Speaker
Description
ALFE2 is a front-end ASIC designed for the ATLAS Liquid Argon Calorimeter (LAr) in the High-Luminosity Large Hadron Collider (HL-LHC) upgrade. It integrates four channels of preamplifiers and shapers, each with two distinct gain outputs that can be read out in parallel to cover a 16-bit dynamic range. Over 80,000 chips have been fabricated in a 130 nm CMOS process. A robotic system performs automated QC at BNL and IJCLab. Over three-quarters of the chips have been tested with a yield of approximately 95%. All QC testing is expected to finish by summer 2026.
Summary (500 words)
ALFE2 is a front-end ASIC developed for the ATLAS Liquid Argon Calorimeter (LAr) as part of the High-Luminosity Large Hadron Collider (HL-LHC) upgrade. It contains four channels of preamplifiers and CR–(RC)^2 shapers with adjustable input impedance, peaking time, and baseline level. Each shaper provides low- and high-gain outputs that can be read out simultaneously, covering a 16-bit dynamic range. ALFE2 also integrates a CR–RC shaper for the layer sum output to be compatible with the current Level-0 trigger system. Over 80,000 ALFE2 chips have been fabricated in a 130 nm CMOS process for production.
A dedicated system has been built for ALFE2 production quality control (QC). The core setup consists of an ALFE2 test board, a calibration pulser, and a Front-End Test Board (FETB). The ALFE2 test board hosts a BGA socket and RF relays that switch between two input impedances (25 Ω and 50 Ω). The calibration pulser injects test current signals. The FETB is a high-speed data-acquisition platform based on the Enclustra Mercury+ XU1 SOM with a Xilinx Zynq UltraScale+ MPSoC running PetaLinux and includes two state-of-the-art 8-channel, 16-bit ADCs (ADS52J65).
To enable automated operation, a custom robotic system has been implemented. The system uses a four-axis stepper-motor-driven robotic arm: the X and Y axes provide horizontal motion, the Z axis opens or closes the socket, and an independent Z′ axis controls a vacuum suction cup for chip handling. A camera reads QR codes etched on chip surfaces for identification. The robot control board integrates motor drivers and two Arduino modules that actuate the robotic arm and the vacuum suction cup.
The QC procedure covers all major performance parameters. Adjustable parameters are swept to determine optimal settings. It takes approximately 1.5 minutes to test a chip. The QC is conducted in parallel at BNL and IJCLab. The two sites combined process about 850 chips per business day.
Over three-quarters of the chips have been tested so far, with an overall yield of approximately 95%. Completion of production QC tests is expected in summer 2026. To evaluate the production quality, the key performance distributions from the QC campaign were examined using the chips tested so far. Narrow distributions indicate highly consistent performance across the tested population. A small fraction of outliers is found and removed by applying the acceptance criteria, which are stricter than design specifications. Additional criteria sensitive to socket degradation were introduced, as socket performance gradually degrades with repeated insertions and requires routine cleaning. ALFE2 demonstrates substantially better‑than‑required performance in its two most critical metrics: integral non-linearity (INL) and equivalent noise current (ENI). The power-dissipation distribution meets the specification with a clear margin. A notable ~1% difference in gain ratio and input impedance among wafers is observed. Variations in tunable baseline and peaking time across wafers are visible but remain smaller than the tuning step sizes.