Automatic NTP converges to wrong frequency on 20m

[zapmaker]

Building wspr from source on May 29, 2019 on Model B V2.0 hardware 000f (512MB) shows NTP converging but external frequency counter for test tone 14e6 shows about 14000370 Hz after convergence, stablizing to -30.34 ppm. The pi had been running about 15 minutes at this point. This frequency prevented detection by wsprnet receivers. I was able to use manual adjustment and success using this: sudo wspr--test-tone 14e6 -f -p -56.21 (I'm sure the value will be different for everyone).

Note: After letting it the pi run over 2 hours and running sudo wspr --test-tone 14e6 the NTP converged very close, bouncing between -53 and -54 ppm. I don't know this issue occurred and then resolved because of the earlier manual ppm adjustment or because the convergence needed a lot longer time (pi crystal needed to settle?), or I'm running older hardware.