SEU tests of an 80Mbit/s optical receiver
F. Faccio, K. Gill, M. Huhtinen, A. Marchioro,
P. Moreira, F. Vasey
CERN, CH-1211 Geneva 23, Switzerland
G. Berger
Cyclotron Research Center, UCL, B-1348 Louvain-la-Neuve,
Belgium
Abstract:
The sensitivity to SEU is presented for a rad-hard 80Mbit/s receiver developed for the CMS Tracker digital optical link. Bit Error Rate (BER) measurements were made while irradiating with 59MeV protons and 62MeV neutrons, for different angles to the beam and for a wide range of optical power in the link. The photodiode is the most sensitive element to SEU. Direct ionisation can explain the SEU rate for protons incident at high angles of incidence and nuclear interactions explain the SEU rate for incident neutrons, as well as for protons for the low angles of incidence and higher optical power.
Summary:
The LHC experiments will use thousands of digital optical link channels for the transmission of timing, trigger and control signals. A large number of optical receivers will seat inside the detector, operating in a radiation environment. Radiation will threaten the correct operation of the optical receivers by degrading the performance of the photodiode, the receiver circuit, or by inducing errors (Single Event Upsets) in the data transmission. To study the sensitivity to radiation-induced errors of the rad-hard 80Mbit/s receiver developed for the CMS Tracker digital optical link, we have exposed the receiver to beams of 59 MeV protons and 62 MeV neutrons.
We performed BER measurements at CERN and at the irradiation facility (CRC, in Louvain-la-Neuve) to verify the reliability of our measurement system in the absence of irradiation. Then, we repeated the measurement in presence of proton and neutron beams. The measurements confirm that the most sensitive element to SEU is the PIN photodiode and not the receiver chip. This was expected on the basis of previous works and of the much greater sensitive volume for charge collection of the photodiode (80 µm in diameter and 2 µm in thickness). Errors induced by ionization in the photodiode occur only during the transmission of a '0' symbol.
By changing the angle of the photodiode to the beam, and by comparing proton and neutron irradiation results, we were able to distinguish between errors induced by direct ionization (from the protons) and by nuclear interactions in the diode (from both protons and neutrons). Protons incident at high angle, that is almost parallel to the diameter of the photodiode, have a sufficiently long trajectory in the sensitive volume to induce an error via direct ionization. Instead, when the angle is low (45o or less), the path-length of the protons in the sensitive volume is only a few microns at most, and the energy deposited via direct ionization is not sufficient to cause SEU. In this case, errors are dominated by ionization from heavy recoils originated from nuclear inelastic interaction of the incident protons in the photodiode. This is confirmed by the neutron irradiation measurements, for which the error cross-section (the number of errors divided by the particle fluence) is equivalent to the one measured for protons at low angles of incidence. Moreover, as expected for processes dominated by inelastic scattering, no angle dependency was observed for the neutron irradiation.
An increase of the optical power raises the threshold for errors and we observe, in all cases, a decrease of the error cross-section at high optical power. We can estimate that, for the optical receiver tested, an optical power of about -12dBm (corresponding to 63 µW) is necessary to achieve a BER below 10-12 in the presence of a particle flux of about 106 cm^-2 s^-1. A similar flux of hadrons above 5 MeV is expected in the silicon tracker of CMS, at a radial distance of about 40cm.
Id:4
Corresponding Author: Federico FACCIO
Experiment: CMS
Sub-system: Tracker
Topic: Optoelectronics and data transfer systems
Status of the 80Mbit/s Receiver for the CMS digital optical link
F. Faccio, C. Azevedo, K. Gill, P. Moreira,
A. Marchioro, F. Vasey
CERN, CH-1211 Geneva 23, Switzerland
Abstract:
The first prototype of the 80Mbit/s optical receiver for the CMS digital optical link has been manufactured in a 0.25µm commercial CMOS process. Its performance satisfies the low power, wide dynamic range, and speed specifications. The required sensitivity (BER of 10^-12 for an optical power of -20dBm) is easily achieved, since this BER is obtained already at -27dBm. The radiation hardness has been verified irradiating the diode with 6 MeV neutrons (up to 6.5·10^14n/cm2) and the receiver circuit with 10KeV X-rays (up to 20 Mrad). Neither type of irradiation did sensibly modify the BER performance of the receiver.
Summary:
The CMS tracker will use approximately 1000 digital optical links for the transmission of timing, trigger and control signals. An optical receiver, made up of a PIN photodiode and a receiver circuit, will convert the digital optical signals into electric signals in a LVDS logic signal. Since the communication channel end sitting inside the CMS detector will work in a harsh radiation environment, radiation hardness is a must for both the photodiode and the receiver circuit. For this reason, and also to decrease the power budget of the receiver, the circuit has been developed as an ASIC in a 0.25µm commercial CMOS technology. The design has been made using enclosed NMOS transistors and guardrings, techniques that have been proven to achieve multi-Mrad TID hardness and to protect the circuit from Single Event Latchup (SEL).
The ASIC has been fabricated and completely characterised before and after irradiation. The power consumption is limited to about 30mW per channel, a factor of ten smaller than for a commercial radiation-soft component with similar performance. To precisely evaluate the sensitivity of the receiver, we have performed Bit Error Rate (BER) measurements at a bit rate of 80Mbit/s using a dedicated setup. The measurements have been made in a configuration similar to the one foreseen for the CMS tracker optical link. Two Fermionics FB80S-7F InGaAs/InP photodiodes have been bonded to two of the receiver channels of the ASIC, one link being dedicated to the clock and the other to the data transmission. In this condition, and with an optical power in the link of -20dBm (which corresponds to 10µW, the minimum specified level), we did not detect any transmission error in more than 15 days, from which we infer a BER below 9.3·10^-15. The CMS digital link requires a BER of 10^-12, which is reached in our configuration for an optical power of about -27dBm (2 µW).
The radiation hardness of the optical receiver has been verified in two ways. First, several photodiodes were irradiated with 6 MeV neutrons up to a fluence of 6.5·10^14 n/cm2. Two irradiated photodiodes were bonded to the receiver circuit, and the BER of the resulting optical receiver channels was measured, revealing no difference from channels where non-irradiated photodiodes were mounted. Second, the radiation hardness of the ASIC was verified with X-ray irradiations up to a TID of 20 Mrad(SiO2). We did not observe any radiation-induced degradation in the BER of the circuit, and the power consumption increase was limited to about 7%.
ID:8
Corresponding Author: Tony WEIDBERG
Experiment: ATLAS
Sub-system: Tracker
Topic: Optoelectronics And Data Transfer Systems
Single Event Upset Studies for the ATLAS SCT and Pixel Optical Links
D.G.Charlton, J.D.Dowell, R.J.Homer, P.Jovanovic,
G.Mahout, H.R.Shaylor, J.A.Wilson
School of Physics and Astronomy, University
of Birmingham,UK
R.L. Wastie, A.R. Weidberg
Physics Department, Oxford University, U.K.
J.K. troska, D.J. White
Rutherford Appleton Laboratory, U.K.
I-M Gregor
Physics Department, Wuppertal University,
Germany.
Abstract:
The readout of the ATLAS SCT and Pixel detectors will use optical links. The radiation hardness of all the components has been extensively studied but this paper discusses the operation of these links in simulated LHC radiation environments. Nuclear interactions can deposit large amounts of energy in electronic components which can cause Single Event Upsets (SEU). The SEU rates have been measured with MIPS from a beta source, low energy neutrons, pions and protons at PSI. The dominant source of SEU effects is from energy deposition in the active region of the PIN diodes.
Summary:
Optical links will be used in the ATLAS SCT and Pixel detectors to transmit data from the detector modules to the off-detector electronics and to distribute the Timing, Trigger and Control (TTC) data from the counting room to the front-end electronics. The radiation hardness of the individual components has been extensively studied. The optical links have been shown to operate at very low Bit Error Rates (BER) in the laboratory. This paper reports on studies of the operation of the links in simulated LHC radiation environments. The flux of pions during high luminosity operation at the LHC will be up to 4 10**7/cm**2/s for the detector closest to the beam line (the Pixel B-layer). The particle flux falls of rapidly with perpendicular distance from the beam line. Nuclear interactions in the detector can deposit sufficient energy in the active volumes of the opto-electronics and electronics to cause bit errors. The rates of Single Event Upsets (SEU) have been studied by measuring the BER while irradiating the opto-electronics with different particles. Minimum Ionising Particles (MIPs) were produced with a Sr90 source, neutrons from deuteron stripping and (d,t) reactions. The SEU rates have also been measured with pions and protons in the momentum range 215 MeV/c to 465 MeV/c at the Paul Scherrer Institute. This momentum range is very similar to that of pions produced in minimum bias interactions at the LHC.
MIPs do not cause any measurable SEU. A significant rate of SEU has been measured for neutrons, pions and protons. The dominant source of this SEU is due to nuclear interactions in the active volume of the PIN diode which deposits sufficient energy to trigger the DORIC4 receiver ASIC. Hence from the point of view of the DORIC4, the energy deposition is effectively a genuine signal. The effective threshold can therefore be raised by increasing the amplitude of the TTC optical signal generated in the counting room. The rate of SEU is found to decrease strongly as the effective threshold is increased. The ASICs have been designed to avoid destructive effects and no evidence for such effects has been found.
The measured SEU data are compared with theoretical calculations. These calculations are then used to extrapolate the measured BER to LHC conditions and hence predict the bit error rate during LHC operation. Even for the highest particle fluxes at the B layer of the Pixel detector, the BER can be reduced to a level below 10**-9, by using a sufficiently large amplitude optical signal for the TTC data. Therefore the problem can be reduced to a rate which is acceptable for ATLAS operation.
ID: 9
Corresponding Author: Gilles MAHOUT
Experiment: ATLAS
Sub-system: Tracker
Topic: Optoelectronics And Data Transfer Systems
Radiation Hard Optical Links for the ATLAS SCT and Pixel Detectors
D. Charlton, J.D.Dowell, R.J.Homer, P. Jovanovic,
G.Mahout, H.R.Shaylor, J.A.Wilson.
School of Physics and Astronomy, University
of Birmingham, Birmingham, B15 2TT, UK
I.M. Gregor, R.Wastie, A.R. Weidberg
Physics Department, University of Oxford,
Keble Road, Oxford, OX1 3RH, UK.
S.Galagedera, M.C.Morrissey, J.Troska, D.J.White.
CLRC Rutherford Appleton Laboratory, Chilton,
Didcot, Oxon, OX11 0QX, UK.
A.Rudge
CERN, Geneva, Switzerland.
M.L.Chu, S.C.Lee, P.K.Teng
Institute of Physics, Academia Sinica, Taipei,
Taiwan 11529, Republic of China.
Abstract:
A radiation hard optical readout system designed for the ATLAS Semi-conductor Tracker (SCT) is described. Two independent versions of the front-end optical package housing two VCSEL emitters and an epitaxial Si PIN photodiode have been irradiated with neutron fluences over 1015 n.cm-2, the level encountered in the ATLAS pixel detector. Environmental tests have been performed down to -20o C. Extensive radiation and lifetime tests have also been carried out on the opto-electronic components and the front-end VCSEL driver and timing/control ASICs. Bit error rate and cross-talk measurements using irradiated devices show that the system easily meets the performance specification.
Summary:
The ATLAS SemiConductor Tracker (SCT) and Pixel detectors will be read out using optical links. The Timing, Trigger and Control (TTC) data are delivered from the off-detector electronics to each detector module by a single optical fibre using an epitaxial Si PIN photodiode as the receiver. The binary data are transmitted using VCSELs operating at 850 nm. Two VCSELs and one photodiode are mounted in a low mass, non-magnetic optical package on each module. Several packaging technologies have been studied and so far two different packages have been successfully developed using different methods for optically coupling the optical fibres to the VCSELs and photodiodes. Radiation hard ASICs have been developed to drive the VCSELs (VDC chip) and to recover the 40 MHz clock and the TTC data from the photodiode signals (DORIC4 chip). Each component, as well as the complete packages, has been extensively tested after irradiation with neutrons and gammas.
Mitel VCSELs have been tested and show good recovery after a short annealing period following irradiation with 2.9x1015 1 MeV equivalent neutrons.cm-2, which is a typical fluence in the pixel detector after 10 years of LHC operation. The total light output before irradiation is typically 1 mW for a current of 10 mA. The main effect of neutron irradiation is to shift the threshold current upwards substantially, which reduces to about 1 mA after annealing, without changing the slope of the light output vs current. The PIN photodiodes (manufactured by Centronic) show a drop in responsitivity of about 30% to 0.3 A/W after irradiation up to 1015 n.cm-2 and an increase in dark current to 60 nA at room temperature which is negligible. Both VCSELs and photodiodes have undergone accelerated ageing tests at elevated temperatures after irradiation. They show excellent reliability (no failure) after several hundred equivalent LHC years corresponding to a failure rate of < 1% after 10 years of LHC operation.
The VDC and DORIC4 use bipolar npn transistors in the AMS 0.8 micron BiCMOS technology. Samples have been exposed to 2.5 1014 1 MeV equivalent neutrons/cm2 and 115 kGy of gamma radiation. All chips work correctly after irradiation. The circuits are designed to work with transistor beta values as low as 10. Measurements of beta before and after irradiation are presented. Accelerated ageing tests have been carried out at 100oC on a sample of DORIC4s under operational conditions without any failure, corresponding to a 0.3% upper limit (90% c.l.) on the failure rate after 10 years of LHC operation.
Bit error rate and cross talk measurements have been carried out using packaged devices after irradiation. The performance is well within specification (< 10-9 BER for an optical power of 200 muW). Single event upset measurements using neutrons and ionising radiation have also been performed and are reported in a separate paper.
The conclusions are that a successful readout scheme has been developed for the SCT, and that the optoelectronic components and packaging are also adequate for the higher radiation levels in the pixel detector.
ID: 12
Corresponding Author: Francois VASEY
Experiment: CMS
Sub-system: Tracker
Topic: Optoelectronics And Data Transfer Systems
Project status of the CMS tracker optical links
F. Vasey, C. Azevedo, T. Bauer(1), B. Checucci(2),
G. Cervelli, K. Gill,
R. Grabit, F. Jensen, A. Zanet
CERN, Geneva (Switzerland)
(1)HEPHY, Vienna (Austria)
(2)INFN, Perugia (Italy)
Abstract:
The development phase of the optical data transfer system for the CMS tracker is now complete. This paper will present the project status and review the preparation for production. In particular, it will focus on the results of the market surveys for front-end components, and on the performance evaluation of a close-to-final readout chain.
Summary:
The development phase of the optical data transfer system for the CMS tracker is now complete. The ~50000 uni-directional analogue links used for data readout are based on edge-emitting laser transmitters and pin photodiode receivers operating at a wavelength of 1310nm. In every single-mode fibre, 256 electrical channels are time-multiplexed at a rate of 40MSamples/s. Two in-line patch-panels allow to fan-in the individual fibres originating from the transmitters, first to a 12-way ribbon, and then to an 8-ribbon cable carrying 96 fibres away from the detector to the counting room. All system components situated inside the detector volume (lasers, fibres and connectors) are radiation resistant and non-magnetic. The laser transmitters and their connectorised pigtails are based on single-channel devices to best fit the distributed nature of the sensor elements, while in the counting room, the receivers are 12-channel arrays. The ~1000 bi-directional digital links used for control and timing distribution are based on almost identical components as the analogue readout system, but with a different modularity. The transceiver modules placed inside the detector include radiation resistant photodiodes and discriminating amplifiers (which are not needed in the readout system), and the transceiver modules located in the counting room are based on standard commercial components.
Apart from the custom designed electronics for the analogue and digital laser-drivers and photodiode-receivers, all optical link components are based on Commercial-Off-The-Shelf products (COTS). Slight deviations from the standard manufacturing process are only allowed to meet specific functionality requirements such as low back-reflection, or environmental constraints such as high magnetic field. This development strategy has the advantage of minimising development and system cost, but dictates the launch of extensive validation programmes to confirm that as wide a range of COTS as possible can be used reliably in the CMS tracker environment. Before invitations to tender can be sent out and orders can be placed to start the production of optical links in large quantities, potential suppliers must be qualified in the framework of open market surveys. In the case of the CMS tracker, optical components suppliers have been grouped in four categories: manufacturers of lasers, connectors, fibre-cables and receiver modules. As long as the tendering process for these components is not complete, it is not possible to know which exact devices will be used in the final system. By mid-2000, market surveys for semiconductor lasers and optical connectors will be complete, while the remaining surveys of fibre-cables and optoelectronic receiver modules will still be ongoing. In our presentation, we will present the status of these market surveys, review the results of the evaluation procedures, and discuss the plans and timescales to enter production.
In parallel to the tendering procedure, tests of readout and control chains are being performed with close-to-final components and architecture. The specifications will be reviewed and a model simulating the effects of components tolerances on full system performance will be discussed. New experimental results obtainted on a very realistic readout chain will then be presented, including for the first time an opto-hybrid transmitter module and a 12-channel analogue receiver module.
In summary, this presentation will review the system architecture and specification, discuss the results of the market surveys for front-end components, and present the performance evaluation of a close-to-final readout chain.
ID: 16
Corresponding Author: Mikhail MATVEEV
Experiment: CMS
Sub-system: Trigger
Topic: Optoelectronics And Data Transfer Systems(poster
report)
Optical Data Transmission from the CMS Cathode Strip Chamber Peripheral Trigger Electronics to Sector Processor Crate
N.Adams, M.Matveev, T.Nussbaum, P.Padley (Rice
University)
J.Hauser, V.Sedov (UCLA)
Abstract:
Data representing three muons will be sent from each sector of the CMS Cathode Strip Chambers to the Sector Processor crate residing in the counting room 100 m apart of the detector. We report on the data transmission scheme based on Agilent HDMP-1022/1024 serializer/deserializer chipset and Methode MDX-19 optical transceivers. Six chipsets and six pairs of optical modules are needed in order to transmit 120 bits of data every 25 ns of the main LHC frequency from the peripheral Muon Port Cards to Sector Receiver modules. Results of prototyping, laboratory tests as well as a possible future options for data transmission are discussed.
Summary:
The CMS Muon System consists of three detectors: Cathode Strip Chambers (CSC), Drift Tubes (DT) and Resistive Plate Chambers (RPC). There are up to four stations of CSC in each CMS endcap. CSC front-end electronics is located on chambers as well as in the VME crates mounted on the periphery of chambers. Trigger Motherboard (TMB) matches anode and cathode tags called Local Charged Tracks (LCT) and sends two best combined LCTs from each chamber to Muon Port Card (MPC). Each MPC collects muon tags from up to nine TMBs, which corresponds to a 60 degree sector for stations ME2-ME4 and a 30 degree sector for station ME1. All TMB and MPC cards are located in the VME 9U crates on the periphery of CSC. The MPC selects the three best muons and sends them over optical links to the Sector Receiver (SR) residing in the counting room 100 meters from the detector.
The main goal of the design was an evaluation of existing commercial solutions for the optical data transmission at 40.08MHz. Proposed design is based on Agilent HDMP-1022/1024 chipset and Methode MDX-19 optical transceivers. Six 20-bit chipsets and six pairs of optical transceivers are needed in order to transmit 120 bits of data representing three muons. Our implementation assumes simplex data transmission from the MPC to the SR without a return path. The transmitters use the main 40.08MHz frequency as a reference clock. At the receiver end, a clock oscillator with slightly different frequency is used for frequency acquisition.
Our prototypes have demonstrated a reliable operation at 40MHz. High power consumption and large board space required for optical modules and serializers or deserializers are drawbacks. Further possible implementations of data transmission circuitry from MPC to SR are discussed. One of these options may utilize the latest Agilent HDMP-1032/1034 chipset for serialization and deserialization. Another approach is to use a multi-channel optical receivers and transmitters. We can also use an appropriate commercial or custom solution adopted by other CMS groups.
Id: 47
Corresponding Author: Gustavo CANCELO
Experiment: CMS
Sub-system: Tracker
Topic: Optoelectronics And Data Transfer Systems
Fiber Optic based readout for BTeV's Pixel Detector
Gustavo I. E. Cancelo*, Sergio Zimmermann*, Sergio Vergara**, Peter Denes*, Guilherme Cardoso*, Bob Downing*, Jeff Andresen* * Fermilab, **University of Puebla, Mexico
Abstract:
The current paper describes the design of BTeV's Fiber Optics Pixel Detector readout. The pixel detectors will be located as close as 6mm from the accelerator's beam into the vacuum pipe. The readout electronics will be located at about 6cm from the beam, imposing strong constrains regarding radiation, mass, power dissipation, vacuum and size. This paper includes an analysis of the convenience of using a fiber optic based readout versus alternative solutions. Since the current design will place several components in a high dose proton and gamma radiation environment the fiber optic based readout will need some radiation hardened custom components, which are here specified. Furthermore, test results on optoelectronic devices are provided along with future plans to complete the design.
Summary:
BTeV's pixel detector consists of 31 double-plane stations of about 100cm² of active detection area. These planes are perpendicular to the direction of the beam. The beam passes through the center of the plane formed by two halves as shown Figure 1. Since BTeV intends to use the pixel detector as part of the lowest level trigger system, one of the most important requirements is hit readout speed [1]. The primary goal is to achieve a data transfer rate high enough to sustain the hit readout rate generated at a luminosity of p/cm² and a bunch crossing (BCO) time of 132 ns at Fermilab's Tevatron. Furthermore, the required readout bandwidth must be achieved while keeping small power and mass budget. In particular, mass is very critical for the Pixel Detector, the most inner part of BTeV's detector where multiple scattering must be minimized.
A fiber optic based design, as proposed in this paper, is the technology that best adapts to BTeV's requirements. Every pixel plane will generate, in average, 4Gb/s of data. The pixel amplifier and discriminator chips, located underneath the pixel detectors will store that information. However, since the pixel detector is the primary component of BTeV's trigger, the data must be readout as soon as possible. A modular design is being proposed for the pixel detector electronics as shown in Figure 1. Every module is autonomous. It groups a certain number of Pixel amplifier/discriminator chips and the readout electronics to transfer the data from the pixel planes to the experiment's counting room. Furthermore, every module must allow for an incoming link to receive several commands to initialize and control the pixel devices and provide them with timing information (i.e. clocks).
The readout electronic serializes the data from the pixel chips into high speed serial links. A 1.06Gb/s link is being proposed based on a custom CHFET GaAs design[2]. This device has undergone functionality, BER and radiation tests. It has shown to be radiation resistant and SEU free. Several test results are provided in this paper. The serializer device has a built in VCSEL driver. VCSELs have proven to operate at very high speeds and be radiation resistant. Test results will be shown.
The control and timing link will have a fiber optic receiver to decode the signal coming from a PIN photodiode. The operative frequency is 53.4MHz. This device must also be radiation resistant. The specifications of this chip are completed. In order to reduce the number of fibers to minimize the total mass, the data and clock signals will be encoded together following a bi-phase encoding. A prototype board has been developed to qualify the bi-phase approach working together and generating a low jitter clock for the gigabit serializer. Performance results will be supplied.
Finally, two new developments are being worked out with a company in the optoelectronic business to generate a very small profile optical assembly for the VCSEL and PIN components, and a hermetic package to exit the accelerator's vacuum pipe. The semi-custom VCSEL and PIN assembly design will allow the Pixel Module to be connected and disconnected from the fiber, allowing the modules to be tested and assembled individually, decoupling the inherent module's yield from the optoelectronic issues. In order to reduce the number of points where the vacuum pipe is disrupted, he hermetic package will handle several fibers. The optimum number is still unknown, but decoupling the modules from the fibers is important to better track the problems and increase the overall yield by being able to change individual pieces.
REFERENCES: [1] BTeV: An Expression of Interest for a Heavy Quark Program at C0, BTeV collaboration, Fermilab, May 18, 1997. [2] Radiation Hard Gigabit Systems. Second Workshop on Optical Readout Technologies for ATLAS, Oxford, January 7-8 1999.
Id: 53
Corresponding Author: Bernard DINKESPILER
Experiment: ATLAS
Sub-system: Calorimetry
Topic: Optoelectronics And Data Transfer Systems
Redundancy or GaAs ? Two different approaches to solve the problem of SEU (Single Event Upset) in a digital optical link optical
From SMU: Ryszard Stroynowski Bernard Dinkespiler
Jingbo Ye Shouxuan Xie
From CPPM: Frederic Rethore
From ISN: Marie-Laure Andrieux Laurent Gallin-Martel
From KTH: Mark Pearce Johan Lundqvist Stefan
Rydstrom
Abstract:
The fast digital optical links for the ATLAS Liquid Argon Calorimeter must survive in a high radiation environment with a total fluence of 2*1013 neutrons/cm2 and 800 Gy. The links based on Agilent Technologies -former HP- Glink chipset show a total dose radiation resistance to neutrons and gammas that would allow for 10 years of operation in the ATLAS detector. We have observed, however, an unacceptable rate of single event upsets (SEU) due to neutrons interacting in the silicon-based serializer. In order to solve this problem, we have developed two link systems. The first one - Dual-Glink-, is based on the principle of redundancy. Data are sent on two independent links. On the reception side, data are analyzed and error recovery is performed without dead time.
The second solution uses a GaAs serializer/deserializer chipset from TriQuint. This technology is intrinsically radiation hard. We expect a minimal number of SEU's and other radiation related problems. High speed of this chipset -2.5 Gb/s- allows for error recovery.
The design of the link, its performance in the laboratory environment and the results of the radiation tests will be presented for both systems.
Summary:
The ATLAS liquid argon calorimeter digital optical links are based on Agilent Technologies (former HP) Glink chipset. Although they show a total dose radiation resistance to neutrons and gammas that would allow for 10 years of operation in the ATLAS detector, they have an unacceptable rate of single event upsets due to neutrons interacting in the silicon based serializer. We developed a Dual-Glink system, composed of 2 independent links, each one sending the complete data set. On the reception side a smart switch realized in FPGA analyzes in real time the data streams from both links, detects errors and selects the link that is not affected by errors. The probability that an error occurs simultaneously on both links is negligible. The total latency of the switch is 20 clock cycles. This solution is more expensive than a simple link since all the hardware components are doubled.
In an attempt to make a cheaper solution, we also developed a digital optical link solution based on the recently available GaAs multiplexer/demultiplexer chipset TQ8123/8223 from TriQuint. The intrinsic radiation hardness of GaAs technology is expected to minimize the impact of radiation effects and of the single event upsets. Since the ATLAS LAr Calorimeter requires the 32 bit data transfer at 40 MHz, the very high speed of 2.5Gb/s provided by this chipset allows for sending the data twice over a single optical fiber. This permits an online error detection and error correction. The design of the link, its performance in the laboratory environment and the results of the radiation tests will be presented.
Id: 68
Corresponding Author: Fredrik Jensen
Experiment: CMS
Sub-system: DAQ
Topic: Optoelectronics and data transfer systems
Statistical performance estimation and optimization of the CMS tracker optical links
F. Jensen, C.S Azevedo, G.Cervelli, K.Gill,
R.Grabit, F.Vasey
CERN, Geneva, Switzerland
Abstract:
A significant number of analogue performance measurements have been carried out on the CMS tracker optical links with components selected to be close the final system. The measurements form the basis for an estimation of the expected analogue performance of the final tracker links. In particular the typical S/N and linearity performance will be estimated. Realistic performance limits based on estimations of the performance spread of the final 50000 links are also deduced. Finally we discuss ways to further optimize the analogue performance of the optical links using offline processing.
Summary:
The CMS tracker optical links project is in a phase where it is possible to identify the final components to be used in the tracker optical link system. Market surveys have been issued and the component selection process is well underway for the whole system with some additional development work remaining mainly for the electronics. These developments have enabled tests to be carried out with links that closely resemble the final system in significant numbers. These measurements form the basis for a statistical estimation of the analogue performance of the final tracker links. In particular the RMS-noise, Signal-to-Noise Ratio and non-linearity distributions are extracted. The analogue performance distributions enable realistic estimations of typical performance and performance spread of the final 50000 links to be deduced, and realistic system performance limits are shown. These results in turn make it possible to carry out a comparison with the link specifications and determine how successful the system and component specifications have been in achieving adequate analogue performance for the links. Finally we discuss ways to further optimize the analogue performance of the optical links using offline processing. In particular we look at ways to reduce system non-linearity by using a modified link calibration scheme.
Id: 73
Corresponding Author: Stefan LUEDERS
Experiment: General Interest
Sub-system: General Interest
Topic: Optoelectronics & data transfer
systems
Compact Bidirectional 2.5 Gbit/s Optical Transceiver for the H1-Experiment
S. Lueders, R. Baldinger, R. Eichler, C. Grab,
B. Meier, S. Streuli, K. Szeker
Institute for Particle Physics, ETH Zuerich,
5232 Villigen PSI, Switzerland
Abstract:
For triggering purposes, 9600 channels have to be read out within 96 ns, i.e. with a rate of 100 Gbit/s, using 40 identical very compact optical transceiver units --- each measuring 130 mm x 45 mm x 9 mm. Taking advance of VCSEL diodes and 90 degree fiber bending, 4x 850 Mbit/s of digitized trigger information as well as two channels with analog monitoring information are transferred to the receiver electronics 40 m away. From there two channels of 10 MHz clock information are received for timing adjustments.
Summary :
The upgrade of the multi-wire proportional chamber (CIP) of the H1-experiment at HERA (DESY) increases the number of chamber-channels to 9600. These channels have to be made available to the z-vertex trigger within the time between two bunchcrossings of 96 ns and thus need be transmitted at a total data rate of 100 Gbit/s. With the extremely tight spatial conditions at the CIP end flange --- an open cylinder with inner and outer radii at 150 mm and 200 mm, respectively, and a length of 130 mm --- a fast and compact bidirectional readout electronic is required, keeping the power consumption and the amount of dead-material in the experiment to a minimum.
40 identical transceiver units, stacked on top of each other in groups of five, were designed to feed the digitized trigger information and selectable analog chamber signals to the receiver electronics 40 m away. In the other direction, clock information is provided.
Using the CIPix chip from the ASIC laboratory, Heidelberg, a 16 fold multiplexing is performed on the digitized information.
A custom made optical hybrid serves as an interface between the optical and electrical world, driving VCSEL diode arrays with four channels of trigger information --- each with a data rate of 625 Mbit/s --- and two channels of analog signals. Aligned to the VCSEL array with a precision of 5 mum, an array of two PIN diodes receives the clock signals needed for internal timing. Firmly attached to the hybrid and positioned with high precision, a 8-fiber ribbon cable deflects the light by 90deg within 5 mm of height and transmits the information to and from the receiver electronics. There, the information is reconverted, demultiplexed and passed on to trigger and monitoring tasks.
Id: 93
Corresponding Author: Gueorgui ANTCHEV
Experiment: CMS
Sub-system: DAQ
Topic: Optoelectronics and data transfer systems
Readout Unit Prototype for CMS DAQ System
G. Antchev, E. Cano, S. Cittolin, S. Erhan,
B. Faure, D.Gigi, J. Gutleber , C.Jacobs,
F. Meijers,
E. Meschi, A. Ninane, L.Orsini, L. Pollet,
A.Racz,
D. Samyn, N. Sinanis, W. Schleifer, P. Sphicas
CERN Div.EP/CMD, Switzerland
Abstract :
In the CMS data acquisition system, the Readout Unit (RU) is a major element of the Readout Column and it is placed between Front-end Devices (FED) and Builder Data Network (BDN). The RU is intelligent fast buffer for intermediate storage of data before transferring between the levels of the DAQ system. Readout Unit prototype is developed to achieve the CMS DAQ requirement for data input bandwidth of 400MB/sec and data output bandwidth of 400 MB/sec. The new RU prototype based on reconfigurable hardware structure and high-speed standard busses is presented in this paper.
Summary :
The RU prototype is implemented in two physical units called Readout Unit Input Output (RUIO) and Readout Unit Memory (RUM) interconnect together via fast PCI busses. Those are long size 64bit at 33/66MHz PCI boards. The RUM unit contains dual-port memory (up to 512 Mbytes on DIMM's) where data events will be stored. The memory can be accessed through two on-board PCI busses. Those busses can receive an extension board to accept one additional PMC/PCI board. The RUIO unit is also connected to them. A Memory Management Unit (MMU) on board operates with memory as a hard disk. A third PCI bus on RUM and RUIO is used to configure and control the units. This allows both units to be plugged in a standard PCI bus environment as (PC, SUN Stations or Macintosh). The interconnection between the busses is done by on-board 4 way PCI Bridge.
In this sense the three PCI busses can work independently from each other at the maximum bandwidth of 533MB/s each. Using FPGA's components latest generation provide possibilities to implement different functions in RU.