[time-nuts] Advanced TBolt tuning using LH
WarrenS
warrensjmail-one at yahoo.com
Sat Jul 14 20:16:04 UTC 2012
Lady Heather is great tool to use for setting up a TBolt to make it much
better than it's default settings.
I've also learned several other tuning tricks to further improve a TBolt's
performance, many of which would only be of interest to the extreme TBolt
Nut.
For example, The LadyHeather plot at:
http://www.febo.com/pipermail/time-nuts/attachments/20120712/03b0a3f0/attachment-0001.gif
shows a well set up TBolt, but it can be made even better with some advanced
fine tweaking of its settings.
Note before being concerned with any of the following advanced tuning items,
It is important that all the things LH can set are done first, like the need
to have a good position set and ... and ... and ...
0) I wonder why the "sat count" plot is turned off. I find that plot to be a
very useful tool.
For one thing, it makes sure the setting are never causing the TB to be
using just one signal and therefore in danger of going into holdover.
1) Lowering the elevation mask from 39 to more like 15 to 25 will increase
the number of satellites used.
This tends to reduce the phase noise a little, and it especially reduces
some small spikes in the phase plot that are caused when satellites are
switched in and out.
My goal is to keep a min of 4 or 5 satellites in use at all times.
I'm guessing if the "sat count" plot was visible, it would show as few as
two satellites are being use at some points,
and that some of the short phase spikes are happening when switching between
say 2 to 3 or 3 to2 satellites.
2) Along with #1, the AMU should be raised from 1.0 to more like 6 to 7.x
to lock out any signal below ~38 dBc
This will reduce the "bad satellite signals" that lower the elevation mask
tends to let in.
3) To reduce that plots ~5ns daily phase ripple which is caused by the
control loop trying to keep up with the daily DAC ripple, reduce the damping
setting from 1.0 to 0.7.
That will speed up the phase correction time by about two to one without
effecting the control loop's TC and allow the control loop to change the DAC
voltage faster as necessary without causing as much phase error to occur.
4) That TB has LH's active temperature controller holding the sensor
temperature constant,
but to further improve the TB's sensitivity to room temperature changes,
which in turn will reduce the ~150uV daily DAC ripple cycles, which in turn
will further reduce the phase variations.
There are additional things that can be done.
a) Lower the room temperature variations at the Tbolt, OR
b) change the relative effect the cooling fan has on the OCXO vs. sensor
location by changing the direction of the air flow or adding a little more
or less heat sink / insulator to one side or the other of the TB's case.
5) Increasing the TC from 500 to 999 will cut the higher freq short term DAC
noise in half, which has a direct effect on the OXCO noise below the TC
frequency.
I find that using the extended TC method with a TC of 1500 to 2000 works
best after doing 1 - 4 above.
Also, increasing the TC can help because the TB will not follow the GPS's hr
to hr noise variations so closely.
But to see the positive effect of that, an external reference source like a
high stable CS or RB is needed.
All LH can do is tell you how close the TB's OCXO is following the GPS.
The problem is that the GPS's ~10ns PP phase noise can still be pretty
noisy at 1000 seconds compared to a good OXCO in a great environment.
6) The most obvious, and maybe the lest possible thing to do is to get rid
of the South West signal blockage or reflection.
Looks like some sort of tall narrow close in obstruction.
If the antenna can not be moved or raised to avoid the obstruction, maybe a
chain saw would help, or a tree trimmer?
Maybe some C4?
7) I find that turning on LH's display filter to about 100 second makes LH's
plots more closely match the external real world data coming out of the TB.
8) I like to see the Osc plot turned on with its trend line visible, but
with its gain turned down and its offset set so the plot is near the top of
the graph and out of the way from the other plots. Sometimes the Osc plot
will show things that I miss in the other plots.
9) I like to use view times set in Hours, i.e. 360, 480, 720 and 1440, and
make the max data saved time longer than the 3 days default.
I use 30 days, to better see which things are daily cycles, and to be able
to view backward into the past.
This is especially important when comparing the long term effect when fine
tuning the things above.
Isn't it fun being nuts with too much spare time?
ws
*****************
More information about the Time-nuts_lists.febo.com
mailing list