For the third running period of the CERN LHC, the ALICE experiment will
undertake several upgrades of its sub-detectors. One of the detectors to be
upgraded is the Inner Tracking System, featuring the new ALPIDE pixel chip.
Control and readout of the 24120 chips are handled by 192 custom FPGA-
based readout units. Each readout unit can forward 9.6Gbps of data to another
custom PCIe card that aggregates the data from several units and transmits it
for further oﬄine/online analysis. Integration and commissioning of the system
is underway and this paper describes the ﬁrst experiences and results of this
During the ongoing Long Shutdown 2 of the CERN LHC, the ALICE experi-
ment will replace the existing Inner Tracking System (ITS), which is based on
silicon strip sensors, silicon drift sensors and silicon hybrid pixel sensors, with
a completely new detector based on Monolithic Active Pixel Sensor (MAPS)
technology. 24120 ALPIDE pixel sensors are mounted onto 192 azimuthally
overlapping staves arranged into seven coaxial cylinders of increasing diame-
ters, providing improved tracking and event-rate capability compared to the
previous detector. The cylinders are separated into inner and outer barrel, con-
sisting of three and four cylinders, respectively. The aim of the upgrade is to be
able to handle 50 kHz Pb–Pb interaction rate and 200 kHz pp interaction rate.
However, the current design is aimed at achieving double the rate for Pb–Pb
and 1 MHz for pp.
Trigger distribution, readout, control, power management, and monitoring
(voltages, currents, and temperatures) of the sensors is handled by 192 custom
FPGA-based Readout Units (RUs) based on a Xilinx UltraScale FPGA, one per stave. The RU can operate both in triggered and continuous readout mode,
whereas in continuous mode the RU generates the triggers to the sensors in-
ternally. Three optical transceivers, each supporting a maximum throughput
of 3.2 Gbps, connects each RU to the Common Readout Unit (CRU) hosted in
the First Level Processor (FLP). Each FLP can host up to four CRUs and each
CRU can receive data from up to eight RUs.
Assembly of the detector, both inner and outer barrel, is well advanced and
the commissioning of the complete readout chain, with all ﬁnal systems, has
started. This paper will present the performance of the full system in terms of
readout-rate capabilities in both triggered and continuous readout mode under
diﬀerent running conditions: at low detector occupancy and high rates, simulat-
ing running with pp interactions at rates up to 1 MHz, and at high occupancy
and low rates, simulating Pb–Pb minimum bias collisions at rates up to 100 kHz.