Gold Deposits: Establishing Sampling Protocols and Monitoring Quality Control

"Abstract - Large differences in gold content between rock fragments or small volumes of unbroken rock, commonly referred to as the nugget effect, are related to sample and subsample weights, crushing and pulverizing sizes, and density, shape and size distribution of gold grains. Gy’s sampling theory, which relates gold grain characteristics to sample weight, fragment size, and error, provides a method of estimating a sampling constant that quantifies deposit characteristics that contribute to the nugget effect.Sampling control tests provide methods for determining the sampling constant. Sampling control charts, comprising nomographs that relate sample weight, error and particle size, use the sampling constant to optimize sample weights and comminution sizes, and to minimize the variability introduced by incorrect or inadequate sampling and sub-sampling procedures. The sampling constant is useful in classifying ores into arbitrary categories according to expected sampling difficulties. Spatial variations in the sampling constant can be used to aid in defining ore types for modeling geological domains, which should have unique sampling protocols. Examples that illustrate the relationship between the sampling constant and expected sampling problems, and the spatial variation of the sampling constant, are given for various deposits. Case histories of sampling programs are presented.Quality control procedures consist of monitoring sample contamination amd accuracy and precision of analytical results. Accuracy, precision and contamination are monitored by including standards, duplicates and blanks into batches of samples being analyzed. Analytical results of standards and blanks are plotted on a batch-and-time basis with control lines at plus or minus two/three standard deviations as confidence limits of the accepted round-robin value. Examples are given that illustrate a laboratory system in statistical control and the various types of problems encountered when interpreting quality control data. These include sample transcription errors, degradation, contamination, sampling errors, detection limit problems, instrumental drift, lack of statistical control, laboratory bias, procedural problems and data tampering.Where bias is negligible, precision, as a function of concentration, can be obtained from duplicates using the Thompson-Howarth precision algorithm, which also allows calculation of the practical detection limit. Examples are given that illustrate variations in precision, with grain size, for different types of duplicates, as well as mineralogical changes and sampling errors."