We've tried both and for the last decade we've exclusively employed usage based trials.
The reason usage based trials convert much better is that a large number of trials are explored in one period of time (the 'research' stage) and actually demoed at a later time (the 'trial' stage). When a user does research they most often look over your product offering, download it, and at least launch it momentarily. What happens very often is that when the user returns to the trial stage days, weeks or sometimes even months later, time based trials are often expired or nearly expired. Expired trials most often go in the trash without contacting the company.
Usage based evals allow for the delay between research and trial with no penalty at all. A greater number of evaluators equals greater conversions to sale.
We have tried some different strategies: Started with a 60 day free trial and found that was too long to encourage decision making and use of the trial. We reduced this to 30 days free so we get a quicker decision on the subscription and can communicate with users during that time to answer questions and suggest different ways of using the system.