James Holton wrote:
> In general, a Bijvoet ratio of 3% or so is needed to solve a structure (the current world record is 0.5% and lots of multiplicity). The above web page will also tell you how many crystals you need if you type in their size in all three dimensions. but this estimate assumes that you don't have high concentrations of heavy metals in your solution! So, if it says you can get away with one crystal but you know your have a dose-doubling concentration of something, then you're going to need to average data from two crystals, etc.
I don't know of any systematic studies, but in my experience if you've
got weak anomalous signal you're better off with multi-crystal phasing
than multi-crystal merging.