I'm not quite sure if you're asking about the method or the content of evals, or both.
I've only used online evals like e-value. They're pretty good, but I'm sure cheaper programs like survey monkey will work just as well, just minus the security/confidentiality precautions maybe.
As far as content, I think it needs to be appropriate to the EM rotation, not just a bunch of generic questions like, "how would you rate teaching on rounds?" Let me know if you are interested in hearing what I think would be good questions.
A mix of Likert for quant and comment for qual options are good for each question you want answered.
For resident/faculty evals, I think it is important to emphasize that the responses will be collated and shown to the individuals at a later time, to maximize confidentiality.
I also agree w/ DocVirk, emphasize that comments should be constructive not a b****-out.
Also, if you want prompt responses, the longer the eval the less likely that will happen.
Sorry if any of this was basic, hope it helps!