Complex gold deposits are generally characterised by variable geometries, strong structural controls on grade distribution and a high-nugget effect. The use of diamond drilling and fire assays in this environment often results in an understatement of grade. Bulk samples are likely to be the closest estimators of true grade, and may be required to evaluate geological/grade risk during resource estimation/feasibility studies. Any bulk sampling program requires appropriate planning and implementation. The planning stage should attempt to delimit the extent and nature of mineralisation; characterise mineralogy and metallurgical properties of the ore; and define bulk sample size and how it will be sampled/processed. The approach to sample processing will be governed by the results of metallurgical testing and geological knowledge. Careful design of sampling protocols must be undertaken through material characterisation, understanding of gold particle sizing and the application of Gy's Sampling Theory.
This paper looks at two ways to estimate the bulk mineralogy of the rocks for assay intervals. The aim is to find an efficient indicator of the most common minerals in the rock. Phase (modal) analysis has traditionally been done using visual methods such as point counting and image analysis. A modern version of this process is the X-ray point counting routine using the SEM-EDS based software. These methods are too slow and expensive for routine analysis of bulk sample mineralogy at the normal assay spacing. Two sources of data were considered that provide information that can be used to determine the mineral abundance in assay samples. The most widely applied method is (semi-) quantitative X-ray diffraction (QXRD). The QXRD method is most applicable to major minerals and has limited application to minerals at low abundance. The nominal detection limit is 0.5 per cent. Values below five per cent have large errors. A second, less common, method is calculation of mineralogy from chemical assay data. Conversion of chemical analyses to mineralogical analyses depends on the unique chemical composition of each mineral. Elements only found in one mineral are easily accounted for, but many compositions are ambiguous. Deciding on the actual mineralogy is not simple. Recalculation of mineral mode from chemical analyses is more accurate than QXRD when the correct minerals, and mineral compositions, are known. Where only a few QXRD analyses are available they can be used to setup a protocol for calculation of mineralogy from assay data. Linear programming works well in this environment. The best results are obtained when both H 2O and CO 2 are directly measured. Loss-on-ignition (LOI) should be included if these are not available. Where both QXRD and chemical analysis are available for all samples, the best results are obtained using the least squared method to merge the data sets assuming QXRD has much higher analytical errors than chemical assays. The combined method provides more robust results because the high abundance minerals are controlled by the QXRD measurements while the chemical assays improve the precision for low abundance minerals.