Re: [time-nuts] Tbolt issues
Bert
a) 1e-10 freq error, Sounds to me like you have a typical TBolt with near
factory default setting.
The most important thing to get good Tbolt frequency performance is the
antenna, with good sky view and correct location setting.
After that there are some 'basic' Tbolt setting and things that can be done
that will get it close to the best, within say 2 to one.
Other comments in text below and attached graph.
Bert Kehren wrote in part:
would you mind sharing 1 second frequency data that you get out of the
tbolt to get an idea what is possible.
Looks like we are not the only ones trying to improve frequency performance
and hopefully some will share settings.
For us it is absolute Frequency, to me it is a measure of true
performance.
a)Attached is data plotted from a modified Tbolt that has a 1 sec ADEV of
1e-12.
This shows that its 100ms "instantaneous" freq error varies from 1 to 3e-11.
As long the frequency counter is not resolution limited, the 0.1sec, 1sec,
10sec, and 100 sec sampled plots all had about the same peak to peak freq
error just with less high freq fuzz,
Tbolt is an excellent time device but not good for frequency reference
past 1E-10 because the frequency is constantly changed to correct time.
a) The Tbolts damping setting is what controls how much 'freq noise' is
added to correct for time error (i.e. Phase error).
You can set it to optimize whatever you want.
Many time-nuts use Phase error, not frequency error as the indicator of
performance, so I tend to set the damping low to minimize the phase error.
When the gain is set correct, damping of 0.707 gives ideal critical damped
phase error correction response.
A damping of 1.2 gives a nice compromise for frequency response correction.
A damping of 0.7 adds ~25% freq overshoot noise and gives you the lowest
phase error.
A damping of 1.0 adds ~10% freq overshoot noise and any time/phase error
takes about 3 times longer to correct.
A damping of 1.25 adds <5% freq overshoot noise and the time error takes ~6
times longer to correct.
With a damping setting of 2, less that 1% of freq noise is added to correct
for time errors and phase errors takes >>10 longer to correct.
With a damping of >=10, the time correcting is so slow that time/phase
correction can take days, and there is no added freq noise.
With the popularity of the Tbolt an analog or digital clean up loop would
make sense.
My Swiss partner Juerg has relied on an OSA F3 for Tbolt clean up but has
had continuous bad results .
The result is that the OSA F3 does not clean up the Tbolt and we see
+-4E-11 changes and old data shows even some +-8E-11 excursions.
a) The Tbolt control loop is already a clean up controller, it is cleaning
up the noisy GPS freq signal.
What I have found is that with a properly setup TBolt, an addition clean up
Osc does not help to give better low frequency stability because the peak to
peak noise output is pretty much constant whether the cleanup bandwidth is
set at 0.1sec, 1sec, 10sec, 30sec or 100 sec.
And of course you'd need a cleanup oscillator that is more stable that the
Tbolt over the cleanup time period.
If you have the low noise clean up oscillator, what works better is let the
Tbolt discipline it directly as an external oscillator.
ws
Warren wrote:
The most important thing to get good Tbolt frequency performance is the
antenna, with good sky view and correct location setting.
Agreed. Nothing will go right if the antenna, sky view, and surveyed
location are not as good as you can get them.
The Tbolts damping setting is what controls how much 'freq noise' is
added to correct for time error (i.e. Phase error). You can set it to
optimize whatever you want.A damping of 0.7 adds ~25% freq overshoot
noise and gives you the lowest phase error.
Meaning, it corrects the PPS position as quickly as practicable but
horses the frequency around quite a bit to do it and overshoots before
settling.
A damping of 1.0 adds ~10% freq overshoot noise and any time/phase error
takes about 3 times longer to correct.
A damping of 1.25 adds <5% freq overshoot noise and the time error takes ~6
times longer to correct.
With a damping setting of 2, less that 1% of freq noise is added to correct
for time errors and phase errors takes >>10 longer to correct.
With a damping of >=10, the time correcting is so slow that time/phase
correction can take days, and there is no added freq noise.
These are good rules of thumb. Someone who is interested solely in a
frequency reference is well advised to set the damping >>2. I think I
ended up around 6-12. There was a minor improvement if I increased it
to 50, but then recovery from holdover took longer than I was prepared
to wait.
Speaking of which -- recovery from holdover will be slow when a Tbolt is
tuned this way, so do everything you can to speed it up. Allow "jam"
setting of the PPS when the error reaches, say, 65-75nS during recovery,
and also allow quite a lot of frequency error during recovery (IIRC, you
can allow the Tbolt to vary the oscillator up to parts in 10e-9 during
recovery). Then, DO NOT USE THE 10MHz OUTPUT WHEN THE TBOLT IS IN
RECOVERY. Wait until it is fully recovered.
And again (back to the first item), make sure everything about your
antenna system and location is in perfect order, so it will only very
rarely go into holdover. ALSO, make sure that the Tbolt sees a nice
thermal environment, either by actively controlling the temperature
surrounding the Tbolt itself, or by isolating it from ambient so the
oven can always keep up with any changes. The latter is my preferred
technique -- see old list messages discussing "cast aluminum boxes",
"thermal mass," "thermal inertia," and "thermal capacitance." But there
is nothing wrong with active control, either, and LH can do that for you
with a few external parts.
Best regards,
Charles