• To participate in the 911Metallurgist Forums, be sure to be Logged-IN
  • Use Add New Topic to ask a New Question/Discussion about Mineralogy or Microscopy .
  • OR Select a Topic that Interests you.
  • Use Add Reply = to Reply/Participate in a Topic/Discussion (most frequent).
    Using Add Reply allows you to Attach Images or PDF files and provide a more complete input.
  • Use Add Comment = to comment on someone else’s Reply in an already active Topic/Discussion.

Sampling method final pulp powder (34 replies)

8 months ago
Standartenfurer 8 months ago

In mining and exploration, geologists and sample preparation technicians take a lot of care to correctly collect, sub sample and prepare a powder (pulp) of the original sample (usually starting a original lot of 1s to 10s of kilograms). Often the last few steps involve pulverizing 1-2 kg to a sub 80-100 micron ‘top size’ powder. Then from this powder 200-300g is collected into a paper packet to serve as the assay source sample. Finally, an amount ranging from say 0.4g (for XRF) to up to say 50g (for Fire Assay) is collected from this final lot for analysis. Now my experience of the first of the sub sample (the 200 to 300 g lot) is that the sample is collected using a scoop. Less frequently, other protocols call fall dumping the lot from the pulverizer onto a matt then collecting a few increments (say up to 10) from the lot. Only once have a seen the use of rotary splitters to divide the pulp correctly. At the laboratory sampling stage the usual process I have seen is to collect the final digestion sub sample (the 0.4 or even the 50 g) from the top of the pulp packet. Now in terms of sampling theory, neither of these methods seem like to be correct, and when questioned, the usual reply is that the pulp is ‘homogenous’ so we should not worry. Certainly duplicate results often confirm the low variability of the replicate samples from the final stage. I would be interested to hear about experiences where the (effectively) grab sampling in the final stage of sub sampling has been identified as a precision problem, or alternatively, what process have been implemented to improve precision at these critical (?) sub sampling stages.

Bob Mathias
8 months ago
Bob Mathias 8 months ago

I've done some testing on pulp material in an African lab where the pulp travels for a fair bit on an uneven, corrugated ground on a shabby carriage. Clearly demonstrated segregation in these and changed approach based on that.

Scooping form the top is done at all major laboratories that I've visited and audited. Terrible practice in my opinion; the argument of "it should be homogeneous.

Only one lab (Antech in Zimbabwe) was spreading on a mat and taking 10 increments to combat GSE. I've never seen a lab using a correctly designed and operated rotary splitter. Doesn't Dominique have his "dipping technique" in some paper or course material? I'd be interested to see others' comments.

Jean Rasczak
8 months ago
Jean Rasczak 8 months ago

Many geologic materials do have qualities which support affordable creation of an effectively homogeneous pulp. The lab has a responsibility to verify procedures are appropriate to achieving the appropriate degree of homogeneous quality. And, when necessary to use one of the many bottle type RSD units sold to create assay portions from materials which are too heterogeneous for creating economically practical homogeneous product.

8 months ago
Sturmbann 8 months ago

The problems you mention are a true issue and generally generate sampling bias.

Powders of minerals are NEVER homogeneous. It is what the laboratory staff has to bear in mind.

Powder is done to reduce the particle sizes in order to reduce the sample mass to the required one for analysis. But in return, there is mineral liberation (which is expected for ore processing). Then the powder is composed of particles of various minerals with various sizes and densities.

Segregation will then take place. For example, heavy minerals will migrate:

To the bottom of the mill (vibrations during mill shut-down favor such segregation);
Close to the matt when spreading material;
To the bottom of the packet during filling, handling, transportation

Conversely, coarse light particles will migrate on the top of the powder heap (Bagnold effect).

When powder is collected on the top, if the analyzed element is mainly in the heavy minerals, it will be underestimated. If it is in the coarse light particles, it will be overestimated.

I observed that in Pb/Zn ore for which many sub-samples have been collected at various locations of the powder heap and then analyzed. There was a clear heterogeneity of distribution. The only way to solve this issue is to use a rotary divider. 

David Kano
8 months ago
David Kano 8 months ago

The pulp duplicates are generally very similar and where they are not you have not achieved a homogenous pulp. Where the pulp is not homogenous you may be dealing with a very high nugget effect. In those cases you would go for a screening off the coarse fraction and doing a duplicate on the fines. If the fine pulp duplicate are now similar the offending nuggets have been captured in the coarse fraction and your problem is solved.

Zander Barcalow
8 months ago
Zander Barcalow 8 months ago

I have also seen numerous cases of pulps being prepared for final analysis by scooping (e.g. with a spatula) from the top of the pulp container

My experience is that laboratories focus more on the repeatability/accuracy of the analytic portions of the process (e.g. particle-sizing, mass-measurement, XRF analysis) and assume (to varying extents) that the "industry standard" or "best practice" sample preparation process is representative and consistent (both internally and with other laboratories).

I have seen a number of "round robin" type comparisons. However, these seem to favor use of a standardized pulp/material that tests the accuracy/repeatability of the analysis step rather than a "standardized" raw sample that tests the whole process including sample management and preparation.

8 months ago
Unterstarm 8 months ago

The answer to your question would depend on the purpose of the analytical results. It would also need to be in the context of your budget. As money is always limited I give priority to addressing sub-sampling stages that give maximum bang for buck.

Recently I was involved in developing a sampling and analytical protocol that involved another party - smell of being commercial. Coarse gold is present in quartz veins as well as alteration haloes and low grade. Resource RC submission of <3kg samples for "Total Prep" using LM5 and FA50 QQ plots indicated Poisson effect at 1.2g/t. It is often forgotten the lower the gold grade and coarser the gold the larger the RC sub-sample is required and total prep is not appropriate as a larger sub-sample is required. I performed a balanced design sampling tree study based on 3 RC holes for grade control purposes submitting 5kg target sub-samples, crushed in a Boyd crusher and 2.6 kg rotary split duplicate sub-sample pulverized in a robotic system.

The LM5 pulp duplicates collected by a trouser leg splitter (two flute Jones Splitter) and laboratory duplicate samples taken from the assay pulp pack replicated well but not homogeneous. The homogeneity will in part depend on how the gold is distributed. Gold encapsulated in hard minerals (quartz, pyrite) may preferentially report in the + 75 micron fraction. This will depend on the contractual particle size, 80% or 90% passing +75 micron. I am not aware of a commercial laboratory using a micro-Jones splitter to produce the assay pulp split. It is considered too time consuming (costly and laboratories are too frequently selected on lowest price- you get what you pay for). Rarely do you see Japanese layer cake sampling. I have seen some laboratories use roll mixing that may introduce segregation of gold and various forms of grab sampling. Generally spooning out a target weight sample from the top of the assay pulp packet into a mixing cup with flux.

In the overall context the errors at the sampling of the assay pulp I would regard as not material in this case compared to the errors incurred in the field using static cone splitters, poor drilling practices, under-size samples and general lack of understanding of TOS and how to apply the rules. There is a general reluctance to improve RC sample quality at the drill rig, especially at the moment considering the cost of the introduction correctly designed cyclone splitters such as Progradex. This could form another discussion. As it indicates with the QA focus on the laboratory we will continue to produce more precise inaccurate assays. Stuff-ups at the laboratories are a bigger issue. I do not regard stuff-ups as sampling errors.

Victor Bergman
8 months ago
Victor Bergman 8 months ago

Density segregation during shipping is clearly an issue, and I support earlier statements for the need of a proper divider. A quick demonstration was proposed using a pXRF handheld, and a mid-size plastic bag of pulp. Shooting both sides of the bag returned results with variable differences. Samples with heterogeneous density mineral compositions showed higher differences than samples with homogeneous densities.

As most desirable commodities have higher densities than host rocks, this demonstration applies to mineral exploration.

8 months ago
Sturmbann 8 months ago

If duplicates are not similar, it indicates the pulp is not homogeneous. If the duplicates are similar, it is not a proof of homogeneity. You have a large probability to have the same bias, specifically if the two samples have been scooped on the top.

In ore sampling, pulp is NEVER homogeneous. It is the more reasonable assumption. Segregation appears very quickly and "homogenization" is generally inefficient. Then scooping generates a Grouping and Segregation Error (GSE) with a not null mean value (bias) which can represent several standard deviation of the Fundamental Sampling Error (FSE).

QA/QC procedures, as they are generally done, are unable to detect this bias as standards are introduced after the scooping stage (and standards are generally more homogeneous than ore) and duplicates are scooped following the same mistake.

Sugar Watkins
8 months ago
Sugar Watkins 8 months ago

Precision/repeatability/reproducibility studies I have done suggest that the practice of scooping the digest sample from the final pulp is not the largest part of the prep error by any means. In my view the initial jaw crush and first volume reduction is where most issues occur. Yes, I agree that it is not best practice, but in value-for-money terms we need to look at where the most error is occurring.

As you have noted, there are different passing rates depending on lab, contract spec, etc. Should we not all be asking for 95% passing rates to be specified so as to use the FSE equation? I am not seeing many labs doing this.
My own personal issue is with accuracy.

8 months ago
Gruppen 8 months ago

Scooping can affect the assay. One test done in Australia involved pulp from the top to the bottom in the bag. The top of the bag returned 30% less Au than the bottom of the bag. This was caused by the vibration when moving the pulp sample bag upright on a trolley. Au tends precipitate to the bottom.

Jean Rasczak
8 months ago
Jean Rasczak 8 months ago

In my current laboratory (not a geologic lab), all assay portions have to be split out to sustain representative quality. The degree of heterogeneity and values involved allow me to apply this costly protocol. Most geologic labs are not permitted to change to this more costly protocol. All they can do is quantify the uncertainty within their QC/QA programs and design prep procedures to minimize this problem as much as possible. The clients (geologists and management) will not pay for the added cost, even when QA testing suggests that, due to sample heterogeneity, assay protocols generate non-representative assays.

The issue is not ignored by all round robins. The Society of Mineral Analysts (SMA) round robin does apply protocols assessing heterogeneity vs. analytic uncertainty in pulps. The larger prep issue which is ignored is the impact of heterogeneity on coarse split accuracy prior to pulverizing to -100 or - 75 micron pulps.

The industry "standard" of weighing directly from blended pulps is driven by the cost conscious client - geologists and management. The process continues because up front immediate costs rule. And, to a surprising degree, the better geologists appear to successfully apply statistical methods and judgment to minimize costly errors which might otherwise be caused by heterogeneous samples. And the opportunities missed? To the accountants, they never existed.

I hope that the laboratory you are using is "screening off the coarse fraction" and removing "the offending nuggets" in the context of applying a screen fire assay. If you are simply discarding the coarse fraction, this practice would, for at least one very profitable mine I know of, have lead to "no mine" and a lost profit opportunity.

Maya Rothman
8 months ago
Maya Rothman 8 months ago

An interesting question, and one that has always bothered me – especially as it is not often discussed in the technical literature (well not the literature I read anyway (no comments thanks)). Whilst I cannot provide any definite examples of where I have overcome issues, I would like to make a few observations:

If there is a precision issue at the pulp stage of sampling, then it is most likely that you have an even larger sampling issue back at the initial sampling stage at the drillrig. I have often been asked by management to look at “poor practices” at the laboratory stages when these are well and truly overshadowed by the inappropriate (wrt to the style of mineralization) sampling processes implemented at the initial sample collection point (i.e. the drill rig);

Taking on board about similar pulp duplicate results (i.e. good precision) not guaranteeing homogeneity - this is a very real issue, and one where a lot of us may become complacent upon receipt of good precision values from pulp duplicate samples;

Some laboratories collect their bulk pulp sample for ensuing assaying directly from the pulverising bowl by (usually) scoping the material directly from the bowl into the craft paper envelopes. The assumption is that the sample is well and truly homogenised in this state. This is a contentious proposition – no laboratory has ever presented to me any studies proving that this is so. I remember Phil Hellman (H&SC) once proposing an interesting experiment that could prove this (and apologies to Phil if I misrepresent him here) – he proposed that a sample containing coarse gold particles be mixed with cement and then pulverised. The cement mixture is allowed to “go off”, and this is removed from the bowl (maybe easier said than done) and carefully carved up with each slice analysed for gold. It would quickly show how homogeneous the mixture. With a bit of judicious sampling it would be possible to map out where in the bowl the gold tends to congregate (if at all). A very interesting Masters topic I would think;

As a young geologist I remember encountering a pulp duplicate dataset from a very nuggetty gold deposit (Rustler’s Roost in the Northern Territory, Australia). The comparison was the best I have ever encountered (the precision even exceeding some base metal deposits I have since evaluated); and the dataset included gold grades in excess of 100g/t. The sample preparation/chemical analyses were carried out by a now defunct laboratory called AssayCorp. The unusual thing about AssayCorp (by Australian standards anyway) was that they used Keegormills for pulverising (apparently these were commonly used in South African laboratories).

[Ray Wooldridge (of Assaycorp) wrote a paper comparing Keegormills to a LM5 and a Mixermill ring mill (Wooldridge, R. 1998: Sample preparation and assaying of coarse gold ores. The problem with Keegormills (as pointed out in Wooldridge’s paper) is that it produces a very segregated product that requires homogenisation prior to sampling for assay. Consequently, the use of Keegormills are labour intensive and not conducive to the fast turn-around times required by the modern laboratory.]

I don’t think Wooldridge’s paper fully explains why the Keegormill produces the excellent precision it does, and it was many years later that I think I discovered why. Years later I was working on another nuggetty gold deposit (in the same area as Rustler’s Roost of all places) when I came across two very interesting datasets. From the same set of field samples there were two datasets of screen fire assay results; one set derived from a laboratory using a ring mill pulveriser and another from Assaycorp (which used the Keegormill). Lo and behold, the amount of gold reporting to the oversize fraction (I can’t now remember the screen size being used) was significantly higher in the samples pulverised in the ring mill than the Keegormill. My conclusion was that the Keegormill was able to physically break the gold grains more successfully than the ring mills.

Anyway, the reason for the long-winded (and perhaps off-topic) yarn is that it is important for us all to be mindful of how effective or otherwise the equipment is at comminution. This may have a bearing on how easy it is to obtain a representative sample for ensuing assaying.

So, ultimately, I think it is up to us as professionals to be mindful of the style of mineralisation we are dealing with, and carefully think through the whole sampling/comminution/sub-sampling protocols/processes in place and:

Envisage if it is feasible that such “heterogeneous” errors could occur;
Devise sampling tests to investigate if there is an issue;
Take remedial action as required (i.e. change the procedures) to eliminate or reduce the error.

8 months ago
Oberstorm 8 months ago

Heavy minerals may migrate to the bottom of the mill is worth keeping in mind, I have seen test work that showed a higher concentration of gold particles were found towards the lower area of the pulveriser bowl.

If this segregation process is continuous and not just restricted to the period of mill shut down, the segregated particles will spend more time being milled under the rings/puck/discuss. So potentially beneficial, as any additional liberation and increased abrasion on larger gold particles is a plus, but negative in that you may now be dealing with segregated material. Either way, it cannot be assumed, that ring/puck/discus mills, always achieve adequate mixing of the milled material.

Ring/discuss mills can achieve some degree of mixing, in contrast to continuous processes such as jaw crushing or continuous ring mills which do not - this is a good trick but not perfect.

The possibility segregation in the bowl and further segregation taking place on removing/pouring the material from the bowl adds up to a case for using a sampling device on milled sample pulps. It may even make QA/QC a bit more convincing.

Rotary sampling devices are slow when operated correctly (slow feed/high cut rate), so you may need a pair of small RSD's to keep up with one mill.

RSD's used for this purpose, will only be required to deal with near constant sample weights (determined by the capacity of mill bowl selected) - in this case, it would be relatively easy to set up a simultaneous collection of two or three, 30 or 50 gram packets for assay.

This would eliminate scooping and associated segregation on the sampling mat. This procedure would also eliminate any effects of potential segregation during a transport phase as the entire contents of the individual packets can be assayed.

Static Riffle splitters are also slower than scooping, have inferior sampling efficiency compared to correctly operated RSD's and tend to block when used for fine powders.

More likely to be acceptable for small bowl, milled samples (250 gram) - larger samples will be hard work, especially when material blocks the chutes/riffles. Again, you have the option of taking multiple assay sized samples to minimize the effects of potential segregation.

Vibratory riffle splitters (using 64 chutes, rather than 12 or 16 chutes, common to smaller static riffle splitters) are faster, easier to operate with fine powders and will not block. Sampling efficiency is probably much closer to that of RSD's than to the static riffle splitters with a low chute count.

There is potential to increase the chute count further for fine milled material, at this time these are custom fabricated or adapted items but worthy of mention as a potential option.

Linear sampling devices may become a practical option in the not too distant future, Rock labs recently made new LSD with a greatly improved cut rate for crushed samples. A similar, scaled down version of this LSD could also provide very high cut rates on milled material.

Carl Jenkins
8 months ago
Carl Jenkins 8 months ago

A really good discussion and starting question. In short I have never seen rotary splitters being used at labs preparing final pulps for assay. Only the grab scooping method has been observed. A further question would be whether world class labs should change to a generic better practice (appropriately sized rotary splitting) or really only on a project by project basis via the technical lead person's guidance. Why open up the possibility for further potential lab error if the rotary splitter can negate it in any event (is cost really a big issue here)

John Koenig
8 months ago
John Koenig 8 months ago

There is no doubt manual sub-sampling of the pulp material at weighing is a potential area where an extraction error can occur, even if not significant in the whole process from sample collection (aka drill rig) to preparation and analysis. However the laboratory shouldn't take pulp homogeneity for granted. The extent to which this is a problem depends on a number of factors, some already mentioned in this discussion (storage and transport history of the pulp packet, size/weight of pulp in packet vs. weight of sample for assay, pulverising technique used, mineral type, element of interest and deportment in pulp, assay weight, TAT/$ and cost of quality etc.). Sampling a DSO for XRF analysis at 0.6g from a 100g packet, is entirely different to a 50g gold ore sample for FA from a 250g packet, which would hardly be considered a "grab sample from the top".

There are a number of quality control mechanisms that can be used in the laboratory to determine the extent of the problem, such as:

Ensure pulverising sizing specifications are met.
Weighing staff are trained to visually inspect pulp material prior to sub-sampling. In-homogeneity can often be easily spotted.
Pulp packets are shaken/mixed prior to sub sampling.
Sampling the pulp towards the middle of the material, and not touching the sides. The intent here is too efficiently and cost effectively gives each particle in the pulp packet an equal opportunity to be sampled for weighing. It is not perfect, but neither are any of the other techniques proposed to extract a specific weight of sample (within tight tolerances) from one vessel/packet/vial, and deliver it to another that has been tarred on a balance. This is not a trivial task when you think about the options.

If pulps have been in storage for long periods of time, or transported over large distances, conduct some analysis of the extent of segregation, by either requesting the laboratory to re-pulverise a random % of samples and assay, or by conducting a re-pulverising exercise of selected samples to determine if segregation has occurred.
Critique the analysis of repeats (analysis from same pulp packet) and compare to results from pulp duplicates (second pulp packet taken from same pulveriser). This is the most effective tool available for assessment of sub-sampling weighing errors.

Critique your submitted standard samples and ensure tolerances are met
Critique the laboratory reported QC samples
Reliable and resubmit pulp samples to the same laboratory for blind duplicate analysis. But be wary of sub-sampling/splitting the original returned pulp packet as a bias may be introduced at transfer.

Bill Fraser
8 months ago
Bill Fraser 8 months ago

Use a miniature RSD –ALWAYS.

Jean Rasczak
8 months ago
Jean Rasczak 8 months ago

Please note that special case course gold ore advantage of disk mills over ring mills is not an endorsement of keegor mills. The vertical shaft design has been documented to associate with significant cross-contamination between samples (carryover). A horizontal shaft design which allows complete break open and direct clean between samples is more appropriate if a laboratory wishes to take advantage of this affect. In addition to this, correct operation of a disk mail requires more technical skill for maintaining appropriate surface contact conditions between disc mails for best outcomes.

8 months ago
Obergruppenfuhrer 8 months ago

"Pulp packets are shaken/mixed prior to sub sampling." Not so sure.

8 months ago
Hauptsturm 8 months ago

Yes we do fire the coarse fraction to extinction. In our case about 60 grams of material and I do use the term homogeneous loosely.

Something that this discussion has made me think about is the sampling of the fine fraction in a typical screening. The fine fraction would almost certainly be segregated after the vibration during screening. Sampling the fine fraction would now be an issue.

Jean Rasczak
8 months ago
Jean Rasczak 8 months ago

I met an owner of a laboratory who had done a study that explains why disc mills on coarse gold can with, proper blending, generate a better homogeneous than ring mills. Ring milled coarse gold particles, as examined under high power.

Bill Fraser
8 months ago
Bill Fraser 8 months ago

A few comments:

The practice of weighing an analytical aliquot to a precise weight is something that went out the window when LIMS came into use. The motivation for this is to make the calculation simple - but the LIMS now does the calculation.

If, at a point in the sample prep, the sample is taken to a specific weight using a splitter designed to do this in a correct manner (I designed and built 'variable split sample dividers' (VSSDs) for Anglo Platinum), then sub-sampling can be carried out using only RSDs, right down to the analytical aliquot. Miniature RSDs (http://is.gd/qUEkq2 and Gilson) are readily available.

Taking an aliquot out of the top of a paper packet after attempting to 'mix' by shaking the packet is a poor practice. Trying to justify this practice because the alternative is a bit more trouble is not what we want.

The manufacture of VSSDs has ceased for the moment - I would be interested to know who might like to have one. If the response is good, I am sure that I can organise something to make them available again. Just to let you know, these machines were designed to take up to about 4 kg of dried concentrate and split out precisely (uncertainty less than 1% of target mass) a target mass of subsample, from 5 to 95% of the starting mass. The machine did this in a few minutes, unattended and took more than 100 increments into the sub-sample. Segregation in the feed hopper was completely overcome. Dust losses were less than 0.01%. The machine could also be interfaced to a LIMS so that target masses could be downloaded automatically.

Anglo Pt uses them to make up weighted composite samples. Their manpower and time for doing this was cut by a factor of 4 or more. The payback time on the machine was months.

Let's try to engender the use of best practice as we know it.

8 months ago
OberstGruppen 8 months ago

Part 3 of Mikli Paper
The LM-1 Laboratory Mill

This is a large capacity Tema-type mill. The tests used the C2000 head, which contains a single puck with no rings.
The entire test used 10 nuggets, average weight 6mg, with various weights of double crushed basalt (DCB). The first test used 1kg DCB, the second 1.75kg DCB, both pulverising for 10 min.

In each case a small residue of +80 meshes remained, consisting mostly of several hundred gold particles.

The third test used 1.5kg DCB and a pulverising time of 20 min. Again, the small amount of +80 meshes contained some five hundred gold particles, many of which had elongated shapes. 99.7% the original material had been reduced to -200 meshes. Given extra time, it comes close to the minimum capacity required to pulverise 2m of split core containing visible gold.

It is concluded that there is a critical thickness of gold sheets, less than 10 microns, at which -200 mesh rock particles can puncture and disintegrate gold sheets into such fine dust that gold is not recoverable by panning the -200 mesh fraction.

In the fourth test an effort was made to recover the thin gold platelets similar to those lost from the previous tests during panning.

The entire 750gm of pulverised sample was screened over 80 and 120 mesh screens, then cyclosized into nominal -75 +45, -45 +34, -34 +24, -26 +16 and -16 +12 micron size fractions. The -12 micron size fraction was lost in overflow. The -120 +200 mesh size fraction was obtained by wet-screening the +45 micron cyclosizer product on a 200 mesh screen. All size fractions were then treated with HF to dissolve rock particles.

4.8mg of gold (8% of total weight) was recovered from the -75 +45 micron size fraction. Gold in un-weighable amounts was also reported in the -80 +120 mesh, -120 +200 mesh and -45 +34 micron fraction.

No gold detected in the -36 micron size fractions or the +80 mesh size fraction.

Screen Tests of LMI-C2000 Experiments
Test Gold Nuggets wt DCB Time +80# % -80 +120# % -120 +200# % -200# %

1 10x6mg 1.00kg 10 min <0.10 0.4 7.2 92.3
2 10x6mg 0.75kg 10 min Tr 0.1 0.7 99.2
3 10x6mg 1.50kg 20 min 0.06 1.2 8.6 90.2
4 10x6mg 0.75kg 20 min TrTrTr 99.9

Part 2 of Mikli Paper

The Ledir Multi-mill (the pulveriser has been modified; it is now marketed as “Multimill”).
Each charge consisted of a single gold nugget with 100g of crushed quartz. Two hours’ pulverising reduced the weight of the nuggets by about 10%, leaving the cores of the nuggets intact.

Four hours’ pulverising left some cores more or less intact, while others were reduced to lakes up to 0.5mm in length.

Ledir mills can handle batches of 60 to 80 samples each up to 100g. Pulverisation times are long, but labour costs are not affected by extending pulverisation times to 8 hours on a normal day or 14 hours overnight. The major drawbacks of the machine are:

Sample weights are limited to 100g
The samples should be very dry
It is not always convenience to accumulate 60 – 80 samples before work can begin

Ring Mills (e.g. Tema, Siebtechnik, “swing mills”).

Each charge was a single gold nugget and 100g crusted quartz. After 2 minutes pulverising residual cores remained. Other gold particles tended to be elongated, up to 1.5 x 6.5mm in area.

5 minutes pulverising destroyed all the nuggets, but gold flakes up to 0.25mm remained.
Pulverisation times can be doubled without increasing labour costs, by setting up a two pulveriser work station for each operator.

Large Capacity Impact Pulveriser (2 ring pulveriser of Tema type, hydraulically vibrated)
This could take charges up to 2.5kg.

Firstly, 2.5kg crushed quartz were pulverised for 6 min with 27 nuggets weighing 216.4mg. Previously experience had indicated that 10 min grinding produced 95% minus 200 mesh products. A time of 6 min was chosen to inspect the nuggets before they were expected to break up.

After 6 min the 27 nuggets still retained 60% of their weight, the largest being irregular plates over 4mm. The remaining particles were elongated, some as large as 2 x 0.5mm. This machine also produced cigars, but they were not solid. The original nuggets had been smeared or flattened into sheets that were next rolled into cigars. The elongated flakes came from cigars being crushed and split down their longitudinal axes.

A second test pulverised 2.5kg crushed quartz with 27 largest nuggets weighed 12mg. A 0.5mg nugget in a 25g assay pulp would report as 20g/t.

The pulverising action has a large abrasive component that the small impact mill. The mill will produce a homogenous sample if pulverisation times are extended.

Part 4 of Mikli Paper


In tests 1 and 3 larger samples weights were used. The +80 size fraction consisted mostly of gold particles. This demonstrates one possible use of impact pulverisers with large bowls.

Pulverising to minus 99.9% minus 200 mesh.

In test 4, 92% of the 60mg of gold present in the sample was unrecoverable in the cyclosizer, 4.8mg of the gold was recovered in the-75 +45 micron size fraction. The weight of gold recovered from other size fraction was too low to weigh on a 5 decimal place analytical balance (0.01mg or less).

The number of gold particles in the -75 +45 micron size fraction was estimated to be 2500. This means that the 8% of gold from each 6mg gold nugget in the +45 micron size fractions consisted of approximately 250 particles per nugget.

In a 25gm assay pulp one such particle would report as 0.08 ppm Au.
The gold which reported in the -45 +34 micron size fraction consisted of four platelets close to 75 x 100 microns and dozens of smaller particles, some only 10 microns in diameter, but nearly equiaxial. The large near -200 mesh size (but thin) gold particles settled in the same size fraction as the equiaxial 10 micron gold particles. All of the remaining 55.2mg of gold particles were either too thin or too small to be trapped in the cyclosizer.

Disc grinders are efficient rock pulverisers but there use is limited to an intermediate pulverisation stage without sample size reduction, if nugget native metal is present.

Impact pulverisers can handle nugget ores better because;

The nuggets can be collected by scalping after a short pulverisation time, orThe entire sample can be homogenised

Increases in labour costs in the low wages sample preparation section are partly compensated by one operator operating several pulverisers.

The main justification for extended pulverisation times is the lowering of cost of labour, chemicals and “housekeeping” costs in the high cost chemical laboratory section and improve accuracy of analytical results.


A brief comparison is made about the action of disc and impact pulverisers on gold nuggets. The performance of LM-1 pulveriser with a C2000 pulverising bowl is discussed in detail.

Examples are given of analytical procedures where improve sample preparation leads to lower cost chemical sample pre-treatment and improve accuracy of analytical results.

1. Pulveriser

Horizontal Axis Disc Pulveriser

Two charges were prepared

1kg crushed quartz with 11 nuggets weighing 114.6mg
1kg crushed quartz with 9 nuggets weighing 80.7mg

After pulverising, the products from (a) and (b) were combined and panned. The twenty largest nuggets weighed 27.5mg and had been rolled into ‘cigars’ up to 2mm long by 0.4mm diameter. The cigars were solid throughout. It appeared that pieces had been torn from the nuggets and then rolled into smaller cigars and spheres. The residual cores were rolled into larger cigars.

The disc pulveriser is not satisfactory for the final reduction stage of a nuggety ore. The 1 and 2mg cigars report as 40 and 80 g/t in a 25g assay pulp.
It would be a useful pulveriser for the initial reduction stage of a large sample.

Vertical Axis Disc Pulveriser

Synthetic samples were prepared, each of 900g crushed basalt and 10 nuggets total weight 40-70mg.

Pulverisation times were varied by altering the gap between the plates. Small gap settings resulted in extended pulverising times. Nuggets tend to be rolled into spheres, the only weight loss arising from the abrasion of projections from the original nuggets, even after 6 min in the machine. A final test confirmed that nuggets smaller than three times the gap between the plates are torn into fragments of less than 0.5mg.

To ensure that nuggets of 4mg or heavier are reduced the plate gap must be 200 microns or more, results as easily obtained with low cost conventional horizontal axis pulveriser. The machine should not be used for the final reduction stage with native metals exceeding 200 microns diameter.

In case of nuggety gold ore the entire sample must be reground at a close plate setting before a split can be taken for final pulverisation stage.

8 months ago
Unterstarm 8 months ago

Whilst I agree with most of the above comments the reality is in this economic climate that commercial laboratories are often chosen by mining and exploration companies on unit price (lowest), turnaround time (fastest) and quality if at all as long it does not cost the client extra. Frequently RC samples especially, or even drill core where recoveries are poor where a smaller diameter core is used to reducecosts are often compromised before they are even reach the laboratory.

Victor Bergman
8 months ago
Victor Bergman 8 months ago

Cost-cutting good practice may lead to temporary savings and also to expensive mistakes in investment decisions, but most of the efforts of budget controlling are focused on the former. The only way to reconcile good practice and cost effectiveness is automation of good practice.

Ace Levy
8 months ago
Ace Levy 8 months ago

I think similar duplicate results should be expected if you follow an identical process.
So perhaps, whether the duplicates are taken by scooping, riffle or RSD - repeating the process, is still perhaps, repeating the same mistake but with better precision.

Going to the extreme - if a RSD can miss some nuggets once, it can surely miss them twice. It’s essentially, doing the same thing over and over, but expecting a different outcome.

I guess the danger is that QAQC may look OK, but if duplicate sampling is capable of hiding sampling issues, as well as flagging them up - alternatives should be considered.

I previously suggested taking simultaneous duplicate samples from a RSD - this is fine if the goal is only to provide additional, back up samples for the fire assayer. Certainly it’s better than scooping from a bag, if segregation has occurred.

For QC samples, it may be better to pass the entire pulp reject, through the RSD process, one more time at a somewhat slower feed rate and take a second check sample made up from more cuts than that of the original.

As many RSD's will work as a mixing device, running the sample a second time through the RSD will also achieve a secondary mix process, to further combat segregation.

Another possibility is to mill the pulp for an additional period, prior to the secondary RSD process. Some type of asymmetric, rather than duplicate check samples may be one way to get more value from the rotary sampling device and QC in general.

The larger sample prep issue is the impact of heterogeneity on coarse split accuracy.
If coarse split accuracy is bad, every cent spent on the sub sample, is money spent, polishing a turf.

The QAQC coarse split duplicates, because it is coarse and unmixed, may also repeat the same mistake with good precision - but with a higher probability of repeating a mistake, if compared to the milled/mixed process, previously discussed.

He suggests that 95% pass rates would be a good idea. I agree, much better than the 70% or 75% minus 2 mm offered by many labs.

The question which should be asked is what is the size range within the 25% to 30% oversize? With a well maintained lab crusher, the oversize may be in the range of 2 mm to 4 mm - with a worn crusher it may be 2 mm to 16 mm. the lab can still meet a 70% or 75% minus 2 mm with a worn crusher - it’s easy.

The promise (QC spec) relates only, to the minus 2 mm product. If 30% of the crushed sample is in the 2 mm to 16 mm range, the efficiency of all sampling devices will be much lower, compared to 30% of the crushed sample, in the 2 mm to 4 mm range.

One 16 mm rock chip has a very good chance of falling into the reject, because the reject is massive in comparison to the sample being taken for milling.

Imagine a rotating pie chart with 80% of the chart representing the reject and 20% representing the sample - now shut your eyes and give yourself one shot at putting a 16 mm chip on the 20% segment. (equivalent to 5 kg with 1 kg taken for milling/pulp)
If you missed, tough luck - each rock chip gets only one chance.

With the same 16 mm chip, crushed into 100 smaller chips, now try to imagine the same test - but this time try to put all of the chips on the 80% segment. It is hard to imagine all 100 chips falling into the reject.

I would expect an exponential reduction in the efficiency of any and all sampling devices, if large heterogeneous rock chips, are effectively ignored prior to the coarse split process.

Perhaps best, not to expect too much from a sampling device - think of it as a game of roulette and consider how best to avoid placing a bad bet - or two with the duplicate.

Victor Bergman
8 months ago
Victor Bergman 8 months ago

I remember investigating on a systematic bias between two labs for gold exploration samples. I demonstrated that digestion, extraction with a different reagent and analysis were not at fault by sharing the pulps between both labs - this moved the bias from one lab to the other.

The core reason was that in one of the labs, in order to increase throughput, the full preparation was led only on the pass fraction of the primary jaw crusher step. Coarse fragments were therefore totally skipped from the analysis.

8 months ago
Oberfuhrer 8 months ago

I want to add on a somewhat tangent course. Reading your comment above, it reminded me of the following:
In the late 1980's, I investigate the effect of various mills on the preparation of pulp samples of Witwatersrand gold exploration drill core.

In comparing the results, I noted strange aspects with regard the vertical spindle pulveriser. Looking at the material, I got suspicious, and by using a paper covered strong hand magnet I managed to extract a significant proportion of iron filings from the pulp (confirmed by inspecting the material under a bi-ocular microscope).

After this, I have always been very sceptical of the use of these mills. In addition to the difficulty in cleaning the mill between samples and the strong possibility of cross contamination between samples, one should also be aware that notwithstanding good repeatability between duplicate pairs of pulp samples, one may dilute your sample significantly with iron filings. A laboratory using such a mill should actually report on the wear of the disks and attribute the mass loss per sample to the pulp sample obtained. It is good practice to consider the wear on any mill carefully.

Carmen Ibanz
8 months ago
Carmen Ibanz 8 months ago

Your tale regarding the Keegor disc mill is interesting, as are the follow up comments regarding disc mills.

Alan Carter
8 months ago
Alan Carter 8 months ago

ISO has a very god set of standards for sampling concentrate (even coarse materials) which are not rocket science to implement. The one thing I have started doing is calculating sampling variance when proposing a sampling scheme so that the client can see what variance can be expected. Even if the client insists on using an improper sampling scheme, they have been warned on the consequences.

I have had several debates on assumed homogeneity of materials with laboratory and country managers of a certain well known testing group up in Africa where the basics of sampling have not been followed:

The sampling equipment could not access the bulk being sampled (only 5-10% was accessible).
The sample was not split correctly
Incorrect sampling equipment was being used

The list goes on.

Tony Verdeschi
8 months ago
Tony Verdeschi 8 months ago

Is there any experience with a miniature linear sampler? We are developing this now for a different sampling application and were wondering if it would be a solution if used to sub sample the grinded gold ore sample from the ring mils?

8 months ago
Sturmbann 8 months ago

Can you provide documentation about this miniature linear sampler?

Paul Morrow
8 months ago
Paul Morrow 8 months ago

You could use a linear sampler to sub sample the feed/bulk per seconds per intervals to comp a represented sample for processing per hourly/hours or days.

Bill Fraser
8 months ago
Bill Fraser 8 months ago

The real solution for sampling finely ground material is a miniature rotary divider - Quantachrome (US) or Gilson.

2 weeks ago
Auggie 2 weeks ago
1 like by David

Some very interesting comments.  I've been in the assaying business for over 30 years and it always starts with the bulk sample, does it representative of the body of ore.  We would get bulk samples of ore which would come in 55 gallon drums from geologist and our challenge was to make sure we can take a 500lb + sample down to an assay ton and have reasonable variation in the assays.  

Using this technique below has worked very well for us on Au and PGMs, but is very time consuming:

  1. Dried and crushed contents to 6mm
  2. Coned and quartered sample 4 times to blend
  3. Coned and quartered sample down to approximately 1K
  4. Pulverized sample to 95% passing 150 mesh
  5. Roll sample on mat to homogenize (corner to corner 100 times)
  6. Cone and quarter, discard opposite corners, repeat #5
  7. Cone and quarter, roll sample repeating #5&#6 until you have approximately 250 -300 grams
  8. Using a Retch rotary sample divider. Divide sample to proper assay quantity, fire assay w/ ICP finish all portions, in our case 8 assays