Larry,
The upper size limit is somewhat arbitrary. Quartz #4 is weak enough to break with finger pressure. I think quartz larger than #4 will be worse in strength following the basic textbook trend for fracture toughness as particles get larger. I could not break the #4 aluminum oxide walter sent me with finger pressure. Once the rocks get much bigger than number 4, I suspect that it will be difficult to mix without pretty large scale equipment and that this will affect us more than strength considerations.
I've done a lot of thinking about aggregate for home use and this is what I've realized might be unclear about the subject:
A single component mixture of angular grained sand has about 45% empty space in it at maximum reasonable compaction. This means that 40%-50% epoxy by volume has to be used to get a void free solid!
For perfectly graded mixtures of angular aggregate with a given number of components, the following table shows the minimum empty space possible in the mixture which governs the minimum amount of epoxy that must be used to fill the voids. For actual gradings, these numbers are worse. (From the paper by gupta et al., we know that more epoxy than the amount of empty space is required to obtain maximum strength in the part.)
<PRE>
Number of Components and Empty Space in a Perfectly Graded Mixture of Angular Aggregate using equation 1.68 in de Larrard's book
# of Mixture components | %empty space
i=1, pmin=0.451000
i=2, pmin=0.250910
i=3, pmin=0.159661
i=4, pmin=0.111375
i=5, pmin=0.082987
i=6, pmin=0.064943
i=7, pmin=0.052769
i=8, pmin=0.044158
i=9, pmin=0.037832
</PRE>
The predicted empty space for the mixture Walter used in the gorgeous part in <A href="http://www.cnczone.com/forums/showpost.php?p=328458&postcount=1893"> post 1893</A> was around 12% which allowed him to use about 20% epoxy by volume as I understand it.
So to answer Larry's question, you can use any source of local materials if you can roughly follow the grading outlined in <A href="http://www.cnczone.com/forums/showpost.php?p=320419&postcount=1824">post 1824 </A>.
In the context of de Larrard's Compressible Packing Model, there is no such thing as a premium filler, only fillers that are more expensive. Each of the aggregate grades has a job to do and must be present in roughly the correct proportion. A mixture of 80% "premium filler" and 20% epoxy will not perform like a mixture of 10% "premium filler" 70% "non-premium filler" and 20% epoxy.
It's quite possible that there is a sandblast sand that is easily available that could serve as basic raw material though I have not worked hard at identifying one. The best candidate I have seen is DuPont Starblast Ultra. It has to be the Ultra grade though as the other grade's distribution is not really what we might like.
One must obtain aggregate from somewhere and for best results, aggregate on the right grading curve seems to work better according to Walter and the models. In buying either graded aggregate mixed to follow the curve approximately or sieved aggregate designed to follow the curve exactly, one has to remember that it is the grading that you are paying for, not the material. While both may be silicon dioxide, only the graded one will pack to high filler density according to the model.
Like I said before, the greatest gain to be had is in using approximately equal parts of two components of different sizes rather than using just one component. Finally, masochists could always call up http://www.dualmfg.com/ or http://www.retsch.com/ or http://www.ascscientific.com and order himself/herself a set of small sieves and sieve it themselves.
Gotta mow the lawn and hammer drill concrete out in the 100 degree sun.
--Cameron