrometer has stabilized. The time to reach stability is likely to be longer
after an instrument is powered up after being turned off or set at low power overnight than when
power changes are made during analyses. It will also be longer when stabilizing for the higher levels of
precision.
The stability test should be carried out as follows.
a) Make 50 consecutive measurements on the test specimen in a static position. During this test, all
the spectrometer parameters should remain fixed and the sample spinner should not be used while
all measurements are made.
b) Calculate the %CoV.
The test should be performed using each of the detectors fitted to the spectrometer.
In this case, the calculated coefficient of variation should not exceed 1,23 times the counting statistical
[2]
error (%CSE) for 50 consecutive measurements. If the calculated value excee
...