Effect > Normalize: [X] Remove DC Offset... and nothing else.
Logically sheering off the bottom of the frequency response should work a treat, but it doesn't. It leaves a pop or spike at the beginning and the end. Audacity Normalize uses a statistical averaging technique without affecting the (audible) response, and it does it cleanly over the whole performance.
We beat this up extensively a while back and over time, we discovered why the developers did it his way. Yes, it is possible to present a very badly damaged performance and fake out the process, but it's unlikely a performance that badly damaged is to be rescued anyway and over the majority of productions and shows, it works very well.
Feel free to High Pass Filter anyway if you think you need it