Tiegel home Talk to a metallurgist

Assay quality control: standards, blanks and duplicates, and interpretation of control charts

Monitoring assay quality with standards, blanks and duplicates: control charts, bias, z-scores, HARD and pooled precision, with examples.

· 5 minute read · By

An assay laboratory is a measuring instrument and, like any instrument, it can drift, suffer contamination or lack precision. These faults cannot be detected from the sample results alone, so control samples of known composition are inserted among them.

A standard, or certified reference material, has a known grade; if the laboratory reports it incorrectly, the samples in that batch are suspect. A blank contains essentially no gold, so gold reported in it indicates contamination. A duplicate is a second portion of the same sample, and a large difference between the two results indicates poor precision. Checking these controls in every batch is quality assurance and quality control (QAQC).

Plant metal accounting, grade control and resource estimates all depend on assays. This article describes the three types of control sample, the charts and calculations used to monitor them, and the response when a control fails.

Control sample types

ControlPurposeApplication
Certified reference material (CRM, standard)Accuracy: is the laboratory biased?Inserted blind; result compared with the certified value and its standard deviation
BlankContamination, during preparation or analysisBarren material inserted after high-grade samples; result should be near the detection limit
DuplicatePrecision at each stage: field (sampling), coarse reject (preparation), pulp (analysis)Second split assayed; the difference between pairs measures the error at that stage

CRMs should be obtained from certified suppliers, with grades in the range of the samples and a matrix similar to the ore. Controls should be inserted blind, with sample numbers that do not identify them. Plant samples require the same discipline as exploration samples: shift composites used in metal accounting warrant as much quality control as drill core.

Standards: control charts, bias and z-scores

bias=x‾−CVCV×100z=x−CVSD

where

x‾
mean of the laboratory results for a certified reference material (CRM)
CV, SD
certified value and certified standard deviation of the CRM
z
how many standard deviations one result is from the certified value

Each CRM result is plotted in batch order, with lines at the certified value and at ±2 and ±3 certified standard deviations. Common rules are that one result outside ±3 SD, or two consecutive results outside ±2 SD on the same side, fails the batch, which is then re-assayed before its results are used. The rules should be set out in a written QAQC procedure.

Certified reference material control chartThirty batch results scatter around the certified 2.48 g/t within the warning limits, except batch 18 at 2.75 g/t, above the upper 3 SD limit.2.22.32.42.52.62.72.8151015202530Assay batchGold (g/t)certified 2.48+2 SD−2 SD+3 SD−3 SDCRM result: 1 → 2.5CRM result: 2 → 2.6CRM result: 3 → 2.5CRM result: 4 → 2.5CRM result: 5 → 2.5CRM result: 6 → 2.5CRM result: 7 → 2.5CRM result: 8 → 2.5CRM result: 9 → 2.5CRM result: 10 → 2.4CRM result: 11 → 2.6CRM result: 12 → 2.5CRM result: 13 → 2.5CRM result: 14 → 2.5CRM result: 15 → 2.5CRM result: 16 → 2.5CRM result: 17 → 2.5CRM result: 18 → 2.8CRM result: 19 → 2.5CRM result: 20 → 2.5CRM result: 21 → 2.4CRM result: 22 → 2.4CRM result: 23 → 2.5CRM result: 24 → 2.4CRM result: 25 → 2.4CRM result: 26 → 2.4CRM result: 27 → 2.5CRM result: 28 → 2.4CRM result: 29 → 2.4CRM result: 30 → 2.5CRM
A Shewhart chart for one CRM (certified value 2.48 g/t, SD 0.08). Batch 18 fails the 3 SD rule: the batch is re-assayed before its results are used. Illustrative figures.

Worked example: a CRM is certified at 2.48 g/t gold with a standard deviation of 0.08 g/t. Ten results average 2.492 g/t, a bias of +0.5%, which is acceptable. One result of 2.62 g/t has a z-score of +1.75, within limits. In the chart, batch 18 reports 2.75 g/t (z = +3.4) and fails.

Trends are as important as failures. A run of results drifting above the certified value indicates a developing bias, often from a calibration or method change, before any single result fails.

Blanks

Blanks are inserted after samples expected to be high grade. A blank reporting gold well above the detection limit, at a threshold set in the procedure (often a small multiple of the detection limit), indicates carry-over in crushers and pulverisers or contamination in the fire assay, and the samples processed immediately before it should be investigated.

Duplicates and precision

HARD=|a−b|a+b×100RSDpooled=1n∑(ai−bi)22mi2×100

where

a, b
original and duplicate assay of one sample
HARD
half absolute relative difference (%)
mi
mean of pair i
n
number of pairs
Pair12345678
Original (g/t)1.200.852.400.425.101.750.603.30
Duplicate (g/t)1.320.792.150.475.351.700.723.05
HARD (%)4.83.75.55.62.41.49.13.9

For pair 1, |1.20 − 1.32| ÷ (1.20 + 1.32) = 4.8%. The pooled relative standard deviation for the eight pairs is 7.1%.

Original against duplicate assayEight duplicate pairs lie close to the one-to-one line; the largest relative difference is at 0.60 and 0.72 g/t.02460123456Original (g/t)Duplicate (g/t)Duplicate pairsPerfect agreementDuplicate pairs: 1 → 1Duplicate pairs: 1 → 1Duplicate pairs: 2 → 2Duplicate pairs: 0 → 0Duplicate pairs: 5 → 5Duplicate pairs: 2 → 2Duplicate pairs: 1 → 1Duplicate pairs: 3 → 31:1 line
Pulp duplicates plotted against the originals. Scatter about the 1:1 line shows precision; points consistently on one side would show bias. Illustrative figures.

Precision targets depend on the stage and the material. Pulp duplicates should agree much more closely than field duplicates, and coarse-gold ores are far less precise than base metal ores. Comparing field, coarse reject and pulp duplicate results shows which stage contributes the most error, and that stage is then improved, often with the aid of Gy's formula. Precision usually deteriorates near the detection limit, so it should be reported by grade range.

The QAQC programme

  1. Write a QAQC procedure covering control types, insertion rates, pass and fail rules, and actions.
  2. Review every batch on receipt, before results are loaded into databases or metal accounting.
  3. Record each failure, the re-assay and the outcome.
  4. Report monthly on CRM bias and failures, blank failures, and duplicate precision by stage.
  5. Send umpire samples (splits of pulps) to a second laboratory periodically to check for bias between laboratories.

Glossary

QAQC
Quality assurance and quality control: the procedures and checks that make assay data reliable.
Certified reference material
A sample with a certified grade and uncertainty, used to check accuracy.
Blank
Barren material used to detect contamination.
Duplicate
A second split of a sample assayed to measure precision.
Bias
A consistent difference between measured and true values.
Precision
How closely repeated measurements of the same sample agree.
HARD
Half absolute relative difference between a sample and its duplicate.
Umpire laboratory
An independent laboratory used to check the main laboratory.

Frequently asked questions

What is the difference between accuracy and precision?

Accuracy is closeness to the true value, checked with standards. Precision is the closeness of repeated measurements to each other, checked with duplicates. A laboratory can be precise but biased.

What action is required when a standard fails?

The batch should not be used. The laboratory should re-assay it, or the part around the failure, and the outcome should be recorded. Repeated failures should be investigated.

How is HARD calculated?

The absolute difference between the original and duplicate results, divided by their sum, multiplied by 100.

Why are blanks inserted after high-grade samples?

Contamination is most likely to appear in the sample processed immediately after high-grade material, in the crusher, pulveriser or fire assay.

Discuss your plant with a metallurgist

Tiegel is a Perth metallurgy company. We help with testwork, plant optimisation, metal accounting and investigations, and we build the Tiegel Suite software.

Talk to a metallurgist

References and further reading

  1. Abzalov, M. (2008). Quality control of assay data: a review of procedures for measuring and monitoring precision and accuracy. Exploration and Mining Geology, 17.
  2. Joint Ore Reserves Committee (2012). Australasian Code for Reporting of Exploration Results, Mineral Resources and Ore Reserves (the JORC Code). Table 1, Sampling techniques and data. Link
  3. Applying the AMIRA P754 Code of Practice for Metal Accounting. AusIMM MetPlant 2008.

Worked examples and charts marked illustrative use invented figures, not data from any operation.