I tried the "official" calibration routine, but turned out to be VERY rough.
Two reasons for this:
a) Zero beating with loudspeaker or earphones requires that these two transducers can reproduce split-Hz tones, plus your ears can detect them.
b) The offered by the firmware calibration values are very wide.
What did I do after this:
1) Selected by ear two calibration values that sounded as the "best" pair in regards to zero beating.
2) Used them both and the WSJT-x software, in calibration mode, to find the two offsets (one positive the other negative if you have done a good zero beating job)
3) With the aid of the CEC manager, read the internal long calibration parameter for both cases.
4) Interpolated parameters of 3) with offsets from 2) and wrote (i.e. programmed) the new value a bit rounded, as it doesn't make any sense to seek more accuracy that the master clock's stability allows.