(PS. Freemium users are more difficult then premium users in most cases, they want a lot for nothing :))
I recently implemented this strategy for my web service. Only about 1% of registered users and 30% of active users donated but the average donor contribution was three times of what I was estimating the price of the service.
It also helped me prioritize which user requests should bubble up quickly. My plan is now to introduce donor exclusive features. Then I want to track changes in number of donors and contribution.
Are you building this for users or customers? If customers, why would you want to test and develop it against people who aren't your customer?
On my current project (https://report4me.com) we're trying the free beta model not only to get beta testers but also to determine a price point.
We give away the product to beta users for the foreseeable future, with the intent to ask them outright what they feel the product would be worth. The idea is that since they can feel comfortable that they won't actually have to pay the price they mention they might actually be more honest about it.