My initial inclination was to stay clear of this one or perhaps be caustic
and say "good luck" It is a topic that gets me started with no perception
of when to stop! Many theses and chapters have been written on extracting
various S-compounds from plants, soil, peat, fossil fuels etc. We have
wrestled with this problem over the years and tried various approaches,
some of which could be detrimental to one's health.
One problem in removing inorganic-S particles is that they might be
resistant to dissolution or chemical reaction by possessing an organic
"skin". Does one address this with organic solvents, perhaps soxhlet
extraction or does one reason that if the sample is ground fine enough,
such skins can be mechanically broken (perhaps use ultrasonic agitation in
a salt solution for sulfate extraction)  But then you must worry whether
some of the organic-S might become oxidized or volatilized in such a
process ( perhaps grinding the sample in liquid nitrogen is an answer) .
One might also consider  reducing agents.. One can read how Kiba reagent (
Sn++ in concentrated H3PO4) was successfully
used to quantitatively convert mineral sulfides, including FeS2 and sulfate
to H2S in coal specimens without apparently attacking organic-S. But then,
one finds an apparent inconsistency since Akira Sasaki was able to convert
organic-S in hair quantitatively to H2S using Kiba reagent. The
complication is that  organic-S comprises different S-functional groups
which vary with the nature of the specimen.  Sulfhydrl, disulfide,
thiophene, etc are quite different in their chemical reactivity, behavior
during solvent extraction, etc
An approach which we have developed with fossil fuels is thermally
programmed reduction (initially, we used pyrolysis). The specimen is heated
while a 20-1 He-H2 mixture is passed through it. H2S is evolved as "peaks"
which we believe originate from different S-functional groups. Aliquots of
the evolved H2S are converted to SO2 for S-isotope analyses. In earlier
times, we had to place grams of the specimen in the reactor to obtain mg
quantities of S for isotopic analyses. We were also concerned that in a few
cm of vertical sample height, secondary reactions generated peaks which
were not in the original sample. Thanks to our
CF-C-GC-IRMS system, we can now analyze a few ug of sulfur and the sample
can be a relatively thin layer in the reactor. In principle, the
temp-programmed reduction can be also used to analyse inorganic-S and/or
organic-S.
One drawback to this procedure is that volatile S-containing compounds
might escape from the reactor before the chemical reduction of the S-groups
commences. Hence, bitumen, heavy oil, coal,etc have been processed very
effectively. I would like to see this approach expanded to deal with a
wider variety of samples and would appreciate suggestions. For fossil
fuels, we are trying to integrate it directly with a CF-IRMS rather than
precipitating the H2S as AG2S which is then converted to SO2.
So how does one address the current question?  Bernard Mayer described an
approach which is as feasible as any. In any case, I think that one must
decide on a protocol, try to become aware of possible complications, and
importantly when publishing, try to describe thoroughly what was done even
under pressures to omit this information to conserve space. As a further
challenge, one could try 2 or more approaches to see how well (if?) they
agree.
Roy Krouse
.