AQC Guideline

ANGLOGOLD AUSTRALIA LIMITED
QUALITY CONTROL &

QUALITY ASSURANCE
GUIDELINES
AUTHORS: K. Kenny and N. J. Crase
DATE: September 2002
REPORT NUMBER: TECH12336
DISTRIBUTION: COPY
NO.
AGA Library 1
URGM 2
SDGM 3
Perth Exploration 4
Darwin Exploration 5
K. Kenny 6
V. Chamberlain - Johannesburg 7
Spare 8
TABLE OF CONTENTS
1. EXECUTIVE SUMMARY 1
2. INTRODUCTION 2
3. THE ASSAY AND QA/QC PROCESS 4

3.1 SCOPES OF W ORK AND CONTRACT LABORATORY PARTNERSHIPS 4
3.2 FIELD AND LABORATORY PROCEDURES 4
3.3 QUALITY CONTROL AND QUALITY ASSURANCE PHILOSOPHY 5
4. TERMINOLOGY & DEFINITIONS 7
4.1 SAMPLE TERMINOLOGY 7
4.2 ASSAY METHODS 9
4.3 ASSAY TERMINOLOGY 11
4.4 STATISTICAL DEFINITIONS 12
4.5 ANALYTICAL FLOWSHEET 12
5. POTENTIAL PROBLEMS 14
5.1 LABORATORY SAMPLE PREPARATION 14
5.2 ASSAYING 15
6. QA/QC OF ORIGINAL ASSAYS 16
6.1 SUBMISSION OF STANDARDS 17
6.2 SUBMISSION OF BLANKS 18
6.3 LABORATORY REPEAT ASSAYS 19
6.4 LABORATORY DUPLICATES 20
6.5 SCREEN TESTING PULVERISED SAMPLES 21
6.6 QUALITY CONTROL FORMAT 22
6.7 LABORATORY QUALITY CONTROL SUMMARY REPORT 22
7. CHECK ASSAYING 24
7.1 CHECK ASSAY SAMPLE CHOICE 24
7.2 SUBMISSION OF CHECK ASSAYS 25
7.3 SUBMISSION OF SCREEN FIRE ASSAYS 26
8. RECOMMENDED EVALUATION APPROACH 29
8.1 STANDARDS AND BLANKS 29
8.2 COMPARATIVE ASSAYS 32
8.3 COMPARATIVE ASSAYS – BIASED DATA 37
8.4 SCREEN TESTWORK 39
9. PRO-FORMA QA/QC REPORT 41
TECH12336
LIST OF APPENDICES
Appendix 1 Statistical Definitions

Appendix 2 Proforma QA/QC Report
Appendix 3 Example QA/QC Report – Analabs Check Assay Results 2000-
2001 Exploration Drilling Programme, Sunrise Dam Project, D.
Stephens, March 2002, Report No. 08.12023 (On attached CD)
Appendix 4 Suppliers of Certified Standards
Appendix 5 Example Exploration Assay Contract – Scope of Work (On
attached CD)
LIST OF FIGURES
FIGURE 4.1: ANALYTICAL FLOWSHEET 12
FIGURE 7.1: CHECK ASSAY FLOWCHART AND POTENTIAL CHECK ASSAY

COMPARISONS 27
FIGURE 8.1: BLANK ASSAYS FEBRUARY – NOVEMBER 2001 30
FIGURE 8.2: CERTIFIED STANDARD STO6/8222 - DECEMBER 2000 TO JULY 2001 31
FIGURE 8.3: CERTIFIED STANDARDS – SRPD BOX AND WHISKER PLOT 32
FIGURE 8.4: SCATTER PLOTS – ANALABS ORIGINAL V ANALABS REPEAT 35
FIGURE 8.5: QUANTILE - QUANTILE PLOT – ANALABS ORIGINAL V ANALABS REPEAT 36
FIGURE 8.6: SRPHD PLOT – ANALABS ORIGINAL V ANALABS REPEAT 36
FIGURE 8.7: SCATTER PLOTS – ANALABS ORIGINAL V SELECTIVELY BIASED ANALABS

REPEAT 37
FIGURE 8.8: QUANTILE - QUANTILE PLOT – ANALABS ORIGINAL V SELECTIVELY BIASED

ANALABS REPEAT 38
FIGURE 8.9: SRPHD PLOT – ANALABS ORIGINAL V SELECTIVELY BIASED ANALABS

REPEAT 38
FIGURE 8.10: ANALABS SCREEN TESTS - 2001 40
FIGURE 8.11: GENALYSIS SCREEN TESTS - 2001 40
TECH12336
LIST OF TABLES
TABLE 6.1: RECOMMENDED MINIMUM FREQUENCY FOR SUBMISSION OF STANDARDS,

DUPLICATES AND BLANKS 16
TABLE 8.1: ANGLOGOLD STANDARDS SUBMITTED WITH ORIGINAL ASSAYS 29
TABLE 8.2: STATISTICAL SUMMARY: ANALABS ORIGINAL (AU) V REPEAT (AUR) 33
TABLE 8.3: SCREEN TESTWORK: ORIGINAL AND CHECK LABORATORY RESULTS 39
TECH12336
1. EXECUTIVE SUMMARY
A key input component of resource estimation and grade control is the basic assay data and
thus significant effort is justified in identifying and minimising sources of error in this data. It
is also a requirement of the JORC reporting code that demonstrably effective assay QA/QC
is carried out. This report provides a recommended process for carrying out gold assay
QA/QC, and for the presentation and interpretation of QA/QC assay results received.
The fundamental issues to be addressed in any quality control and quality assurance
program involve the question of accuracy and precision. Accuracy is a measure of how
close assay results are to the “true” value. Precision, or repeatability, is the consistency with
which an assay result can be repeated. These concepts are illustrated in the diagram
below.
The diagrams above illustrate the potential for results to be “precisely wrong”.
Internal laboratory checks and internal and external check assays enable assessment of
precision. Contamination between samples is checked for by the use of blank samples.
Assessment of accuracy is carried out by the use of certified Standards. Lack of accuracy is
referred to as a bias in the data, which is a measure of how far results differ from the true
result. Unidentified bias and lack of precision may result in poor grade estimates and poor
tonnage assignment with significant economic consequences.
Effective assay QA/QC should be viewed as a management tool to build a partnering

relationship with the contract assay laboratory to effect continuing improvement in assay
procedures and reliability of results.
TECH12336 1
2. INTRODUCTION
To quote Carras (1990):
“Successful resource estimation (and grade control process) is about minimising the
sources of error associated with the gathering of data and its analysis. The approach
should be to minimise the sources of error and in particular the maximum potential error for
that orebody.”
Carras also summarised his view (supported, at least in part, by the authors’ experience) of
the relative impact on a resource estimate of potential error sources as follows:
Factor Potential Error Range

in Resource Estimate
Geological Interpretation
Drilling Orientation 0-100%
Drill hole Spacing 0-50%
Sampling Grade Bias
Drilling (eg smearing) 0-25%
Size Fraction 0-15%
Assaying 0-15%
Bulk density 0-30%
Dilution
In-situ to Recoverable Reserves 0-40%
High Grade cutting factors 0-25%
Calculation of resource estimate 0-10%
Production rate 0-30%
While geological interpretation factors far outweigh other potential error sources, a key input
component into most of the other potential error sources is the basic assay data. Thus
significant effort is justified in identifying and minimising sources of error in this data as
assay errors may have significant economic impact on a project investment decision and an
operating mine’s economics and business plan.
Gold by its very nature is difficult to sample, assay and evaluate with confidence. Samples
that are anomalous in gold may typically contain extremely small amounts of gold that are
often measured in the parts per billion range. In many sample-types all the gold may be
only a few microns in diameter, in others only coarse gold may be present. Ore grade
samples assayed for evaluation purposes may be inhomogeneous and difficult to sample
reliably. Large gold particle size may lead to erratic and unreproducible analytical data.
Gold is malleable and pulverising of samples requires special care to avoid aggregation and
rolling of gold particles.
This document is aimed at generating a Guideline which will outline the procedures for
carrying out Quality Assurance and Quality Control (QA/QC) of assay data for Resource
estimation or mine grade control at Mine Sites and Advanced Exploration projects.
The following definitions have guided the development of this report and the overall
approach to quality management within AngloGold Australia.
TECH12336 2
Quality
A trait or characteristic used to measure the degree of excellence of a product or service.
Quality Assurance (QA)
QA is the process of evaluating overall project performance on a regular basis to provide

confidence that the project will satisfy the relevant quality standards.
Quality Assurance Plan
This is a plan that guarantees a quality approach and conformance to all customer
requirements for all tasks in a project.
Quality Control (QC)
QC is the process of monitoring specific project results to determine if they comply with
relevant standards and the identification of ways to eliminate causes of unsatisfactory
performance.
Quality Criteria
Quality criteria are the characteristics of a product that determines whether it meets certain
requirements.
This document describes the process for the efficient carrying out and monitoring of results
received as well as suggested course of action given spurious results.
The prime focus of this report is QA/QC of RC and drill core samples that are used in the
resource estimation and grade control process. QA/QC of exploration samples such as
RAB, AC, soil and stream samples and multi-element geochemistry is excluded from this
report, however similar principles apply.
TECH12336 3
3. THE ASSAY AND QA/QC PROCESS
3.1 Scopes of Work and Contract Laboratory Partnerships
Any resource-related and grade control assaying should be carried out under a detailed and
auditable contract or written scope of work which describes the sample preparation and
assay technique. Most assaying is carried out under an assay contract, generally awarded
on an annual basis under a formal “call for tender” process. Each assay contract includes a
specific Scope of Work which should describe in detail the work to be carried out, the level
of accuracy required and the QA/QC measures to be used by the laboratory. An example
Scope of Work from an assay contract (from WA Exploration) is attached on CD as
Appendix 5.
Choice of assay lab is generally decided using a combination of 3 inter-related variables

listed in rank order below:
• Quality
• Turn-around
• Price
Clearly, exploration and grade control assaying have different drivers, with turnaround being
a major issue for grade control and of lesser importance for exploration. All contract labs
are in business to make a profit and over-emphasis on any one of these three variables in
the choice of lab will result in a non-optimum outcome. If price is cut to the bone,
turnaround or, more usually, quality will suffer.
No matter how good the choice of lab, it is the authors’ view that all laboratories will “drop
the ball” from time to time, irrespective of how long the turn-around is made or how high a
price is paid for the assaying. The QA/QC program is aimed at identifying these lapses and
enabling corrections to be made. The QA/QC program is not designed to provide a means
of punishing the laboratory. It is the authors’ view that the development of a “partnership”
relationship between AngloGold and the assay contractor is critical to the exploration and
mining business. Regular visits to the laboratory as well as regular, minuted meetings with
laboratory staff should be used to build this partnering relationship with the service provider.
3.2 Field and Laboratory Procedures
Assays produced for Grade Control and Resource Evaluation are the product of processes
that involve both Field Procedures and Laboratory Procedures.
Field Procedures include the following:
1. Application of geological and geostatistical knowledge and interpretation to determine

an appropriate sampling methodology, drilling orientation and spacing,
2. An understanding of statistical concepts to enable the number and size of samples to
be determined. This includes sampling theory, basic statistics and geostatistics,
3. The physical processes of collecting samples, either as diamond drill core, RC drill
cuttings or rock chips and
4. The documentation and quality control of each of these processes.
These procedures are generally under the direct or indirect daily control of AngloGold staff.
TECH12336 4
Laboratory Procedures include:
1. Sample preparation that involves drying, crushing, pulverising and splitting to produce
an assay sample, and
2. Analytical processes that include digestion and analysis to produce the final assay.
These procedures are usually carried out remote from AngloGold sites and under the
control of contract assay laboratory personnel. Control by AngloGold staff is via the agreed
Scope of Work, the Assaying Contract and intermittent laboratory audits and inspections.
The final assay received results from this two-stage process. Each stage has associated
potential errors and a total QA/QC program needs to capture potential sources of error
throughout each of these stages.
This report focuses mainly on the techniques used to assess, minimise and
potentially quantify errors at the Laboratory Procedure stage.
It is the authors’ experience that a large proportion of any error in the laboratory generally
arises at the Sample Preparation Stage and the Scope of Work for the laboratory
procedures needs to specifically address this stage, as does the QA/QC program.
QA/QC programs can be thought of as two-stage processes:
• Ongoing daily QA/QC via submission of Standards and Blanks that enables immediate
feedback on a batch basis as to whether results are of an acceptable standard. Initial
check assays may be included at this stage via re-submission of previously split pulps.
• Intermittent systematic check assays on a monthly or quarterly basis via submission of
duplicate sample splits to both the original assay laboratory and an adjudicator
laboratory. The sample choices for this program need to be designed to check both
the sample preparation and assaying components of the Analysis stage.
Standards include samples submitted by AngloGold and the laboratory (internal standards).
The client and the laboratories generally submit blank samples to check for contamination
within the laboratory.
Duplicate assays cover a range of repeat assays by the original laboratory on the original
pulp or on the residue sample, assays by the original laboratory on field duplicates, assays
by a check laboratory on material already assayed in the first laboratory. A clear description
and definition of the terminology to be used for duplicate assays is included in
“Terminology & Definitions” (Section 4).
3.3 Quality Control and Quality Assurance Philosophy
To be effective, any QA/QC program needs to have a clear set of objectives. These
objectives can be site-specific, but in general the underlying philosophy should be the same
for all programs.
The QA/QC program is not just the submission of samples to the laboratory, or the
evaluation of check assaying some time after the event. The QA/QC program must include
the quality control loop
TECH12336 5
In addition, the responsibilities of Competent Persons (JORC Code definition) and the
requirements for public reporting need to be considered. The JORC Code check list of
assessment and reporting criteria suggests consideration be given to, amongst other things,
sub-sampling techniques and sample preparation, and quality of assay data and laboratory
tests. These criteria apply to exploration results and to resource assessment programs.
TECH12336 6
4. TERMINOLOGY & DEFINITIONS
4.1 Sample Terminology
Assay Sample is a representative portion of a drill interval, for example a 12.5% split of a 1
metre RC interval or half sawn core from a one metre length of diamond drill core. The
Assay Sample is the final product of the Field Procedures and is delivered to the laboratory
in a batch. Assay sample mass is usually in the range 2.0 to 4.5kg.
Assay Pill is a commercially available pill or coarse sample of known gold grade used for
“spiking” barren coarse samples to check on components of the sample preparation stage.
This allows generation of samples of known gold grade (by accurate weighing of the barren
coarse sample and pill, and weight-averaging) to be submitted through the entire sample
preparation process. The pill may be of two types, a capsule of fine material which may be
emptied into the sample and finely dispersed throughout as well as a solid pill (~0.5cm in
diameter) which is dropped into the bag. Alastair Inglis, Principal of Assay Solutions Pty Ltd
is the contact for supplies of the Assay Pill. (Assay Solutions, 1 Karrinyup Place, Marrara,
NT 0812; telephone O8 8927 9353, facsimile 08 8927 9435, e-mail
[email protected])
Batch is a group of samples submitted to a laboratory for assay or analysis in one run.
Generally batches contain 50 to 200 samples. A pre-agreed scope of work, or procedure,
specifies the work to be done and the standards to be applied by the laboratory.
Blank samples are included in a batch to check for contamination in the sample preparation
and analytical phases of the work. These may be of two types - pulp samples or coarse
samples that appear as a typical field sample. Pulp blank samples are usually easily
identifiable by the laboratory and removed from the sample preparation process (as are
Standards) and thus check for cross-contamination in the furnace during firing. Coarse
samples check for cross contamination during both the sample preparation process as well
as in the analytical phase.
Field Duplicate samples are generally a second assay sample selected from material likely
to contain some gold mineralisation. These may be the second half of diamond drill core,
or, more usually, a second 12.5% riffle-split RC sample. They are included in the batch to
check the sample preparation and analytical phases of the work and are generally not
identifiable by the laboratory. Assay results received from these samples are may show a
high variance due to nugget effect, however they should show no bias.
Comparison of screen sizing test-work results from the field duplicates and the original
samples will check for any problems in the sample splitting process with frequency
distributions obtained for screen size weights showing similar patterns and no bias between
the field duplicate and original sample.
Field Duplicate Bulk Samples are the bulk “reject” RC samples (usually ~20kg, 75% split)
obtained after riffle splitting the drill sample to produce the original assay sample and field
duplicate samples. Screen sizing test-work on these samples will check for any problems in
the sample splitting process with frequency distributions obtained for screen size weights
showing similar patterns between the original, field duplicate and field duplicate bulk sample.
Pulp is the portion of a sample (usually 250g to 500g) that has been through the complete
sample preparation process in the laboratory and is ready to be used in the actual analytical
TECH12336 7
process. The splitting-out of this pulp sample frequently is a source of error in the Analysis
process.
While the introduction of the mills such as the LM5 has reduced pulp/residue splitting errors
(see “Residue” below), milling can result in Au segregation (where the dense gold particles
or gold-bearing sulphides tend to accumulate towards the bottom of the bowl) and Au
agglomeration (where coarse gold particles “accumulate” free gold within the sample). Both
of these may add additional sampling complications that require checking. For example, if
the pulp sample is taken by “scooping” the sample packet through the LM5 bowl (the
general practice) any sample segregation problems would potentially produce pulp assays
lower than the residue assays.
Residue is the material that remains after splitting-off the Pulp (generally 1-2kg). Check
assaying of residues examines the reliability of the pulp splitting process.
Historically, due to equipment limitations, a sub-sample of approximately 500 grams was

split off from a coarse product (say +110 micron) for production of the 75-micron pulp.
Errors introduced by this splitting process have largely been overcome by the introduction of
equipment such as the LM5 that is theoretically able to reduce the entire sample of ~2kg to
a particle size of less than 75 micron before splitting.
Standard is a pre-prepared sample or pulp that has a reliably known or certified expected
grade and standard deviation. Commercially produced pulp Standards are homogenous
and these can be used to test the accuracy of the analytical process.
Certified standards may be obtained from:
Geostats Pty Ltd., Peter Hayes; 4 Stack Street, Fremantle, WA, Australia, 6160, Phone
(+61 8) 9430 9696, Fax: (+61 8) 9430 9695). 68, Watkins Street, White Gum Valley WA
6162; e-mail: [email protected]; website: www.geostats.com.au 4 Stack Street,
Fremantle, WA; Australia, 6160, Phone (+61 8) 9430 9696, Fax: (+61 8) 9430 9695).
Rocklabs Ltd., Manager: Ian Devereux, P.O. Box 18-142, Auckland 6, New Zealand, Phone
(+64 9) 634 7696, Fax: (+64 9) 634 6896. [email protected]. Web site:
www.rocklabs.com. Local Agent: Sietronics WA, Fred Hoetmer tel: 08 9354 4581 Web site:
www.sietronics.com.au
Gannett Ltd., 43 Fredric St. Naval Base, WA, 6165 tel 08 9410 2356; fax 9437 1212; PO
Box 329 South Perth WA 6951; tel 08 9376 1164, fax 08 9368 1636.
Further details are attached in Appendix 4.
Pulp Standards are usually readily identifiable by the laboratory and thus are often removed
and treated separately from the remainder of the assay batch. As the pulp Standards are
already pulverised and homogenised they do not give any information about the quality of
the sample preparation process and only provide information on the analytical or assay
phases of the process. Laboratories also regularly include their own Standards in the
analytical process and, in terms of most assay contracts, make results of these analyses
available to the client.
Standards should be chosen carefully to cover a range of values appropriate to the grades
expected in the deposit or assay batch. For example, it would be inappropriate to use a 10
g/t Au Standard if the values are expected to be in the 1 g/t Au range.
TECH12336 8
Coarse sample Standards may be used to test both the sample preparation and analytical
processes but are rarely available as Certified Standards as they are difficult to prepare if
sample segregation and bias are to be avoided. The Assay Pill overcomes some of these
difficulties and also enables additional checking of the sample preparation process.
4.2 Assay Methods
The assay method used in any routine or check assay program should be clearly stated in
any reports dealing with these data. Either Fire Assay or Screen Fire Assay is commonly
used for gold assays. Less commonly, due to lower costs, Aqua Regia and Leachwell
assaying may be used; however care needs to be taken in use of these techniques. All
laboratories provide a detailed description of the procedure used for the particular assay
method.
Aqua Regia: Aqua Regia is a mixture of hydrochloric and nitric acid that has the ability to
dissolve gold and forms the basis of most “wet digestion” techniques of gold analysis. Its
oxidising properties make it suitable for dissolution of sulphide minerals and iron oxides,
however silicates may slowly or not fully dissolve. Determination is by AAS. The technique is
generally quite good for oxide ores, however care needs to be taken if it is suspected that
gold may be encapsulated in silica. Aqua Regia assays are less expensive than Fire Assay
and are often used as an initial screen to determine samples for follow up fire assay. The
technique is generally 10-20% less expensive than fire assay.
BLEG: BLEG is an acronym for Bulk Leach Extractable Gold which is a method that
detects gold at low concentrations in large samples of up to 2-5kgs by cyanide extraction.
The technique is generally used at the reconnaissance stage and is a partial extraction
technique in that only cyanide-soluble gold is extracted. Gold needs to be accessible to the
solution and adequate oxygen is required in the leach solution. Most laboratories carry out
the leach with cyanide strength of 0.1-0.2% w/v NaCN. Leaching is carried out generally in
the pH range 9.5-10.5, as poisonous HCN gas may be produced at a lower pH. Iron
sulphides (pyrrhotite and to a lesser extent, pyrite) may form ferrocyanide complexes. Other
minerals (cyanicides) may also form cyanide complexes that reduce the gold-dissolution
rate or consume the available cyanide. These include many gold-associated minerals
including minerals of copper, arsenic and antimony. Gold particle size also may affect the
rate of dissolution as the usual 24-hour leach time may be insufficient to dissolve larger
grains. Determination is by AAS after a solvent extraction, activated carbon or zinc pre-
concentration stage. Detection limits down to 0.05ppb are commonly available. The
technique is generally 20-40% more expensive than fire assay.
Fire Assay (FA): Fire assay is a form of quantitative chemical analysis by which noble
metals are separated and determined in ores. The objective is to form a melt of at least 2
phases- a liquid borosilicate slag and a liquid lead phase. The high degree of solubility of
the noble metals in molten metallic lead plus the large specific gravity difference between
the lead and slag permit separation of the noble metals from the slag as lead alloys. Lead is
separated from the noble metals during cupellation leaving a metallic bead for noble metal
analysis.
Stage 1. The crucible fusion: A charge (usually 50 gram) is extracted from the pulp packet
using a method that does not introduce bias or error, with allowances to be made for high
sulphide or graphitic samples. The charge is weighed (the catch weight is usually reported
digitally with analytical results) and mixed intimately with 150 grams of assay flux that
contain soda ash, litharge, borax, flour, silver nitrate and fluorspar in varying concentrations
as required. Fluxes may be varied by the assay laboratory depending upon material type
and sample matrix (e.g. oxide versus sulphide) and range from an acid, through neutral to a
TECH12336 9
basic flux. Approximately 2mg of silver are added to the charge before the fusion allowing
any trace amounts of noble metals present to be easily transferred to the cupellation stage.
Problems may occur with use of inappropriate fluxes causing splattering during firing and
cross contamination within the fire batch.
The charge and flux mixture is transferred to a fire assay crucible. Samples are fused at
1050°C in a gas or electric fired crucible furnace for approximately 40 minutes until fusion is
achieved. Lead droplets alloy with the noble metals (including the added silver) and collect
in a lead button at the base of the crucible with a slag containing the balance of the sample
and fluxes above it. The assay crucible is poured and cooled and the lead button is
separated from the slag.
Stage 2. Cupellation: The sample is then cupelled at ~1000°C in a magnesia cupel in an

electric muffle furnace to separate the noble metals from the lead. The lead oxidises and
either is absorbed by the cupel or volatilised leaving behind the noble metals as a bead or
prill.
Stage 3. Analysis: Traditionally, the gold content of the bead is determined by weighing
after silver is removed using nitric acid (“parting”).
More usually, gold is analysed via AAS with the prill transferred to a pyrex test tube. Nitric
acid and hydrochloric acid is added and the solution heated to 90°C. When the reaction is
complete the solution is cooled and the volume made up with deionised water. The solution
is then analysed by Atomic Absorption Spectrophotometry using gold standards prepared
from 99.999% pure gold. The gold content of the sample is calculated using the sample
weight, solution volume and gold content of the solution and reported in ppm (ie grams per
tonne).
A detection limit of 0.01ppm is usual with an accuracy of ± 10% at 5 times the detection
limit. All duplicate results are usually required to be reported.
Particularly high-grade gold samples may cause cross contamination within the batch and
even carry-over into subsequent fires. The assay lab should be informed if extremely high-
grade results are expected in a sample submission.
Ores high in chromite are very difficult to fire and may produce “shotting” where globules of
lead remain in the slag phase after pouring.
Other impurities may cause problems in the cupellation process. The most troublesome of
these is copper which is more difficult to oxidise than lead and so tends to concentrate in the
bead. If present in sufficient quantities, copper will arrest the cupellation process. High
concentrations of nickel, antimony, arsenic, bismuth, tellurium and selenium may also lead
to cupellation failure. If the assay lab is informed, steps can be taken to eliminate or reduce
the associated problem, eg if high Ni values are expected, additional lead would be added to
the flux to avoid poor prill formation.
Problems may occur with issues such as AAS and solution dispenser miss-calibrations that
may result in systematic errors. AAS machines also generally only read up to the 5-6g/t Au
range before samples require dilution to allow AAS-reading. Higher grade samples may
require gravimetric checking (where the prill is weighed) to obtain reliable results.
Screen Fire Assay (SFA): Screen fire assays are carried out determine whether coarse
gold (depending upon sample preparation milling specification, usually > 75 micron) is
contributing to poor assay repeatability. The technique, while expensive, also improves
TECH12336 10
precision, as assay variability due to the presence of coarse gold is addressed. A 1kg
portion of pulverised sample is dry-sieved through a disposable 75-micron screen. The
+75µm fraction (approx 50g) is weighed and the total portion (including the 75µm screen-
cloth) is assayed for gold using lead collection fire assay techniques. The -75µm fraction is
weighed, and assayed in duplicate for gold by the same technique. All weights, all assays
plus a calculated (weighted average) gold value for the sample are reported.
Note: If a stainless steel screen is used there is potential for coarse gold to be caught in the
screen. The disposable screen is a superior technique as it is fired to capture any coarse
gold particles caught in the screen.
At times it may be worthwhile to carry out total digestion using hydrofluoric acid to allow
examination of gold grain morphology to reveal whether the coarser gold grains are primary
or “milling-induced”.
Screen Fire Assays are expensive and time consuming and are thus inappropriate for
routine assaying. The technique is usually applied as a check on known mineralised
intervals (from Fire Assay) in resource drilling where coarse gold is expected or in Grade
Control as occasional checks to provide data in the examination of Grade Control to Mill
reconciliation issues.
Leachwell: Leachwell is a form of BLEG analysis using a proprietary leaching agent in the
form of a tablet that is added to the pulp. The process can be used to circumvent a
separate milling stage of sample preparation with raw ore and tablets added to a steel
cylinder containing milling balls which is then used to both mill and leach the ore
simultaneously (PAL process). This process is generally used only in the production grade
control situation (usually oxide ore only) to increase analytical sample size (100g to 5kg),
reduce costs (between aqua regia and fire assay) and to reduce sample turnaround. All
quality issues relating to BLEG analysis apply, as well as grind size/liberation issues (partial
leach) due to a less controlled grind.
Assay to Extinction: Multiple 50g-fire assays of the sample are carried out until no sample
remains and the weighted mean grade of the original sample is calculated. This process
can be used either for the original pulp to check variability (checking laboratory precision
and presence of coarse gold) or for the SFA fines assay.
Screen Tests are sizing tests carried out on assay pulps on a routine basis by the assay
laboratory to check comminution of samples to assay contract specification (usually 90-95%
passing 75 microns). QA/QC controls involve regular submission of remaining pulps and
residues to an adjudicator lab for check screen tests. These are usually wet screening
techniques. An example of comparative screen test data is shown in Section 8.4.
Screen tests may also need to be carried out on field samples to examine possible sample
splitting problems. Frequency distributions of weights obtained for each screen size should
show similar distributions between the original, field duplicate and field duplicate bulk
sample (see Field Duplicate and Field Duplicate Bulk Samples in Section 4.1 above).
Example sizes used at SDGM for this type of test work are: 4mm, 2.36mm, 2.00mm,
1.20mm, 850µm, 600µm, 300µm, 150µm and 53µm. For very fine fractions cylcosizers,
ultrasonic or laser devises may be required.
4.3 Assay Terminology
TECH12336 11
The terminology used to describe various assays, and the headings used in assay reports
are outlined below. Clear abbreviated alphabetic codes can also be used for each
laboratory and enable concise terminology for comparative analysis.
The first 50g fire-assay of a 250g original pulp is reported as assay Au.
Laboratory Repeat Assay is an assay carried out on a second 50g fire-assay charge of a
sample taken from the original 250g (or larger) pulp. Repeat Assay No. 1 on the original
pulp is reported as assay AuR (alternative is AuR1). A second repeat assay on an original
pulp is reported as assay AuR2.
Laboratory Duplicate Assay is an additional assay carried out on a second, 250g split from
the residue after the initial 250gm pulp is removed from the LM5. The assay sample may be
referred to as a split and the assay is designated AuS.
Check Assay is a blind re-assay of a previous assayed sample (pulp or residue) either at
the original lab or at an adjudicator lab or more usually, at both (see Figure 4.1). Data from
these samples are a key component of QA/QC assessment. The sample choices for this
program need to be designed to check both the sample preparation and assaying
components of the Analysis stage.
4.4 Statistical Definitions
Evaluation of the accuracy and precision of the analytical techniques and the assessment of
the level of confidence to be given to the reported assay results requires the use of a
number of standard statistical functions. These functions are defined and discussed in
Appendix 1.
4.5 Analytical Flowsheet
Figure 4.1 illustrates the recommended analytical flow sheet including Sample Preparation
(in blue), Analysis (in red), Screen Testwork (in orange) and Check Assays (in green).
Figure 4.1 Analytical Flowsheet (overleaf)
TECH12336 12
TECH12336 13
5. POTENTIAL PROBLEMS
The recommended laboratory sample preparation and assay protocol is outlined in Figure
4.1. Potential problems, pitfalls and issues associated with steps within this protocol are
discussed below.
Errors can start before the actually submission of the sample to the assay laboratory. Care
needs to be taken by AngloGold staff in choice of sample numbering systems, efficient use
of Sample Despatch Orders (SDO) as well as implementation of sample reconciliation
techniques. A reconciliation system, which crosschecks the SDO with the samples actually
received by the laboratory to provide immediate feed-back on missing samples to the site, is
required.
5.1 Laboratory Sample Preparation
Potential sample preparation problems include incomplete jaw crushing of the sample, poor
splitting practice or device, incomplete pulverising of the sample and cross-contamination of
samples (usually high-grade carryover). Sample number or sample order mix-ups are also
possible but difficult to detect in samples of unknown grade, but are usually obvious in
assays of Standards or Blanks.
Incomplete jaw crushing (if required) and pulverising of samples may be due to sample bag
choking (designed to reduce the workload of the operator, or speed up the process) or to
jaw crusher settings that are too coarse to meet the safe sub-sampling regime determined
for the material. Quality control measures required to detect these practices include use of
the Assay Pill, as well as retrieving all sample material from the laboratory, weighing all
products and comparing these with the original sample weights, screen testing all material
and comparing size distributions, and spot (unannounced) laboratory audits.
The LM5 mills have capacity to pulverise in one pass a sample of maximum size of 2.5 to
3.5 kg. Ensure samples submitted are not larger than this range, otherwise the laboratory
will usually reduce the size of the sample by riffle splitting, potentially introducing additional
sampling problems.
Incomplete pulverising of sample (in LM5 mills) due to sample bag choking (designed to
reduce the workload of the operator, or speed up the process by not presenting some of the
sample to the mill) can be detected using the Assay Pill, checking product and sample
weights and size distributions and spot laboratory audits.
Generally, pulverisation times are in the range of 2-7 minutes, depending upon sample
hardness. If the entire sample is fed to the pulverising unit, incomplete pulverising may be
due to the pulverising time being too short or due to puck wear. This can be determined by
screen testing 1 in 20 samples to ensure that >90% of the sample passes 75 microns or
required contract specification. If 2 samples per batch fail this screen test the entire batch
should be re-processed through the LM5. This procedure needs to be included in the Scope
of Work.
Cross-contamination of samples occurs when housekeeping practices are shoddy.

Laboratory inspections are important in assessing the potential for contamination to be a
problem. Blank samples, which pass through the sample preparation and analytical
process, provide some degree of control but the information is not available until the
completion of the assay stage.
TECH12336 14
Collection of residues (which may be later used in expensive check assay programmes)
may also require checking by comparing the weight of the sample submitted with the sum of
the weights of the residue, the pulp plus the material used in one or more fire assays. While
there will almost always be some sample loss due to dust losses, there should be
reasonably good agreement if correct procedures are being followed.
5.2 Assaying
Potential problems at the assaying or analysis stage usually involve poor laboratory
procedures. These can include sample number mix-ups, incorrect fluxes and furnace
temperatures, cross-contamination during firing, incorrect dilution of samples, incorrect
instrument calibration and incorrect application of the assay methodology. Standard
samples included in a batch by the laboratory and by AngloGold provide the control on this
phase of the work. If two Standards in a batch give results outside pre-specified limits, the
entire batch should be re-assayed.
TECH12336 15
6. QA/QC OF ORIGINAL ASSAYS
The recommended laboratory protocol for sample preparation and analysis is outlined in
Figure 4.1.
The Assay Contract must include, in the Scope of Work, clear instructions regarding the
number and frequency of internal (laboratory) QA/QC tests to be carried out during sample
preparation and analysis and the appropriate follow-up action required when the results of
these checks do not meet pre-defined criteria. Reporting requirements for all results and
check assays must be specified.
Regular monthly (or more frequent) meetings to discuss performance with the assay
laboratory are recommended. These meetings should be minuted and action items
followed-up promptly.
External QA/QC tests (controlled by AngloGold) include the submission of Standards,

Blanks and Duplicate assays. The recommended frequency for inclusion of these additional
samples or assays is shown in Table 6.1.
Each of the following sections of this report discusses one of the QA/QC steps identified in
Figure 4.1.
Table 6.1: Recommended minimum frequency for submission of

Standards, Duplicates and Blanks
Grade Control Resource Advanced

(1) (2)
Assessment Exploration
No. of Samples per 250 100 100
Batch
No. of Standards per 3 5 3
Batch
No. of Duplicates 2 per batch 1 in 100 2 in 300
Location of Duplicates mineralised mineralised mineralised

material material material
No. of Blanks 2 per batch 2 in 300 2 in 300
(1) Resource Assessment includes Resource Definition, Resource Infill, and Resource
Extension.
(2) Advanced Exploration covers exploration targeting identified mineralisation.
As a variant, AngloGold Africa combines blanks, standards and initial early check assays
into sample submission by generating a “box-of-tricks” comprising a large box of pre-
prepared and carefully numbered and sequenced pulp samples of blanks, standards and
previously assayed pulps. This large box is despatched to the assay lab, and one sample
per batch retrieved and assayed. With the sample type blind to the assay lab, early check
assay data can be obtained along with the more usual standard and blank data.
TECH12336 16
6.1 Submission of Standards
Objective
Standards are to be inserted in all assay batches to enable continuous monitoring of the
laboratory. Internal standards used by the laboratory are an adjunct to this control but these
data are often not available in a timely manner.
Methodology
Pre-prepared standards from Gannet, Geostat or OreSearch are to be inserted in each

batch of samples. A minimum of five standards in every 100 samples is to be used for
resource evaluation work and 3 in 100 samples is the recommended frequency for
advanced exploration work. For grade control a frequency of two in 250 is recommended.
Care needs to be taken in sample numbering, recording and batch preparation to avoid
sample mix-ups. Consideration should be given to pre-numbering all Standards under direct
supervision of senior personnel.
Pre-prepared samples do not pass through the sample preparation phase, either at the rig
site, the core processing facility or in the laboratory and as such provide no information on
these parts of the sampling and assaying process.
Standards are the only way of providing an assessment of accuracy of the assaying
process. Standards should be chosen carefully to cover a range of values appropriate to the
grades expected in the deposit or assay batch.
Evaluation of results
Results of analyses of Standards should be presented in the form discussed in Section 8.1
and shown in Figures 8.2 and 8.3. Pre-determined assay tolerances should also be shown
on these plots. Alternatively the running mean and standard deviation should be shown.
Most laboratories shown assay tolerance limits of the order of ± 2 to 3 Standard Deviations
for their internal standards.
If sample mix up has occurred incorrect values will be returned. Crosschecking against other
Standards may indicate that another Standard was incorporated in the batch, either at the
time of submission or at the laboratory. All instances of Standard mix-up should be
investigated. If the correct standard can be identified, the assay should be incorporated in
the database for that standard.
The data should be checked to ensure that if two or more consecutive assays of a standard
fall outside the 2 x Standard Deviation tolerance limits the entire batch is re-submitted for
assay. This evaluation needs to be carried out as soon as the results are available to
ensure that unacceptable delays to the flow of assay data do not occur.
Longer-term evaluation should also be completed to provide information on the overall

laboratory assay performance. This evaluation should use information from the laboratory
and AngloGold Standard samples.
The long-term mean and relative standard deviation should be calculated and the bias of the
batch standards against the long-term mean should be calculated and monitored and
compared to the certified value.
TECH12336 17
Interpretation
The short-term action plan involves assessing the Standards in a way that is consistent with
the analytical contract and the laboratory internal practices.
Laboratory Internal Quality Control
Internal standards are used by Analabs as follows:
“Reference materials (Standards) incorporated into batches must, to enable

acceptance of sample data, produce results which are within the action limits (± 3.09 x
std. dev). Where one out of two standards fall outside the action limits, then 25% of
the samples must be retested to confirm validity – unless the majority (ca. 80%) of
sample results have values less than 25 x the detection limit then 10% of the samples
need retesting. Where all standards fall outside the action limits, then all the batch
data is rejected and the batch repeated.”
Potential problems
As Standard samples are generally not submitted to the sample preparation phase, they
only provide information about the assaying process.
Recommendations
Longer-term evaluation may identify a need for an increase in the frequency of laboratory
audits, the introduction of other controls, and a critical evaluation of the laboratory or a
change to another laboratory.
6.2 Submission of Blanks
Objective
Blank samples are intended to check laboratory hygiene.
Methodology
Pre-prepared blank pulps or blank samples can be included in the assay batch. It is
recommended that Blank Samples be prepared and inserted immediately after a sample
that is likely to contain mineralised material. Alternatively a sample of known grade is
prepared using the Assay Pill and this sample is immediately followed by the blank in the
batch.
Reported results should be inspected and any values above a pre-determined threshold
should be investigated.
Plots of Blanks against time, with the tolerance level shown (eg 4x detection limit), should be
prepared. An example is included in Figure 8.1.
Laboratory Internal Quality Control
Laboratories typically reject blank results when the value obtained exceeds 4 x the detection
limit.
TECH12336 18
Recommendations
• Blank samples to be crushed rock not sand. Fresh crushed rock is to be used with fresh
samples and oxidised material is to be used with oxidised samples.
• Blank samples to pass through as much as possible of the sample preparation and
assay processes.
• Blank samples to be “invisible” to the laboratory, to the extent practical. This could be
attained at least from time to time by the use of the AngloGold Africa’s “box of tricks”
approach (see section 6- Introduction).
• Blank samples to include the Assay Pill on an irregular basis. This will counter
recognition of the Blank by the laboratory. While the Assay Pill is mainly used in
checking the sample preparation process for sample “choking”, it also presents an
opportunity for “spiking” samples that the laboratory may have learned to be blank.
6.3 Laboratory Repeat Assays
Repeat assays are a second 50g-fire assay of a sample taken from the 250g pulp by the
laboratory. These assays are referred to as AuR (see Assay Flowchart – Figure 4.1).
Objective
Laboratory repeats are intended to test the assay repeatability of each 50g-fire assay
charge selected from a single 250g pulp. This can be termed the “within pulp variability”.
The two sample sets, taken from the same pulp, allow the precision of the analytical
procedure at the laboratory to be assessed without any masking or bias introduced by
sample preparation procedures.
In general, laboratory internal pulp assay repeats usually show good repeatability, with tight
distribution of assays displayed on scatterplots. No bias should be displayed ie points
should scatter evenly about the 1:1 line. Basic statistics such as mean, median, correlation
coefficient would usually show reasonable agreement. This may in part be due to the
reduction in the areas of the potential sources of error in comparison to later check assays,
such as sample miss-numbering and the laboratory not reporting inconsistent assays.
Methodology
The laboratory routinely selects these samples on a 1 in 20 basis from the 250g pulp. This
produces 5 repeat assays per batch.
Evaluation involves comparison of the original assay (Au) and the repeat assay (AuR). Data
should be tabulated in the form shown in Table 8.2 for each batch. On a longer term basis,
scatter plots, Quantile – Quantile plots and box and whisker plots of sRPHD as discussed in
Section 8.2 should be prepared. These should be considered as standard plots. Examples
are included in Table 8.2 and Figures 8.4 – 8.6.
If it is assumed that the two assays, Au and AuR should return the same value, the
analytical precision (expressed as sRPHD) can be calculated for each batch, or each group
of batches.
TECH12336 19
Interpretation
The sRPHD plots should show a symmetrical distribution about the zero percent difference
line. The mean sRPHD for individual grade ranges gives an indication of the average
deviation of assay 1 from assay 2 within the grade range. Positive values indicate assay
one is greater than assay 2.
At low grades, small variations in grade (eg 0.01, 0.02, 0.03g/t) related to detection limit
issues, result in large changes in sRPHD. These grades are of limited economic
significance and should be incorporated into a bin range of zero to 0.15g/t.
At high grades (say >30 g/t) outlier values can have a large impact on the sRPHD, and if
necessary outlier values should be excluded and the sRPHD recalculated.
Gold analysis generally has a maximum tolerance of ±15% relative between values. If the
results fall outside this range, consideration should be given to wet screening the pulps to
ensure the sample is sufficiently fine (90% <75µm).
If the wet screen test gives satisfactory results, then it is probable that the gold distribution is
spotty. Such poor repeatability gives early warning that Screen Fire Assay checks need to
be done.
6.4 Laboratory Duplicates
A Laboratory Duplicate assay is an additional assay carried out by the laboratory on a

second 250g split sample taken from the residue after the 250g pulp has been removed
from the LM5. By convention these assays are referred to as AuS assays. (See Flowchart-
Figure 4.1).
Objective
Laboratory duplicates (splits) are intended to test the assay repeatability of each pulp
separately selected from the pulverised sample. This can be termed the “between pulp
variability”.
The two sample sets, taken from the same pulverised sample but from different pulps, allow
the pulp sampling process at the laboratory to be assessed. In general, laboratory internal
assay duplicates usually show good repeatability but with higher scatter than the pulp repeat
assays. Points should scatter evenly around the 1:1 line. No bias should be evident. If bias
is present, early warning is given that check assays on residues are required to further
check the pulp splitting process.
Methodology
The laboratory routinely selects these samples on a 1 in 20 basis and submits them to the
analytical process. This produces 5 duplicate assays per batch.
To evaluate this step in the process it is necessary to have previously completed an

evaluation of analytical precision based on the repeat assays (see Section 8.2).
TECH12336 20
A comparison of the original assay (Au) and the duplicate assay (AuS) should be made.
The two assays, Au and AuS should return the same value, within the limits of the laboratory
precision.
Data should be tabulated in the form shown previously in Table 8.2 for each batch and
Standard plots (see Section 8.2) should be produced.
Interpretation
Interpretation of these results should be carried out after the assessment of the within-pulp
variability and the laboratory precision (evaluation of Repeat Assays – see Section 8.2).
If the laboratory precision as determined in Section 8.2, is significantly different from the
precision calculated using the duplicate assay data (Au and AuS), there is an indication of
splitting or sub-sampling problems at the pulp selection stage.
The evaluation of precision and bias should be carried out in a similar manner to that
discussed in Section 8.2.
If there is any bias in these results, further work involving residue check assaying should be
initiated. If a significant problem is identified, this needs to be fully pursued and may require
a change in laboratory procedures used to produce the pulp split.
In grade control assaying, the residue sample is discarded and therefore not available for
independent sub-sampling. This practice may need to be changed on an interim basis if
further testing of residue samples appears warranted.
6.5 Screen Testing Pulverised Samples
Check screen tests are part of the ongoing laboratory checks and controls, and should also
be included in adjudicator laboratory QA/QC programs.
Objective
Screen testing of pulverised samples is designed to ensure that >90% of the sample passes
75 microns.
Methodology
As a routine check, one in 20 samples is to be wet screened through 75-micron nylon cloth.
The weight of the original sample and the over- and under-size is to be recorded and
compared to the Scope of Work specification.
As part of an adjudicator laboratory QA/QC program, complete batches of samples (taken

from both Pulps and Residues) should be submitted and tested. Recommended
presentation of results is discussed in Section 8.4.
If two samples per batch fail to meet this specification the entire batch is to be re-ground.
Checks on 1 in 20 samples of the re-ground material are to be carried out, to determine if
the batch is acceptable prior to assaying.
TECH12336 21
6.6 Quality Control Format
A typical internal laboratory QA/QC procedure for the analytical process is outlined below.
A routine analytical batch consists of fifty samples made up of one analytical blank, two
standards, two duplicates, two replicates and 43 client samples. This regime enables
comparisons to be made between replicates and duplicates and gives the opportunity of
selecting appropriate standards that cover the analytical range. Common standards are
used where sample numbers are such that they must be split into several analytical batches.
This allows inter-batch comparisons to be made. The analytical blank follows the samples
through the entire process in order to monitor contamination from all sources external to the
sample.
The quality control format will vary for some techniques where set-up or autosampler
configurations are not compatible with the fifty-sample format.
The criteria for acceptance and rejection of data based on the quality control measures
(standards, replicates, duplicates and blanks) are fully documented in the on-site Site
Details and Quality Control Manual. Sample repeats are performed on part or whole
batches based on the performance of these quality control measures or in cases where
results appear anomalous.
6.7 Laboratory Quality Control Summary Report
Many laboratories offer clients the opportunity to receive statistical reports based on sound
and properly implemented quality assurance program. These reports provide the client with
added confidence in the laboratory’s ability to achieve the desired accuracy and precision of
analytical measurement.
As an example, one of the laboratories used by AngloGold produces a report detailing the
following:
• Control charting using the centre line to define the certified value of a control standard,
these plots incorporate warning (2σ) and control (3σ) limits and as such can have
considerable diagnostic value. Normal QQ Plot and a Distribution Plot for control
standard.
• Histograms of Turnaround Performance and Sample Submission.
• Blank Performance
• Repeatability Plots for Second Splits and Replicates
Analysis performed on major exploration programs is supported on a monthly basis by “The

Laboratory Quality Control Report” designed to report on all laboratory quality criteria
monitoring the work as performed.
These documents show all quality control information relevant to a client’s samples, whether
they pass or fail. Failure results in sample re-analysis, which will be evident in the report.
The report provides information on:
• Result turnaround - average, longest, shortest, 90%

• Submission rate on a daily basis
• Standard performance
• Blank performance
TECH12336 22
• Duplicate performance
• Replicate performance
• Sample preparation size distribution
All statistics are derived directly from the laboratory reporting software and are commented
upon by laboratory staff. Each statistical anomaly is related back to a job batch in which it
occurred, enabling close customer scrutiny.
This document will act as a permanent record of all internal laboratory QC history for the drill
program, and should be incorporated into the Assay Contract Scope of Work.
TECH12336 23
7. CHECK ASSAYING
Check Assaying involves blind re-assaying of previously assayed samples (pulp or residue)
either at the original laboratory or at an adjudicator laboratory, or more usually, at both.
Data from these samples are the key components of the assessment. These programs
need to be designed to check both the sample preparation and assaying components of the
laboratory procedures. QA/QC of these programs is required (as for all programs), and
Standards and Blanks are to be submitted in the batches of check assays and the
procedures outlined in Section 6 are to be applied to these batches.
Figure 4.1 shows the recommended check assay protocol, as part of the overall assay
flowchart. Figure 7.1 summarises all the potential comparisons of various check assays and
original samples.
7.1 Check Assay Sample Choice
As mineralised ore samples are usually only a small proportion of the total drilling
population, check assay samples need to be chosen carefully targeting grade ranges of
economic significance (or in some cases, geology) rather than being chosen on an arbitrary
proportion-of-samples submitted basis. It is best if ore zones are targeted, running from
waste into ore and back into waste as if only ore is targeted for check assay, statistically the
overall check assay result will be lower grade than the original ie some waste needs to be
check assayed. Grade control and resource check assay programs may well have different
critical grade ranges.
Grade control sampling is dominantly aimed at resolving the definition of Ore and waste
destinations. While stockpile grade definition is of short-term significance, in the long term,
stockpiles will be fed to the mill and any grade-assignment issues within varying stockpiles
will be resolved by the mill. Thus the key grade control issue is the correct definition of the
ore/waste boundary, as miss-classification results in ore loss to the waste dump and waste
dilution being fed to the mill. Check assay samples should thus be chosen focussing
around the waste/ore cut-off grade.
While resource check assay sampling should also pay attention to samples around the ore-
waste cut-off, the critical investment decision is to decide to develop the project or not. In
many gold deposits, material that is critical to this key economic decision is usually the high
grade end of the sample population as these samples often carry a significant proportion of
the contained gold. Check assay sampling for resource work should therefore show a wider
spread than grade control sampling, examining samples covering a grade-range from below
the possible economic cut-off to the high grade end of the distribution.
In deposits where the resource average grade is close to the possible economic cut off
grade, even more attention needs to be given to check assaying and QA/QC of assay data.
It is more important to be sure of assay quality in a low grade, marginal deposit than it is in a
clearly economic deposit where the average grade is well above the cut-off (“grade is king”
and high grades will “carry” a lot of mistakes),
TECH12336 24
7.2 Submission of Check Assays
Objective
Check assaying programs should provide an independent and auditable trail of checks on all
aspects of the sample preparation and analytical stages of the assay process. Check
assaying enables assessment of the primary laboratory and provides the information that
enables further improvements to be made.
Methodology
Check assays can be completed on the Residue Sample and on Pulps. It is essential that
Standards and Blanks be submitted as part of the Check Assay program to ensure Quality
Control is maintained.
The check assay program should be designed to focus on the key potential problem areas
and issues, as discussed above. The check program should be representative of all
batches submitted in the evaluation period, and should include enough samples for
statistical significance to be maintained.
Having chosen the grade ranges of interest across the batches to be Check Assayed, and if
both pulp and residue are to be checked, samples should be chosen that match pulp and
residue numbers. AngloGold staff should carry out sample splitting of the samples returned
from the laboratory. Precision laboratory-scale riffle splitters should be used to provide a
50:50 split of the matching pulp and residue. Strictly speaking, this additional splitting stage
introduces another potential error source; however in the authors’ experience this is rarely
significant if care is taken in splitting the sample.
One of the 50% splits should be sent to the adjudicator laboratory, with the other 50% split
being re-numbered and resubmitted “blind” to the original assay laboratory. This will allow
checking and comparison of all components of the original sample preparation and
assaying.
As the reliability of the results of any check assay program is critical to any assessment of
laboratory performance, extreme care must be taken in the splitting, sample numbering,
inclusion of standards and blanks and general sample submission. Check assay
programs where sample numbers become mixed only create more problems than
solutions, and waste money and time.
Field Duplicate samples may be submitted as part of a check assay program. These should
be prepared with the same degree of care as any other check assay samples. Where
possible the field duplicate samples should match the pulp and residue samples submitted
as part of the check assay program. For RC drill cuttings a riffle splitter should be used to
split out two representative samples of 2.5 – 3kg for submission to an adjudicator laboratory
and resubmitted “blind” to the original assay laboratory. This will allow checking and
comparison of all components of sample collection, sample preparation and assaying.
Due to nugget effect, it is reasonable to expect much greater scatter in “field duplicate to
original assay” comparisons than in check assays of the original sample. Sizing test work,
rather than check assaying, is often more useful in determining whether the initial sample
split is statistically valid as both the original and field duplicate splits should show similar size
distribution by screen.
TECH12336 25
Field duplicates for core samples, involve the submission of the second half of core (or
quarter core) for assay. As this will result in the destruction of the core sample, such
sampling is usually limited to small numbers of samples. Metallurgical testwork results may
also be examined to obtain comparisons between the original sample and composites.
Both internal and client-submitted Standards and Blanks must be examined to ensure that
QC of the check assay program is maintained at an acceptable standard.
Evaluation of the check assay results requires tabulation and plotting of the data in the form
discussed in Section 8.2. Check assay data are shown in Table 8.2 and Figures 8.4-8.6.
Interpretation
Comparison of check results from Laboratory 1 with results from Laboratory 2 should be
made for all Pulp assays. A comparison can also be made between Lab1 Original and
Lab1 Check Pulp assays. Analytical precision will be shown by the sRPHD values.
Comparison of sRPHD values for Residue assays from two sets of data will show the
precision for the splitting of the LM5 product and the analytical stages of the laboratory
process.
7.3 Submission of Screen Fire Assays
Screen fire assays have been included in the Check Assay component of the work to be
carried out although they can also be considered part of routine QA/QC of the original
assays. They check for the presence of coarse gold, i.e. +75 micron gold that will create
sample-splitting problems when taking an assay pulp sample. Annealing with other grains
during the milling process may also enlarge the naturally occurring coarse gold grain.
Objective
Screen fire assays are used to check for the presence of coarse gold.
Methodology
The methodology used by Analabs to carry out screen fire assays is described in Section
4.2. Methods used by other laboratories are similar to the Analabs procedure.
Comparison of screen fire assays and normal fire assays give an indication of the effect that
any coarse gold is having on the repeatability of assays.
The weight and grainsize of the coarse fraction gold gives a good indication of the extent of
the “coarse gold problem”. This may be due to naturally occurring coarse gold or poor
sample preparation in the laboratory. Some sample preparation equipment is known to
produce flaky gold particles due to smearing – for example Keegor Mills, LM5s which may
exacerbate a coarse gold problem.
TECH12336 26
Potential problems
Screen fire assays cannot provide a reliable total gold content for a sample where the
repeatability of the assay on the fine fraction is poor.
Recommended action
Screen fire assays should be completed at an early stage in the investigation of a property
and if coarse gold is present an appropriate program of routine screen fire assays should be
included in the sampling and assaying protocol.
Figure 7.1: Check Assay Flowchart and Potential Check Assay

Comparisons
HALF CORE/ RC CUTTINGS
ANALABS
Original Au, Repeats (AuR) & Residue Split (AuS)
Pulp & Residues Returned to

PROJECT
Analabs Original
Analabs Analabs
ANR Residues Pulps ANP
Genalysis Genalysis
GPR Residues Pulps GPP
TECH12336 27
Potential Check Assay Comparisons
Assay 1 Assay 2 Parameter Checked

Analabs Genalysis Interlab. Original Check
Precision Precision Precision Split Assay
Split
Analabs Analabs X X
1
Original Pulp
Analabs X X X
Residue2
Genalysis X X
Pulp
Genalysis X X X
Residue
Analabs Analabs X X X
Pulp Residue
Genalysis X X
Pulp
Genalysis X X X
Residue
Genalysis Genalysis X X X
Pulp Residue
Analabs X X X
Residue
Analabs Genalysis X X
Residue Residue
Note 1. Equivalent to Au and AuR comparison, with the addition of a check assay split.
Note 2. Equivalent to Au and AuS comparison, with the addition of a check assay split.
TECH12336 28
8. RECOMMENDED EVALUATION APPROACH
It is recommended that a series of standardised Tables and Plots are generated on a

routine basis to facilitate rapid examination of the data for potential problems. The following
sections display and describe the recommenced standard tables and plots. It should be
made clear that the Standardised Plots are designed only to facilitate routine examination of
the data. Any problems found via the standardised output would require additional
examination of the data that may need production of other plots and statistics applicable to
the problem identified.
8.1 Standards and Blanks
8.1.1 Standardised Tables
The recommended standard table for Standards and Blanks data is shown below as Table
8.1.
Table 8.1: AngloGold Standards Submitted with Original Assays
Certified Standard No. of Calculated Values

StandardCode Value SD Samples Mean Au SD CV Mean sPD
BLANK 0.00001 0 345 0.01 0.005 0.00001 -19.65
ST70/5156 0.106 0.03 63 0.12 0.027 0.229 -9.76
ST16/7181 0.49 0.05 20 0.47 0.022 0.048 4.90
ST226 1 0.06 96 1.00 0.112 0.112 0.21
ST06/8222 1.06 0.06 16 1.01 0.034 0.033 4.66
ST227 1.2 0.06 87 1.15 0.056 0.049 4.10
ST42/9272 1.33 0.08 137 1.28 0.146 0.115 4.02
ST228 1.5 0.07 66 1.42 0.079 0.056 5.08
ST09/8170 1.99 0.13 17 1.83 0.459 0.251 8.13
ST49/8242 2.06 0.1 114 2.01 0.089 0.044 2.67
ST43/7194 3.64 0.24 22 3.65 0.115 0.032 -0.40
ST04/8193 4.72 0.2 24 4.65 0.272 0.058 1.44
ST274 6.27 0.31 5 6.42 0.217 0.034 -2.39
ST18/8239 9.7 0.52 29 9.67 0.466 0.048 0.30
ST28/8240 35.6 1.2 2 18.62 24.438 1.312 47.70
ST136/8243 49.9 1.96 1 49.30 1.20
Columns one contains the name and batch number of the Certified Standard.
Columns two and three contain the certified grade and Standard Deviation of the Certified
Standard as reported by the supplier of the standards. The SD of the standard can vary
from batch to batch, even though the certified grade remains the same. Not all of the
information on SDs of the standards has been recorded in the AngloGold database.
Columns four to eight contain information on Standards submitted during the time period
under review.
Column four of the table shows the number of assays completed on each standard. This
column should be examined to ensure that there are an adequate number of samples
present before drawing conclusions about the performance of the Standard.
TECH12336 29
Column five of the table show the mean gold grade of the Standard calculated from the
available data.
Column six of the table shows the standard deviation for the data set for each certified
standard. These data may be used in standardised plots (see Section 8.1.2).
Column seven of the table shows the coefficient of variation (CV) for data set for each
Certified Standard. The (CV) is the standard deviation of a data range normalised by
dividing by the mean of the data range. This normalisation enables the CV of many different
data ranges to be compared. In comparative data, higher CVs indicate increased spread
and or outliers in the data. Further investigation of these data may be required.
Examination of Table 8.1 indicates a number of Standards that have high CVs – ST28/8240
(2 samples only), ST09/8170 (17 samples) and ST70/5156 (63 samples). These high CVs
should be investigated and the future performance monitored as these may be
unsatisfactory standards and their continued use may have to be terminated.
Column eight of the table records the mean signed Relative Percent Difference (sRPD) for
each Certified Standard. This is a measure of the departure of the assayed value from its
certified value. The relative percent difference is used as the certified value is taken as the
“correct” value. A positive sRPD indicates that the certified value is greater (on average)
than the returned assay value. Data spread and outliers can impact on this average.
Outliers may represent sample mix-ups and should be investigated and potentially may
need to be rejected from the statistics. Often, cross checking against grades of other
Standards assayed in the batch may identify the mix up. Figure 8.2 shows these data as a
standardised plot. Data distributions need to be inspected to determine if there is a high
degree of skewness in the data (large sRPD values) or an even distribution of positive and
negative individual sRPD values which could result in a mean sRPD close to zero.
8.1.2 Standardised Plots
The recommended standardised plot for blanks is shown in Figure 8.1.
Figure 8.1: Blank assays February – November 2001
Blanks
0.07
0.06
0.05
Au ppm
0.04
FaAu
0.03
0.02
0.01
0
10/28/01
11/28/01
2/28/01
3/28/01
4/28/01
5/28/01
6/28/01
7/28/01
8/28/01
9/28/01
TECH12336 30
The recommended standardised plots for Standards are shown in Figures 8.2 and 8.3.
Figure 8.2: Certified Standard STO6/8222 - December 2000 to July 2001
ST42/9272 - 1.33 ppm
2.2
2
1.8 FaAu
Au ppm
1.6 Cert Au
1.4 2SD
1.2 Linear (FaAu)
1
0.8
2/28/01
3/ 28/01
4/ 28/ 01
5/28/01
6/28/01
7/28/ 01
8/28/01
9/28/01
10/28/01
Figure 8.2 shows all assays of a particular Certified Standard (ST42/9272) completed during 11/28/01
the specific time interval under review. Additional plots using data for other Certified
Standards should use exactly the same time period to ensure that any consistent adverse
laboratory trends can be identified.
Other features included in the chart are the certified value (column two of Table 8.1), a
warning level set at ±2 SD (the certified SD, column three, or the calculated SD, column six
of Table 8.1) and a linear regression of the actual assay value reported for the certified
standard. Ideally the data points should be symmetrically distributed about the certified
value, and should lie within the warning limits. Decreasing or increasing linear trends may
indicate potential laboratory problems, especially if several standards behave in the same
way over the same time intervals. The certified SD value and SD should be shown on these
plots unless there is evidence from a considerable (>100) number of assays of an individual
standard that there are problems with the certified values. If this is the case further
evaluation should be completed, and if warranted, a recommended new assay value and
standard deviation should be used on subsequent control plots. All users of the standard
should be advised of the change in values.
Interpretation of Figure 8.2 shows good agreement between the certified value and the
values returned by the laboratory. Most values fall within ±2 SD and there is a negligible
positive drift. The two values falling well outside these warning limits should have been
investigated. It is possible that they represent a mix-up of the submitted standard.
TECH12336 31
Figure 8.3: Certified Standards – sRPD Box and Whisker Plot
sPD and sRPHD of Standards and Blanks respectively
10
-5
-10
-15
-20
No. of Samples & Standard No's Mean sPD

+ve sRPHD/sPD : Cert Au > Au
25% sPD Positive skew in white (Median > Mean)
75% sPD Negative skew in black ( Median < Mean)
Median sPD
Figure 8.3 shows a box and whisker plot of mean sRPD for each Certified Standard (and in
column one for the Blank) used during the period. The legend on this plot is self-
explanatory.
Ideally, box and whisker plots for Certified Standards should have short whiskers and a
narrow box. This result for all Certified Standards would indicate the laboratory is producing
a consistent, unskewed result (as is generally the case above). The box should also plot at,
or very near to, the zero line indicating the laboratory is accurate. The plot above has most
of the boxes plotting above the zero line which potentially indicates the laboratory has a
slight “under-reading” problem across the grade range. This would require further checking
by examination of internal laboratory Standards results and comparison to Certified
Standard results from the check laboratory.
The above box and whisker plots also show the potentially poor performance of standards
ST70/5156 and ST09/8170 as they have large boxes and long whiskers. This indicates
there is a high degree of skewness and scatter in the data (large positive or negative sRPD
values-long whiskers, mean and median widely separated - large box). This poor
performance would need to be quickly confirmed by a similar poor result from the check
laboratory to allow a decision to discontinue use of the Standard.
8.2 Comparative Assays
Standardised tables and plots may be used to evaluate pairs of assays which may include
internal laboratory data or check assays. The data used in these examples are from the
Sunrise Dam 2001 drilling campaign. Analabs original and repeat assays have been tabled
and plotted as Au1 v Au2. The data are considered to be of high quality, with low CVs and
TECH12336 32
sRPHDs in most grade ranges. These data have also been used to show the effect of
assay bias and this is discussed in Section 8.3
The recommended standard table for comparison of check assay data (Au1 v Au2) is shown
in Table 8.2. The table title must show the laboratory and type of sample analysed. Table
8.2 shows, as an example, Analabs original assay v Analabs repeat assay.
Table 8.2: Statistical Summary: Analabs Original (Au) v Repeat (AuR)
Range No. of mean SD CV sRPHD

Au g/t Samples Au1 Au2 Au1 Au2 Au1 Au2 (mean)
0 - 0.15 2046 0.02 0.02 0.03 0.03 1.15 1.18 -0.56
0.15 - 0.5 240 0.27 0.27 0.10 0.11 0.36 0.42 0.63
0.5 - 1.0 128 0.71 0.74 0.15 0.22 0.21 0.30 -1.05
1.0 - 1.5 106 1.24 1.31 0.14 0.57 0.12 0.43 -1.12
1.5 - 5.0 245 2.72 2.77 0.92 1.34 0.34 0.48 0.70
5.0 - 10 120 7.27 7.31 1.52 2.02 0.21 0.28 0.55
> 10 240 77.40 75.81 229.70 238.80 2.97 3.15 2.02
TOTAL 3125 6.54 6.43 66.76 69.04 10.20 10.73 -0.17
All samples 1064 19.17 18.84 113.38 117.36 5.91 6.23 0.58
>0.15
Note: Au1 is Analabs Original (Au); Au2 is Analabs Repeat (AuR)
Column one of the table contains seven grade ranges into which the data should be divided.
Evaluation of a cumulative frequency distribution to determine grade ranges is not
recommended on a routine basis, given that it can be time consuming, that the number of
data points is often low, and there are advantages in comparing the same grade range, as
defined in this standardised table, across a number of check assay sample type
comparisons. Column one also enables the statistics of the total dataset to be shown
(TOTAL) and for the statistics of the data with grade greater than 0.15g/t to be shown.
It is recommended that the statistics for the low grade bin (0 – 0.15 g/t) are not evaluated in
any detail. Grades in this range are below any economic criteria and grade variation can be
influenced by many factors including analytical detection limits, contamination and AAS
machine precision limits. Measurements at or near the detection limit generally have poor
precision.
Column two of the table shows the number of samples within each grade range. This
column should be examined to ensure that there are an adequate number of samples
present before drawing conclusions about the data range. If there only a few samples in
one or more grade ranges it may be necessary to base any interpretations on the grouped
data (grade > 0.15 g/t).
Columns three and four of the table show the mean gold grade for data set one and two
respectively, within each of the grade ranges. With an adequate number of samples in each
grade range there should be very little relative difference in the mean grades of Au1 and
Au2 within each of the grade ranges.
Columns five and six of the table show the standard deviation for data set one and two
respectively, within each grade range. With an adequate number of samples in each grade
TECH12336 33
range there should be little difference in the standard deviation of Au1 and Au2 in each of
the grade ranges. Where differences do occur, the data set with the higher standard
deviation is demonstrating that there is more spread in the data, and possibly some outliers.
Further investigation of these data may be required.
Columns seven and eight of the table show the coefficient of variation (CV) for data set one
and two respectively, within each grade range. The (CV) is the standard deviation of a data
range normalised by dividing by the mean of the data range. This normalisation enables the
CV of many different data ranges to be compared. Care is needed to ensure that there are
an adequate number of assay samples in the grade range before any conclusions are
drawn, based on differences in CVs. In general the CV will be higher in the low-grade bin (0
– 0.15 g/t) and in the high-grade bin (>10 g/t) than in the middle grade ranges. At the low
grade end of the data distribution the impact of analytical detection limits, contamination and
AAS machine precision limits combine to increase the CV. At the high grade end of the
data distribution, coarse-grained (nuggetty) gold, poor grind quality and sample segregation
can combine to increase the CV. In comparative data, higher CVs indicate increased
spread and or outliers in the data. Further investigation of these data may be required.
Column nine of the table records the mean signed Relative Percent Half Difference
(sRPHD) for each grade range. The half difference is used in this case because neither
assay 1 nor assay 2 may be correct. If there is only a small difference between assay one
and assay two the sRPHD will be low (zero when there is no difference between the two
assays). A positive sRPHD indicates that assay one is greater than assay two. The sRPHD
values for each grade range represent the average for that range. Data spread and outliers
can impact on this average. Data distributions need to be inspected to determine if there is
a high degree of skewness in the data (large sRPHD values) or an even distribution of
positive and negative individual sRPHD values which could result in a mean sRPHD close to
zero.
The recommended standard plots are shown in Figures 8.4, 8.5, and 8.6.
TECH12336 34
Figure 8.4: Scatter Plots – Analabs Original v Analabs Repeat
Analabs Original Vs Analabs Reassay Au(r) Analabs Original Vs Analabs Reassay Au(r)
(>0.15ppm Au) (0.15 - 0.5 ppm Au)
40 0.6
0.55
0.5
30
Analabs reassay Au(r) g/t

0.45
0.4
20 0.35
0.3
0.25
10
0.2
0.15
0 0.1
0 10 20 30 40 0.1 0.2 0.3 0.4 0.5 0.6
Analabs Original Au g/t Analabs Original Au g/t
(0.5 - 5 ppm Au) ( > 5ppm Au)
6 40
30
3 20
10
1
0 0
0 1 2 3 4 5 6 0 10 20 30 40
Analabs Original Au g/t Analabs Original Au g/t
Figure 8.4 shows a series of scatter plots for the data included in the standard table (Table
8.2). These scatter plots cover the following grade ranges: all grades above 0.15 g/t, 0.15
to 0.5 g/t, 0.5 to 5.0 g/t, all grades above 5.0 g/t. These scatter plots can be used to obtain
a visual impression of the distribution of the data. All data points should be evenly
distributed about the 1:1 line (45˚ slope) i.e. the same number of points above and below the
1:1 line. Data comparisons between assays from the same laboratory on the same pulp
(pulp to pulp comparisons) can be expected to have less spread than comparisons between
the original assay and a residue check assay (pulp to residue comparison). Laboratory to
laboratory comparisons may introduce more scatter. When the data set contains sufficient
data points differences of this type may be apparent in the scatter plots.
TECH12336 35
Figure 8.5: Quantile - Quantile Plot – Analabs Original v Analabs Repeat
Quantile Quantile
Analabs Original Vs Analabs Reassay Au(R)
40
35
30
25
20
15
10
0
0 10 20 30 40
Figure 8.5 shows a quantile – quantile plot in which the rank order of assay one is plotted
against the rank order of assay two for all data. Departure from the 1:1 line indicates
differences between assay laboratory results and direction of the difference. The Q-Q plot
is used as it displays the overall relationship and sign of any differences between the two
datasets being compared more effectively than the scatter plots. Departure from the ideal
relationship may be exaggerated at high grades. This is less of an issue than a systematic
small difference throughout the data range. Q-Q plots are scale-sensitive and should be
examined at a number of different scales to ensure that any bias at low grades is not
masked.
Figure 8.6: sRPHD Plot – Analabs Original v Analabs Repeat
Analabs Original Vs Analabs Reassay (Au1 v Aur) sRPHD
5 Mean sRPHD
4 25% sRPHD
3 75% sRPHD
2 Median sRPHD
1 +ve sRPHD :
sRPHD
0 Au1>Au2
-1 Positive skew in
-2 white
-3 (Median > Mean)
-4
Negative skew in
-5 black
2046 240 128 106 245 120 240 3125 1079
(Mean > Median)
0.0 - 0.15 0.15 - 0.5 0.5 - 1.0 1.0 - 1.5 1.5 - 5.0 5.0 - 10 gt 10 gt 0.15
No. of Sample s & Au Range s
Figure 8.6 shows a box and whiskers plot for sRPHD for each data range. The legend on
this plot is self-explanatory. Data spread and outliers can impact on mean sRPHD. These
plots may show that there is a high degree of skewness in the data (large positive or
TECH12336 36
negative sRPHD values, mean and median widely separated, uneven length whiskers) or an
even but wide distribution of positive and negative sRPHD values (equally long whiskers).
In these cases the data distributions would need to be further examined.
The above plot (in conjunction with earlier plots and Table 8.2) demonstrates significant
scatter but limited signs of any bias. There is a tendency, as grades increase, for the
original result to be marginally higher grade than the repeat assay.
8.3 Comparative Assays – Biased Data
The data from Section 8.2 have been factored to show the effects of bias. Grades of Assay
2 (the Analabs Repeat assay) have been increased by 20% in the range 0.5 to 4.99 g/t and
15.0 to 19.9g/t Au.
The introduced bias is apparent in the scatter plots and the Q-Q plots (Figure 8.7 and 8.8).
Figure 8.7: Scatter Plots – Analabs Original v selectively biased Analabs

Repeat
(>0.15ppm Au) (0.15 - 0.5 ppm Au)
40 0.6
0.55
0.5
30
0.45
0.4
20 0.35
0.3
0.25
10
0.2
0.15
0 0.1
0 10 20 30 40 0.1 0.2 0.3 0.4 0.5 0.6
Analabs Original Au g/t
(0.5 - 5 ppm Au) ( > 5ppm Au)
6 40
30
3 20
10
1
0
0
0 1 2 3 4 5 6
0 10 20 30 40
TECH12336 37
Figure 8.8: Quantile - Quantile Plot – Analabs Original v selectively
biased Analabs Repeat
At the scale of Figure 8.8 the biased data in the grade range 0.5 to 4.99g/t Au is barely
Quantile Quantile
Analabs Original Vs Analabs Resplit Au(r)
40
30
20
10
0
0 10 20 30 40
discernible, although the biased data in the 15.0 to 19.9g/t range can be clearly seen.
The sRPHD plot (Figure 8.9) shows the impact of the biased data in the lower grade range.
Comparison of Figure 8.8 and 8.9 shows the impact of the bias at the higher grade, and the
overall impact.
Figure 8.9: sRPHD Plot – Analabs Original v selectively biased Analabs

Repeat
Analabs Original Vs Analabs Resplit (Au1 Vs Aur) sRPHD
5
Mean sRPHD
25% sRPHD
0
75% sRPHD
Median sRPHD
-5
sRPHD
+ve sRPHD :
Au1>Au2
-10 Positive skew in
white
-15 (Median > Mean)
Negative skew in
black
-20 (Mean > Median)
2046 240 128 106 245 120 240 3125 1079
0.0 - 0.15 0.15 - 0.5 0.5 - 1.0 1.0 - 1.5 1.5 - 5.0 5.0 - 10 gt 10 gt 0.15
No. of Samples & Au Ranges
TECH12336 38
8.4 Screen Testwork
Results of routine Screen Tests on assay pulps by the assay laboratory are used to check
comminution of samples to assay contract specification (usually 90-95% passing 75
microns). QA/QC controls involve regular submission of remaining pulps and residues to an
adjudicator lab for check screen tests. An example of comparative screen test data is
shown in Table 8.3 below. Screen tests performed on residues would add an additional five
columns. The overall monthly average sizing is not a good measure of the performance of
the laboratory. The key performance measure is the number of samples which fail to meet
the criteria of “> 90% of the individual sample passing 75µm” (i.e. Columns 6 and 11).
Table 8.3: Screen Testwork: Original and Check Laboratory Results
Analabs Sieve Data - 2001 Genalysis Check Sieve Data - 2001

Percentage of samples meeting criteria Percentage of samples meeting criteria
Month No. of Overall > 95% >90 <95% < 90% No. of Overall > 95% >90 <95% < 90%
samples Monthly passing passing passing samples Monthly passing passing passing
Average 75µm 75µm 75µm Average 75µm 75µm 75µm
January 249 92.2% 23% 34% 43% 357 90.8% 28.9% 28.3% 42.9%
February 121 91.7% 26% 32% 42%
March 178 93.2% 52% 42% 6% 243 91.0% 21.4% 45.7% 32.9%
April 513 97.7% 82% 18% 0% 23 96.3% 69.2% 26.1% 4.3%
May 214 96.8% 77% 23% 0% 21 94.0% 52.4% 23.8% 23.8%
June 367 96.5% 63% 37% 0% 15 92.0% 26.7% 53.3% 20.0%
July 763 96.0% 70% 30% 0% 32 92.0% 37.5% 31.3% 31.3%
August 766 96.1% 71% 29% 0% 48 91.0% 25.0% 39.6% 35.4%
September 959 97.0% 78% 22% 0% 60 90.0% 11.8% 56.9% 35.5%
October 443 95.3% 55% 45% 0% 51 90.4% 21.0% 43.5% 31.4%
November 559 96.6% 69% 30% 1% 59 91.3% 37.3% 27.1% 35.6%
December 777 96.3% 71% 27% 2% 74 95.5% 54.1% 28.4% 17.6%
YTD 4573 95.3% 59.7% 31.2% 9.1% 850 91.9% 32.7% 38.7% 28.6%
Columns two to six present results for the original assay laboratory while columns 7 to 11
present results of the check laboratory.
Column one of the table contains the time period under examination.
Columns two and three and seven and eight of the table contains the number of samples
screened and the average percent passing for the period respectively.
Columns four and nine detail the percentage of samples that achieved better than 95%
passing 75 µm.
Columns five and ten detail the percentage of samples that achieved above 90% passing 75
µm and less than 95% passing.
Columns six and eleven detail the percentage of samples that failed to achieve the required
“90% passing 75 µm” requirement.
Examination of the above data indicates Lab1 had significant problems in achieving the
required grind specification in January and February, with the problem apparently resolved
for the remainder of the year. Check screen testwork at Lab 2 however indicates that this
TECH12336 39
may not be the case, with generally poor results continuing for the entire year. Conflicting
results require immediate follow up and resolution of the issue.
Presentation of screen test data in graphical format clearly demonstrates the laboratory
performance as shown in Figures 8.10 and 8.11 below.
Figure 8.10: Analabs Screen Tests - 2001
Sieve Data Analabs - 2001

100%
90%
80%
Percentage of samples with
% passing 75 microns
70%
60%
50%
40%
30%
20%
10%
0%
July
May
January
April
June
March
October
February
August
November
December
September
> 95% 90 - 95% < 90%
Figure 8.11: Genalysis Screen Tests - 2001
Check Sieve Data Genalysis - 2001

100%
90%
80%
Percentage of samples with
% passing 75 microns
70%
60%
50%
40%
30%
20%
10%
0%
July
May
April
January
March
June
October
February
August
November
December
September
> 95% 90 - 95% < 90%
TECH12336 40
9. PRO-FORMA QA/QC REPORT
All aspects of the AngloGold QA/QC program need to be documented. These programs
plus laboratory internal QA/QC reports form a vital part of company records. Any external
audit of the QA/QC process would need to access this information.
To assist in the documentation of the QA/QC process a pro-forma QA/QC report is included
as Appendix 2. An example QA/QC Report is attached as Appendix 3.
TECH12336 41
Appendix 1 – Statistical Definitions
TECH12336
Statistical Definitions
The following statistical definitions are commonly applied to an assessment of the accuracy
and precision of the analytical techniques and hence the level of confidence given to the
reported assay results.
1. Measures of Central Tendency
Sample Mean
The mean is probably the most widely used measure of central tendency and is simply the
sum of the assay values divided by the number of assays.
Sample Mean N = Σx
n
The Running Mean is the mean of a series of assays completed over an extended period.
E.g. assays of a standard over several batches or over a three-month period.
Median
The median is the number that appears in the middle of a set of assay data when they are
arranged in ascending or descending order.
Mode
The mode is the most commonly occurring in a data set. The number of times the number
occurs is its frequency.
2. Measures of Accuracy
Accuracy
A measure of how close the assay result is to the expected value. Commercially produced
Standards are used to assess accuracy as their expected assay result, or preferred or
certified value is considered to be very reliable and the samples are homogeneous.
Accuracy of a series of assays of Standards can be gauged by the bias of the mean from
the Certified Value.
Bias
When measuring Bias between a result and a Certified Standard (usually expressed as a
percentage of the Certified Value) the following formula is used:
Bias = Certified Value - Value1 x 100

Certified Value
When measuring Bias between the mean of a number of results for that Certified Standard
result and the Certified Standard (usually expressed as a percentage of the Certified Value)
the following formula is used:
TECH12336
Bias (of a set of Standard Assays) = Certified Value - Mean of results x 100
Certified Value
While it is appropriate to use the bias formula to measure the performance of the laboratory
against a Certified Standard result, it is not appropriate to apply the bias formula to two
individual assays from separate laboratories. The relative half difference between the
assays should be used in these circumstances (see below).
3. Measures of Dispersion
Range
The range is a measure of the spread of the data
Range = largest value – smallest value
Quartiles and Percentiles
Quartiles are the values in the data range that coincide with one, two or three quarters of the
data. A Percentile is the value in the data range that coincides with a specified percentage
th
of the data. The 20 Percentile is the value in the data range that is greater than 20% of the
data (and obviously less than 80% of the data).
The Range and Percentiles only supply limited information about the distribution of values in
a data set.
Precision
Precision, or repeatability, is the consistency with which an assay result can be repeated for
a single sample. The smaller the difference between assays the higher is the precision.
Precision can be expressed as the difference between two values relative to the mean of the
two values and is often expressed as a percentage.
Precision = Difference between the values x 100

Mean of the values
= (Value1 – Value2) x 100

(Value1 + Value2)
The relationship between accuracy and precision is shown in the following figure.
TECH12336 2
Signed Relative Percent Half Difference (sRPHD)
This is another measure of precision equivalent to half the precision as previously defined.
This is a robust statistic for comparing different assays of a given set of gold samples since
it is independent of the actual gold grade and gives equal weighting to high and low grade
samples.
If an original assay and a check assay have been completed on the same material we do
not know which assay is “correct”. In reality, the best estimate of the “correct” result
probably lies close to the mean of the two assays. Thus a measure of the distance away of
one of the assays from this “true” result is half of the difference between the two results.
This can be expressed relative to the mean value giving the following formula:
Relative Half Difference = Difference between the values

Mean of the values
This is usually expressed as a percent, giving the following formula:
sRPHD = (Value1 – Value2) x 100

Mean
= (Value1 – Value2) x 100

(Value1 + Value2)
This can be simplified to:

sRPHD = (Value1 – Value2) x 100
(Value1 + Value2)
(i.e. The difference between the values over the sum of the values)
The mean sRPHD for two populations where the data is paired gives an indication of a bias
between the two populations. Positive mean sRPHD values indicate that on average assay
1 is greater than assay 2. This is also a robust statistic since it is independent of the actual
gold grade and gives equal weighting to high and low grade samples.
Unsigned Relative Percent Half Difference (uRPHD):
TECH12336 3
This gives a measure of the size but not the direction (sign) of the difference between two
values. The formula is derived in the same way as the sRPH.
uRPH = abs (Value1 – Value2) x 100

(Value1 + Value2)
The mean uRPHD for two populations where the data is paired gives an indication of the
average difference between sample pairs. It is a useful statistic for investigating repeatability
but is often exaggerated by precision errors between low-grade sample pairs.
For a data set consisting of paired samples (Assay 1 v Assay 2), the sRPHD value relative
to the uRPHD value provides an indication of bias within the data set.
For individual pairs of assays the ratio uRPHD/sRPHD will be plus1 if assay 1 > assay 2 and
minus1 if assay 1 < assay 2.
The average of the individual uRPHD/sRPHD ratios should be zero if there is no bias.
Positive values indicate that more assay 1 values are greater than assay 2 values.
A data set demonstrating these effects is included below.
Assay 1 (g/t) Assay 2 (g/t) AVGRADE sRPHD uRPHD uRPHD/sRPHD

AN #1 GEN #1 A#1 V G#1 A#1 V G#1 A#1 V G#1
3.8 3.45 3.625 4.83 4.83 1

3.8 3.73 3.765 0.93 0.93 1
3.9 4.3 4.1 -4.88 4.88 -1
4 3.66 3.83 4.44 4.44 1
4.1 1.8 2.95 38.98 38.98 1
4.15 4.04 4.095 1.34 1.34 1
4.25 3.71 3.98 6.78 6.78 1
5.05 4.29 4.67 8.14 8.14 1
5.45 6.06 5.755 -5.30 5.30 -1
6.25 9.2 7.725 -19.09 19.09 -1
10.3 14.76 12.53 -17.80 17.80 -1
11.7 9.89 10.795 8.38 8.38 1
12.5 7.6 10.05 24.38 24.38 1
12.7 13.18 12.94 -1.85 1.85 -1
13.8 9.67 11.735 17.60 17.60 1
15.8 12.67 14.235 10.99 10.99 1
16.5 11.22 13.86 19.05 19.05 1
16.9 19.95 18.425 -8.28 8.28 -1
17 9.58 13.29 27.92 27.92 1
35.4 24.35 29.875 18.49 18.49 1
Average 10.37 8.86 9.61 6.75 12.47 0.4

Min 3.8 3.45 3.625 -19.09 0.93 -1
Max 35.4 24.35 29.875 38.98 38.98 1
TECH12336 4
Variance
Variance is a measure of dispersion of the data that uses each value in the data set. It
provides a much better understanding of dispersion than the previously discussed
measures.
The variance for a sample containing n measurements is equal to the sum of each of the
2
squared deviations from the mean divided by (n - 1). The symbol s is used to represent the
sample variance. We calculate the variance of a series of assays (the sample variance) to
help provide us with some information about the variance of the population (all assays). If
we divide by (n - 1) the sample variance is a better estimate of the population variance. On
the other hand, if we have data constituting the whole population, then we divide by n. As
the number of samples increases the difference between the sample variance and the
population variance decreases.
Sample Variance s2 = Σ( x - N )2
(n-1)
Population Variance σ2 = Σ( x - N )
2
These two expressions can be rearranged and presented in the following form
Sample Variance s2 =
2
nΣx – (Σx)
2
n(n-1)
Population Variance σ2 = nΣx2 – (Σx)2

n2
The form of these expressions is the same as the functions VAR and VARP used in Excel
spreadsheets, where VARP is used for the population variance.
Standard Deviation:
The variance calculated using the formulae derived above is in different units from those
that were initially measured. In the case of assay data the variance is expressed in units of
2
(g/tonne) and the assays are in g/tonne. This makes interpretation of variance a little tricky.
Taking the square root of the variance easily solves this and brings the units back to g/tonne
allowing easier interpretation.
The standard deviation of a set of assay data is the square root of the variance. Recall that
2
the symbol for the sample variance is s . The symbol for standard deviation is then s.
Sample Standard Deviation s = √ s2 =

¶ [nΣx2 - (Σx)2]
[ n(n-1) ]
TECH12336 5
Population Standard Deviation σ = √ σ2 =
¶ [nΣx2 - (Σx)2
[ n2 ]
Standard Deviation (s or SD) is a measure of dispersion about the mean. It is directly

proportional to grade differences between sample pairs and is measured in the same units
as the variable under study. Very low values (close to the assay detection limit) may show
very large differences between pairs of assays (eg 0.01, 0.03g/t) and will produce large
standard deviations. However, grades in this range are not economically significant and
these large standard deviations should be ignored.
For a Normal distribution 68% of the data should occur within ±1SD of the mean, 95% of the
data should occur within ±2SD of the mean and 99.7% of the data should occur within ±3SD
of the mean. An alternative, and often more useful way of stating this is as follows.
• Data should fall outside the range ±1SD 1 in 3 times

• outside the range ±2SD 1 in 20 times, and
• outside the range ±3SD 3 in 1000 times.
With increased sample grade, acceptable differences between assay results also increase,
together with the standard deviation. Simply comparing the standard deviation of two sets of
data will not provide meaningful information about the degree of dispersion in the data. The
coefficient of variation enables comparison of sets of data with different means and
standard deviations.
Coefficient of Variation
The coefficient of variation provides a measure that allows us to compare the spread of
different sets of data, and in particular data with mean values that are quite different. The
coefficient of variation is found by dividing the sample standard deviation by the mean. The
larger the coefficient of variation the greater the relative spread of the data.
CV = Standard Deviation
Mean
Some examples of the coefficient of variation for different types of ore bodies are as follows:
Placer Deposits CV = 6; Kalgoorlie Vein systems (1m drill samples) CV = 2 to 3; Mt.
Charlotte CV = 1.2; Nickel/Copper deposits CV = 0.7 to 1.0. The Archaean gold ore bodies
of the Kalgoorlie area have CVs in the range 2 to 4.
Coefficient of Relative Variation
Sometimes the coefficient of relative variation (CRV) is calculated.
This is the coefficient of Variation expressed as a percentage.
CRV = CV x 100 = Standard Deviation x 100

Mean
This parameter is also known as Pearson’s Coefficient of Variation or the relative

standard deviation.
TECH12336 6
Correlation analysis
Correlation analysis allows determination of the degree or strength of a relationship between

two sets of data. Pearson’s correlation coefficient is used for measuring the relationship
between quantitative data. The correlation coefficient has a range of values from –1 to 1. A
perfect positive relationship would have a correlation coefficient of 1 and a perfect negative
relationship would have a correlation coefficient of –1. If there is no correlation between the
data set the correlation coefficient will be zero. The correlation coefficient is strongly
influenced by outliers and thus is not a robust statistic and should be used with care.
Before calculating Pearson’s correlation coefficient the data should be plotted as a scatter
plot (Data Set 1 v Data Set 2) to check for outliers and clustering.
Check or repeat assay data are expected to show strong positive correlations with the
original assay data (ideally a correlation coefficient of +1), and simple arithmetic scale plots
are usually used. If the data range is very large, a log transformation of both sets of data
can be applied before the scatter plots are produced or, better still, the distribution can be
subdivided into a number of scatter plots of various grade ranges.
Pearson’s Correlation Coefficient:
Pearson’s Correlation Coefficient r is defined as
r = 1/(n-1)
å [x – mean x] [y – mean y]
[ sx ] [ sy ]
where sx and sy are the standard deviation of x and y respectively.
The term 1/(n-1) å(x – mean x ) (y – mean y ) is called the covariance of x and y
and sx is √[variance (x)].
The correlation coefficient can be expressed as
covariance of x and y
√[Var (x)] [Var (y)]
Note that this expression has been developed using the notation that s is the Sample
Standard Deviation. If the Population Standard Deviation σ is known the component 1/(n-1)
should be replaced be the component 1/n. See the discussion on Variance and Standard
Deviation above for more details.
The function used in Excel is CORREL and is based on a known population standard
deviation.
At this point it is worth quoting Moroney (1953, p303).
“At no point are statistical methods more of a sausage machine than in correlation analysis.
The problem of interpretation is always very much more difficult to deal with than the
statistical manipulations, and for this side of the work there is no substitute for detailed
practical acquaintance with every aspect of the problem. The statistician can only help out
TECH12336 7
the specialist in the field not replace him. The man (person) who plays carelessly with sharp
tools is asking to be cut.”
Regression analysis
Regression analysis of two sets of assays of the same samples should produce a
regression line passing through the origin (0,0) with a slope of +1. Poor precision of the data
will not be identified in this regression equation, but will be shown in a scatter plot. Bias in
the data will result in a regression equation that departs from the ideal stated above. There
is little point in calculating and plotting regression expressions for data that includes some
bias. It is more meaningful to show on the scatter plot the line of slope +1 passing through
the origin (ratio of assay1:assay 2 is 1:1), which will give indications of the degree of scatter
in the data.
TECH12336 8
Appendix 2 – Proforma QA/QC Report
TECH12336
ANGLOGOLD AUSTRALIA LTD
……………………..Project
Check Assay Results

2000-2001 Exploration Drilling Program
Author Report No: ……………
Date Copy … of …
Keywords: …………………………………..
Location: ……………………………………
Map: 1:250,000 ………………………………..

1:100,000 ………………………………..
Distribution
1. AngloGold - …………..
2. AngloGold - ………….
3. AngloGold - …………
TECH12336
EXECUTIVE SUMMARY
EXAMPLE TABLE OF CONTENTS

1. INTRODUCTION
1.1. REPORT PURPOSE
1.2. PROJECT LOCATION AND GEOLOGY
1.3. STATISTICS APPLIED
2. SAMPLING AND ASSAY PROCEDURE
2.1. SCOPE OF RESAMPLING AND ASSAY PROCEDURE
2.1.1. Screen Fire Re-assays
2.1.2. Pulp and Residue Re-assays
2.2. DATA ANALYSIS
2.3. QUALITY CONTROL
3. ANGLOGOLD SUBMITTED STANDARD AND BLANKS
3.2. BLANKS AND STANDARDS
4. ANALABS INTERNAL QUALITY CONTROL

4.1. REPEATS AND SECOND SPLITS
4.1.1. First Repeats.
4.1.2. Second Splits.
5. CHECK ASSAY RESULTS
5.1. OUTLIERS IN THE DATA
5.2. SCREEN FIRE RE-ASSAYS
5.2.1. Interpretation
5.3. LABORATORY COMPARISONS
5.3.1. Analabs Original v Genalysis Pulp
5.3.2. Genalysis Pulp v Analabs Pulp
5.3.3. Analabs Original v Genalysis Residue
5.3.4. Genalysis Residue v Analabs Residues
5.4. COMPARISON OF DIFFERENT SAMPLE SETS SUBMITTED TO ANALABS
5.4.1. Analabs Original v Analabs Pulp
5.4.2. Analabs Original v Analabs Residue
5.4.3. Analabs Original Vs Analabs Screen Fire (restricted dataset)
5.5. COMPARISON BETWEEN SPLITS
5.5.1. Analabs Pulp v Analabs Residue
5.5.2. Genalysis Pulp v Genalysis Residue
6. CONCLUSIONS & RECOMMENDATIONS
6.1. CONCLUSIONS
6.2. RECOMMENDATIONS
7. REFERENCES
TECH12336 11
List of Figures
List of Tables
List of Appendices
TECH12336 12
INTRODUCTION
This report discusses QA and QC of assays completed during the [insert quarter or
year or other period] ended [insert date] for [insert grade control or resource
evaluation or exploration] at [insert minesite or project name].
The report should provide the following information and any other relevant data,
commentary or discussion.
SAMPLING AND ASSAY PROCEDURE
• Type and number of original samples submitted during the period,

• Name and location of the original laboratory,
• Name and location of all check laboratories,
• Analytical techniques used,
• Sampling & Assaying flowsheet,
• Type and number of check assays carried out.
QUALITY CONTROL
Results of each laboratory’s quality control using AngloGold submitted Standards and
Blanks,
Results of each laboratory’s internal quality control programs,
Tabulation of all data in standardised tables.
Standardised plots of Blanks and Standards against date,
Standardised plots of sRPHD against date.
CHECK ASSAY RESULTS
Laboratory Comparisons
Compare Lab 1 Original, AuR and Aus with Lab. 1 Check pulp and residue and Lab. 2
Check pulp and residue.
Compare Check Lab 1 pulp and residue with Check Lab 2 pulp and residue.
CONCLUSIONS AND RECOMMENDATIONS
Draw appropriate conclusions
Make recommendations
Flag any follow action and set timetable. Refer to these action points in future reports.
TECH12336 13
Appendix 3 - Example QA/QC Report
(CD attached)
TECH12336
Appendix 4 – Suppliers of Certified Standards
TECH12336
GANNET STANDARDS
Adrian Knowles Tel: 08 9367 1164 Fax: 08 9368 1636
Gannet Holdings Pty Ltd PO Box 329 South Perth WA 6951
43 Fredric St. Naval Base WA 6165 Tel : 08 9410 2356 Fax: 08 9437 1212
e-mail [email protected]
STANDARDS CURRENTLY AVAILABLE 8th March 2002
Designation Material Recommended Ore
Value (g/t Au) Type
ST01/7212 Very high grade ore 1.49 Oxide
Drill Cuttings
Feldspar/Quartz dilution
ST02/0286 Mine Ore 2.36 Free Milling Oxide
ST04/9210 VHG DC FQ Dil. 5.16 Oxide mine ore
ST05/9280 VHG DC FQ Dil. 2.52 Free Milling Oxide
ST06/0250 VHG DC FQ Dil. 1.05 Oxide
ST17/8171 Mine Ore-Diluted 0.76 Oxide
ST18/8239 V.H.G.O.QFD. 9.7 Oxide
ST28/9240 V.H.G.O.QFD. 35.6 Oxide
ST37/ 8229 Mine Ore-Diluted 1.73 Oxide
ST39/6167 Mine Ore-Diluted 0.87 Oxide / Transition
ST42/9272 Mine Ore 1.33 Sulphide
ST48/9278 Transition Ore 4.55 Transition
ST70/5156 Mine Ore-Diluted 0.106 Oxide
ST73/1292 Mine Ore 1.48 Minor Sulphide
ST92 Mine Ore-Diluted 8.18 Oxide
ST136/8243 Mine Ore 49.9 Oxide
ST147 Mine Ore-Diluted 2.66 Sulphide Minor Cu Present
ST148 Mine Ore 8.67 Sulphide Minor Cu Present
ST267 Mine Ore 13.2 Minor Sulphide Ni and Cu
ST274 Mine Ore 6.27 Low Sulphide
ST 44/0294 Mine Ore 13.9 Low Sulphide Dolerite
Price 8 X 4 Soil bags ( 120 to 150 gm.) $4.50 ea. lots of 50
Ex Factory 1Kg. Multi-Layered sealed bags $18.00/Kg. in 25 Kg. lots
1 Kg. Single $ 25.00 / Kg.
Plus GST if applicable
BLEG SAMPLES FOR SOIL & STREAM SURVEYS
Designation Au ppb. Coarse stream sand
BLG 5B 0.02 2 Kg. bags
BLG 6B 0.5
BLG 7B 5.6
BLG 8B 9.4
BLG 9B 12.4
BLG 11 0.99
BLG 12 0.53
BLK 0.03
BLG 10 B 28.3
Price $ 16 / 2 Kg. Bag + GST if applicable Ex Factory
Designation Au ppm.
ST122 0.062 Fresh Diorite
ST138 0.022 Approx. 40 % Fe Ox.
ST140 0.023 Oxide Surface
ST154 0.003 Laterite
ST155 0.084 Oxide Waste
ST195 0.029 Pulverised oxide
Price 119 gm 8 X 4 soil bags $ 4.50 ea. in lots of 50
1Kg. Plastic bags $18 / bag in 25 Kg. lots
0 Kg mixed lots $ 25 /Kg ea.
Plus GST if applicable Ex Factory
TECH12336 2
ROCKLABS
Rocklabs Ltd., Manager: Ian Devereux; P.O. Box 18-142, Auckland 6, New Zealand, Phone
(+64 9) 634 7696, Fax: (+64 9) 634 6896. [email protected]. Web site:
www.rocklabs.com. Local Agent: Sietronics WA, Fred Hoetmer tel: 08 9354 4581 Web site:
www.sietronics.com.au.
Currently available standards are maintained on the websites.

AngloGold has negotiated a worldwide price discount for Rocklabs Standards.
TECH12336 3
GEOSTATS
Geostats Pty Ltd., Peter Hayes; 4 Stack Street, Fremantle, WA; Australia, 6160, Phone (+61
8) 9430 9696, Fax: (+61 8) 9430 9695). 68, Watkins Street, White Gum Valley WA 6162
email: [email protected]; website: www.geostats.com.au
Example Standard Listing (check web site):
GEOSTATS PTY LTD

Sample and Assay Monitoring Services
50g Fire Assay Statistics Aqua Regia Statistics

Product Code Au ppm Std.Dev. Count Confidence Int. p DESCRIPTION OF SOURCE / MATRIX
G01 0.02 0.02 56 0.005 Gossan / Ochre in the Murchison gold field WA
G301-6C 0.03 0.01 155 0.002 Fresh basic milled waste ex Eastern Goldfields
G999-10 0.05 0.02 90 0.010 Low Grade Oxide, Eastern Goldfields
G300-2 0.06 0.02 89 0.003 Oxide ex Eastern Goildfields
G396-1 0.09 0.02 66 0.005 Eastern Goldfields blended Laterite and Oxide.
G396-2 0.12 0.03 65 0.007 Eastern Goldfields blended Oxides.
G399-1 0.22 0.02 96 0.004 Eastern Goldfields oxide waste.
G997-7 0.31 0.04 84 0.009 Laterite/Kaolin waste.
G398-9 0.33 0.03 84 0.006 Diluted Waste. Oxide with quartz and feldspar added.
G998-9 0.38 0.03 95 0.006 Eastern Goldfields basalt with minor Sulphide.
G997-1 0.41 0.04 82 0.009 Fresh basaltic Ore in the South West Mineral Field.
G997-2 0.41 0.04 81 0.009 Oxide ore in the Eastern Goldfields.
G07 0.43 0.04 60 0.010 Western Pilbara Oxide ore.
G397-10 0.49 0.05 64 0.012 Blended Oxides ex Eastern Goldfields.
G398-2 0.50 0.04 84 0.009 Cu/Au Ore,South west mineral field,carbon present,Aqua Regia requires pre-roast
G300-4 0.50 0.03 92 0.006 Fresh Ore ex Kalgoorlie region
G996-4 0.51 0.04 70 0.009 Low grade laterite waste, South West mineral field.
G999-2 0.63 0.06 93 0.010 Oxide Gold Ore ex N.E. Goldfields
G398-4 0.66 0.05 85 0.011 Cu/Au ore ex South West. Mineralised sulphidic waste. Basaltic.
G397-1 0.76 0.05 64 0.012 Oxides in the Eastern Goldfields
G900-1 0.78 0.03 81 0.006 Trace Sulphide Ore ex Eastern Goldfields
G998-6 0.80 0.06 98 0.012 Oxide ore with Kaolin and minor Iron.
G998-3 0.81 0.05 93 0.010 Basalt ore ex Eastern Goldfields, Minor Sulphide.
G999-1 0.82 0.06 90 0.010 Cu/Au Ore ex Pilbara Region
G301-1 0.85 0.05 91 0.010 Gold Ore ex Eastern Goldfields
G996-2 0.87 0.07 70 0.016 Blended low grade, low sulphur ores.
G399-5 0.87 0.05 94 0.010 Fresh rock. South West Mineral Field
G999-3 0.95 0.07 95 0.010 Low Grade Mine Tail.Ore contains carbon, Aqua Regia requires pre-roast
G300-7 1.00 0.04 88 0.008 Oxide / Cu / Au ore ex Murchison
G399-3 1.05 0.07 101 0.014 Oxide in the Eastern Goldfields
G300-8 1.07 0.06 95 0.012 Transition Ore ex Eastern Goldfields
G900-3 1.18 0.06 79 0.014 Oxide Mine ore. Diluted
G301-8 1.19 0.09 87 0.019 nPyrite Concentrate
G02 1.20 0.08 59 0.020 SWAustralia Gold field oxide ore
G300-3 1.26 0.06 92 0.012 Fresh Basic Ore ex Eastern Goldfields
G399-8 1.33 0.08 98 0.016 Laterite with Kaolin profile. South west mineral field
G997-3 1.41 0.08 82 0.017 Oxide ex Eastern Goldfields.
G399-2 1.46 0.09 100 0.018 Oxides in the Kalgoorlie Region
G301-4C 1.46 0.08 189 0.011 Gold Ore ex Eastern Goldfields
G900-2 1.48 0.06 81 0.013 Trace Sulphide Ore ex Eastern Goldfields
G300-9 1.53 0.06 94 0.012 Ore ex Eastern Goldfields ( Fresh Rock)
G997-6 1.68 0.08 80 0.018 Diorite ore ex Indonesia.
G397-3 1.73 0.12 68 0.029 Laterite / Kaolin ores.
G398-3 1.73 0.11 87 0.023 Oxide gold ore ex Eastern Goldfield. Quartz and feldspar added
G301-3 1.96 0.08 90 0.017 Gold Ore ex Eastern Goldfields
G999-9 1.98 0.11 89 0.020 Supergene Ore ex Kalgoorlie Region
G996-5 1.99 0.13 72 0.030 Blended oxide ore.
G300-10 1.99 0.07 90 0.014 Composite oxide Ore with Minor sulphide.Trace carbon,Aqua Regia requires pre-roast
G998-7 2.06 0.10 94 0.020 Sulphide ore with minor Copper ex Pilbara region.
G300-5 2.34 0.11 95 0.022 Transition Ore ex Eastern Goldfields
G399-7 2.37 0.14 100 0.027 Sulphide in the Pilbara Region. Some Copper.
G900-8 2.45 0.12 81 0.025 Mill Tail in the South West Mineral field
G399-6 2.52 0.12 97 0.024 Fresh rock. South West Mineral Field
G999-7 2.52 0.12 89 0.030 Fresh Basic Ore ex Kalgoorlie Region
G900-6 2.56 0.09 76 0.020 Supergene Ore in various locations
G995-3C 2.66 0.14 132 0.024 Sulphide Mine Ore in the Pilbara region.
G398-7 2.71 0.14 86 0.030 Minor sulphide/Cu ex Pilbara region.
G995-1 2.74 0.18 66 0.043 Pilbara Sulphides.
G998-1 2.95 0.12 91 0.025 Oxide with high iron content.
G398-6 2.95 0.17 87 0.036 Basalt ore ex Eastern Goldfields. Minor Sulphide
G999-4 3.02 0.16 89 0.030 Oxide Ore ex Eastern Goldfields
TECH12336 5
Appendix 5 – Example Exploration Assay Contract
(CD Attached)
TECH12336 6

AQC Guideline

Uploaded by

Copyright:

Available Formats

AQC Guideline

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

AQC Guideline

Uploaded by

Copyright:

Available Formats

ANGLOGOLD AUSTRALIA LIMITED

QUALITY CONTROL &

AUTHORS: K. Kenny and N. J. Crase

DATE: September 2002

REPORT NUMBER: TECH12336

3. THE ASSAY AND QA/QC PROCESS 4

Appendix 1 Statistical Definitions

FIGURE 4.1: ANALYTICAL FLOWSHEET 12

FIGURE 7.1: CHECK ASSAY FLOWCHART AND POTENTIAL CHECK ASSAY

FIGURE 8.1: BLANK ASSAYS FEBRUARY – NOVEMBER 2001 30

FIGURE 8.2: CERTIFIED STANDARD STO6/8222 - DECEMBER 2000 TO JULY 2001 31

FIGURE 8.3: CERTIFIED STANDARDS – SRPD BOX AND WHISKER PLOT 32

FIGURE 8.4: SCATTER PLOTS – ANALABS ORIGINAL V ANALABS REPEAT 35

FIGURE 8.5: QUANTILE - QUANTILE PLOT – ANALABS ORIGINAL V ANALABS REPEAT 36

FIGURE 8.6: SRPHD PLOT – ANALABS ORIGINAL V ANALABS REPEAT 36

FIGURE 8.7: SCATTER PLOTS – ANALABS ORIGINAL V SELECTIVELY BIASED ANALABS

FIGURE 8.8: QUANTILE - QUANTILE PLOT – ANALABS ORIGINAL V SELECTIVELY BIASED

FIGURE 8.9: SRPHD PLOT – ANALABS ORIGINAL V SELECTIVELY BIASED ANALABS

FIGURE 8.10: ANALABS SCREEN TESTS - 2001 40

FIGURE 8.11: GENALYSIS SCREEN TESTS - 2001 40

TABLE 6.1: RECOMMENDED MINIMUM FREQUENCY FOR SUBMISSION OF STANDARDS,

TABLE 8.1: ANGLOGOLD STANDARDS SUBMITTED WITH ORIGINAL ASSAYS 29

TABLE 8.2: STATISTICAL SUMMARY: ANALABS ORIGINAL (AU) V REPEAT (AUR) 33

TABLE 8.3: SCREEN TESTWORK: ORIGINAL AND CHECK LABORATORY RESULTS 39

Effective assay QA/QC should be viewed as a management tool to build a partnering

Factor Potential Error Range

A trait or characteristic used to measure the degree of excellence of a product or service.

Quality Assurance (QA)

QA is the process of evaluating overall project performance on a regular basis to provide

Quality Assurance Plan

Quality Control (QC)

3.1 Scopes of Work and Contract Laboratory Partnerships

Choice of assay lab is generally decided using a combination of 3 inter-related variables

3.2 Field and Laboratory Procedures

Field Procedures include the following:

1. Application of geological and geostatistical knowledge and interpretation to determine

QA/QC programs can be thought of as two-stage processes:

3.3 Quality Control and Quality Assurance Philosophy

4.1 Sample Terminology

Historically, due to equipment limitations, a sub-sample of approximately 500 grams was

Certified standards may be obtained from:

Further details are attached in Appendix 4.

4.2 Assay Methods

Stage 2. Cupellation: The sample is then cupelled at ~1000°C in a magnesia cupel in an

4.3 Assay Terminology

4.4 Statistical Definitions

4.5 Analytical Flowsheet

Figure 4.1 Analytical Flowsheet (overleaf)

5.1 Laboratory Sample Preparation

Cross-contamination of samples occurs when housekeeping practices are shoddy.

External QA/QC tests (controlled by AngloGold) include the submission of Standards,

Table 6.1: Recommended minimum frequency for submission of

Grade Control Resource Advanced

Location of Duplicates mineralised mineralised mineralised

Pre-prepared standards from Gannet, Geostat or OreSearch are to be inserted in each

Longer-term evaluation should also be completed to provide information on the overall

Laboratory Internal Quality Control

Internal standards are used by Analabs as follows:

“Reference materials (Standards) incorporated into batches must, to enable

Relative Half Difference = Difference between the values

sRPHD = (Value1 – Value2) x 100

= (Value1 – Value2) x 100