<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Archiving and Interchange DTD v1.0 20120330//EN" "JATS-archivearticle1.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink">
  <front>
    <journal-meta />
    <article-meta>
      <title-group>
        <article-title>Robert Wolffa on behalf of the ATLAS Liquid Argon Calorimeter group</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <string-name>Aix-Marseille Université</string-name>
        </contrib>
        <contrib contrib-type="author">
          <string-name>CNRS/IN</string-name>
        </contrib>
        <contrib contrib-type="author">
          <string-name>Marseille</string-name>
        </contrib>
        <contrib contrib-type="author">
          <string-name>France</string-name>
        </contrib>
      </contrib-group>
      <pub-date>
        <year>2017</year>
      </pub-date>
      <fpage>53</fpage>
      <lpage>60</lpage>
      <abstract>
        <p>The upgrade of the Large Hadron Collider, scheduled for 2019-2020, will increase the instantaneous luminosity by more than three, hence the ATLAS trigger rates. To cope with this increase, the trigger signals from the ATLAS Liquid Argon Calorimeter will be rearranged in 34000 so-called super cells to get a 5 to 10 times finer granularity. This will improve the background rejection performance through more precise energy measurements and the use of shower shape information to discriminate electrons, photons and hadronically decaying tau leptons from jets. The new system will process the super cell signal at 40 MHz and with 12 bit precision. The data will be transmitted at 5.12 Gb/s to the back-end system using a custom serializer and optical transmitter. To verify full functionality, a demonstrator set-up has been installed on the ATLAS detector and operated during the LHC Run 2. This document gives a status on hardware developments towards the final design readout system, including the performance of the newly developed ASICs. Their radiation tolerance, the performance of the prototype boards, results of the high-speed link test with the prototypes and the performance of the demonstrator with collision data are also reported.</p>
      </abstract>
      <kwd-group>
        <kwd>ATLAS</kwd>
        <kwd>Calorimeter</kwd>
        <kwd>LAr</kwd>
        <kwd>Liquid Argon</kwd>
        <kwd>Phase-1 upgrade</kwd>
      </kwd-group>
    </article-meta>
  </front>
  <body>
    <sec id="sec-1">
      <title>1. Introduction</title>
      <sec id="sec-1-1">
        <title>1.1. The ATLAS experiment</title>
        <p>
          The ATLAS experiment [
          <xref ref-type="bibr" rid="ref1">1</xref>
          ] is a multi-purpose high energy physics experiment at the Large
Hadron Collider (LHC) with the goal to precisely measure the Standard Model of particle physics
(SM) and its extensions. The ATLAS detector consists of sensor layers surrounding the collision
point: an inner detector made of pixel, strip and transition radiation detectors, followed by
electromagnetic and hadronic calorimeters, and then, at the outermost, the muon chambers. Due to a
solenoid and toroidal magnet system of magnetic fields of up to 2 T, momenta of charged particles,
e.g. muons can be precisely measured. More than 80 million channels are read out for each triggered
event. Proton bunches collide in the experiment at a rate of 40 MHz with an instantaneous luminosity
of up to 1.74 × 10−34 cm−2s−1 and on average 25 interactions per bunch crossing (&lt;μ&gt; = 25). The
ATLAS data acquisition is selecting proton-proton collision events using a two-step trigger system.
The first level trigger (L1) selects events based on information from coarse-granularity calorimeters
and muon spectrometers with a limited bandwidth of 100 kHz and a fixed latency of less than 3 μs.
The high level trigger (HLT) further decreases the event rate further down to about 600 Hz to
1.5 kHz using inputs from the whole detector with full granularity.
        </p>
      </sec>
      <sec id="sec-1-2">
        <title>1.2. The Liquid Argon Calorimeter</title>
        <p>The Liquid Argon Calorimeter (LAr) is a sampling calorimeter with liquid argon as the
active medium. Its scheme is shown in Figure 1. In the electromagnetic barrel and end-caps (EMB
and EMEC) it consists of accordion shaped lead absorbers and copper/kapton electrodes. The
hadronic end-caps (HEC) use copper absorbers and copper/kapton electrodes and the forward
calorimeter (FCAL) involves copper absorbers in the electromagnetic section and tungsten absorbers
in the hadronic section. The LAr Calorimeter plays a crucial role in the electron and photon
reconstruction as well as jet identification and missing transverse energy measurement. It consists of
182000 readout channels with 1600 front-end boards (FEBs) and 200 readout driver boards (RODs)
at back-end.</p>
        <p>The LAr Calorimeter is currently read out at two stages, which are the regular and the trigger
readout. In the regular readout the calorimeter cell signals are amplified, shaped and sampled in the
FEBs for each bunch-crossing at 40 MHz and digitised and transmitted upon a L1 trigger decision at
100 kHz. The trigger readout uses a reduced granularity by summing signals from calorimeter cells
on the Layer Sum Boards (LSBs) and on Tower Builder Boards (TBB) to form analog trigger tower
signals, which are sent off-detector to the L1 Calorimeter Trigger System (L1Calo) for each
bunchcrossing at 40 MHz.</p>
      </sec>
      <sec id="sec-1-3">
        <title>1.3. The Phase-1 upgrade</title>
        <p>The plan for LHC and high-luminosity LHC (HL-LHC) is presented in Figure 2. The
currently on-going Run 2 will be followed by the long shut-down 2 (LS2) during 2019/2020, in
which the Phase-1 upgrade system will be installed. The Phase-2 upgrade is foreseen for 2024–2026
during LS3.</p>
      </sec>
      <sec id="sec-1-4">
        <title>1.4. The Phase-1 LAr Calorimeter readout electronics upgrade</title>
        <p>
          The current L1 trigger system uses trigger towers with a 0.1 × 0.1 pseudo-rapidity × polar
angle size in the EMB as input from the LAr Calorimeter. The Phase-1 upgrade of the LAr
Calorimeter readout electronics [
          <xref ref-type="bibr" rid="ref3">3</xref>
          ] foresees an increase of granularity by a factor of 5 to 10, resulting
in the so-called super cells. Longitudinal shower information is added by separating the four layers in
the new readout format. This is illustrated for a shower of an electron in Figure 3. New discriminant
variables can improve to distinguish interesting physics objects as electrons and hadronic decaying
tau leptons from jets at the L1 trigger level.
        </p>
      </sec>
    </sec>
    <sec id="sec-2">
      <title>2. The front-end electronics upgrade</title>
      <sec id="sec-2-1">
        <title>2.1. The base-plane, the FEBs and the LSBs</title>
        <p>The connectivity, cross-talk and noise performance has been verified for the standard
baseplanes for EMB and EMEC with prototypes and the production has started. The development of the
special base-planes for EMEC, HEC and FCAL is progressing. The procurement of common parts
such as ground springs and alignment pins is ongoing. The FEBs will remain unchanged until
Phase2, but the LSBs will be replaced to comply with super cell signal summing. Their production is
ongoing.</p>
      </sec>
      <sec id="sec-2-2">
        <title>2.2. The LTDB</title>
        <p>There are a total of 124 LTDBs reading 34 thousand super cells to be installed. Each LTDB
reads up to 320 super cells. It digitizes analog signals with 12 bits at 40 MHz using custom
analogto-digital converters (ADCs). The digital signals are transmitted using 40 optical links at 5.12 Gb/s
with custom application-specific integrated circuits (ASICs). It has 5 GigaBit Transceiver (GBT)
serializer-de-serializer links for trigger, timing and control (TTC) signals. Its power distribution
board (PDB) has to comply with the total power consumption of around 125 W.</p>
        <p>The whole system is on the detector and thus, all components need to be radiation-tolerant. It
has full compatibility with the Phase-2 upgrade. The final design has been fixed and prototypes are
being produced.</p>
      </sec>
      <sec id="sec-2-3">
        <title>2.3. The LTDB custom ADCs</title>
        <p>Each LTDB requires 80 custom ADCs that continuously process the sampling and
digitization of four super cell signals at 40 MHz. The power dissipation is less than 50 mW per
channel and the latency must be less than 200 ns. The dynamic range is 11.7 bits per sample.</p>
        <p>The Nevis13 ADC fulfils all requirements. Its layout is based on a 130 nm IBM CMOS 8RF
of 3.6 mm × 3.6 mm with 72 quad-flat no-leads (QFN) pins. It uses four multiplying
digital-toanalog converters (MDACs) for the most significant bits and one successive approximation ADC
(SAR) for the lower 8 bits. It is tested for radiation up to 10 Mrad which corresponds to 100 times
more than currently expected at HL-LHC.</p>
      </sec>
      <sec id="sec-2-4">
        <title>2.4. The LTDB optical links</title>
        <p>Two kinds ASICs for optical links on the LTDB are developed. The serializer (LOCx2) is
based on a 250 nm silicon-on-sapphire technology with a die of 6.0 mm × 3.7 mm and 100 QFN
pins. Its output is at 5.12 Gb/s at a latency of less than 75 ns. It has about 1 W of power consumption.
The laser driver (LOCld) uses the same technology as the LOCx2 on a die of 2.1 mm × 1.1 mm and
40 QFN pins. It is a dual-channel vertical-cavity surface-emitting laser (VCSEL) driver.</p>
        <p>Both ASICs have been tested on radiation-tolerance. Only few change in the output eye
diagrams has been observed after about 200 kHz. The wafers are produced and tests on the LTDB
prototypes are ongoing.</p>
      </sec>
    </sec>
    <sec id="sec-3">
      <title>3. The back-end electronics upgrade</title>
      <sec id="sec-3-1">
        <title>3.1. The LDPS</title>
        <p>The LDPS receives the digital 12-bit data from the front-end system. It calculates the
transverse energy of each super cell and various sums of super cells at fixed latency and transmits the
results at 40 MHz to the L1Calo feature extractors. Additionally it buffers the calculated results for a
readout upon L1 trigger decision for debugging and monitoring. Because the back-end LDPS is
offdetector, there is no need to be radiation-tolerant as the front-end system.</p>
        <p>
          Its main component are about 30 LAr Digital Processing Blades (LDPBs) which read the
outputs of 124 LTDBs. Each LDPB consists of one LAr carrier board (LArC) with four advanced
mezzanine cards (AMCs) and a rear transition module (RTM). It is a custom advanced
telecommunications computing architecture (ATCA) board and uses a Xilinx Virtex7 FPGA. It carries
four AMCs and drives the communication to the FELIX system [
          <xref ref-type="bibr" rid="ref4">4</xref>
          ] and the data acquisition for the
data monitoring. The AMCs have 48 input fibers at 5.12 Gb/s and the same amount of output fibers
at 11.2 Gb/s. They are called LATOME, acronym for LAr trigger processing mezzanine.
        </p>
      </sec>
      <sec id="sec-3-2">
        <title>3.2. The LATOME</title>
        <p>
          The LATOME receives super cell data from the LTDBs at 5.12 Gb/s on up to 48 optical
links. It is responsible for the computation of transverse energies of super cells and sums of super
cells using optimal filtering algorithm (OF) [
          <xref ref-type="bibr" rid="ref5">5</xref>
          ] at fixed latency. It assigns bunch crossings by timing
measurement. Finally it sends the data at 11.2 Gb/s on up to 48 optical links to the L1 trigger.
Furthermore it monitors the data and sends report to the DAQ system upon request. In Figure 5 a
picture of the LATOME board prototype is shown.
        </p>
      </sec>
    </sec>
    <sec id="sec-4">
      <title>4. The LAr demonstrator</title>
      <p>The LAr demonstrator is a pre-prototype of the calorimeter readout of the L1 trigger
processors installed in 2014 during LS1 at the ATLAS experiment. It covers 3.1 % of EMB and is
located in pseudo-rapidity 0 &lt; η &lt; 1.4 and azimuthal angle 1.77 &lt; φ &lt; 2.16. Before its installation no
disturbance to the current system has been verified. It validates the energy reconstruction and
bunchcrossing identification development. The system is successfully calibrated and data from
protonproton (and heavy-ion) collisions is taken during the on-going Run 2. Two LTDB prototypes provide
the analog summing of calorimeter cell signals to super cells at the rate of 40 MHz. The AMC
prototypes are the so-called ABBA (ATCA board for a baseline of liquid argon acquisition) boards. It
is planned to replace the two LTDBs and the LDPS by final prototypes in early 2018.</p>
      <sec id="sec-4-1">
        <title>4.1. The LTDB prototypes</title>
        <p>The two LTDBs handle each up to 320 super cell signals (284 super cells in EMB). The super
cell signals are digitized with a commercial, not radiation-tolerant 12 bit ADC (TI ADS5272). On
one 4.8 Gb/s optical link eight super cell signals are multiplexed. Two prototypes with different
technology have been developed by several institutes. The first one uses an analog mezzanine and a
digital main board and the second one uses a digital mezzanine together with an analog main board.
Both versions are successfully operated during Run 2.</p>
      </sec>
      <sec id="sec-4-2">
        <title>4.2. The ABBA boards</title>
        <p>The ABBA board receives the digital signals of up to 320 super cells from one LTDB on up to
48 optical links at 4.8 Gb/s. It stores the ADC super cell data in circular buffers and sends it using the
IP-based protocol IPbus over user datagram protocol (UDP) on a 10 Gb Ethernet upon a L1 trigger
accept in the so-called “monitoring mode”. It contains three Intel (former Altera) Stratix4 FPGAs,
which is shown in Figure 6.</p>
      </sec>
      <sec id="sec-4-3">
        <title>4.3. The LAr demonstrator operation</title>
        <p>In calibration electronic pulses are sent by calibration board and measured by the
demonstrator readout. Figure 7 shows ADC pulse measurements and noise level for different super
cells. Good linearity has been observed up to DAC 8000, beyond that analog saturation is expected.
The noise level is well below 1 ADC count for all super cells and consistent with test bench
measurements.</p>
      </sec>
    </sec>
    <sec id="sec-5">
      <title>5. Conclusion</title>
      <p>The ATLAS LAr calorimeter electronics will be upgraded during the LS2 (2019-2020). The
trigger path will be digitized at the front-end level with increased granularity. New LTDB (front-end)
and LDPS (back-end) systems have been developed. The digitization and readout of this system will
be done at 40 MHz. Specific radiation tolerant ADCs and optical links have been designed and tested
for the LTDBs. The production will be started in 2018.</p>
      <p>A demonstrator system has been installed and successfully run since 2015 for data-taking
from proton-proton and heavy-ion collisions. This gives valuable data to study the filtering algorithm
development for super cell energy measurement. It is planned to replace the LTDB and LDPS with
final prototypes in early 2018 to test the full pre-production readout chain with proton-proton
collision data in 2018. The Phase-1 upgrade is a stepping stone towards the full readout upgrade in
Phase-2.</p>
    </sec>
  </body>
  <back>
    <ref-list>
      <ref id="ref1">
        <mixed-citation>
          [1]
          <string-name>
            <surname>The</surname>
            <given-names>ATLAS</given-names>
          </string-name>
          <string-name>
            <surname>Collaboration. The ATLAS</surname>
          </string-name>
          <article-title>Experiment at the CERN Large Hadron Collider, JINST 3</article-title>
          ,
          <issue>S08003</issue>
          (
          <year>2008</year>
          ), doi.org/10.1088/
          <fpage>1748</fpage>
          -0221/3/08/S08003, cds.cern.ch/record/1129811.
        </mixed-citation>
      </ref>
      <ref id="ref2">
        <mixed-citation>
          [2]
          <string-name>
            <given-names>G.</given-names>
            <surname>Apollinari</surname>
          </string-name>
          et al.
          <article-title>High-Luminosity Large Hadron Collider (HL-LHC)</article-title>
          .
          <source>Technical Design Report V. 0</source>
          .1,
          <string-name>
            <given-names>CERN</given-names>
            <surname>Yellow</surname>
          </string-name>
          <article-title>Reports: Monographs</article-title>
          , Vol.
          <volume>4</volume>
          /2017, CERN-2017
          <string-name>
            <surname>-</surname>
          </string-name>
          007-M (
          <year>2017</year>
          ), doi.org/ 10.23731/CYRM-2017-004, cds.cern.ch/record/2284929.
        </mixed-citation>
      </ref>
      <ref id="ref3">
        <mixed-citation>
          [3]
          <string-name>
            <given-names>M.</given-names>
            <surname>Aleksa</surname>
          </string-name>
          et al.
          <source>ATLAS Liquid Argon Calorimeter Phase-I Upgrade Technical Design Report, LHCC</source>
          <year>2013</year>
          ,
          <string-name>
            <surname>CERN-LHCC-</surname>
          </string-name>
          2013-
          <volume>017</volume>
          (
          <year>2013</year>
          ), cds.cern.ch/record/1602230.
        </mixed-citation>
      </ref>
      <ref id="ref4">
        <mixed-citation>
          [4]
          <string-name>
            <given-names>J. T.</given-names>
            <surname>Anderson</surname>
          </string-name>
          et al.
          <article-title>FELIX: a PCIe based high-throughput approach for interfacing front-end and trigger electronics in the ATLAS Upgrade framework</article-title>
          ,
          <source>TWEPP</source>
          <year>2016</year>
          , JINST 11,
          <issue>C12023</issue>
          (
          <year>2016</year>
          ), doi.org/10.1088/
          <fpage>1748</fpage>
          -0221/11/12/C12023, cds.cern.ch/record/2229597.
        </mixed-citation>
      </ref>
      <ref id="ref5">
        <mixed-citation>
          [5]
          <string-name>
            <given-names>W. E.</given-names>
            <surname>Cleland</surname>
          </string-name>
          and
          <string-name>
            <given-names>E. G.</given-names>
            <surname>Stern</surname>
          </string-name>
          .
          <article-title>Signal processing considerations for liquid ionization calorimeters in a high rate environment</article-title>
          ,
          <source>Nucl. Instr. and Meth. Phys. Res., A338</source>
          <volume>467</volume>
          -
          <fpage>497</fpage>
          (
          <year>1994</year>
          ), doi.org/10.1016/
          <fpage>0168</fpage>
          -
          <lpage>9002</lpage>
          (
          <issue>94</issue>
          )
          <fpage>91332</fpage>
          -
          <lpage>3</lpage>
          .
        </mixed-citation>
      </ref>
    </ref-list>
  </back>
</article>