I'm not entirely sure you correctly understood Brian's point...he's referring to "backing off" the Tx gain to ensure that the RF power amp remains entirely linear throughout the power envelope of the OFDM signal. The PAPR is directly proportional to the number of carriers employed and is one of the problematic aspects of OFDM. Most USRP daughter cards (or any radios) tend to loose linearity as they approach the top end of there gain range. The GMSK signal however being FSK based has a constellation that remains on the unit circle and hence has constant amplitude allowing it to operate well in the non-linear region of an RF amplifier (hence it's popularity in low cost, high efficiency single chip radios). It's easy therefore to see why empirically it might appear that GMSK is "out performing" OFDM here, given that this is a simple PHY demonstration, rather than an application providing a robust link with appropriate channel coding. Tom has addressed similar questions as yours in the past here, I suggest you search the archive, I think you'll find more useful information.
On May 17, 2012, at 9:00 PM, Alex Zhang wrote:
What's the EVM of your OFDM signal coming out of the RF daughterboard?
Are you backing off enough to allow for OFDM's high PAPR to remain in
the linear region?
I think backoff time is enough, as I actually did not see the conflict occurs.
GMSK has the nice advantage of being constant envelope, so compression
doesn't matter. OFDM, on the other hand, doesn't get that luxury and,
in fact, requires a significant backoff.