[time-nuts] Tbolt instability

[John Green]

Tbolts seem to have a very good reputation for accuracy and stability. Most
comments I have seen compare them favorably to the HP/Symmetricom Z3801.
Mine may be OK in the very long term but doesn't look too good in the short
term. Let me explain. First, I'll describe my set up. I have a Z3801 which
has been running continously for at least 6 weeks. The 10 MHz output feeds a
resistive 2 way splitter with one output going to the external timebase
input of a Racal Dana 1992. The other goes to input B of said 1992. The 10
MHz output of the Tbolt goes to input A. What I am seeing, this is after a
20,000 point self survey and being powered on overnight, is that the time
difference between the Tbolt and Z3801 varies by as much as 10 nS over a
period of a couple of minutes. Sometimes, it moves 4 or 5 nS in a few
seconds. I don't believe this is either the 1992 or Z3801 because when I
connect the Z3801 to A and B through different length cables, I see a
constant, non changing time difference. I also don't believe it is the Z3801
because I have compared several TCXOs and one Rubidium to it in the past few
weeks. Without exception, the time difference moved, either up or down, in a
constant fashion. In the case of the Rubidium, it would slip a nS in just
under an hour. This was always constant and in the same direction with an
occasional dither of 1 nS. I can't believe this is normal. I would assume
that two GPS receivers fed from the same antenna would see the same
satellites and if they had an error, it would be the same error so no
difference in timing between the two signals. It seems to me that the way
the two receivers discipline their respective oscillators is responsible.
The Tbolt adjusts quickly, while the Z3801 adjusts very slowly. Lady Heather
is currently showing an ADEV  of 1.5 x 10-11. This doesn't seem all that
good to me considering that the receiver has been on for 16 hours. Shoud I
give it a couple of days and try again? I had plans to use the Tbolt instead
of the Z3801 as a cal standard here at work but I am now inclined to use the
LPRO and cal it against the Z3801 maybe twice a year. Is there a way I can
improve the Tbolt?
Sorry for the lengthy post but I felt I should explain everything as
completely as possible.