Page 232 - Introduction to Mineral Exploration

10: EVALUATION TECHNIQUES 215

rapidly under gravity and homogenization of which are attributed to analytical laboratories
such material is difficult (Austr. Inst. Geosci- arise from the selection of the assay portion
entists 1987, 1988, Royle 1995). and not, on the whole, from actual analytical
techniques (section 10.1.2). Errors occur, how-
ever, and vary from laboratory to laboratory.
Calculation of the variance of the An example is given in Table 10.5 where an
fundamental error
artificial sample was made from gold in silica
−1
Under such circumstances, and with the sand, at 2.74 g t and was sent to four different
very low concentrations of precious metals laboratories for analysis (Springett 1983b). The
(≈1 ppm), Gy (1992) suggests a modification of following conclusions can be drawn:
his formula for calculating the fundamental Laboratory 1 Results are both inaccurate and
error. imprecise.
Normally, the factor C = cβfg (Box 10.1) but Laboratory 2 Results are accurate but
for gold and platinum Gy (1992) recommends a imprecise.
factor of: Laboratory 3 Results are inaccurate but
precise.
. 24 d 3 C Laboratory 4 Results are accurate and
=
2
C and S = precise.
a L M
Obviously only the results from Laboratory 4
where a L is the concentration of the valu- are acceptable.
able constituent expressed as a fraction (i.e. The scope of the control of error in chemical
−6
1 ppm = 1 × 10 ). analysis is beyond this chapter but these errors
An example of the use of this formula is must be considered (see section 8.2). External
given in Box 10.2. control is exercised by submitting duplicate
samples and reference materials of similar
matrices to the unknowns (see section 8.2.1). It
10.1.6 Analytical errors
is essential that the geologist is aware of the
According to Gy (1992) analytical errors (AE), internal control procedures of the laboratory.
in the strict sense, are negligible and errors These usually employ the analysis of replicate

BOX 10.2 Variance of fundamental error [S (FE)] with liberated gold mineralization.
2
Gold mineralization contains 1 g t (a L = 10 ) and is liberated from its gangue (i.e. β = 1) with the
−6
−1
coarsest particles about 2 mm in size (d = 0.2 cm.). Under such circumstances Gy (1992) recommends a
modification of his standard formula for the calculation of the fundamental variance. Normally the
3
heterogeneity factor C equals cβfg (Box 10.1), but for free gold he recommends C = 2.4d /A L and
S (FE) = C/M.
2
3
−2
2
−1
With a 1 tonne sample (M = 10 g), C = 2.4(0.2) /10 − 6 = 19,200 and S (FE) = 1.92 × 10 , S = 1.4 × 10 , or
6
14%. If a normal distribution is assumed the 95% confidence level double-sided interval is ±28%, which
is a low precision. If, say, a 10% precision is required at 95% confidence then a larger sample (M) is
needed. At this level S(FE) equals 5 × 10 , and S (FE) equals 25 × 10 ; then
−4
−2
2
4
2
M = C/S (FE) = 19,200/25 × 10 −4 = 768 × 10 g or 7.68 tonnes.
The treatment of such large samples presents a problem, but Gy (1992) recommends the use of stratified
sampling. In this process the sample is reduced to a grain size of a few millimeters and then concentrated
on a shaking table, such as the Wilfley table (Wills 1985). This produces a high density concentrate that
contains all the coarse gold, an intermediate fraction, and low density tailings which is largely gangue.
The intermediate fraction is recycled on the table while the tailings are sampled again on a 1% sampling
ratio. In this way the 8 tonne sample is reduced to about 80 kg which is then processed as in Figs 10.6
and 10.7.