Hi everyone,

In message 20081124152247.DCDB0E91529@mail.ebirds.it, Marco IK1ODO -2
writes:

I'm looking into something similar: transmitting an H-Maser signal
(probably 10MHz) over some 34km using CWDM SFPs. At first glance this
seems fairly uncomplicated: get some SFPs, and SFP connector + cage. Use
a fast opamp/differential driver to drive the transmitting SFP, and use
a similar setup at the other end to transform the received data back to
50 ohm unbalanced. How feasible would such a setup be?
Possible problems might be that a 10MHz squarewave is simply too 'slow'
to be transmitted by an SFP, which expects 1.25Gb/s 8/10 encoded data.
Another interesting question would be how much jitter/noise such a setup
would add?

Regards, Paul Boven.

In looking at the off-the-shelf boxes, jitter seemed to vary from ps range
to tens of ns.  Also, how stable is the 34km of fiber...  One of the
manufacturers I looked at had a 1550nm option, so it's probably not a stretch
to get it on a 100GHz ITU grid channel or other CWDM channel.

Scott

Paul Boven wrote:

Presumably, you're using a maser because you need good stability over significant time spans (seconds or greater)? Then you probably need some way to compensate for the changes in propagation through the fiber.  That's what the JPL boxes do. They send signals both ways (I don't have it in front of me, but it's possible it's on the same fiber, or at the least, two fibers in close contact, so they're in the same environment) and measure the round trip time.

By the time you get the signal from the maser through 34 km of fiber optic,
what you get at the end might be not better than a decent GPSDO...

Make sure you go through your drift and jitter budget before commiting.

Didier

Hi Didier, everyone,

Didier wrote:

That's exactly what I'm trying to work out before I'll actually start
building something like that. The JPL device (thanks for the hint, James
P. Lux) is very interesting, see:
http://tycho.usno.navy.mil/ptti/1988/Vol%2020_02.pdf

The JPL system aims to achieve better than 1E17 in 1000s, whereas the
H-maser in their publication is quoted as 1E15 in the same averaging
time. They also state that: 'when stabilities higher than 1E15 are
required the link must be actively stabilized'. So, in a nutshell - If I
were to build such a system, the first incarnation would be without such
a compensation system.
If such a compensation does turn out to be necessary I'd probably not
use a 50% mirror but simply loop the received data back to the
trasmitter of the SFP at the receiving end, and use a fiber pair instead
of a single fiber.

Regards, Paul Boven.

All:  Sorry for jumping in abruptly on this ongoing thread.  I have been meaning to chime in ever since the discussion got started.
 
I apologize in advance if I am dragging up an issue that has already been discussed and put to rest.
 
That being said, have you given consideration to the effects of chromatic dispersion in the fiber?
 
A good, telcom grade, dispersion shifted (designed for the 1550nm wavelength band, where absorption of the light by water is small), single mode fiber has a chromatic dispersion of approximately 1 to 10 ps/nm/km.  The concern therefore is changes in timing induced by changing wavelength content of the laser or LED used to drive the fiber - for example, due to temperature changes, etc.
 
If you've already considered this, then please ignore this.  Otherwise, it would be interesting to work through a "what if"?
 
jeff

Jeffrey K. Okamitsu, PhD, MBA
+1-609-638-5402

--- On Wed, 11/26/08, Paul Boven p.boven@xs4all.nl wrote:

From: Paul Boven p.boven@xs4all.nl
Subject: Re: [time-nuts] 10 MHz over optical fiber?
To: "Discussion of precise time and frequency measurement" time-nuts@febo.com
Date: Wednesday, November 26, 2008, 12:02 PM

Hi Didier, everyone,

Didier wrote:

optic,

That's exactly what I'm trying to work out before I'll actually
start
building something like that. The JPL device (thanks for the hint, James
P. Lux) is very interesting, see:
http://tycho.usno.navy.mil/ptti/1988/Vol%2020_02.pdf

The JPL system aims to achieve better than 1E17 in 1000s, whereas the
H-maser in their publication is quoted as 1E15 in the same averaging
time. They also state that: 'when stabilities higher than 1E15 are
required the link must be actively stabilized'. So, in a nutshell - If I
were to build such a system, the first incarnation would be without such
a compensation system.
If such a compensation does turn out to be necessary I'd probably not
use a 50% mirror but simply loop the received data back to the
trasmitter of the SFP at the receiving end, and use a fiber pair instead
of a single fiber.

Regards, Paul Boven.

time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.

w3kl@w3kl.com wrote:

Without compensation, the delay tempco of 34 km of fiber will be
somewhat more problematic unless the temperature of the entire run is
constant and/or expensive low delay tempco fiber is used.

Bruce

The biggest concern with temperature effects in the fiber would be thermal expansion.  Potentially, as well, changes in waveguide shape, which could add more waveguide dispersion as well as potential polarization mode dispersion.
 
Jeff

Jeffrey K. Okamitsu, PhD, MBA
+1-609-638-5402

--- On Wed, 11/26/08, Bruce Griffiths bruce.griffiths@xtra.co.nz wrote:

From: Bruce Griffiths bruce.griffiths@xtra.co.nz
Subject: Re: [time-nuts] 10 MHz over optical fiber?
To: "Discussion of precise time and frequency measurement" time-nuts@febo.com
Date: Wednesday, November 26, 2008, 2:13 PM

w3kl@w3kl.com wrote:

meaning to chime in ever since the discussion got started.

discussed and put to rest.

dispersion in the fiber?

wavelength band, where absorption of the light by water is small), single mode
fiber has a chromatic dispersion of approximately 1 to 10 ps/nm/km.  The concern
therefore is changes in timing induced by changing wavelength content of the
laser or LED used to drive the fiber - for example, due to temperature changes,
etc.

Otherwise, it would be interesting to work through a "what if"?

Without compensation, the delay tempco of 34 km of fiber will be
somewhat more problematic unless the temperature of the entire run is
constant and/or expensive low delay tempco fiber is used.

Bruce

time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.

Hi Group;
It seems to me that the increase in noise introduced with an
optoelectronic device would not matter in most applications if a
cleanup oscillator is added.

Best Wishes;
Thomas Knox
NIST
4475 Whitney Place
Boulder Colorado 80305
1-303-554-0307
tomknox@nist.gov

Quoting "Paul Boven" p.boven@xs4all.nl:

If your concern is simply a stable frequency reference, that's true, even
though I am not sure what kind of cleanup oscillator would match the short
term stability of a maser. But also if you want to use it as a time
standard, the phase shift in the fiber has to be compensated, and it's
variations over temperature/humidity/gravity and whatnot must be accounted
for.

This is time-nuts, we don't simply want to make things work, we want to make
them work good :-)

Didier

Didier wrote:

Didier

A linear temperature ramp will create a linear ramp (equivalent to a
frequency shift) in the fiber delay.
Such temperature ramps may be expected at least twice a day. In other
words the the rate of change of temperature needs to be low to preserve
the frequency accuracy at the receive end of the fiber.

Bruce

I think for singlemode LEAF fiber you see something around 100ps/km per degree C.
Hopefully the fiber is buried and the temperature changes are more gradual.

Scott

Bruce Griffiths wrote:

Scott Mace wrote:

Most of it may be buried, however the ends of the fiber run may not be.
There are expensive fibers available with much smaller tempcos, at least
over limited temperature ranges.
The thermal expansion tempco of fused silica fibers is relatively low
(0.5ppm/C or so depending on exact composition) so most of the
propagation delay tempco is due to the refractive index tempco.

Bruce

From the corning leaf spec:

Environmental Test Induced Attenuation
Condition (dB/km)
1550 nm /1625 nm
Temperature Dependence
-60°C to +85°C* ? 0.05
Temperature – Humidity Cycling
-10°C to +85°C* and up to
98% RH ? 0.05
Water Immersion, +23°C ? 0.05
Heat Aging, +85°C* ? 0.05
*Reference Temperature = +23°C
Operating Temperature Range: -60°C to +85°

I just checked the daily variation of optical loss, and it seems to be about +-0.15dB
over a 90km DWDM system that operates in the NY,NJ metro area.  It's slightly larger
from summer to winter.  Clearly measuring delay through a loop would be a more accurate
metric, but this should give you a ballpark of real-world environmental influence.  All
of the electronics are in cooled rooms (typical CO or datacenter), all the fiber was
buried and in building risers.

And don't forget about back-hoe induced phase shifts.

Scott

Bruce Griffiths wrote:

Bruce,

Over a 34 km path, one should expect temperature variations not to be
uniform (unless the fiber is buried, but over 34 km, that would be
expensive), so simple open loop pre-programmed compensation would probably
not work well, some form of active, dynamic compensation should probably be
used.

Didier

Paul Boven skrev:

Not if you choose STM-1 SFPs. STM-1 has a baudrate of 155,52 MBd so a
cycle will be 15,552 symbols, which is certainly not a problem as a
STM-1 must be able to handle runs as long as 72 symbols of 0 or 1, and
then the problem is really in the clock recovery PLL. The AGC will
certainly work well. These SFPs are of the shelf components. You would
like to drive them diffrentially and all that, but it is not too hard.
There are nifty test boards with SMA connectors which would make it easy
to hook up to form a nice rig.

Consider that for SDH links as such, the jitter must be less than 0,07
UI, where a UI is the length of a symbol. This turns out to be 450 ps
for a STM-1. However, that includes ISI which would be essentially
canceled out for a 10 MHz signal, so you should expect less than that.

It's off the shelf stuff. 34 km is not a big problem. You could pick up
a pair of 70 km SFPs for better specs, but be sure to not overload the
input, you might need to insert a damping pad in that case.

STM-4, GE or STM-16 SFPs can also be used, with increased requirements
on jitter... GE SFPs will probably work very well even if they expect
8B10B data according to standard, but they are not significantly
different to the SDH SFPs these days. The expected jitter on STM-16
would be about 28 ps, but can again be expected to be less...

So, jitter-wise it is a valid solution and off the shelf components will
probably work well for you. I recommend you to use SFPs with monitoring
functions, so you can monitor levels, temperatures etc. It's fairly
simple I2C stuff so it is not big magic there.

Cheers,
Magnus

Paul Boven skrev:

You'd get a fiber-pair anyway, and the SFP provides both transmitter and
receiver, so use them!

Infact, you should loop back the signal to the original site and monitor
the signal there. If you use standard SFP testboards, you can take one
of the signals from the SFP and loop it back to the SFP using a short
semi-rigid SMA cable and put a 50 Ohm terminator on the other input
signal. That's all the loopback you would need. Then at the transmitter
end you use the output and send it to a counter and measure 10 MHz phase
against the H-maser.

Logg against time and log outdoor temperature as well as lab temperature
in both ends. Preferably SFP temperature in addition.

Cheers,
Magnus