Data Transmission and Acquisition in NEMO

1
Data Transmission and Acquisition in NEMO
Gabriele Bunkheila INFN - Rome
2
NEMO Information Transmission main tasks
Information streams
LAB
UNDERWATER TELESCOPE
100 km
Discussion will be primarly focused on Sensor
Data Transmission !
3
NEMO Topological distribution of sensors
A future apparatus will be composed of several
detection elements, distributed over an
underwater surface.
FLOOR
OPTICAL MODULE
Under an à la NEMO perspective, these can be
referred to as towers
Each tower is made of up to 18 floors.
Each floor supports up to 8 Optical Modules (OM -
4 in the figure).
The Optical Module is the individual data
source What kind of data are generated here?
TOWER
4
Single Optical Module Data Definition
Voltage pulses produced by a PhotoMulTiplier,
when hit by one or multiple concurrent photons
Data

Some numbers
80 bits, or 10 Bytes, on avg
If Poissonian... E(RE)3s(RE) 70 kHz
5
The Single Subsystem Paradigm for NEMO
1st Link from each Optical Module (OM) to a
floor concentration stage (x4 / x8)
2nd Link from the off-shore floor concentration
stage, to an on-shore deconcentration station
OM0
OM2
OM1
OM3
...
3rd Link from the on-shore deconcentration
station to a topological redistribution of OM data
6
FEM0
FEM1
The Single Subsystem a Hardware-oriented point
of view
Optical Module data come from 4 (up to 8) Front
End Modules (FEM from now on)
Off-shore floor data concentration is carried out
by a Floor Control Module (FCM from now on)
FEM2
FEM3
One floor data is received on-shore by a twin FCM
board, plugged on a host machine (FCM Interface,
or simply FCMI in the following)
At the FCMI, data are made available on memory
buffers (Front End Buffers, or FEBs). Each FEB
contains formatted data issued from the
corresponding Front End Module
FCMI
7
Information links analisys. Link 1 FEM ? FCM
(1/2)
Effective capacity 15.55 Mbps, or 1.944 MB/s
Very low probability of channel saturation and
consequent data loss
To be eventually noted...
Data packets contain a header actual load is
always greater than previously estimated
Target (RARE) physical phaenomena photon
production adds up to poissonian background
Additionally taken care of by an internal FIFO
queue on FEM board
8
Information links analisys. Link 1 FEM ? FCM
(2/2)
Conclusion
FEM0
FEM1
Each FEMs may produce an incoming instantaneous
load of up to
2 MB/s
x 4 8 MB/s x 8 16 MB/s
FEM2
FEM3
9
Information links analisys. Link 2 FCM
Off-shore? FCM On-shore
Physical medium totally passive Fiber Optics link
Prodocol STM-1 (part of SDH standard)
Existing Stardardized TLC Self-Synchronous Serial
Protocol
Main features
Transmission organized in frames, used here as
simple data containers
Frame clock
8 kHz 19.44 MHz 155.52 MHz
Byte clock
bit clock
10
Link 2 STM-1 protocol utilization
Facts on STM-1 frames (125 ms ? 8 kHz)
19.44 MB/s
Raw load 2430 B
(155.52 Mbps)
Available Payload
18.72 MB/s
2340 B
Room for accomodating up to 8 OM STATIC data
channels
Raw load per frame 9 x 270 2430 Bytes
No need for complex statistical multiplexing
mechanisms !
Usable PAYLOAD 9 x 260 2340 Bytes
11
Link 2 STM-1 Related hardware on each of the
two FCM boards
SDH Interface DWDM SDH compliant Transceiver
SDH STM-1 Mapper
Spartan-3 FPGA bridges SDH and OMs
SDH Compliant PLL
12
Information links analysis. Link 3 FCM
On-shore? Front End Buffers
Recall the main goal of this local link is to...
... Make sensor data packets available to
data-acquisition applications, by writing them on
a local machines memory
FCMI
Major constraints include...
... Supporting the load of incoming data from
off-shore
... Letting data be formatted in easily readable
packets
... Allowing packets to be topologically
organized with respect to their origin (Optical
Module number in the floor)
13
Information links analysis. Link 3 FCM
On-shore? Front End Buffers
FCM Board is a PCI 32 bit / 33 MHz - compliant
device. This means
PCI is a parallel bus for high speed data
transfers between two agents
Direct Memory Access functionality is provided,
thereby offering, in our case, a direct data path
from FCM board to FCMI RAM, without CPU
intervention
FCMI
32 bit / 33 MHz bus configuration supports a
nominal capacity of 132 MB/s
Actual transfer capacity may depend on different
factors
Latencies in bus-granting procedures (bus arbiter
_at_FCMI)
Address-specification mode (single word vs.
burst transfer)
Latencies due to PCI peripheral management
Latencies due to data elaboration prior to
transfer
Number of agents sharing the same PCI bus
14
Information links analysis. Link 3 FCM
On-shore? Front End Buffers
Supported actual transfer rate for present board
release allows
Up to 4 FCM boards (4 OM channels each) on the
same bus with static transfer time allocation and
no need for data queuing
Even more than 4 FCM boards, if strong data
queueing was implemented
PCI bus 32 bit / 33 MHz
NevertheIess, in these conditions bus capacity
saturation could become a critical issue for
other processes
A wiser solution seems to be
Just 1 FCM board per bus
Remaining bus capacity capitalized for active
data elaboration and external data transfer
through a LAN
Migration towards a faster PCI stardard will of
course allow a larger scale data concentration on
a single FCMI machine
15
What is DWDM ?
DWDM stands for Dense Wavelength Division
Multiplexing
Wavelengths distributed around l 1550 nm, where
a dense carrier spacing (down to 50 GHz) is made
possible by a low attenuation factor in fibers
and sharp monochromatic laser behaviour
16
DWDM single stage multiplexing/demultiplexing
techniques
f
f
f
f
DEMUX
f
f
DWDM - compliant fiber
f
f
17
DWDM add/drop techniques
TRANSPORT
- 0.4 dB
0 dB
- 0.2 dB
0 dB
- 0.4 dB
DESTINATION
- 0.2 dB/km fiber contribution neglected
18
Simplified version of a fiber network for a NEMO
tower
Main simplifications
- Information flow is only upwards (only sensor
data are considered)
- Only four floors per tower (floor 3/0 relative
power loss due to ADD devices negligible)
FCM 3
- No redundancy need accounted
Floor 3
FCM 2
Floor 2
FCM 1
Floor 1
Floor 0
FCM 0
19
Additional features in actual NEMO underwater
fiber network
Recall the single-backbone fiber structure used
for upwards sensor and control data transmission
(knots are add modules)
A second backbone fiber provides a medium to
downwards control information flows (knots are
drop modules)
Doubling the number of backbones is desirable as
the number of floors grows, to limit the losses
due to add and drop modules insertion
Redundancies are provided for the long haul link
All leaving signal are muxed onto the same fiber.
Same thing for the incoming ones
Incoming signals two fibers reach the tower
only one is used
Leaving signals the signal is split into two
fibers both are used, with half power each
2-BAND DEMUX
From ON-SHORE LAB
2-BAND MUX
To ON-SHORE LAB
20
DWDM System certfication at italian ISCTI
(Ministry of TLC - July 2005)
SDH Protocol Analyzer
DWDM Transceivers
155 Mbps eye diagram
On-Shore Off-Shore FCM Boards
Add Drop Modules
Real test fiber link over ?100 km ? bit error
rate lt 10-9
21
DWDM System certfication at italian ISCTI
(Ministry of TLC - July 2005)
On-Shore DWDM Transceivers
On-Shore Demux stage
22
DWDM System certfication at italian ISCTI
(Ministry of TLC - July 2005)
SDH Protocol Analyzer
Us
On-Shore Off-Shore FCM Boards
Add Drop Modules
DWDM Transceivers
23
Conclusions Why DWDM?
The long haul transmission network is completely
passive
Fewer components
Existing, standardazed, TLC-oriented commercial
material
SDH Standard makes system intrinsicly synchronous
24
Thank you