[time-nuts] 10 MHz over optical fiber?
magnus at rubidium.dyndns.org
Fri Nov 28 01:50:30 UTC 2008
Tom Van Baak skrev:
>> Yes. I know of several commecial systems. If you only need to do a short
>> jump, then using fairly basic E/O-O/E equipment should work well
>> enought. It all depends if you want/can to roll your own or need to buy
>> a finished product (aka "buy this, and you will be fine!").
> Magnus, what's the typical noise floor, tempco or drift of cheap
> (i.e., non JPL-level) fiber distribution systems like this? Is it less
> than regular coax, or phase stabilized heliax? At 100 m lengths?
The temperature stability is fairly good, but what Bruce posting didn't
expose is the temperature stability of the laser. The problem is that
the laser changes frequency (not wavelength as the optics people say, as
that depends on the refraction index of the material) with temperature.
This is a fairly dramatic effect over distances. The fiber delay changes
by three factors: thermal expansion of fiber (lengthwise), change of
wave-solution as dielectric and physical expansion occurs, and
wavelength of light. The lasers for WDM networks typically has a peltier
cooler for stabilizing temperature.
Choice of fiber and laser is also important. Multi-mode fiber is not
good at all since the dispersion provided by the many different
wavepaths in the fiber will shift in a not so forgiving fashion. For
shorter runs like 100 m and with not too high requirements it will be
fine, but tune up your requirements and you want to go single-mode.
The laser-type is also important. Cheap lasers exhibit a huge number of
frequencies and when looking at an optical spectrum analyzer it can be
hard to decide which has most energy... there is typcially a shape but
the spikes are so many and close that you can't say point out with
certaintly which will be the strongest peak... it will shift. Also, the
dispersion that is gained from the aggregate of those peaks is not nice.
WDM lasers like those for 1550 nm has a much cleaner spectrum which
will gain you alot in the dispersion field.
Long distance communication is very dispersion-sensitive and both
chromatic and non-chromatic dispersion is being considered. While
inter-symbol interference as such is not that important to a pure sine
distribution, the stability aspect of dispersion is.
If you want optimum performance for longer taus, there is really no
option but to do two-way time transfer compensation. Typically the
process is fairly slow on properly laid fiber. For a pure frequency
transport, just looping the signal back to the source and do phase
measure and phase-adjust so that the sent phase compensate the phase
error will work. This way you can continuously compensate the variations
out. Such systems is used for instance in DSN and similar systems with
multiple arrays of antennas at microwave frequencies.
Anyway, I beleive I had the sum of errors become about 85 ps/nm/K/km.
Could be wrong, it's just a number stuck in my head.
Does this effect occur in real life? Well, naturally I have an anecdote
to share on that. In Sweden the national power company had put fibers on
their high voltage cables running through the country. They needed it
for their internal use, but could of course provide service to others,
which they do. They have a SDH network on top of that and can then
provide SDH based services, alongside WDM channels. However, about 7 in
the Monday mornings in February their SDH network failed as the sync
went totally bananas. They called in a synchronisation expert which I
happend to know. He measured the wander and picked up a very nice
deviation with a huge swing right there in the morning hours. What
happend? Well, what happens 7 in the Monday mornings which is
exceptional? That's when they start up all the heavy machinery in the
industry, they increase heating since it is cold, people wake up and
turn TVs on etc. etc. etc. Many of the things which have been off or
turned down during the weekend is starter up. This makes the current go
up, this makes the power cables go warmer (they will expand quite
noticably!) and the fiber being pulled on the cable will change
temperature. As the heater and fiber run over a considerable length
together, many 100s of km, and the shift in temperature is several C
then it is not strange that they experienced a huge shift in delay.
Having heard my comments on this, the synchronisation expert pointed
this out. They ended up to move the fibers to the relatively unloaded
ground wire at the top, and have a much more quieter system.
This shows that uncompensated delay even in a good fiber system can eat
you for lunch.
The depth by which you have buried your fibers is another issue. Deeper
means higher stability in temperature, as mother wander (the sun) will
contribute alot at tau = 43,2 ks.
For shorter runs, one way transfer can work well. The requirements on
stability may be an issue if you have high requirements. If you do a 10
m run, few would have a problem on an uncompensated system. Off the
shelf components could probably satisfy most needs even for 100 m or so.
Interestingly enought, there is a Siemens patent for a fiber based
oscillator. The short term stability of silica fiber is actually quite
good. It is the long term aspects which can eat you out if you do not
deal with it.
Consider also things like laser gyroscopes. The stability of the fiber
is sufficient to detect even small shifts frequency due to Sagnac effect.
More information about the time-nuts