Page 222 - Introduction to Mineral Exploration

10: EVALUATION TECHNIQUES 205

from hammers, and that derived from the of this on the concept of their random distribu-
crushing and grinding equipment. tion is a matter of debate.
5 Unintentional sampling errors must be
avoided such as mislabeling and the mixing of Sample volume–variance relationship
fractions of different samples. Preferably samples should be of equal volume,
In essence preparation errors result from based on similar cross-sectional area. If vol-
the careless treatment of samples and can be umes are unequal there must be no correla-
caused by the use of inadequate crushing and tion between this and their analytical results.
grinding equipment, insufficient drying cap- Although the mean of observations based on
acity, and inappropriate sample splitting. large volumes of rock should not be expected
Obviously, PE values can be estimated to be different from those based on small
only from the results of an operating sample volumes (e.g. larger or smaller diameter drill
reduction system and, like SE values, cannot be core), the corresponding variance of samples
estimated directly for either preliminary cal- from the larger volumes might be expected to
culations or the design of such a system. For- be smaller. This is because the variability of
tunately, the totality of both can be estimated the sample depends both on the variance and
from the relationship: the number of samples taken (n):

2
2
S (TE) ≤ 2S (FE) S 2
∝
variability
n
where the variance of the total sampling error
(TE) does not exceed twice that of the funda- and instinctively we feel that there should be a
mental error (FE) (Gy 1992).
similar relationship with the volume of the
sample taken. Indeed, it has been advanced
Summary that:
Sampling consists of three stages: 2
1 Extraction of the original sample from in situ variability S
∝
material (i.e. the population), including its volume
delimitation and collection.
2 Preparation which involves a reduction in It appears in practice, however, that there is no
both mass and grain size of the original sample such simple, direct, relationship. This can be
to an assay portion for chemical analysis. tested by splitting samples in half (e.g. mineral-
3 Chemical analysis of the assay portion. ized parts of a drill core), analyzing them separ-
Each of the above three stages is discussed in ately, and then combining the results as if
turn. the whole core sample had been tested. If the
values from the two halves are statistically
independent values from a single population,
10.1.3 Sample extraction
then the ratio of the variance of the values for
the whole sample to the average values for
Random sampling the two halves should be 0.5. Results show a
Correct sampling technique requires a random calculated ratio of 0.9 or greater so that little
selection of each sample from the popula- appears to be gained by analyzing the whole
tion. At an operating mine, it is possible that sample rather than half of it.
sampling is nonrandom since it is completed If the sample volume–variance relationship
from mine development which is planned on were important it would be desirable to analyze
a systematic basis. Provided, however, that all portions of a core from a drillhole rather
the mineral particles which form the popula- than follow the common practice of splitting
tion are themselves randomly distributed this the core longitudinally to save a half as a record
objection is overcome. Many mineral deposits and use half for analysis. In sampling, however,
display trends and variations in the spatial dis- this relationship should always be considered
tribution of their grade variables and the effect and, if possible, tested. This is particularly