It really depends on what you are trying to achieve.We used the Niton directly on wet RC chips in the original bag off the rig and got perfectly usable As, V, Mo, Cr, Zr, and Th numbers that we needed to evaluate both mineralization and stratography on the Livengood Project in Alaska. We had a sample turn around of 12 hours from the time a hole was drilled to have all the data back and it could be faster if required. If there was a question on the TD then the Niton was used in real time to assay the samples. We did tests comparing these results against 4-Acid ICP MS data before setting up our protocol. The accuracy and precision was sufficient for the tasks we used it for which were determining the level of mineralization, (100 vs 1000) ppm arsenic and distinguishing our volcanic suite from the intrusive suite (Zr and Th) and mapping the organic- rich marker horizon through the alteration (Mo, V) and identifying our ultramafic and basaltic units (Cr and Ni). Some sample bags have Ti as a whitener so watch out for that. For more on XRF sample prep see this. http://www.911metallurgist.com/blog/sample-preparation-methods-for-xrf-analysis

I have done a case study on iron ore from Pilbara area, Western Australia showing effects of sample preparation. For light elements (Al, Si, Mg, S and P), the best method is to have the sample pressed in the field (as a pellet) and do not use sample cup (which uses a 4 micron file; the film can absorb light elements signal). So if these light elements are important for you, use press method. If these elements are not that important, you can use cups. To pulverize samples on site and make pellets or cups, you may use "Sample Prep Tools" offered by ThermoFisher Scientific (which manufactures Niton pXRF analyzers).

We met this challenge during a big drilling project in Saudi Arabia (3000+ samples from RC cuttings) and Thermo Niton's Hammermill field preparation device helped us a lot, even if it was slightly under capacity. I know that Bruker has a nice alternative, and I trust Olympus people for offering something too. The key point is: sample prep is the most important issue for your results.

It is great to now see people “finally” taking the time to do sample prep and vastly improve their results. Modern pXRF analysers are very precise and accurate given the right sample and calibration considerations. Your point on capacity is also a very valid point, given the large sample that Reverse Circulation (RC) drilling delivers and as such, splitting and sample reduction is also a critical consideration for sample representivity. For large volume, exploration/grade control/production drilling, these small hammer mills just don’t cut it, not only in terms of throughput, but also ruggedization and pricey consumable wear. As such, we have partnered with a company that has designed and procured a fit-for-purpose, single pass crushing-grinding, production-grade solution.

Additionally, for the members of this forum, the above mentioned OEM field sample prep solutions are now available directly and non-exclusively from the original designer and manufacturer, Bear Claw Scientific. They have the full range including the modified Kinematica hammer mill, angle grinder - direct rock sampler, crushing tool, pellet press and various other handy accessories to greatly assist your sample preparation in the field.

The most important point is what do you aim to use the pXRF data for? You mention this is an on-rig exercise, presumably to support geological interpretation/logging more so than to provide an assay for ore grade/contaminants? If this is the case I have found that if the geos collect the RC fines when sieving their chips and take a pXRF reading on the (dry) fines, you get a reasonably representative pulverised sample which is ideal for alteration & litho-geochem. This will also give a good indication (although potentially slightly biased) grade estimate for Fe and other payable elements detectable with the pXRF. The only real bias introduced by this method is related to elements preferentially represented in the fine or coarse fractions. My experience from an 8000 sample orientation program in an orogenic gold setting is that this potential sampling bias had no material impact for litho/alteration geochem interp. and the simple process of sieving the fines had almost no impact on geo productivity on the rig and provided immediate pXRF feedback to the geo.

I need to disagree, but beware DSO iron ore samples with hematite/goethite in the fines, and most of the silica in the chips. I think a representative split and pulverize is a good idea to start with, then check against the RC fines. If you get the same answer, good, but I'd guess you'll have most of the iron and aluminum in the fines, and most of the silica in the chips as a direct RC product.

Well, the representativity of "biased" fines depend actually of the lithology. The bias will be negligible for sediment-hosted targets or for homogeneous formations, for disseminated or invisible commodities. The bias may be significant for quartz-hosted or vein-type commodities - personal experience about this.

About the equipment suggested, it will be adequate for reconnaissance tasks and on-the-rig spot decisions, not for a continuous job. It nevertheless constitutes a great leap forward for sample quality and subsequently for data quality.

Good points. I am by no means an Iron Ore expert. I have found the method provides good repeat-ability with ICP-OES (within the limits of the pXRF) for litho/altn geochem in hydrothermal base & precious metal systems (even vein related systems) so long as you are thinking about the bigger mineral system rather than individual meter analyses. It is particularly useful in providing a quantitative lithological and alteration logging guide which can help standardize logging across the geological team and of course the pXRF data must be integrated with textural, mineralogical and other critical "geological" observations. I think I have now got off the point of Tony's query. Best of luck finding an effective on-rig pulverizing method, sorry but I cannot help with this one, although it seems Bob has some good suggestions. 

You might find this application summary helpful: Iron Ore Mining and Grade Control Using Thermo Scientific Portable XRF Analyzers at http://www.niton.com/docs/literature/iron-ore-mining-and-grade-control-app-summary.pdf?sfvrsn=2 It gives methodology, results, and conclusions, as well as charts showing Fe, Al, and Si correlations between values measured by portable XRF and lab in the pile and the prepared pulp samples.

If you are dealing with systematic RC samples, with siliceous lithologies or with heterogeneous ores, you will end up with the need to build on site a small grinding/milling unit. This can be achieved using benchtop lab equipment (Retsch for instance) but also using redundant small equipment from your mining operations.
However, you must be careful with splitting and homogenisation, and with sample contamination by equipment active parts. Be careful when setting up a cheap grinding line, sampling bias may be more expensive in biased decisions.

Thanks everyone for all the conversation and comments, it's been incredibly helpful! This is for greenfields exploration mainly, looking at developing geochemical exploration vectors and trying and differentiate between BIF units in structurally complex areas. I'm in the very early stages of working any of this out so all the different insights have been awesome, thanks again and any further advice is always welcome!