Representative Sampling

Gold Refining Forum

Help Support Gold Refining Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

4metals

Moderator
Staff member
Supporting Member
Moderator
Joined
Apr 5, 2009
Messages
6,084
Location
northeast USA
This topic needs to be addressed because we often ask the question "What is the assay?" and all assayers hope that the sample is representative of the entire lot. So it is imperative to do a good job sampling.

It is my hope that this thread can become an informative collaboration from some of our members to help both to explain some of the science behind the sampling and detail the different methods to effectively sample as many different scrap types as we can.

This thread grew out of some comments made on a thread about gathering data from yields into a database. If there is any chance that such a database can be successful, it will be necessary to assure ourselves that the materials are being sampled properly.

I started this by re reading an old text named The Sampling and Assay of the Precious Metals by E.A. Smith. Written in 1913, second edition in 1947 and reprinted in 1987 by Met-Chem Research.

I mentioned this in the previous thread as did Chris (GSP) and had to chuckle when I read this in the textbook.

E.A. Smith in 1913 said:
To select a bona-fide sample may seem an easy matter, but those who have had experience testify that it is not so easy as it looks, and generally regard the selection of the sample as a more difficult matter than the actual assay of it.

One of the nice things about a fire assay and even an AA analysis is they are done in duplicate. For valuable lots I prefer running samples in triplicate. What this does for the assayer is it assures him or her that the sample is homogeneous if there is good agreement between the different duplicates of the same sample lot.

To put this all into some perspective to get us all on the same page, let's start off by discussing an easy material. Jewelers sweeps. Generally processed and blended into samples which are representative of anywhere from a 5 gallon bucket up to 5 full 55 gallon drums of material. Based on high grade sweeps that I have assayed, I preferred to use a 5 gram sample of the powder. A multiple drum blend of the prepared sweeps can represent as many as 2000 ounces of gold. So this lot would be run in triplicate. So every individual crucible would be loaded with 5 grams of sample and it would be done three times. Or to put this differently, every 5 gram fraction that i chose out of the 1 ounce package of powdered sample that I received for that lot had to be very very close to each other and if all samples are within close agreement, the sample is considered representative and the assay result is used. To take this analogy 1 step further, every individual ounce of the material in the entire lot has to be essentially identical to every other ounce.

So how is that accomplished?

The material has been incinerated and ball milled and passed through a screen so the largest particles are -40 mesh. I always sifted to -60 mesh because my in house assays were in better agreement for settlements with individual customers. For settling large multiple drums blended together to be sold by a primary refiner, I always liked the grab sample taken after the blend with a sample thief to be further milled in a puck mill to -80 mesh.

thief sampler.jpg

The material I brought for sampling was blended in a large double cone blender which effectively mixes the contents of all 5 drums into a powder that is thoroughly blended and homogenous. This material is returned to empty drums by dumping it from the blender filling one connected drum at a time until the contents are back in the drums. Once back in the drums a thief sampler is used to draw a sample from multiple locations in each drum. The sampler is inserted into the powder and pushed to the bottom of the drum. Then the sampler is opened up by twisting the inner sleeve, and powder will flow into the hollow tube from every level of the drum full of powder. The sampler is twisted closed and removed from the drum and what is inside is a sample of every level in the drum, top to bottom. This sample is dumped out of the sampler and the process is repeated 2 or 3 times per drum. All of these samples are mixed together in the same bucket and taken to the lab to prepare.

in the lab, the sample is either spread on a tray, leveled and quartered and manually mixed by an age old process known as cone and quartering. Another quicker option is to pass the powder through a splitter which divides all of the powder into equal portions. Which ever method is used, the purpose is to break down a large sample into more manageable samples without producing a systematic bias. The smaller lot now weighs about a pound and it should be the same as any other pound in the lot being sampled.

Splitter.jpg

The last thing that is done is the pound sample is sifted through a -80 mesh screen. Most of the powder is small enough to pass through the screen already, remember every screen classifies the material and holds back the bigger materials passing the smaller through. Whatever does not pass through is pulverized in a puck mill. The goal is to have all of the material pass through the -80 mesh screen.

Particle size is very important in producing a homogeneous powder sample for precious metal assay. For prepared jewelers sweeps in this example I have found the agreement between assays is much closer with smaller particle size. A lot of these ratio's of particle size to sample weights were worked out for mining ores and successful sampling and assaying. And most of the work was done in the 1800's.
With high grade polishing sweeps the results at the final weighing can vary by 1 milligram and the results are considered an acceptable standard deviation.

But nobody is going to take down a mountain and grind it into dust just to assay to see if it is worth processing. So over the years standards have evolved using larger particle sizes and allowing for greater standard deviations because the larger the particle the more probability that a single high yield particle will skew the entire sample higher. For the high yield powders we eliminate that possibility be crushing it significantly small enough that every fraction of a gram contains 42 begillion particles (translation = very very many) so a lot of small and evenly dispersed values are much less likely to cause a high standard deviation. We can apply those allowable deviations from hands on practical experience to develop sampling protocols for many different materials.
 
When I was at Sel-Rex, in the dark ages, I recall a huge batch of sweeps that we ended up fire assaying 96 samples on, if I remember right. Statistics was used on the results of preliminary samples to determine approximately how many samples were required to get a certain standard deviation. I might note that no amount of sampling/assaying can be done to give a 100% guarantee that any lot contains a certain amount of gold. The only way to guarantee this is to process the entire lot (as one big sample). With those 96 samples, we may have ended up with, say, a 97% probability that such-and-such amount of gold would be in the lot. About any $15-$20 technical calculator has statistical functions and they are not that difficult to use and understand, with a little study. I'm not saying that statistics is needed for most run-of-the-mill lots - it isn't. I just wanted to show how complex things can get.

The reason assaying is simpler than sampling is that, once you learn assaying it's pretty rote, in that you do essentially the same thing every time with minor differences. With sampling, the idea is to select representative samples, those that will give a reasonably true representation of how much gold is actually in the entire lot. Sampling require ingenuity and knowledge and various types of material can require different approaches As 4metals does, I prefer 3 samples on anything important like karat gold. If you only run 2 samples and one is high and the other is low, you don't know which is more right.

I have done a lot of cone and quartering (C&Q) on things like shredded circuit boards or drums of pins. Like the sample splitter pictured in 4metals' post above, for best results, I feel that each sample requires the entire lot to be put through the C&Q or the sample splitter. This can be very laborious. I've mentioned several times that I ran about 6000 film assays representing about 50,000,000 pounds of film over a 4 year period. The film had been chopped into about 3/8" size pieces and then put into 40 gal fiber drums. Each drum was thief sampled with a grain sampler (see 4metals' 1st photo above) until about the equivalent of 2 or 3 gallons was obtained and put into a plastic tray. At first, I ran all of those samples through a sample splitter to get about a 200g baggie full, but that took forever. I then experimented and found that all I had to do was mix the film in the trays well with my hands and then hand grab a couple of samples and put them in baggies. My assayed result were quite accurate and compared within a percent or two of the actual processed yields of the lots. At first, I used fire assay to run the samples. This was expensive and very slow. After a month or two, I started using the Volhard titration method which, in this case, was more accurate than fire assaying. The reason being was that the film samples had to be incinerated before they were fire assayed and losses occurred during the incineration. Since the lots were eventually mixed together during production, my assays were used to settle with the customers.

On the forum, people with ore are constantly told the get an assay. The people that say this are very uneducated about the real problems involved to do this with any accuracy. This it good advice, but no one tells the person how to properly sample the ore in order to get meaningful assays. Ore is extremely difficult to sample and entire books have been written on the subject. To do it right requires expensive equipment and a helluva a lot of knowledge. When a person pulls a grab sample of ore and has it analyzed, the only thing accomplished is he now knows, with fair accuracy, how much gold is in the sample baggie. That's it. As far as the gold content in the ore body is concerned, the grab sample is usually meaningless. Here again, statistics is commonly used when sampling and assaying orebodies. Also, to do it right, a ton of assays is usually necessary.
 
Since Chris mentioned coning and quartering as well, maybe I should go into a little more detail here.

Coning and quartering can be a useful tool in getting the material into a quasi consistent state. And for forum members who do not have the access to all the fancy sampling equipment the big boys have , all you need to cone and quarter is a shovel.

Basically if you had say 100 pounds of ground up material that contained your values, and you feel that all of the material is relatively the same size, you could use the technique to reduce the volume to a small enough sample to process in your own lab or to send out a sample for assay.

Start by shoveling the material out on a smooth flat floor, usually concrete. Every shovel of material is placed on top of the previous shovel full and in short order you will have a cone of material. Then the cone is smoothed out and shovels of the material are taken from all around the cone and again used to form another cone. This is done 2 or 3 times. What you have essentially done is mix the contents of the drum completely by shoveling it multiple times into new piles.

Next you flatten the pile again and make a line down the center and then a second line running the other way like a giant plus which divides the pile into 4 roughly equal sections. Now you carefully use the shovel to scoop out all of the contents of one of the 4 quarters you just made with the giant plus. This is shoveled into a new pile. Now you shovel the contents of the opposite quadrant on to the same pile you just started. So you will be left with a new cone which is half the size of the cone you started with and the 2 remaining quadrants of the quartered cone you started with. You can now shovel the 2 remaining quadrants back into the drum as you have effectively divided the original sample in half and eliminated any sampling bias by doing it this way.

You can now do the same exact process over again, and at the end you will have a pile which is representative of 25% of what you started with. And every time you do this same process in succession, you will have again halved the sample size.

On large scale it is possible to reduce a large lot to a manageable lot for processing with this method. I have witnessed large 20 plus drum lots processed and sampled this way to end up with a sample the size of a 5 gallon pail which is used to determine the value of the lot.

This is a You Tube video I found online showing a guy doing this on a lot of powders

https://www.youtube.com/watch?v=RkBgp-A5nX4

So how can a small guy use this methodology to produce a sample of a larger lot? Say I'm a smallish refiner and my capacity is a plastic lined 3 1/2 cubic foot mixer for acid digestions. My customer brings in a skid full of drums of 486 processors like Goran mentioned towards the end of the HYDRA SIEGE thread in the Data section. http://www.goldrefiningforum.com/phpBB3/viewtopic.php?f=33&t=24949&p=266993&hilit=random#p266993 You know you can make a profit but the guy wants to be paid up front. Since you don't want the guy living with you for a month, your only option is to sample.

Keeping in mind the problem Goran realized, the best solution would be to break up all of the processors into pieces that will pass through a 1/4" screen. In a perfect world breaking them to dust is even better but 1/4" pieces is attainable with a granulator pretty quickly. So you granulate the lot and screen it to 1/4" and now you have the material ready to cone and quarter.

You can effectively reduce this pile of granulated chips down to a manageable size which you can digest and recover the values from in your shop and have a high degree of confidence that what you processed (which you weighed before so you know the percentage of the total you have to base your payout on) will be recovered when you process the entire batch.

To gain confidence in the sampling method you can do this with 20 or 30 pounds of processors and reduce it down to 1 pound samples which you can process 2 separately to see how well they compare.

This process has been completed many many times by large refiners to determine just how close they can come to actual content by processing a representative sample. And a rather basic technique using nothing other than an everyday shovel can take you to the next level in evaluating the material you are processing.
 
Before Cone and Quartering, you should sweep the floor. Some people use a real canvas tarp. Generally, the idea behind cone and quartering, is to dump each shovelful in the exact same spot. Ideally, you want a cone to form, and you want the material to slide down the cone, equally all around the cone. For some stuff, you can mount a big cutoff plastic funnel and use it as a hopper to dump the material more perfectly.

In the video, which is quite good, he only went through 2 cycles. I've always split it down until I had a quart baggie about half full and then started all over with the full amount for the 2nd sample. For a 1# sample when starting with 100# lot, it will take about 8 splits. Then, when you get it in the lab, you can split it further if you want.
 
During the gold rush when karat was king I had a client who had over 50 drums of silver chlorides skimmed from a Miller process and he had to send them out to refine. He had no assays and the chunks were all dry and from football size down to dime size. They didn't even want to assay them. No time. Big profits back in those days!

I convinced them how insane this was and talked them into passing all of the material through a jaw crusher and pile it on the floor in this large warehouse. They had a Bobcat on site for something that was going on construction wise and I happen to be pretty good with a Bobcat. (Many years of four times a year cleaning out the horse corrals at home to consolidate the manure.)

It took me about 3 hours in that Cat to cone and quarter all of it and retain 3 pails of sample and ship the rest to separate refiners. It was crude but better than nothing. We also told the refiners that the lot was split and had been coned and quartered and sampled.

All the lots came back within 5% of each other. Unwitnessed.

In keeping with my motto, never ship a lot when you have no idea of assay if it will be unwitnessed.

I had asked them to mill and screen each pail and assay the powder and melt the oversize to see how closely each compared. They said they would, don't know if they did or just yessed me to death.

Sorry Chris I don't remember if we swept the floor! My bad!
 
goldsilverpro said:
On the forum, people with ore are constantly told the get an assay. The people that say this are very uneducated about the real problems involved to do this with any accuracy. This it good advice, but no one tells the person how to properly sample the ore in order to get meaningful assays. Ore is extremely difficult to sample and entire books have been written on the subject. To do it right requires expensive equipment and a helluva a lot of knowledge. When a person pulls a grab sample of ore and has it analyzed, the only thing accomplished is he now knows, with fair accuracy, how much gold is in the sample baggie. That's it. As far as the gold content in the ore body is concerned, the grab sample is usually meaningless. Here again, statistics is commonly used when sampling and assaying orebodies. Also, to do it right, a ton of assays is usually necessary.
The only way to know within a certain margin what an ore contains is to drill core samples of the ore body and record the position of each sample. That is the only way to know if an ore is going on or just ending close to the surface. Based on this knowledge you can then decide if it is worth setting up a mining operation.
This is probably prohibitively expensive for the small guy and a method reserved for the big mining companies.

But I don't agree that a single assay is worthless. Most people coming here have no clue and if they pick something that looks good and sends it for assay then he will know if it even contains any precious metals at all.

What a single assay can tell you is if you can recognize an ore rock and in some cases what composition an ore have, if there is arsenic or not for example.

The main reason to keep telling people to get an assay is that most of them haven't the faintest clue about what an assay is, and they are starting to mine just because it looks good and there was a gold mine there a hundred years ago.
If they get an assay then we can tell them how to make the next ten assays in a representative way... but I can't remember one case where a guy was told to get an assay and came back afterwards with the result.

The geologist in me gently reminds that it should actually be called a mineralization until you have analyzed it and found it to be economically viable to mine. It is at that point it becomes an ore. :wink:
https://en.wikipedia.org/wiki/Ore

Göran
 
Here are a couple videos that though not detailed &/or specific to sampling - they give a good general over view of how large companies sample/assay for customer pay out

http://www.jmrefining.com/video-page/

http://advchem.com/index.php/videos/19-videos/64-acc-incineration-process

Kurt
 
4metals said:
I started this by re reading an old text named The Sampling and Assay of the Precious Metals by E.A. Smith. Written in 1913, second edition in 1947 and reprinted in 1987 by Met-Chem Research.
I have that book in my digital library. It was scanned by Google Books. You can download a free copy at The Sampling and Assay of the Precious Metals by E.A. Smith.

Dave
 
FrugalRefiner said:
4metals said:
I started this by re reading an old text named The Sampling and Assay of the Precious Metals by E.A. Smith. Written in 1913, second edition in 1947 and reprinted in 1987 by Met-Chem Research.
I have that book in my digital library. It was scanned by Google Books. You can download a free copy at The Sampling and Assay of the Precious Metals by E.A. Smith.

Dave
For people outside of the US, google books is just a preview with snippets of the books, even free ones.

The book can be downloaded from archive.org in several different formats.
https://archive.org/details/samplingandassa00smitgoog

Göran
 
Google chrome does have few addons which can let you choose where your login appears to be from. If I see some video which is available for viewers of specific country only then I go and set it to that country, reload browser and I can see it then without problem. It should work for the books too.
 
Back
Top