source: Selma/PARISROC/parisroc-jinst.tex @ 480

\documentclass{JINST}
\usepackage[pdftex]{graphicx}
\graphicspath{{figures/}}
\usepackage[figuresright]{rotating}
%\usepackage{graphicx}
%\usepackage[T1]{fontenc}
\usepackage{eurosym}
%\usepackage{rotating}
%\usepackage[dvips]{color}


%used explicitly in the text
\newcommand{\refTab}[1]{Tab.~\ref{#1}}
\newcommand{\refFig}[1]{Fig.~\ref{#1}}
\newcommand{\refSec}[1]{Sec.~\ref{#1}}




\title{PARISROC, a Photomultiplier Array Integrated Readout Chip.}
%

\author{S. Conforti$^a$, Second Author$^b$\thanks{Corresponding
author.}~ and Third Author$^b$\\
\llap{$^a$}Laboratoire de l'Accélérateur Linéaire, IN2P3-CNRS, Université Paris-Sud 11,
Bât. 200, 91898 Orsay Cedex, France\\
\llap{$^b$}Name of Institute,\\
  Address, Country\\
  E-mail: \email{conforti@lal.in2p3.fr}}




\abstract{
PARISROC is a complete read
out chip, in AMS SiGe 0.35 \begin{math}\mu{}\end{math}m technology
\cite{Genolini:2008uc}
%[1]
, for photomultipliers array. It allows triggerless acquisition for
next generation neutrino experiments and it belongs to an R\&D program
funded by French national agency for research (ANR) called
PMm2: "`Innovative electronics for photodetectors array
used in High Energy Physics and Astroparticles"'
\cite{PMm2Site:2006}
%[2]
(ref.ANR-06-BLAN-0186). The ASIC integrates 16 independent and auto
triggered channels with variable gain and provides charge and time
measurement by a 12-bit ADC and a 24-bit Counter. The charge
measurement should be performed from 1 up to 300 pe with a good
linearity. The time measurement allowed to a coarse time with a 24-bit
counter at 10 MHz and a fine time on a 100ns ramp to achieve a
resolution of 1 ns. The ASIC sends out only the relevant data through
network cables to the central data storage.
}%end of abstract

%\pacs{13.30.a,14.20.Dh,14.60.Pq,26.65.t+,29.40.Gx,29.40.Ka,29.40.Mc,95.55.Vj,95.85.Ry, 
%97.60.Bw}

%\submitto{Journal of Instrumentation}

\keywords{Keyword1; Keyword2; Keyword3}

\begin{document}
%use BST file provided by SPIRES for JHEP and modify it to forbid "to lower case" title
\bibliographystyle{Campagne}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Introduction}
\label{sec:Intro}
%%%%%%%%%%%%%%%%%%%%%%
The PMm2 project: "`Innovative electronics for
photodetectors array used in High Energy Physics and
Astroparticles"' \cite{PMm2Site:2006}
%[2]
proposes to segment the large surface of photodetection in macro
pixel consisting of an array of 16 photomultipliers connected to an
autonomous front-end electronics () and powered by a common High
Voltage. These large detectors are used in next generation proton decay
and neutrino experiment (i.e. the post-SuperKamiokande detectors as
those that will take place in megaton size water tanks) and will
require very large surfaces of photo detection and a large volume of
data. The micro-electronics group's (OMEGA from the LAL at Orsay)
purpose is the front-end electronics conception and
realization. This R\&D \cite{PMm2Site:2006}
%[2]
involves three French laboratories (LAL Orsay, LAPP Annecy, IPN
Orsay) and ULB Bruxells for the DAQ. It is funded for three years by
the French National Agency for Research (ANR) under the reference
ANR-06-BLAN-0186.


LAL Orsay is in charge of the design and tests of the readout chip
named PARISROC which stands for Photomultiplier ARrray Integrated in
Si-Ge Read Out Chip.

\begin{figure}[!htbp]
\begin{center}
\includegraphics[width=0.7\columnwidth]{img1.jpg}
\caption{Principal of PMm2 proposal for megaton scale Cerenkov water
tank.}
\label{fig:1}
\end{center}
\end{figure}

The detectors such as SuperKamiokande, are large tanks covered by a
significant number of large photomultipliers (20"),
the next generation neutrino experiments will require a bigger surface
of photo detection and thus more photomultipliers. As a consequence the
total cost has an important relief \cite{Genolini:2008uc}.
\begin{itemize}
        \item A smaller number of electronics, thanks to the 16 PMTs macropixel with
a common electronics, even if it induces more electronic channels;
        \item A common High Voltage for the 16 PMTs so a reduced number of
underwater cables, cables  that are also used to brought the DATA to
the surface;
        \item The front-end closed to the PMTs that allow a suppression of
underwater connector.
\end{itemize}

The general principle of PMm2 project is that the ASIC and a FPGA
manage the dialog between the PMTs and the surface controller (\refFig{fig:2}).

\begin{figure}[!!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img2.jpg}
\caption{Principle of the PMm2 project.}
\label{fig:2}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{PARISROC architecture}
\label{sec:PARISROCArchi}
The ASIC Parisroc is composed of 16 analogue channels managed by a
common digital part (\refFig{fig:3}).

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img3.jpg}
\caption{PARISROC global schematic.}
\label{fig:3}
\end{figure}

Each analogue channel is made of a low noise preamplifier with
variable and adjustable gain. The variable gain is common for all
channels and it can change from 8 to 1 on 4 bits. The gain is also
tuneable channel by channel to adjust the input detector's gain, up to
a factor 4 to an accuracy of 7\% with 8 bits.

The preamplifier is followed by a slow channel for the charge
measurement in parallel with a fast channel for the trigger output.

The slow channel is made by a slow shaper followed by an analogue
memory with a depth of 2 to provide a linear charge measurement up to
50~pC; this charge is converted by a 12-bits Wilkinson ADC. One follower
OTA is added to deliver an analogue multiplexed charge measurement.

The fast channel consists in a fast shaper (15~ns) followed by 2 low
offset discriminators to auto-trig down to 50~fC. The thresholds are
loaded by 2 internal 10-bit DACs common for the 16 channels and an
individual 4bit DAC for one discriminator. The 2 discriminator outputs
are multiplexed to provide only 16 trigger outputs. Each output trigger
is latched to hold the state of the response until the end of the clock
cycle. It is also delayed to open the hold switch at the maximum of the
slow shaper. An "`OR"' of the 16 trigger gives a 17th output.


For each channel, a fine time measurement is made by an analogue
memory with depth of 2 which samples a 12-bit ramp, common for all
channels, at the same time of the charge. This time is then converted
by a 12 bit Wilkinson ADC.

The two ADC discriminators have a common ramp, of 8/10/12 bits, as
threshold to convert the charge and the fine time. In addition a bandgap bloc provides all voltage references.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img4.jpg}
\caption{PARISROC Layout.}
\label{fig:4}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{Analogue Channel description and simulations}
\label{ssec:AnalogChannel}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
\refFig{fig:5} represents, in a schematic way, the detail of one channel analogue
part.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img5.jpg}
\caption{PARISROC one channel analogue part schematic.}
\label{fig:5}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{Preamplifier}
\label{ssec:Preamplifier}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
The input preamplifier is a low noise preamplifier with variable gain
thanks to the switched input ($C_{in}$) and feedback ($C_f$) capacitors that
can be adjusted (\refFig{fig:6}).

This gain can vary changing $C_{in}$, which is
common to the 16 channels, over 4 bits and $C_{f}$, to adjust preamplifier
gain channel by channel. This adjustment allows correction of the PMT
gain dispersion due to a use of a common HV.


\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img6.jpg}
        \caption{PARISROC preamplifier schematic.}
        \label{fig:6}
\end{figure}


The preamplifier is designed as a voltage
preamplifier in p-type Cascode structure to allow the acquisition of a
fast input signal with a large dynamic range.

The input transistor is a PMOS in common source
configuration: $W = 800~\mu$m; $L = 0.35~\mu$m; the big input transistor is
chosen to keep the preamplifier noise contribution low and to achieve a
high gm. It supplies the output (the drain terminal) to the input
terminal (source terminal) of the second stage transistor: $W = 100~\mu$m;
$L = 0.35~\mu$m; the output transistor must be small to reach preamplifier
high speed performances. The utility of the cascode preamplifier is in
the large input impedance of the common source (with also the
characteristic of Current Buffer) and better frequency response of a
common Gate. An output buffer stage is designed in order to adapt the
output impedance to the loaded impedance. The input dc level is high
(about 2.6~V) while the output dc level is low (about 1~V). Because of
the single side structure of preamplifier, it is hard to use the
external reference voltage to set the dc operating point; the idea is
to use an OTA as the dc feedback amplifier.

In  \refFig{fig:7} are shown preamplifier's output waveforms
for fixed gain and different input signal (left panel) and for fixed
input signal and different preamplifier gain (right panel).

\begin{figure}[!htbp]
\centering
\begin{tabular}{rl}
\includegraphics[width=0.5\columnwidth,height=6cm]{img7a.jpg} & 
\includegraphics[width=0.5\columnwidth,height=6cm]{img7b.jpg}
\end{tabular} 
\caption{Simulated preamplifier output waveforms for different input
signals with fixed gain (left panel) and for fixed input
signal at different gain (different input capacitor values (right
panel).}
\label{fig:7}
\end{figure}

The input signal, used in simulation, is a triangle signal with 4.5~ns
rise and fall time and 5~ns of duration as shown in \refFig{fig:8}. This current
signal is sent to an external resistor (50~Ohms) and varies from 0 to 5~mA 
in order to simulate a PMT charge from 0 to 50~pC which represents 0
to 300 photo-electrons when the PM gain is $10^{6}$.


\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img8.jpg}
\caption{Simulation input signal.}
\label{fig:8}
\end{figure}

The \refFig{fig:9} displays the input dynamic range allowed to the preamplifier
linearity performance. \refTab{tab:1} lists the residuals obtained for different
gains and shows a good linearity (better than $\pm 1\%$).

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img9.jpg}
\caption{Preamplifier linearity.}
\label{fig:9}
\end{figure}


\begin{table}
\centering
        \caption{TO BE COMPLETED}
        \label{tab:1}
\begin{tabular}{|c|c|c|c|}
\hline
$G_{pa}$ &  $V_{out-max}$ &  $Qi_{max}/n_{pe}$ & Residuals (\%) \\
\hline
 8 & 1.394~V  & 40~pC/250~pe & -0.6 to 0.2 \\
 4 & 0.841~V  & 48~pC/300~pe & -0.1 to 0.3 \\
 2 & 0.417~V  & 48~pC/300~pe & -0.2 to 0.3 \\
\hline
\end{tabular}
\end{table}


 
The \refFig{fig:10} displays the preamplifier noise with an
rms value of 13~fC and a Signal to Noise ratio of $\approx 12$. 
\refTab{tab:2} summarizes the results obtained.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img10.jpg}
\caption{Preamplifier noise simulation; $G_{pa}=8$; $C_{in}=4$~pF and
$C_{f}=0.5$~pF.}
\label{fig:10}
\end{figure}

\begin{table}
\centering
\caption{TO BE COMPLETED}
\label{tab:2}
\begin{tabular}{|c|c|c|}
\hline
RMS  & SNR & $V_{out}(1 p.e)$  \\
\hline
$468~\mu$V ($\approx 1/12$~p.e, $\approx 13$~fC ) & 11.6 & 5.43~mV\\
\hline
\end{tabular}
\end{table}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{Trigger output}
\label{ssec:Trigger}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
The PARISROC is a self-triggered device. The fast channel has been
conceived for this purpose.The amplified signal flows in a fast shaper that is a CRRC filter with
a time constant of 15~ns. Its high gain allows to send high signal to
the discriminator and thus to trigger easily on 1/3 of photo-electron.
It has a classical design: differential pair is followed by a buffer.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img11.jpg}
\caption{Fast shaper schematics.}
\label{fig:11}
\end{figure}

The \refFig{fig:12} represents the fast shaper output
waveforms for a variable input signal. The \refTab{tab:3} lists the fast
shaper principal characteristics obtained in simulation.

\begin{figure}[!htbp]
\centering
\begin{tabular}{rl}
\includegraphics[width=0.5\columnwidth,height=6cm]{img12a.jpg} &
\includegraphics[width=0.5\columnwidth,height=6cm]{img12b.jpg}
\end{tabular}
\caption{Simulated fast shaper outputs ($G_{pa} = 8$ with input from 1-10~pe (left panel)  
and from 1/3~pe to 2~pe (right panel).}
\label{fig:12}
\end{figure}

\begin{table}
\centering
        \caption{To be completed}
        \label{tab:3}
        \begin{tabular}{|c|c|c|c|}
        \hline
RMS  & SNR & $V_{out}(1 p.e)$  & $T_p$  \\
\hline
$2.36~\mu$V ($\approx 1/16$~p.e, $\approx 10$~fC ) & 16 & 37.85~mV & 8~ns\\
        \hline
        \end{tabular}
\end{table}

The fast shaper (15~ns) is followed by a low
offset discriminator to auto-trig down to 50~fC (1/3~pe at $10^6$ gain). 


The two discriminators can be used alone or
simultaneously. Their outputs are multiplexed to ease the choice. Both
are simple low offset comparators with the same schematic. The
difference comes from the way to set the threshold. The first
discriminator has the threshold sets by one 10-bit DAC, common to all
16 channels, and one 4-bit DAC for each channel. The second
discriminator has the threshold sets by only the 10 bit common DAC.
Each output trigger is latched to hold the state of the response in SCA
channel. In  \refFig{fig:13} are shown the triggers and the zoom of the triggers rise
time in order to see the time walk of around 4~ns.


\begin{figure}[!htbp]
\centering
                \begin{tabular}{rl}
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img13a.jpg}&
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img13b.jpg}
                \end{tabular}
\caption{Simulated trigger output (input charge from 0 to 10~p.e;
threshold at 1/3~p.e). Zoom of trigger rise time on right
pannel.}
\label{fig:13}
\end{figure}

Each output trigger is latched to hold the
state of the response in SCA channel.  SCA channel is the also called
"`Analogue memory"'. The SCA has a
depth equal to two; this means that there are two T\&H for time
measurement as well as for charge measurement.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img14.jpg}
\caption{SCA (switched capacitor array) scheme.}
\label{fig:14}
\end{figure}

The voltage level of the signal coming from
slow shaper or ramp TDC cell is memorised in the T\&H capacitor (500~fF) 
so "`Track \& Hold Cell"' allows
to lock the capacitor value only when a calibrated trigger (from fast
channel) occurs within the selected column. The SCA column is selected, read and erased by
the digital part. 

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img15.jpg}
\caption{Operation of T\&H cell.}
\label{fig:15}
\end{figure}

On  \refFig{fig:15} is illustrated the T\&H cell mode of
operation: when a signal arrives in the discriminator cell is detected
and the output trigger signal is sent to the T\&H cell.
The output trigger is delayed and calibrated before being sent.


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{Charge channel}
\label{ssec:Charge}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
The charge channel is the slow channel: the signal amplified by the
variable gain preamplifier is sent to the slow shaper, a typical
$\mathrm{CRRC}^2$ filter with variable peaking time. The
peaking time can be set from 50~ns (default value) to 200~ns thanks to
the switched feedback capacitors.

On left part of \refFig{fig:16} are represented the slow shaper waveforms for
different shaping times and the same input signal. The noise value (\refTab{tab:4}
and right part of \refFig{fig:16}), from $980~\mu$V to $1.6$~mV (simulation results), foresee
good noise performance.

\begin{figure}[!htbp]
\centering
                \begin{tabular}{rl}
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img16a.jpg}&
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img16b.jpg}
                \end{tabular}
\caption{Slow shaper output waveforms simulation (left panel). Slow shaper
output noise simulation (right panel).}
\label{fig:16}
\end{figure}

\begin{table}
\centering
\caption{TO BE COMPLETED. $G_{pa} = 8$}
\label{tab:4}
\begin{tabular}{|c|c|c|c|}
\hline
Time constant & RMS  & SNR & $V_{out}(1 p.e)$ \\
\hline
50~ns & \parbox[t]{20mm}{$1.68$~mV \\ $\approx 1/17$~p.e \\ $ \approx 9$~fC}
     &  11 
                        & \parbox[t]{20mm}{$29$~mV \\ $T_p = 48$~ns } \\
100~ns & \parbox[t]{20mm}{$1.26$~mV\\$\approx 1/12$~p.e \\ $ \approx 20$~fC}
     &  8 
                        & \parbox[t]{20mm}{$15$~mV \\ $T_p = 78$~ns }\\
200~ns & \parbox[t]{20mm}{$0.98$~mV\\$\approx 1/5$~p.e \\ $ \approx 32$~fC}
     &  5 
                        & \parbox[t]{23mm}{$8$~mV \\ $ T_p = 141.5$~ns } \\
\hline                  
\end{tabular}
\end{table}

The \refFig{fig:17}  and \refTab{tab:5} illustrate the linearity performance for
different time constants.  Simulations show a good linearity with
residuals from -0.5\% to 0.2\% at $T_p = 50$~ns, from
-1\% to 0.3\% at $T_p =100$~ns and -0.7\% to 0.3\% at
$T_p=200$~ns.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img17.jpg}
\caption{Slow shaper linearity simulation.}
\label{fig:17}
\end{figure}

\begin{table}
\centering
\caption{TO BE COMPLETED}
\label{tab:5}
\begin{tabular}{|c|c|c|c|}
\hline
Time constante & $V_{out-max}$ & $Qi_{max}/n_{pe}$ & Residuals (\%) \\
\hline
 50~ns &  1.437~V &  13~pC/80~pe &  -0.5 to 0.2 \\
100~ns &  1.493~V &  24~pC/150~pe &  -1.0 to 0.3 \\
200~ns &  1.385~V &  48~pC/300~pe &  -0.7 to 0.3 \\
\hline
\end{tabular}
\end{table}

The Slow shaper maximum value, therefore the charge value, is then
memorized in the analogue memory, with a depth of 2, thanks to the
delayed trigger. \refFig{fig:18} gives the simulated slow shaper and SCA
signals.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img18.jpg}
\caption{Slow shaper \& SCA simulation.}
\label{fig:18}
\end{figure}
This charge, stored as a voltage value, is then converted in digital
value thanks to the 8/10/12 bit Wilkinson ADC.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{Time measurement}
\label{ssec:Timemeas}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
For each channel, a fine time measurement is performed by the analogue
memory with a depth of 2 which samples a 12 bit ramp (100~ns), common
for all channels, at the same time of the charge.

In \refFig{fig:19} is represented the TDC Ramp general schematic. The current,
which flows in feedback, charges the capacitance $C_f$ when the switch is
off. When the switch is turned off, $C_f$ discharges. Signals \verb|start\_ramp| and
\verb|start\_ramp\_b| manage the switches. The rising signal starts the ramp
and the falling signal stop the ramp (\refFig{fig:19}).

\begin{figure}[!htbp]
\centering
\begin{tabular}{rl}
\includegraphics[width=0.5\columnwidth,height=6cm]{img19a.jpg}&
\includegraphics[width=0.5\columnwidth,height=6cm]{img19b.jpg}
\end{tabular}
\caption{TDC Ramp general schematic.}
\label{fig:19}
\end{figure}
In order to avoid the large falling time of the ramp due to the $C_f$
discharge time and the problem of non linearity at the start and the
end of ramp signal (\refFig{fig:20}), the real ramp is created from two
ramps.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img20.jpg}
\caption{TDC Ramp.}
\label{fig:20}
\end{figure}

The signal start ramp, coming from the digital
part, enters in two delay cells. The two delayed signals create the
first and second ramps. Commutating alternatively two switches the 100~ns ramp TDC is created 
(\refFig{fig:21} and \refFig{fig:22}).

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img21.jpg}
\caption{TDC Ramp scheme.}
\label{fig:21}
\end{figure}

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img22.jpg}
\caption{TDC Ramp simulation.}
\label{fig:22}
\end{figure}

This time value, stored as a voltage value, is then converted in
digital value tanks to the 8/10/12 bit Wilkinson ADC.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{ADC ramp}
\label{ssec:ADCramp}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
In \refFig{fig:23} is represented the Ramp ADC general scheme. It is the
same as TDC ramp one, the difference is in a variable current source
which allows obtaining 8bit/10bit/12bit ADC according to the injected
current. \refTab{tab:6} gives, for each ramp, the time duration to reach 3.3~V.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img23.jpg}
\caption{ADC ramp schematic.}
\label{fig:23}
\end{figure}

\begin{table}
\centering
\caption{TO BE COMPLETED}
\label{tab:6}
\begin{tabular}{|l|l|}
\hline
 Header 1      & Header 2 \\
 12 bit ADC & From 0.9~V to 3.3~V in $102.0~\mu{}$s \\
 10 bit ADC & From 0.9~V to 3.3~V in $25.6~\mu{}$s \\
 \phantom{ }8 bit ADC & From 0.9~V to 3.3~V in $6.4~\mu{}$s \\
\hline
\end{tabular}
\end{table}

Then the ADC ramp is compared thanks to a Discriminator to the voltage
values, which corresponds to charge and fine time values, stored in the
SCA. The digital converted DATA are then treated by the digital part.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{Digital part}
\label{ssec:Digital}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
The digital part of PARISROC is built around 4 modules which are "`acquisition"', "`conversion"', "`readout"' and "`top manager"'. Actually, PARISROC is based on 2 memories. During acquisition,
discriminated analog signals are stored into an analog memory (the SCA:
switched capacitor array). The analog to digital conversion module
converts analog charges and times from SCA into 12 bits digital values.
These digital values are saved into registers (RAM). At the end of the
cycle, the RAM is readout by an external system. The block diagram is
given on \refFig{fig:24}.


\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img24.jpg}
\caption{Block diagram of the digital part.}
\label{fig:24}
\end{figure}

This sequence is made thanks to the top manager module which controls
the 3 other ones. When 1 or more channels are hit, it starts ADC
conversion and then the readout of digitized data. The maximum cycle
length is about $200~\mu$s. During
conversion and readout, acquisition is never stopped. It means that
discriminated analog signals can be stored in the SCA at any time of
the sequence shown in on \refFig{fig:25}.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img25.jpg}
\caption{Top manager sequence.}
\label{fig:25}
\end{figure}

The first module in the sequence is the acquisition
which is dedicated to charge and fine time measurements. It manages the
SCA where charge and fine time are stored as a voltage like. It also
integrates the coarse time measurement thanks to a 24-bit gray counter
with a resolution of 100~ns. Each channel has a depth of 2 for the SCA
and they are managed individually. Besides, SCA is treated like a FIFO
memory: analog voltage can be written, read and erased from this
memory.


\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img26.jpg}
\caption{SCA analogue voltage}
\label{fig:26}
\end{figure}

Then, the conversion module converts analog values stored in
the SCA (charge and fine time: cf. \`refFig{fig:26}) in digital ones thanks to a 12-bit
Wilkinson ADC. The counter clock frequency is 40~MHz, it implies a
maximum ADC conversion time of $103~\mu$s
when it overflows. This module makes 32 conversions in 1 run (16
charges and 16 fine times).

Finally, the readout module permits to empty all the registers
to an external system. As it will only transfer hit channels, this
module will tag each frame with its channel number: it works as a
selective readout. The pattern used is composed of 4 data: 4-bit
channel number, 24-bit coarse time, 12-bit charge and 12-bit fine time.
The total length of one frame is 52 bits. The maximum readout time
appears when all channels are hit. About 832 bits of data are
transferred to the concentrator with a 10~MHz clock: the readout takes
about $100~\mu$s with $1~\mu$s between 2 frames.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{ASIC Laboratory tests}
\label{sec:ASICLAbTest}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
The PARISROC has been submitted in June 2008; a first batch of 6 ASICs
has been produced and received in January 2009 (a second batch of 14
ASICs in May 2009.

The ASIC test has been a critical step in the PARISROC planning due to
the ASIC complexity.A dedicated test board has been designed and realized for this purpose
(\refFig{fig:27}). Its role is to allow the characterization of the chip and the
communication between photomultipliers and ASIC. This is possible
thanks to a dedicated Labview program that allows sending the ASIC
configuration (slow control parameters; ASIC parameters, etc) and
receiving the output bits via a USB cable connected to the test board.
The Labview is developed by LAL.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img27.jpg}
\caption{Test Board.}
\label{fig:27}
\end{figure}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{General tests}
\label{ssec:GeneralTest}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
On  \refFig{fig:28} is shown the Test Bench used in laboratory. It is composed by a
test board, a signal generator, an oscilloscope, multimeters and PC to
run labview program.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img28.jpg}
\caption{Test Bench.}
\label{fig:28}
\end{figure}

The signal generator is a TEKTRONIX single
channel function generator. It is used to create the input charge
injected in the ASIC. The signal injected has the shaping as similar as
possible to the PMT signal. On \refFig{fig:28} is represented the generator input
signal and its characteristics.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img29.jpg}
%%%% NOT USED \includegraphics[width=0.5\columnwidth,height=6cm]{img34.jpg}
\caption{Input signals}
\label{fig:29}
\end{figure}

At the beginning all the standard electrical
characteristics have been tested: DC levels, analogue output signals,
the analogue part characteristics and then the pedestals, the DAC
linearity, S\-curves (trigger efficiency as a function of the injected
charge or the threshold), the ADC linearity. The first purpose is the
comparison between simulation results and test measurements; most of
them are in agreement with the ASIC characteristics, obtained in
simulation.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{Analogue tests}
\label{ssec:AnalogueTest}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
The DC level characterization is the first step in ASIC
characterization; in particular the DC uniformity of the analogue part
DC level for the different channels has to be measured.

In \refFig{fig:30} are represented the preamplifier, slow
shaper and fast shaper DC uniformity plots. The DC uniformity test has a small dispersion
of 0.4\%, 0.1\% and 0.05\% respectively for the preamplifier, the slow
shaper and the fast shaper (\refTab{tab:7}).

\begin{figure}[!htbp]
\centering
\begin{tabular}{c}
\includegraphics[width=0.7\columnwidth,height=6cm]{img30a.jpg}\\
\includegraphics[width=0.7\columnwidth,height=6cm]{img30b.jpg}\\
\includegraphics[width=0.7\columnwidth,height=6cm]{img30c.jpg}
\end{tabular}
\caption{DC uniformity.}
\label{fig:30}
\end{figure}

\begin{table}
\centering
\caption{TO BE COMPLETED}
\label{tab:7}
\begin{tabular}{|l|c|c|c|}
\hline
DC level & RMS \\
Preamplifier & 3.8~mV (0.40~\%) \\ 
Slow shaper  & 1.3~mV (0.10~\%) \\
Fast shaper  & 1.0~mV (0.05\%)  \\
\hline
\end{tabular}
\end{table}

The second step is the analogue part output signals: Injecting a
charge equivalent to 10~pe, and setting a preamplifier gain at 8, are
observed and compared with simulation results all the output waveforms.

There is a good agreement in preamplifier results ( \refFig{fig:31} and \refTab{tab:8}), the
amplitude has the same value while time rise value has a difference of
3~ns. This difference is due to the output buffer placed in the test
board.

\begin{figure}[!htbp]
\centering
                \begin{tabular}{rl}
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img31a.jpg}&
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img31b.jpg}
                \end{tabular}
\caption{Measurement and simulation of the preamplifier output for
an input charge of 10~pe.}
\label{fig:31}
\end{figure}

\begin{table}
\centering
\caption{TO BE COMPLETED. Preamplifier parameters.... $G_{pa} = 8$. WHY not same parameters 1~pe and 10~p.e}
\label{tab:8}
\begin{tabular}{|l|c|c|}
\hline
             &  Measurement    & Simulation \\
\hline
Maximum voltage (10~pe) & 50.00~mV  & 50.83~mV \\
Rise time (10~pe) & 7.78~ns & 4.79~ns \\
RMS noise &       1~mV       & 0.47~mV \\
without USB cable & 0.66~mV  &         \\
Noise in pe   & 0.2  & 0.086 \\
without USB cable & 0.132 &         \\
Maximum voltage (1~pe) & 5.00~mV  & 5.43~mV \\
SNR (1~pe ????) & 5 & 11.6 \\
without USB cable & 7.5 &         \\
\hline
\end{tabular}
\end{table}

The slow shaper waveforms are shown in \refFig{fig:32} while \refTab{tab:9} 
summarizes the results. The first differences appear: a different value
in amplitude for slow shaper signal and fast shaper signal that is
probably associate, also, to the Output Buffer. The second relevant
difference is in noise value, in particular in slow shaper noise
performance (\refTab{tab:9}).

\begin{figure}[!htbp]
\centering
                \begin{tabular}{rl}
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img32a.jpg}&
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img32b.jpg}
                \end{tabular}
\caption{Measurement and simulation of the slow shaper output for an
input charge of 10~pe.}
\label{fig:32}
\end{figure}

\begin{table}
\centering
\caption{TO BE COMPLETED. $G_{pa} = 8$ and $RC = 50$~ns.}
\label{tab:9}
\begin{tabular}{|l|c|c|}
\hline
             &  Measurement    & Simulation \\
\hline
Maximum Voltage (10~pe) & 117~mV & 290~mV \\
Rise time (10~pe) & 18.0~ns & 19.1~ns \\
RMS noise &  4.0~mV & 1.7~mV \\
Noise in pe &  0.3 & 0.08 \\
Maximum Voltage (1~pe) & 12~mV & 19~mV \\
SNR &  3 & 11  \\
\hline
\end{tabular}
\end{table}

The Fast shaper results are shown in \refFig{fig:33}
and \refTab{tab:10}.
\begin{figure}[!htb]
        \centering
                \begin{tabular}{rl}
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img33a.jpg}&
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img33b.jpg}
                \end{tabular}
        \caption{Measurement and simulation of the fast shaper output for an
input charge of 1 pe.}
        \label{fig:33}
\end{figure}


\begin{table}
\centering
\caption{TO BE COMPLETED. $G_{pa} = 8$.}
\label{tab:10}
\begin{tabular}{|l|c|c|}
\hline
             &  Measurement    & Simulation \\
\hline
RMS noise &  2.5~mV & 2.4~mV \\
Noise in pe &  0.08 & 0.05 \\
Maximum Voltage (1~pe) & 30~mV & 42~mV \\
SNR &  12 & 18  \\
\hline
\end{tabular}
\end{table}
Another important characteristic is the
linearity. The output voltage in function of the input injected charge
is plotted for the different analogue signals. \refFig{fig:34} gives few examples for
the preamplifier at different gains. \refTab{tab:11} summarizes the fit
results of these linearities. Good linearity performances are shown by
residuals (better than $\pm 2~\%$) value but for a
smaller dynamic range than simulation.

\begin{figure}[!htbp]
\centering
                \begin{tabular}{c}
                        \includegraphics[width=0.7\columnwidth,height=6cm]{img34a.jpg}\\
                        \includegraphics[width=0.7\columnwidth,height=6cm]{img34b.jpg}\\
                        \includegraphics[width=0.7\columnwidth,height=6cm]{img34c.jpg}\\
                \end{tabular}
\caption{Preamplifier linearity for different gains.}
\label{fig:34}
\end{figure}

\begin{table}
\centering
        \caption{TO BE COMPLETED}
        \label{tab:11}
\begin{tabular}{|c|c|c|c|}
\hline
Preamplifier Gains & Maximum voltage & Charge/Nb of pe & Residuals \\
\hline
8                  &   0.52~V        & 12~pC / 78~pe & -1.0~\% to 0.8~\% \\
4                  &   0.64~V        & 32~pC / 198~pe & -1.0~\% to 1.0~\% \\
2                  &   0.51~V        & 50~pC / 312~pe & -2.0~\% to 1.5~\% \\
\hline
\end{tabular}
\end{table}


\refFig{fig:35} represents an example of slow shaper
linearity for a time constant of 50~ns and  a preamplifier gain of 8
with residuals better than $pm 1~\%$.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img35.jpg}
\caption{Slow shaper linearity; $RC =50$~ns and $G_{pa}=8$.}
\label{fig:35}
\end{figure}

\refFig{fig:36} gives an example of the fast shaper linearity until an injected
charge of 10~pe. Residuals better than $ \pm 2~\%$
are obtained.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img36.jpg}
\caption{Fast shaper linearity up to 10~pe.}
\label{fig:36}
\end{figure}

The preamplifier linearity in function of
variable feedback capacitor value with an input charge of 10~pe and
with residuals from $-2.5~\%$ to $1.4~\%$ is represented on \refFig{fig:37} . The gain
adjustment linearity is nice at 2~\% on 8 bits.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img37.jpg}
\caption{Preamplifier linearity vs feedback capacitor value.}
\label{fig:37}
\end{figure}

On \refFig{fig:38}  is given the gain uniformity. For the
different preamplifier gains is plotted the maximum voltage value for
all channels in order to investigate the homogeneity among the whole
chip, essential for a multichannels ASIC. Residual dispersion of 0.05~\%,
0.013~\% and 0.012~\% have respectively been obtained for gain 8, 4 and
2.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img38.jpg}
\caption{Gain uniformity for $G_{pa}=8, 4, 2$.}
\label{fig:38}
\end{figure}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{DAC linearity}
\label{ssec:DAClinearity}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
The DAC linearity has been measured and it consists in measuring the
voltage DAC ($V_{dac}$) amplitude obtained for different DAC register
values. \refFig{fig:39} gives the evolution of $V_{dac}$ as a function of the register for the two
DACs and residuals from $-0.1~\%$ to $0.1~\%$.

\begin{figure}[!htbp]
\centering
                \begin{tabular}{rl}
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img39a.jpg}&
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img39b.jpg}
                \end{tabular}
\caption{DAC linearity; DAC1 and DAC2 respectively.}
\label{fig:39}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{Trigger output}
\label{ssec:TriggerMeas}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
The trigger output behavior was studied scanning the threshold for
different injected charges. At first no charge was injected which
corresponds to measure the fast shaper pedestal. The result is
represented on \refFig{fig:40}  for each channel. The  S-curves 
are superimposed meaning good homogeneity. The spread
is of one DAC count ($LSB DAC = 1.78$~mV) or 0.06~pe.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img40.jpg}
\caption{Pedestal S-curves for channel 1 to 16.}
\label{fig:40}
\end{figure}

The trigger efficiency was then measured for a
fixed injected charge of 10~pe. On \refFig{fig:41} are represented the S-curves
obtained with 200 measurements of the trigger for all channels varying
the threshold. The homogeneity is proved by a spread of 7 DAC unit (0.4~pe) and a noise of 0.07 pe ($RMS =2.19$).

\begin{figure}[!htbp]
\centering
                \begin{tabular}{rl}
                        \multicolumn{2}{c}{\includegraphics[width=0.5\columnwidth,height=6cm]{img41a.jpg}}\\
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img41b.jpg}&
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img41c.jpg}
                \end{tabular}
\caption{Fast shaper and trigger (top panel); S-curves for input of 10~pe (left panel);
uniformity plot for channel 1 to 16 (right panel).}
\label{fig:41}
\end{figure}

The trigger output is studied also by scanning
the threshold for a fixed channel and changing the injected charge. On \refFig{fig:42}
on the left panel  is shown the trigger efficiency versus the DAC unit and on
the right panel is plotted the threshold versus the injected charge but only
until 0.5~pC. From these measurements a noise of 10~fC has been
extrapolated. Therefore the threshold is only possible above $10~\sigma$ of the noise due to the discriminator coupling
(\refFig{fig:43}).

\begin{figure}[!htbp]
\centering
                \begin{tabular}{rl}
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img42a.jpg}&
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img42b.jpg}
                \end{tabular}
\caption{Trigger efficiency vs DAC count up to 300~pe (left panel) and
until 3~pe (right panel).}
\label{fig:42}
\end{figure}

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img43.jpg}
\caption{Threshold vs injected charge up to 500~fC. It is shown the 1~p.e threshold for a PMT gain of $10^6$.}
\label{fig:43}
\end{figure}

The trigger coupling illustrated in \refFig{fig:44} with the
injected charge in channel 1 and output signal observed in channel 2,
shows a coupling signal around 25~mV (10~fC). This coupling signal is
due, probably, to the input power supply ($V_{dd-pa}$ and $V_{ss}$).

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img44.jpg}
\caption{Trigger coupling signal.}
\label{fig:44}
\end{figure}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{ADC characterisation}
\label{ssec:ADCMeas}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
The ADC performance has been studied alone and with the whole chain. Injecting to the
 ADC input directly a DC voltage by the internal DAC,
in order to have a voltage level as stable as possible, were measured
the ADC values for all channels (\refFig{fig:45}).

The measurement is repeated 10000 times for
each channel and in the first plot of the LabView front panel window (\refFig{fig:45}). The
minimal, maximal and mean values, over all acquisitions, for each
channel are plotted. In the second plot there is the rms charge value
versus channel number with a value in the range $[0.5, 1]$ ADC unit.
Finally the third plot shows an example of charge amplitude
distribution for a single channel: a spread of 5 ADC counts is
obtained.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img45.jpg}
\caption{ADC measurements with DC input 1.45~V (middle scale).}
\label{fig:45}
\end{figure}

The ADC is suited to a multichannel conversion
so the uniformity and linearity are studied in order to characterize
the ADC behaviour. On \refFig{fig:46} is represented the ADC transfer function for the
10-bit ADC versus the input voltage level. All channels are represented
and have plots superimposed. 

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img46.jpg}
\caption{10  bits ADC transfer function vs input charge.}
\label{fig:46}
\end{figure}

The good homogeneity observed is confirmed by
the linear fit parameters comparison. In  are plotted the slope and the
intercept distributions for all channels. The RMS slope value of 0.143
and the RMS intercept value of 0.3 confirm the 10-bits ADC uniformity
(\refTab{tab:12}).

\begin{figure}[!htbp]
\centering
                \begin{tabular}{rl}
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img47a.jpg}&
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img47b.jpg}
                \end{tabular}
\caption{Evolution of the fit parameters (slope on the
left panel and intercept on the right panel) as a function of the channel
number.}
\label{fig:47}
\end{figure}

\begin{table}
\centering
\caption{TO BE COMPLETED. 10 bits ADC parameter fits.... 25 acquisitions per channel, $LSB = 1.06$~mV...}
\label{tab:12}
\begin{tabular}{|l|c|c|}
\hline
   & Slope      & Intercept \\
Mean & 936.17   & 859.8 \\
RMS  & 0.14     & 0.3   \\ 
\hline
\end{tabular}
\end{table}

In \refFig{fig:48} are shown respectively the 12, 10 and 8 bits ADC
linearity plots with the 25 measurements made for each input voltage
level. The average ADC count value is plotted versus the input signal.
The residuals from $-1.5$ to $0.9$ ADC units for the 12-bits ADC; from $-0.5$
to $0.4$ for the 10-bit ADC and from $-0.5$ to $0.5$ for the 8-bit ADC. This prove
the good ADC behaviour in terms of Integral non linearity. 

\begin{figure}[!htbp]
\centering
                \begin{tabular}{c}
                        \includegraphics[width=0.7\columnwidth,height=6cm]{img48a.jpg}\\
                        \includegraphics[width=0.7\columnwidth,height=6cm]{img48b.jpg}\\
                        \includegraphics[width=0.7\columnwidth,height=6cm]{img48c.jpg}
                \end{tabular}
\caption{12, 10, 8 bit ADC linearity.}
\label{fig:48}
\end{figure}
In terms of Differential non linearity, the
value from $-1.0$ to $0.65$ for the 10 bit ADC and from $-0.3$ to $0.2$ for the 8
bit ADC, show us a good behaviour even if the plots are the results of
preliminary measurements.

\begin{figure}[!htbp]
        \centering
                \begin{tabular}{rl}
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img49a.jpg}&
                        \includegraphics[width=0.5\columnwidth,height=6cm]{img49b.jpg}
                \end{tabular}
\caption{Differential non linearity.}
\label{fig:49}
\end{figure}

Once the ADC performances have been tested
separately, the measurements are performed on the complete chain. The
results of the input signal autotriggered, held in the T\&H and
converted in the ADC are illustrated in  where are plotted the 10-bit
ADC counts in function of the variable input charge (up to 50~pe). A
good linearity of $1.4~\%$ and a noise of 6 ADC units are obtained. In \refTab{tab:13}
are listed the setting value for measurements.

\begin{table}
        \centering
        \caption{TO BE COMPELTED. $G_{pa}=14$ ($C_{in}=7$~pF , $C_f=0.5$~pF),
Slow shaper $RC=50$~ns,
DAC delay: $bit<0> = 1$ \& $bit<2> = 1$.
}
        \label{tab:13}
\begin{tabular}{|l|c|c|c|}
\hline
Parameters & 12 bits ADC & 10 bits ADC & 8 bits ADC\\
\hline 
LSB         & $0.27$ & $1.06$~mV  & $4.26$~mV\\
Min ADC count at 3~pe& $509$ &  $132$ & $33$  \\
Max ADC count at 50~pe & $3873$ &  $989$ & $241$ \\
Residuals in ADC units &$[21,54]$ & $[6,14]$  & $[2,3]$ \\
\hline
\end{tabular}
\end{table}

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img50.jpg}
\caption{10 bit ADC linearity.}
\label{fig:50}
\end{figure}

On \refFig{fig:51} is plotted the 8-bit linearity at $1.4~\%$
and a noise of 1.53 ADC unit. In \refTab{tab:13} are listed the setting value for
measurements.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img51.jpg}
\caption{8 bit ADC linearity.}
\label{fig:51}
\end{figure}

On  \refFig{fig:53} is plotted the 12-bit linearity
at $1.4~\%$ and a noise of 23.69 ADC unit. In \refTab{tab:13} are listed the setting
value for measurements.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img52.jpg}
\caption{12 bit ADC linearity.}
\label{fig:52}
\end{figure}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Measurements with PMTs}
\label{sec:MeasWithPMT}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
The first measurements with a photomultiplier at input are started in
IPNO at Orsay.

\begin{figure}[!htbp]
\centering
\includegraphics[width=0.7\columnwidth]{img53.jpg}
\caption{TO BE COMPLETED}
\label{fig:53}
\end{figure}

\acknowledgments
%\begin{acknowledgments}
This work, especially one of the author, is supported by the National Reasaerch Agency under contract ANR-06-BLAN-0186.
%\end{acknowledgments}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\newpage
%\section*{References}
\bibliography{campagne}
\end{document}