To clear up the issues with licensing and LAME - I found this via the cwenc faq(
http://cwenc.sourceforge.net/faq.html ):
"Q: What about MP3 licensing? I have read that you need to have a license from Fraunhofer/Thomson if you use an MP3 encoder?
A: Strictly speaking this is correct. According to the licensing conditions put forward by Thomson (who owns the MP3 patents), anyone who develops software with MP3 encoding or decoding must purchase a license for each copy of the software sold. But for free software they have unofficialy accepted that no such fee is required."
--
Thus, it's OKAY to use the free LAME encoder, and you're not going to get pegged by Thomson for it. Cakewalk's implementation of the LAME encoder is a pay-for product probably mostly because 1) Sonar isn't free and 2) Thomson could change their license down the road. File that under "cross that bridge when it happens" as far as I'm concerned.
The big problem I have with Cakewalk's LAME encoder version is that they're obscurring the settings that you get if you download it and set it up as an external encoder. The people who use that one get better results, it seems, and I think it's worth a try. Definitely use Variable Bit Rate, as it will help compensate for quality problems, especially at lower bit rates. Also - I hope you're not sending the encoder a stream that peaks at 0db or 100% - mp3 will introduce some artifacts if there's no headroom. Better to compress/limit the program content a bit for the podcast and encode it when the max level is somewhere around -3db give or take. You may also want to make sure you don't have any dc offset, extreme low frequencies, etc, as these can rob some fidelity. Better still would be to do external ogg's but I guess that's out for podcasting. Ogg-Vorbis is well known to be better than mp3 at basically any bitrate, but it's more audible with lower bitrates, obviously.
pete