Speaker
Description
Summary
The inner detector (ID), for tracking the charged particles in the 2-Tesla
solenoidal magnetic field, of the ATLAS detector at LHC is made of 2 detecting
media: silicon and gas. The latter is the transition-radiation tracker (TRT) in
the outer regions. The former is further made of 2 types: the pixel for the inner
region (PIXEL) and the microstrip detectors for the middle region (SCT). The
detector unit of the central barrel region of the SCT is made from the so-called
ATLAS SCT barrel module as shown in Figure 1. The strips run horizontally and 6
ASIC’s, 128 channels each, sense the strips on the top and the bottom side of the
module. The hybrid that carries the ASIC’s is the ATLAS SCT barrel hybrid.
The hybrid is set in the centre region in the module such that the geometry allows
clear overlapping of the adjacent modules along the strip direction, reducing the
input resistance to the front-end transistor to 1/4 compared with the sensing at
the end of the strips, thus reducing the input noise. The hybrid is only glued to
the cooling tabs at ends and not glued to the surface of the sensors for the
minimum damage to the sensors.
The hybrid for one module is made of one-piece of Cu-polyimide flex circuit, from
the Input/Output (IO) connectors to the end of the circuit in the bottom side. The
flex-circuit was bent 180 degrees for wrapping around the sensors from the top to
the bottom side. The area of the ASIC’s are reinforced with a carbon-carbon
substrate for excellent thermal conductivity, strong mechanical strength, good
electrical conductivity, and lightweight. The excellent thermal conductivity is
required for removing the heat from the ASIC’s. The strong mechanical strength is
required for allowing the wire-bonding from the ASIC’s to the sensors where no
direct supporting underneath the ASIC’s exists. The good electrical conductivity
reinforces the electrical stability of the ASIC’s operation, together with the
intrinsic stability of the Cu-polyimide flex circuit. The lightweight means the low
radiation length.
The ATLAS uses 2,112 SCT barrel modules in the experiment and required 2,600
hybrids including spares and losses. With 50 pre-series production hybrids, the
final design and production readiness review (FDR/PRR) were carried out, with the
results of mechanical, thermal, and electrical performances. At around the 10% of
series production, the production advance review (PAR) was carried out for
monitoring the quality and performance in the series production. The quality and
performance of the pre-series and series production were as expected and no serious
flaw or trouble was surfaced.
Whole through the production of 2,600 pieces, the Cu-polyimide flex circuit did not
encounter a major problem. The design followed the industry’s design rules
maximally. Neither breakage of traces nor delaminating of the layers was
encountered. There were two minor problems: one was the residue of adhesives and
the other was the residue of the solvent on the surface of the circuit, such that
these residues weakened the sticking of the wire-bonds. The wire-bond strength was
monitored continuously and the weakening of the wire-bond strength fed back to the
production process and the problem was quickly removed.
The major problem appeared in a small piece in the hybrid, the so-called pitch-
adapter. Since the pitches of the strips (80 microns) and ASIC’s (48 microns) were
different, the fanning traces were made with pure-aluminium traces on glass
substrates. Up to the PAR quantity, the wire-bond strength on these pitch adapter
was as expected. At about 15% production, the wire-bond strength started to
decrease gradually. Reduction of the strength also associated with the generation
of the “whiskers” around the wire-bond feet. The problem was traced to the quality
of the aluminium metalisation and the metalisation process was improved
subsequently. At around 50%, and again around 70%, of the series production, the
yield of acceptable generation of whiskers was decreased. In order to solve the
problem, the series production was stopped for a few months each.
A through investigation of the problem traced the source of the problem to the deep
inside the aluminium metalisation process. Also the yield varied depending on the
wire-bonding machines, possibly due to the shape of the wedges and the frequency
and power of the ultra-sonic. The key parameter of the problem seemed to be the
hardness of the metalised aluminium. The hardness was revealed to be required
harder than the pure-aluminium. However, when it became too hard, it did not allow
the wire-bond to stick. The hardness was identified to be a function of the
temperature at metal-deposition. After identifying the mechanism and tuning the
temperature to be the right range, there were no more problems till the end of the
series production.
