For this experiment (conversion rate of varying sucrose concentrations to microbial cellulose), we could easily tolerate a few percent error in accuracy — but we needed pretty high precision to distinguish between individual samples because the total yields are relatively small.

I’m kind of obsessed now with designing an open-source analytical balance that is compact, elegant, precise and easy to build — probably a hybrid with a micro-controller and a galvanometer (the galvanometer can be used for active damping and also pick off the milligrams portion of the measurement).

I would be cautious about using coins as reference weights. I’ve weighed 3 different pennies and found 3 different weights, differing by a couple tenths of a gram. I assume all the coins from a given year / style are pretty similar to each other.