Hello Jörg
Nice to see all your measurements. Especially the long term.
Seems that your BG7TBL GPSDO is better in hold than in run
In your reply #596 April 13 you lost "locked" and my guess would be that the OCXO made a jump of a few tenths of a ppb. I have seen jumps after days, after months and even more frequent on my OCXO's bought from eBay. I have one MV102 OCXO that is so bad that I don't get a useful ADEV graph between jumps. Of course I have some OCXO that I haven't seen a jump on (maybe have run them to little
)
What is more interesting with your graph is the step wise frequency changes. For me it seems that the BG7TBL GPSDO you have has an FLL. Like the first BG7TBL GPSDO's. So it counts the frequency for about a minute and when corrects. If the error gets smaller it counts longer and make smaller steps. As has been pointed out earlier in the thread and by KE5FX tests, FLL is not the optimum for a GPSDO. This is also the reason you even after 40000seconds averaging has only about 10 digits accuracy.
Even a very simple GPSDO like mine (
https://www.eevblog.com/forum/projects/lars-diy-gpsdo-with-arduino-and-1ns-resolution-tic/?all )
has a integral term that normally hold the output phase within 20-50ns (even with a DOT050V TCXO!). 40ns in 40000second should be 1E-12 so 12 digits.
Making a PI-loop is not difficult. A brief explanation is on page 8 in my instruction. On page 15 I have several good links to GPSDO control loops also. But it seems BG7TBL hasn't tried for several years.
Another thought is why nobody (except KE5FX) do time interval measurements between the BG7TBL and other professional GPSDO's with their excellent counters. I really like timelab by KE5FX. Also if the 1PPS comes from the GPS module and not is a divided down 10MHz it would be interesting to see if the phase (time) goes away. 0.1ppb would be 4us in 40000seconds. If I remember correct KE5FX showed this.
Lars