The LNGS UTC system
D.Autiero, Strasbourg 22/1/03
• Some notions concerning GPS
• The LNGS UTC clock
• Interface to the OPERA timing system
The Global Positioning System (GPS)
The GPS system was built and it is
controlled by the US militaries (DOD), it
is composed by at least 24 satellites
orbiting at 20000 Km, the budget for
maintenance is 500 M$/year
Depending on the place and
the time of the day on
average between 5 and 8
satellites are visible by an
observer on the earth
Each satellite is equipped with a Cs atomic clock
and transmits its time and position (computed
with its ephemeris):
e.g.: x1,y1,z1,t1
The observer on the earth,
getting the quadrivectors of
4 satellites has to solve a
system of 4 equations in
order to find his own time
and position: x0,y0,z0,t0
(x0-x1)2 + (y0-y1)2 + (z0-z1)2 = c2(t0-t1)2
(x0-x2)2 + (y0-y2)2 + (z0-z2)2 = c2(t0-t2)2
(x0-x3)2 + (y0-y3)2 + (z0-z3)2 = c2(t0-t3)2
(x0-x4)2 + (y0-y4)2 + (z0-z4)2 = c2(t0-t4)2
Then x0,y0,z0,t0 can be converted in the more familiar
Latitude,Longitude,Altitude and time
o At least 4 satellites are needed, if only 3 satellites are visible one can
still compute the position by imposing the constraint of being on the
surface of the earth (geoid model), in this case the altitude cannot be
measured
o In case n (n>4) satellites are visible the system becomes
overconstrained and the precision improves as 1/sqrt(n)
o If the observer knows already his position x0,y0,z0 (fixed observer)
then just one satellite is needed in order to measure t0
o One assumes the speed of the ligth in vacuum (c) but this is not
exactely the case since the ionosphere is a medium:
The satellite emits on two frequencies:
L1=1.575 GHz (C/A code)
L2=1.2276 GHz (P(y) code) in order to model and correct for the real
speed of the ligth in the ionosphere (differential GPS)
o The typical resolutions of a GPS measurement (single time fit) are:
22 m horizontal accuracy
200 ns UTC time
The satellites needs to know their ephemeris and the atomic clocks
need continously to be readjusted from the ground stations with a
complex procedure. Main control at Colorado Springs.
Relativistic corrections are not neglegible and are correctly taken
into account
The system has been running for many years unchanged, the main change
happened in 2000 by disabling the SA (about a factor 10 improvement)
GPS signals can be
anytime denied on
regional basis
Typical commercial high precision GPS system (4Keuro)
• Has an accuracy of about 1 microsecond (limited
by the quality of the local clock)
• People do not need to distribute the signal
underground at 8 km distance
• It has a square wave output at 1 Hz (1 PPS) to tell
when each second start
The date (down to the second) is sent out in the
IRIG-B format
The LNGS UTC clock
For the physicists who like to have an accuracy of 100 ns and like to
work underground at many Km from the antenna something more
sophisticated is needed …..
The LNGS UTC system exists since the MACRO experiment started
data-taking (1987). It is among the duties of the lab to provide UTC
signals to the experiments in the caverns
GPS antenna
GPS receiver
Slave clock(s)
Master clock
8 Km opticql fiber
Underground LABs
External LAB
The GPS receiver is coupled to an atomic clock (rubidium oscillator at 10
MHz) called master clock
The atomic clock generates the local time scale in between GPS
syncronizations
Stability:
Short term fluctuations 3x10-12/100s
Long term drift (aging, temperature) 1 microsecond/day
Each second the atomic clock is readjusted on the GPS, the UTC time
scale is known with a precision better than 100 ns
Two working modes:
1) Position: averages over a sufficient number of measurements to find
the position of the antenna within a few m
2) Time only: once the position is known can be used as input in order to
get a very precise measurement of the time
If the GPS signal is not available one relies only on the atomic clock
The LNGS system was built by
the italian company ESAT
Signal distribution to the underground labs
The UTC time scal is known at better than 100 ns. Every ms (1000
better than the 1 PPS standard) a synchronization pulse and the
time/date string are sent through the 8 Km long monomodal optical
fiber (1310 nm) to 6 slave clocks in the caverns. The ligth pulse is
converted in electric serial pulses.
Slave clocks have to regenerate the local time with TCXO oscillators
with a stability of 1 PPM (100 times better than what needed to keep
in track within 1 ms) The time is then available in a 80 bits formatand
it is known at the level of 100 ns
MACRO was reading it with CAMAC modules
BOREXINO with VME
How to interface to the existing system ?
The original master card for the time distribution was foreseeing
a 1 PPS input and IRIGB input (then one needs onboard a very
high stability oscillator stable at the level of 100 ns/1s
The slave clock has a 1 PPS output, then one has to convert the
date string in IRIGB format
The other possibility is to bypass the slave clock and build a card
directly reading the fiber, this would allow to use a less
sophisticated oscillator
We asked ESAT which proposed a new slave clock for us for 10
keuros
ESAT is also available in helping us to develop the master card
directly interfaced to the fiber (we are missing of experience
on the high stability oscillator part). We are going to meet at the
beginning of february.
Clock distribution
Node card i
SM2
SM1
Master card 0
MLVDS
MLVDS
O/E converter
1:2
splitter
Optical fiber
Master clock
PCI card
PCI card architecture
Inputs from GPS receiver
(pps, 10Mhz, analog Irig B, digital Irig B)
Optional for Lab tests
O/E
Converter
Must be define
Master clock date
10Mhz
5.10E-11
OC-050
Vectron Int.
EPC2
Local
bus
PLX
9080
EEPROM
To the station
APEX
20KE
Hot
Link
923
Hot
Link
933
TX
HFBR1116T
Optical fiber
from the master
clock
Optical fiber
to the O/E
converter
Data+clock
mixed
PECL
RX1
HFBR2116T
Optical fiber
From the SM1
RX2
HFBR2116T
Optical fiber
From the SM2
Schéma optique
Rx=? nm
Fibre optique Master Clock
O/E converter
Rx=1310 nm
HFBR1116T Tx=1310 nm
1:2
Tx=1310 nm
MLVDS
Vers SM1
HFBR5205T
HFBR2116T Rx=1310 nm
HFBR2116T Rx=1310 nm
Fibre optique
Multimode
62.5/128
O/E converter
HFBR5205T
MLVDS
Rx=1310 nm
Carte PCI
Vers SM2
Tx=1310 nm
Master card architecture
Rx Tx
Data + clock
mixed
Bus n0
deserializer
HOT
Link
933
SN65MLVD
202D
clock
data
Clock
From node
data
data
RJ45
clock
HOT
Link
923
serializer
EPLD
EPM7256
Bus n1
clock
data
Clock
From node
Differential signals
Node Card architecture
Clk
data
Clk from node
RJ45
On the Node card
Input clock x 10 = 100Mhz
Clock 10Mhz
Node address
EPLD
EPM7128
Commands sent
Address Address
serialized requested
to the FPGA
(reset, reboot, incr cptr, …)