Speaker
Description
The study of Single Event Upsets (SEU) for the ATLAS ITk Pixel front-end (FE) is crucial for reliable detector operation. In early prototypes, SEU occurring at high data rate could block the FE readout until a reset is given.
Two SEU campaigns were carried out at the Centro Nazionale Adroterapia Oncologica with a 228 MeV protons beam, reaching the hit rate of the innermost ITk pixel layer, with ITkPixV1.1 (RD53b) and ITkPixV2 (RD53c) FEs, bump-bonded to 3D silicon sensors.
These are the first measurements of ITk pixel FEs with hadronic beams on detector assemblies at the nominal data taking rate.
Summary (500 words)
The ATLAS ITk pixel detector for HL-LHC is equipped with front-end chips (FE) designed by the RD53 Collaboration. The study of Single Event Upsets (SEU) for the production FE is crucial to verify its robustness and reliable operation at the HL-LHC, where the hit rate, dominated by charged hadrons, reaches up to 240 hits/FE/trigger.
Prototype FEs showed a failure mode in which SEU can completely block their readout until a reset command was issued.
To measure the actual rate, two SEU campaigns have been carried out at the Centro Nazionale Adroterapia Oncologica (CNAO) in Pavia (Italy). The CNAO synchrotron can delivere proton and carbon ion beams to an experimental hall dedicated to R&D activities. SEU measurements were taken with a 228 MeV protons beam of intensity up to 10^9 p/s.
ITk pixel modules, consisting of ITkPixV1.1 and ITkPixV2 FEs bump bonded to 3D silicon pixel detectors by FBK, have been tested at the nominal 1 MHz trigger rate with hit rates up to 226 hits/FE/trigger, with high occupancy tails due to hadronic interactions, providing a realistic test bench for SEU studies and measurement of operational performance, compared to measurements with X-rays or bare chips.
To maintain the high data rate and reduce DAQ dead time, the YARR firmware had to be modified by adding hysteresis to the synchronization of the FEs data lines, reducing deadtime from 8% to 1%, and registers were added to detect SEU blocking events and data line desynchronization.
The results show a mean time between failure for the readout-blocking event of 6 s for the prototype ITkPixV1.1 FE and of 3400 s for the production ITkPixV2 demonstrating the success of the mitigation strategies implemented for ITkPixV2. The FE resetting options were also verified. The designated one for ITkPixV2 was verified to allow to reset the least amount, improving the recovery time compared to the hard reset required for ITkPixV1.1.
The second campaign also tried to give an estimation on the Single Event Transients (SET) occurrence, which causes the momentarily disconnection of the Aurora links, resulting in data losses from the FE.