3.1 The offshore timing calibration system
3.1.1 The system architecture
3.1.2 The optical pulser
3.1.3 The fiber
3.1.4 The optical network
3.2 The Cable Calibration Station
3.2.1 The goal of the cable calibrations Station
3.2.2 The implementation
3.2.3 The measurements and performance
3.3 The onshore clock system
3.3.1 Introduction
3.3.2 The project
3.3.3 The implementation hardware
3.3.4 The implementation software
3.1.1 THE SYSTEM ARCHITECTURE
The philosophy chosen to get the time offsets for each sensors is that to stimulate the OM by known
optical signals at known time. As indicated in paragraph 1.2, a good signal candidate to stimulate
the OMs is blue light pulse, in fact the water transparency is maximum for that wavelength, then the
blue optical pulses got from OMs will be more numerous then other wavelengths.
To measure the real arrival time of photons of these light pulses it is needed that the sources
produce the pulses at the same time or a different but known time. In fact, if the pulsers have an
offset each other, the time offsets measured at OMs will be affected from the differences of starting
time of the pulses. Moreover, supposed that it is possible to line up the pulses starting, it does not
guarantee that this alignment is preserved for the whole life of the detector. At contrary, we expect
ourselves that the start time of light sources will drift sooner or later for the ageing of the devices
but it is not known how this effect is sensitive related to the whole life of the apparatus.
Then, to take into account the possible drifts of the light sources the strategy chosen is that to have
few common sources where the signals are split to provide all the OMs. In this way the optical
signals of a group of OMs are originated from the same light source. Consequently, if drifts are
observed in the time offset, they will have to be the same for all the OMs connected to that source.
Alternatively, the drifts are not originated from the pulser but from the PMT involved.
Optical Module
Optical fibre
splitter
Optical source
FCM
FCM electronics interface
Figure XXX. Splitting principle of calibration signals at floor level.
In figure XXX is schematized the principle of the common light sources. As the figure shows, the
common sources are implemented at floor level. In fact, as described in previous chapter the
synchronism is extracted at floor level.
3.1.2 THE OPTICAL PULSER
The first step to follow this planning is to produce adapt optical signals. The only constraint met so
far is the wavelength of the light source, the blue that has range from 400 to 500 nm.
To get the other important time constraints it is better to refer to a typical signal of a single
photoelectron given for a PMT. In figure XXX it is possible to see an example.
20 – 30 ns
Figure XXX. Shot of a single photoelectron signal got from sensor.
This signal is negative, it has an fall time of several ns and long tail. The duration is 20-30 ns. The
multiphotoelectron signals have a bigger peak and are longer then it.
The choice of the duration of light pulse, it has to compared with the duration of the single
photoelectrons signal. In fact, if the duration of the light pulse is comparable with the output signal
of PMT, then the time behaviour of the signals produced from calibration system will be different
from the signals we aspect to get. Contrarily, if the duration of light pulses are well below that
duration, the high performances of pulser are useless because the PMT is slower.
Another primary feature requested to the light source is the highest time stability or the lowest
standard deviation of threshold time. Because, obviously, the uncertainty of measured offsets will
be directly affected from the jitter of the rising edge of the light pulses.
The commercial blue light pulse sources available are laser and LED. The light pulses produced
from laser have very tight shape, well defined wavelength and high levels of optical power.
Unfortunately they are very expensive compared with LED.
A quite good compromise it has obtained adopting a LED source drived by a selected circuit layout.
The circuit adopted is shown in figure XXX. The schematic of this circuit was evolved from the
circuit discussed in [5]. Moreover, especially care it was paid to the placement of components and
designing of routing. After implementation a long characterization activity was done.
LED
Power
R1
Start
C2
R3
C1
Pulse
Q2
R4
Q1
R2
Pulse Intensity
Fig. XXX. Schematic of the pulser circuit.
This circuit features a LED Agilent HLMP CB15, which was carefully selected to work as the
source of the fast light pulses with a blue wavelength as we need. The peak wavelength is 470nm
[datasheet] in figure XXX is shown the intensity spectrum vs wavelength.
Figure XXX. Intensity spectrum of blue LED HLMP CB15.
As it is possible to notice in figure XXX the focus of LED is not good enough, in fact, the
angular displacement, where is comprised the 99% of the intensity, is inclusive between  = -15°
and + 15° (figure YYY). Therefore, to couple the pulser with the fiber it was introduced a
commercial optical collimator Thorlabs F230FC-A to increase the optical power picked up by the
fibre.
Figure XXX and YYY. Angular dispersion of intensity LED HLMP CB15 on left side. On
right side the same dispersion represented in space.
In figure XXX a photograph of the chosen collimator is shown. To mate a robust and stable
coupling between LED and fibre a special mechanic support was realized. A design view is
provided in figure YYY.
Figure XXX and YYY. Photograph of Thorlabs F230FC-A collimator on left side.
Mechanical view of collimator and fibre support on the right side.
In order to measure the fast and faint light pulses emitted by the pulser, fast photomultipliers
Hamamatsu H6780 were used in the functionality tests (see figure XXX).
5.5 cm
Figure XXX. Compact PMT Hamamatsu H6780 adopted to test the optical pulser.
These devices offer very appealing characteristics: they are compact, fast (rise-time less than
1 ns), quite stable (TTS < 0.28ns) and reliable; in addition, they are internally equipped with a HVsupply. The coupling of the photocathode to the fibre can be performed easily with a commercial
adapter.
An illustration of one of the pulser signals as viewed by the test PMT is shown in
Figure XXX.
Fig. XXX. Light pulse from the pulser as viewed with a test Hamamatsu H6780 PMT (yellow
trace: trigger signal; blue trace: PMT output).
The pulses detected by the PMT are characterized by:
Table 1. Time performance of the LED pulser measured by H6780
Falling time
~ 2.5 ns
FWHM
< 7 ns
Standard deviation of pulse time from trigger
< 100 ps
It should be noted that the FHWM of these optical signals is larger than for the current pulse
flowing through the LED. The result of a simulation made with PSpice is shown in Figure XXX
(calculated FWHM: about 1.6 ns), while in Figure 7 we show the measurement performed on the
board (measured FWHM: about 2.8 ns).
Figure XXX. Simulated current pulse flowing through the LED of the pulser board.
Figure XXX. Current pulse flowing through the LED as measured on the pulser board (yellow
trace: trigger; red trace: current pulse).
In order to cope with the attenuation introduced by the cable and connectors we have
developed a remote control to manage the pulse intensity of the LED by means of the regulation of
the diode voltage. This voltage is set from two levels of regulations respectively for the coarse and
fine tuning of the intensity of the light pulses. The regulation is realized by digital potentiometers.
The chosen devices are X9C103S by Xicor [datasheet]. They are non volatile, the full range
resistance is 10k0 and the range is divided in 100 steps.
In figure XXX is shown the regulation part of circuit. As possible to notice the voltage ranges
applied to the digital potentiometers (between Vh and Vl terminals) are different. The upper device
is the course regulation where between Vh and Vl it is applied the full range of 5V, the lower device
realizes the fine regulation where the range applied is 2.5V. The regulated outputs (Vw) supply an
adder opamp that at the same time adapt the range to the needs of the analogic pulser circuit.
The managing of these device is quite simple, it needs to chose the crescent/decrescent versus
(UpD* pin) and send how many pulses we want to step up or down the output. The devices are
sensitive to the falling edge. The choices of this kind of simple but effective control follows the
criteria of high reliability due to the inaccessibility of site to repair or re-program the devices.
Figure XXX. Intensity regulating section of pulser circuit layout.
In Figure XXX it is possible to see a photograph of the board equipped with the pulser, digital
circuit controller and fibre-collimator support.
Fig. XXX. LED pulser complete of collimator and fibre coupling.
3.1.3 LE FIBRE
Another critical aspect in the calibration strategy chosen is the fibre performance. It is known
that optical fibres have a very low attenuation especially in range of distance of few hundred
metres, as we have. Nevertheless, the common signals adopted for them are in near or far infrared,
quite far from wavelengths we chosen. Then a long fibre characterization activity was performed to
assure low optical power losses and low timing degradations of signals produced by optical pulser.
The fibres examinated and their main features were resumed in the table 2:
Table 2. Fibres examinated
Fibre
AFS 50/125
Cordis single window 850 nm
BFH/L 200m
GIF625
460HP
Mode
50m core 125m cladding – multimode
50m core 125 cladding – multimode
200m core 230m cladding – multimode
62.5m core 125m cladding – multimode
9m core 125m cladding – monomode
NA
0.22
0.22
0.37
0.27
0.13
BFH/L 200m is a large core fibre that is good because Numeric Aperture (NA) is the
maximum value and more is the light picked up form LED.
GIF625 is a graded index fibre where the different paths of photons are equalized with
velocity and the intermodal dispersion in minimized.
Cordis is a commercial low cost fibre, currently used in data transmission as Ethernet on fibre.
AFS 50/125 and 460HP are specialized for visible wavelength, the first is multimode the
second is monomode.
Considering the advantages and the disadvantages of each of them, the choice was fallen on
AFS 50/125 because the losses are minimized, the optical power picked up was quite good and the
degradation time performances were no sensitive for the working lengths. In figure XXX is shown
the curve of attenuation vs wavelength of the chosen fibre.
Figure XXX. Spectrum attenuation of AFS 50/125 fibre.
Another critical aspect is the fibre termination. In fact, it is needed that the light is injected on
photocathode of PMT without big optical power losses due to shadow of objects or reflections and
also mechanically stable. Unlike the PMT chosen for apparatus shown in figure XXX is much less
handy of H6780, then, several surveys and attempts on field were performed. The solution was
found designing a new support where the edge of the fibre is properly oriented toward the sensitive
part of the sensor. As Figure XXX shows, the light is injected from the backside of the
photomultiplier with a specific angle to maximize the signal amplitude. Consequently, the light
pulses can illuminate a wide spot on the photocatode given from a cone of 12.7° from the axis.
10°
=12.7°
Fig. XXX. Photo of photomultiplier complete of fibre support.
3.1.4 THE OPTICAL NETWORK
As discussed at beginning of this chapter, the offshore time calibration strategy is that to send good
quality of optical pulses to groups of OMs at same time. In figure XXX the groups of OMs were
individuated as all OMs of each floor. That initial chosen was performed in natural way, due to the
cabling arrangement discussed in 2.3. Nevertheless, it is essential extract an unique time offset
matrix where the OMs offsets of each floors are consistent each other. For this it is necessary to link
the offsets retrieved from pulsing of each floor together. A way to reach this goal is to extend the
pulse splitting concept to multiple floors. To obtain the maximum number of OMs illuminated at
once, the splitting scheme has to taking into account the amount of optical power the pulser can
supply and the losses of cables along the dorsal, connectors, splitters, etc. The final scheme reflects
a compromise between the maximum number of OMs to provide and a good optical signal quality
at final level. The range between the light intensity minimum and maximum supplied from pulsers
it was fixed to obtain the range 1 – 10 photoelectrons at OM levels. Where the pulsing to multiple
photoelectrons allows to study the calibration at lower levels of TTS.
To obtain the final scheme many simulations and tests were performed. A simulation system is was
assembled to study the required devices and the ratio of the splitters.
The system was composed by a PMT 10”, a power supply board to permit PMT to work in nominal
conditions, a dark box to screen the environmental optical noise, a FEM board, PC to retrieve the
data, an electrical pulser to trig the event, the oscilloscope and a reference optical pulser. Between
the optical pulser and the PMT an variable attenuator was placed obtained from several splitter in
succession, to simulate the presence of several losses declared from datasheet of cables and
connectors. In figure XXX is shown part of the setup.
Reference
optical pulser
Splitters chain
Dark box
Fig. XXX. Photo of setup up used to simulate the optical losses in the optical network.
The devices chosen to split the optical signals are commercial splitters multimode 50/125 m for
data transmission in 850 nm window [go4fibre]. They are build melting more edge of a optical
fibres. In particular the kind of optical fibre adopted for the process is the Cording multimode fibre
described in Table 2. The advantages of this solution are the availability and costless considering
the prospective of km3.
The final scheme is represented in figure XXX
IV floor
III floor
II floor
I floor
1x2 10/90
2x2 50/50
1x2 50/50
Hybrid connector
Figure XXX. Final layout of the fibre network for minitower
Following the scheme it is possible to noticed that lighting an intermediate floor there are
illuminated all the OMs of the same floor, 2 OMs of the upper floor and 2 of the lower floor. In
brief see table 3:
Table 3. Floors pulsed from floor pulsing
Floor pulsing
I
II
III
IV
Floor pulsed
I,II
I,II,III
II,III,IV
III,IV
The topics of the scheme are:
 Redundancy respect the “floor” configuration;
 Lighting bidirectionality along the mini-tower;
 The splitting configuration is the same for all floors.
The groups of splitters are packed together to facilitate the integration operations and outride the
deployments shocks and vibrations. A photograph of a splitting pack integrates in mechanical crate
of FCM is shown in figure XXX.
Figure XXX. Splitting pack integrated in FCM mechanical crate.
3.2 THE CABLE CALIBRATION STATION
3.2.1 THE GOAL OF THE CABLE CALIBRATIONS STATION
The scheme described from figure XXX shows the different paths that the pulses have to follow
starting from pulsers to each OMs. Because the pulses have different paths, there are associated
different times to get there. Moreover, if the paths were so little to neglect the time to get them and
the pulsers are synchronized, the different times registered from OMs give without further
corrections, the offset required, because the pulses should arrive in the same time and in the same
time when emitted. Unfortunately the time exploited to run over the fibre affects strongly the final
offset because this term is added, or better subtract, to the time measurement from OMs of
calibration signals. Then also the uncertainty of the knowledge of these terms contributes to the
uncertainty of the final offset.
In formula:
offseti  t peack ,i  corri
Where tpeack,i is the mean of the times where the signal pulses overcome a given threshold, instead
corri is the correction of the time exploited from the pulses to run over the fibre.
3.2.2 THE IMPLEMENTATION
The way adopted to measure the different paths described is based on time measurement between
start and stop signals. The device chosen to make that measurement is the Time to Digital Converter
(TDC) SR620 by Stanford Research System, that has very performing feature. Some these are here
summarised [http://www.thinksrs.com/products/SR620.htm]:



25 ps single-shot time resolution
11-digit frequency resolution
GPIB and RS-232 interfaces
It useful to underline the firsts two: “25 ps single-shot time resolution” because well under 1 ns as
discussed in previous chapter and “11-digit frequency resolution” because it possible to achieve
measurements of delay reasonably large or fibres very long. The figure XXX shows the front panel
of SR620.
Figure XXX. Front panel of SR620 Stanford Research System TDC
As is explained in 3.1.4 the whole path of the light pulses comprises connectors, pieces of fibres
more or less long and characterized by different reflective index, then it is necessary that the optical
signal carrier chosen to perform the measurements is the same of calibration signals. The block
scheme that represent the cable calibration station is shown in figure XXX.
Discriminator
negative
pulse
Pulses detector
(H6780)
pulse
Fibre to calibrate
pulse
stop
TDC
SR620
result
PC
Reference fibre
command
pulse
start
start
Interface
setting
Reference
Pulser
Figure XXX. Block scheme of cable calibration station.
The measurements are executed for difference. Before there is a calibrations phase where only the
reference fibre is connected from pulser to detector. In that phase it is measured the time between
start and stop signals. In the second phase is inserted the fibre to calibrate. At this time, the time
measured between the start and stop signals will include the time spent to pass through the fibre we
want to calibrate. Then, the difference between the time measured in the last time and the time
measured in calibration phase will give only the time spent to pass through the fibre to calibrate. In
fact, the all other contributes as the response time of pulser or detector, the cables will be collected
by the first phase.
The Interface block is made up of a microcontroller PIC17C756 connected to PC by RS232. The
controller is included in a multifunction board that include also some ADCs and DACs called
UNIV1. The commands are sent from PC by MODBUS protocol [modicon modbus reference
guide] thought a LabView control panel application (in LabView the applications are called VI,
Virtual Instrument) that is able to set up the light intensity required managing the coarse and fine
settings, retrieve the LED bias and send the trigger pulse. In figure XXX is shown the VI control
panel of pulser manager.
Figure XXX. VI control panel of pulser manager.
The Interface board, also, fan out the starting command from PC to reference pulser, described in
the previous paragraph of this chapter and to the TDC start connector. The start signal is in TTL
standard lever accepted from TDC instrument, but not so for PMT signal get out from H6780. In
fact, as seen in figure XXX the shapes of PMT signals are negative and out of common standards.
To apply the stop signal to TDC, then, it was necessary implement a discriminator circuit to fix the
time of pulse arrival for a given threshold. The device chosen for this application is ADCMP563 by
Analog Device [www.analog.com/UploadedFiles/Data_Sheets/ADCMP563_564.pdf] configured in
as a comparator.
Finally, it was realized another VI application to manage and store the measurements done by TDC
SR620. The TDC is connected by GPIB cable to PC.
Figure XXX. VI panel able to manage and log the measurements of the cable calibration station.
3.2.3 THE MEASUREMENTS AND PERFORMANCE
Due to cabling arrangement and integration, as discussed in chapter 2, the optical paths of
calibration pulses are split in several contributes and the measurements performed by cable shore
station are stored in a big database. In following part we want to represent the ending computation
that report of the mean corrections to each path.
To follow the table with the results it is better refer to the figure XXX.
A0
B3
A1
Optical pulser
B2
Figure XXX. Scheme of OMs tagging.
The OMs in each floor are tagged from zero to three, moreover the tag contain the side where is
placed, A side is on left and B side is on right. From figure XXX it is also possible to notice that A0
and B3 OMs look horizontally and A1 and B2 look down.
In table 4 is illustrated the summary of the estimated values and the accuracy for each path of fibre
network. The mean values are obtained adding each contribute that take part to the path and the
standard deviation of the path is obtained from geometrical mean of each standard deviations.
To read correctly the table 4 it is needed to start from a floor pulser and end to the specific OM and
floor. As it is described in the previous paragraph not all the paths are available. For example the
delay introduced by the fibre network from floor 1 pulser to OM A0 of floor 1 is 136.62 ns with
standard deviation of 110 ps. The delay from floor 2 pulser to OM B3 of floor 3 is 461.2 ns,
standard deviation of 380 ps.
Table 4. Summary of the delay introduced by fibre for all the paths
Floor 1 pulser lighting on
Floor 2 pulser lighting on
Floor 3 pulser lighting on
Floor 4 pulser lighting on
OMs of floor 1
OMs of floor 2
OMs of floor 3
OMs of floor 4
Standard
deviation
(ns)
OM
Mean
(ns)
136,62
0,11
A0
A1
131,79
0,10
B2
103,65
B3
104,84
OM
Mean
(ns)
A0
Standard
deviation
(ns)
OM
133,36
0,30
A0
A1
133,15
0,22
0,12
B2
104,38
0,11
B3
106,9
OMs of floor 2
Standard
deviation
(ns)
OM
Mean
(ns)
136,27
0,51
A0
135,29
0,21
A1
136,07
0,12
A1
136,44
0,22
0,22
B2
103,66
0,16
B2
104,79
0,14
0,18
B3
109,36
0,17
B3
107,23
0,19
OMs of floor 1
Mean
(ns)
OMs of floor 2
Standard
deviation
(ns)
OMs of floor 3
A0
490,54
0,27
A0
496,13
0,26
A0
487,91
0,24
A0
482,64
0,56
B3
464,17
0,36
B3
461,46
0,23
B3
461,56
0,31
B3
456,05
0,32
OMs of floor 3
OMs of floor 4
A0
488,42
0,60
A0
483,82
0,29
B3
461,2
0,38
B3
456,16
0,24
As it possible to notice from the table 4 the accuracy has some good values and some values are not
very well. That is due to a poor accuracy for few contributes. For example the OM A0 of floor 3 has
a big standard deviation if the path start from floor 3, floor 2 or floor 4 pulser. It means that in the
last part of the fibre path the measurements have a poor accuracy but that affect also other paths.
3.3 THE ONSHORE CLOCK SYSTEM
3.3.1 INTRODUCTION
From point of view of timing alignment, inside each OM there is a clock that run and restart
continuously measuring the time each time there is an over threshold event on photocatode.
Nerveless, the clocks are not in free running but they are synchronized by FCM main boards in
each floor. In turn, the FCM main boards receive data and clock from the symmetric FCM onshore
boards, as described in the previous chapter. The onshore synchronization is an important issue, in
fact, if the floors are not synchronized together the arrival time for PMTs of different floors is not
comparable each other. Only synchronization can guarantee that time offsets calculated once do not
change for little time intervals.
Part of this these work was dedicated to realize the item indicated.
Another need that it was raised associated to this purpose is the transmission of the absolute time. In
fact, to compare operations or results done in different parts of the apparatus is comfortable refer to
an unique time flow. For example the environmental parameters like temperature, pressure or
humidity are tagged with a time data when they are performed. Also, to archive the events received
from OMs is necessary to refer an unique time flow. Nerveless, it needs to pay attention to not
confuse with the requirements that need for the relative time measurements. In fact, the relative
time requirements are more strict than those of absolute time. That because the relative time is the
tagged time of the physical events and it needs the maximum time accuracy to get the particles
tracks. As specified before, in the FEM boards that counter is restarted each 500 s. That counter
has to be precise well under 1 ns as previously required. Indeed, absolute time refer to data
management and slow control operations. The minimum unite of absolute time chosen is the
duration of data frame packet used for the synchronous protocol between the onshore and offshore
FCM mains boards, as is specified in previous chapter, that time is 125 s.
Another concept that it is needed to clarify is that the absolute time is related to a local place.
It is quite obvious that if we need to transmit a time flow to a far place, the far place receive the
time flow later than the transmitter. Then, the far place will be always a way back respect to the
transmitter. How much it will be the difference it depends on the distance between transmitter and
receiver and the velocity of the vector information. For the case of light pulses into the fibre support
the light velocity is about 2  105 km/s, then for 25 km of the distance from shore to the apparatus
the offset will be about 125 s. After that, whatever clock of the apparatus will be a way back 125
s (at least) respect to the shore. The same concept it is possible to extend to the time offsets of
timing calibration. In fact, the time offsets are relative times between two any OMs but it is possible
also to express them respect to a common reference and if the common reference change, it will
change also the offset matrix. Nerveless, the common reference of offsets matrix is not a crucial
item because in any case the relative difference between each couple of OMs will be unchanged.
3.3.2 THE PROJECT
Some requirements are already discussed and some are not yet explicated. It is possible to
summarize the functionalities demanded to timing shore station in list:
•
•
•
•
•
•
•
Recovery clock from GPS receiver
Data and time recovery
UTC data flow sent to FCMs onshore in synchronous protocol
Clock Fan-out toward FCMs onshore
Format the clock and time flow signals to be acceptable to FCM on shore boards
Go-and-Back time measurements FCM onshore/offshore by TDC
Network Time Protocol (NTP) server
The first requirement refers to the possibility to compare physical events as muon or neutrino tracks
with other experiments around the world as, for instance, those listed in the first chapter. The data
manager has to able to tag the physical events with Coordinated Universal Time (UTC) time flow.
Then, the absolute time also has to be sourced from an Global Position System (GPS) device. To be
reproducible and scalable with KM3 project it was chosen a commercial device GPS receiver. The
device chosen is XLi from Symmetricomm [Symmetricomm] because it has very appealing
features, some of these are listed here:











1 Pulse Per Second (PPS) Output
Rate Output 1/10/100 PPS, 1/10/100 kPPS, 1/5/10 MPPS
Code Output (IRIG-A, B, and NASA 36)
Code Input (AM or DC: IRIG-A, B, and NASA 36)
RS-232/422 Serial I/O Port
Standard Network Port (10 Base-T)
GPS Time and Frequency Reference Card (+/-30 ns RMS)
Time Interval - Event Time Output (TIET)
Frequency Measurement Input
Programmable Pulse Output (PPO)
NTP server
Most of listed features will be recall later but it is useful to point out that the accuracy of GPS
receiver, declared to +/-30 ns RMS. It means that the error on PPS signal got from GPS is well
under the absolute time unit of 125 s. That accuracy has to compared with absolute time accuracy
required for apparatus that usually is declared less then 1ms.
The second requirement refers to extraction of data and time from GPS receiver to be sent to each
FCM onshore boards. The XLi supports several code outputs: IRIG (Inter Range Instrumentation
Group) and NASA[symmetricomm]. The code chosen for retrieve the current data and time is IRIG
B because it is based on 1s frame. In figure XXX it is shown the protocol of the code.
Figure XXX. IRIG B protocol frame.
The frame protocol is composed by 100 pulses. The rising edges start each 10 ms, in this way they
divide the second in 100 equal parts, then counting the rising edge it is possible to know the current
part of frame. The starting and the ending of the frame is marked with a pulse 8 ms long (marker),
each 10 pulses there is a marker, so it is possible to distinguish the starting of frame because there
are two consecutive markers. Between two markers 8 ms long there are pulses that represent bits: a
short pulse (2ms) represents 0, a medium pulse (5ms) represents 1. Between the first and second
marker there is information of the seconds in Binary Code Decimal (BCD) code. Between the
second and third marker there is the minute, between the third and forth there is the hour. Because
the day of year in BCD code need two blocks, that information is split between the forth and fifth
and between the fifth and sixth marker. The next three blocks not are defined and finally the last
two blocks contain the current number of the second of the year in standard binary code. As it is
possible to notice the code does not contain the current year, so it will be repeated at the same every
year.
The third requirement listed refers to the transmission data protocol agreed to be read from FCM
onshore board. A scheme of the protocol is shown in figure XXX.
Data absolute time
UTC
start sec. min hous
days
frames since 01/01/2006
Acoustic absolute time
sec since 01/01/2006 Idle Idle Idle Idle
…
Idle stop
500 bit = 1/8000 seconds
Figure XXX. Absolute timing data protocol, transmitted from onshore timing station to FCMs
onshore boards
This protocol is characterized by:
•
•
•
•
•
8b/10b code (revealing error, scrambling 0/1, byte function embedded)
Time upgraded each 125 s
Data (event and SC) absolute time
Acoustic data absolute time
More functions and information are possible to add
Adopting the same frequency of transmission clock of 4 MHz and the minimum unite of absolute
time of 125 s, it derives a synchronous protocol of 500 bit framed. To assure a revealing error
function, each byte is encoded by 8b/10b encoding. Adopting 8b/10b code there are also defined
some standard byte function as start and stop frame and idle. Because the 8b/10b codes each byte
by 10 bits, the frame useful to transmit the data are reduced to 48 bytes (excluded start and stop
words). All the bytes are binary standard with Most Significant Bit (MSB) on most left part. As the
figure XXX represents the first bytes reflect the IRIG B protocol: the first byte contains the current
second, the second byte contains the minute, the third contains the hour, the forth and fifth contain
the day of the year. The following six bytes contain the number of data frame or the current
absolute time expressed in 125 s unit. As pointed out before, the IRIG B does not contain the year
information, so it was necessary to fix a zero data from which it is possible to count the absolute
time. The zero data chosen is midnight of January 1th of 2006. The six byte are enough to count
more the 1115 years! The last 4 meaning bytes contain the same information of absolute time but in
seconds unit. Those are useful for acoustic positioning boards that tag the measurements in that
format. The last part of frame constituted by 33 bytes are without of meaning and they are free for
more applications.
The forth and fifth requirements, listed in YYY page, refer to the signal adaptation with receive
parts of FCM onshore boards. As mentioned before, the clock frequency required is 4 MHz and the
electrical format chosen is LVDS, because that format guarantee a better noise immunity also for
cable of some metres. Moreover, it was applied the ground decoupling to avoid ground loops.
The following requirement request to the Clock Onshore Station is the go-and-back time
measurements. As clarified in 3.3.1 the absolute time refers to local position, then, it can be useful
to know how much is the temporal distance from the shore to each floors or more important
monitoring eventually problems can occurs on synchronization between the floors. Moreover, that
information may confirm the inter-floors offshore timing calibration, if the accuracy is enough. For
this application a signal of start frame and one of arrival frame are available from each FCM
onshore board. Then, the duty of the shore station is to address the proper start and stop signal to
TDC time by time for all floors, managing the TDC, logging and storing the measurements.
The last requirement is the distribution of synchronism signals over Ethernet to other PCs of shore
station. This is an useful item that allow to synchronize the computers clocks to share jobs,
applications or results knowing those are time aliened. This synchronization is realized by a server
that uses the protocol NTP to send time packets on port 123 of UDP (User Datagram Protocol)
[http://www.ntp.org/]. The accuracy of this mechanism is quite lower respect the others accuracy
described because it involves long distances and processor activities but enough to the data
management.
3.3.3 THE IMPLEMENTATION HARDWARE
The block scheme of functionalities involved in the clock onshore station is shown in figure XXX.
Antenna
GPS
GPIB
XLi – GPS
Symmetricom
IRIG B
RS232
PC Cronos
Clock 10MHz
Startx
TDC
SR620
Stopx
Interface
board
SCSI
cable
Crate PXI
FPGA
4 x LVDS 4 x LVDS
Dati 4MHz Clock 4MHz
Figure XXX. Block scheme of clock onshore station
As mentioned before Symmetricom XLi realizes the GPS receiver, the GPS signals come from a
GPS Antenna installed on the roof of shore station. XLi supply IRIG B and clock to the device that
manages the clock fan out and the absolute timing transmission. An Interface board is able to
collect the cables and different connectors and it multiplexes start and stop signals to provide with
TDC. The PC manage all the devices as it will be clear later.
Some of blocks are known because discussed previously. Here we want to describe the device that
realize of the clock fan out and the transmission of the absolute timing data protocol.
The choice was addressed to reconfigurable logic devices, in particular FPGAs because the
operations to need to implement are fixed in time. Nerveless, the reconfigurable choice it was also
necessary due to split the function implementation and testing. Moreover, the possibility to re-use
the same hardware to different project as KM3.
The FPGA chosen it was Virtex II Pro – Xilinx, installed in a National Instrument development
board NI – PXI 7811. An advantageous of the Virtex II pro device is the availability of Rocket I/Os
[www.virtex.com]. These are special I/Os where the jitter is very well controlled (40-50 ps) because
of using special care to the routing, often they are used to distribute the clock. In figure XXX it is
shown the mechanical view of NI board.
Figure XXX. Mechanical view of NI – PXI 7811R. In front of the board are available four
connectors where are split the 160 I/Os.
The main features of this board are summarized in table 3:
Table 4. Main features of NI PXI – 7811R
Bus
Digital I/O
FPGA
I/O formats
PXI
160
Virtex II pro – 1M gates
TTL input, LVTTL, LVCMOS input/output
PCI eXtensions for Instrumentation (PXI) is a modular instrumentation platform originally
introduced in 1997 by National Instruments. PXI is designed for measurement and automation
applications that require high-performance and a rugged industrial form-factor. It uses PCI-based
technology and an industry standard governed by the PXI Systems Alliance (PXISA)
[http://www.pxisa.org/] to ensure standards compliance and system interoperability.
PXI is based on CompactPCI, and it offers all of the benefits of the PCI architecture including
performance and industry adoption. An industry consortium defines hardware, electrical, software,
power and cooling requirements, leaving nothing to chance. Then PXI adds integrated timing and
synchronization that is used to route synchronization clocks, and triggers internally. In table 5, it is
listed the signals that are distributed on PXI backplane.
Table 5. Backplane signals of PXI divided from original specification
System
Signals
PXI
CompactPCI
PCI
PXI_BRSV
PXI_CLK10
PXI_CLK10_IN
PXI_LBL[0:12]
PXI_LBR[0:12]
PXI_STAR[0:12]
PXI_TRIG[0:7]
BD_SEL#
BRSV
CLK[0:6]
DEG#
ENUM#
FAL#
GA0-GA4
GNT#[0:6]
HEALTHY#
INTP
INTS
IPMB_PWR
IPMB_SCL
IPMB_SDA
PRST#
REQ#[0:6]
RSV
SYSEN#
SMB_ALERT#
SMB_SCL
SMB_SDA
UNC
ACK64#
AD[0:63]
C/BE[0:7]#
CLKDEVSEL#
FRAME#
GND
GNT#
IDSEL
INTA#
INTB#
INTC#
INTD#
IRDY#
LOCK#
M66EN
PAR
PAR64
PERR#
REQ#
REQ64#
RST#
SERR#
STOP#
TCK
TDI
TDO
TMS
TRDY#
TRST#
V(I/O)
3.3 V
5V
+12 V
-12 V
The crate chosen to house the FPGA board is NI PXI-1042Q also from National Instruments
because in that crate the signal PXI_CLK10_IN, indicated in table 5, is accessible from connector
on the back side of the crate as is shown in red circle in figure XXX. This crate can house seven
PXI multi-user boards like PXI – 7811R.
Figure XXX. Backside mechanical view of NI – PXI 1042Q crate.
The last block of the scheme in figure XXX to need to describe is the interface board. A photograph
of this board is shown in figure XXX.
FCMs - Start
and Stop
NI FPGA
board - SCSI
cable 68 pin
XLi
IRIG B
FCMs - Clock
and data with
LVDS driver
FCMs - Clock and data
without LVDS driver
Figure XXX. Interface board.
The board is realized to adapt different connectors, electrical formats and house the asynchronous
multiplexer. It is possible to chose if the LVDS format of clocks and data is implemented directly
on FPGA chip or inside the board by a fast LVTTL-LVDS driver. The RJ45 connectors visible in
figure XXX contain a transformer for ground decoupling. They are commonly used for fast
Ethernet data communication.
A photograph of the timing shore station is shown in figure XXX.
TDC SR 620
Symmetricom XLi
NI PXI crate
+
7811R FPGA board
GPS Antenna
PC Cronos
Figure XXX. Photograph of timing shore station.
3.3.4 THE IMPLEMENTATION SOFTWARE
Another advantageous of the NI – PXI 7811R board regards the easy programming. In fact,
LabView has realized an optional part that permit the project implementing graphically as
“LabView sytle”. This option part is called LabView FPGA Module. In spite of the high level
language it is possible to manage the I/Os, the trigger events, hardware counters, etc. up to hardware
details. It is also possible to include HDL (Hardware Description Language) functions.
LabView FPGA Module uses specialized functions to implement the hardware, two function panels
are shown in figure XXX and YYY.
Figure XXX and YYY. Master panel on the left side, I/O functions panel on the right side.
Because LabView FPGA Module uses only specialized functions, the VI FPGA program has to be
separated from other VIs. This kind of VI is called FPGA VI. The figure XXX shows the FPGA VI
front panel implemented to obtain the function needed.
Figure XXX. Front panel of FPGA VI that realize the hardware functions requested to the timing
shore station
As the figure XXX illustrates, while the FPGA run, there are shown the fields of absolute timing,
also the panel include the managing of multiplexing of start and stop signals, the clocks switching
and an useful reference pulser manager.
To share the several functions implemented the last LabView versions integrate a project approach.
In figure XXX is shown the project tree of the software implemented. In the project are shared also
the hardware (see FPGA Target in figure), the clocks, I/Os, FIFOs and other resources available.
Figure XXX. LabView project tree panel of software implemented.
To merge FPGA VI functions and ordinary VI LabView FPGA Module adds to the ordinary VIs
some interface functions. As the figure XXX shows the panel has functions to open, close, call, read
and write, etc. an FPGA VI. In this way the speed of VI executing (on PC processor) is adapted to
the operations developed from FPGA. In other words, these functions can open a window on the
part of PC memory that receive results and transmit parameters to FPGA.
Figure XXX. Functions panel of Host VI where LabView FPGA module is installed.
The VI that call an FPGA VI is called Host VI and is able to use all the functions available to
ordinary VI as file, string, graphical function that were not available to FPGA VI. In figure XXX is
shown the main console developed to manage the several functions requested to timing shore
station.
Offshore time calibration
FPGA Monitoring/management
On shore time calibration Satellite signal GPS locking
Go and Back time measurements TDC Management
Figure XXX. Main console of timing shore station.
The main console includes several functions: the FPGA monitoring and management and onhore
timing calibration are obtained calling the FPGA VI, the TDC interface realizes the automatic
multiplexing and logging of the go and back time measurements, the offshore time calibration
function will be explicated in the next chapter. The last function to describe is the satellite signal
GPS locking. It is an useful function that monitoring the alarms of GPS receiver in particular the
GPS locking. If the receiver does not receive enough strong signals from unless three satellites, it
will raise an unlock alarm. The VI periodically query the XLi by RS232 port. In case of unlock
alarm the console sends a warning to the Data Manager (DM) of the shore station. In that situation
the data got from apparatus are potentially misaligned with UTC.