Take a look at the Lee Perfect Powder Measure.
It gets no respect, because it's plastic and it only costs about $25. I ran some tests on it a few years ago, and with ball powder the standard deviation of the random error is around .04 grains. With 4831SC, it's around .1 grain. I didn't do any tests with the old long cut 4831.
It beat out the Hornady powder measure and the Lyman digital scale for repeatability. It's a surprisingly good little powder measure.
Another vote for the Lee, only measure on my bench, works so well I've never thought of "upgrading" to a less precise measure...although Lee is making a new version of this design made of metal now...