fine frequency error estimate

[jmfriedt]

The FFT on the squared signal provides a rough estimate of the frequency offset while phase slope should provide a fine frequency offset estimate.
1/ the squared signal exhibits many additional spectral features beyond the expected 0-\pi at 2.5 Mchips/s. Where do these additional spectral components come from?
2/ phase unwrapping does not continuously connect rough frequency estimates but still exhibit some "jumps" from one coarse estimate to the next.
3/ why is the phase of the correlation not rotating at the rate of the fine frequency offset?