This is, I think, the number 1 application for A/B testing that makes it worthwhile, despite so many people loudly screaming that it's worthless.
The simplest suggestion I can make is to test the living daylights out of pricing. Set your base at something reasonable -- definitely more than your rock bottom price, and should be well above break-even. If you can scale your fixed costs and estimate the average cost per user (bandwidth, disk space, etc.) then you can apply a margin. Make sure you're covering overhead (which is insanely complicated if it's Freemium), and then play around.
Offer the 'base' price as a special deal, while testing out prices all over the place. Double that price, and set it as the 'non-discounted' deal to some users. Someone else recommended recently (I can't remember who, sorry) that if you want to be equitable, you can double the price on the sign-in page, then if they bite on that, only CHARGE them what you're charging as the base price -- but if you track the results on that signup page, you'll know how many people were willing to pay double.
Lastly, as a piece of obvious advice, if you can at all help it, don't raise the prices on existing users. Grandfather them in unless you absolutely can't, and if you can't, figure out some way to take really good care of them.