fine pitch acf for medipix/timepix · 2019-02-08 · •fine pitch in 2d -> small contact...
TRANSCRIPT
Fine pitch ACF for Medipix/Timepix
Mathieu Benoit
The problematic
• The cost of a Timepix/Medipix assembly is dominated but the cost of interconnect • Cost of UBM on sensor wafer• Cost of UBM+Bump on ASIC wafer • Cost from flip-chip process
• Cost of an assembly from a vendor ~1500CHF
• To make medipix/timepix more accessible for education and outreach, system price must go down• Readout board can be made for ~100CHF• Software is one-time cost• With ACF technology, assembly cost could go down to ~100CHF
• ASIC ~ 20CHF• Sensor (mass production, 8” with Infineon, Hammamatsu) 30-40CHF• Bonding (ENIG+ACF+Bonding +Staff) ~30CHF
What is ACF• Anisotropic Conductive Film
• Adhesive film filled with conductive micro-particle
• ENIG contact pads is grown on Al (maskless process)
• Force and heat is applied between sample
• Statistically, a few conductive balls get stuck under the pads and crushed
The particles
• Typically , polymer core of 1-5 µm coated with Ni-Au or other metals
• Used by flat screen industry to connect chip driving the LED, OLED etc• 1 connection per column,row of
pixels• Pitch of 25µm, but only in one
dimension, ~600µm2 of contact surface
• The industry is moving toward smaller pitch and smaller contact surface
Use in industry
The particles
The particles
Challenges for Timepix application
• Fine pitch in 2D -> Small contact surface • Need to ensure the presence of at least 1-2 balls under
each pad
• Need to optimize ENIG process for ACF
• Need to obtain good low resistance with relatively modest force < 200N
• Optimize metal coating on particles for fine pitch interconnect
• Develop methods for accurate placement of ACF on ASIC• Might be more efficient to do with jigs that inside the FC
Collaboration UNIGE/CERN/SINTEF/CONPART• Development on a fine-pitch ACF for Timepix
application • R&D on particle and film manufacturing
(SINTEF/Conpart)
• ENIG deposition optimization (CERN/UNIGE)
• Optimization of bonding process (UNIGE/Conpart)• Quick iteration in our clean room
• bonding -> Probe station ->bonding etc…
• Characterization of assembly in lab and TB (CERN/UNIGE)
ATTRACT submitted !
First steps• A first report from CONPART on the available ACF
was delivered to CERN at the end of 2018 • Choice between ACP,ACF ?
• Due to fine pitch of our application, a careful control of the particle density distribution is needed. • Pastes has too little viscosity, particles are simply mixed, leading
to large variation of their distribution
• In Films, high viscosity resins are used to hold the particles
• A solvent is used to reduce the viscosity of the resin
• Particles are mixed to the resin
• Resin is printed in a film, solvent evaporate
• The particle distribution is controlled by the film manufacturer during the printing process.
ACF the clear candidate for our application
First steps (2)• Due to the large area of our sensors
compared to typical LCD screen application, planarity, good control of particle distribution is required : • CONPART polymer particles have better
elasticity than pure metal particles, helping to cope with possible variation in planarity in the sample
• Need to investigate optimal size of particles vs yield
• Most suitable ACFs are build with a thin layer of resin containing the particles and a thicker upper layer, the so called squeeze layer
• Squeeze layers thickness need optimization wrtinter-pad spacing
• Force and time for the bonding process need to be optimized to ensure good contacts and sufficient squeeze of the upper layer in the adjacent gaps
• The viscosity of each materials need to be optimized for limiting the mobility of particles while promoting the mobility of the squeeze layer.
ACF candidates procured by CONNPART for first tests at UNIGE
Timepix samples• Old samples from the first trials with Medipix/Timepix found in
the cupboards for first trials • Bump height maybe not sufficient for ou application• Planarity of the bump not satisfying, due to the small thickness of the
underlying metal in 250nm IBM process• Should be much better with thick top metal on 130nm
Timepix3 + Sensor samples
• Sensors from the 2012 CLIC Micron production were still available • 2 x 300 µm thick 6’ wafers with Timepix
compatible sensors
• Two Timepix3 wafers were provided by Medipix/Timepix Collaboration
• Sensors and Timepix3 wafers shipped to PacTech for ENEPIG deposition (10 µm)
• About 6-8 weeks processing
• Timepix3 wafers will come back diced
• Sensor wafer need more ”Manual” dicing• Possibility to do this with Optim WS
First trials
• Dived Timepix3 samples will be back in beginning of March, diced• More samples on the wafers than on sensor wafers
• Plans are to use identified bad ASIC to do Timepix3-Timepix3 assemblies using ACF for mechanical inspection and cross-section and optimize bonding process
• Good ASICs to be used for trial bonding to identified good sensors from the sensor wafers • Possibility to have samples ready for TB by mid-2019
• Further step dependent on ATTRACT proposal
A look at cost
• ENEPIG deposition on Timepix3/sensor wafers • 4500 euros NRE for setup cost • 110 euros per wafers
• Dicing • 1500 euros of setup NRE• 90 euros per wafer
• Timepix3 bumping, sensor UBM (as of 2012)• 2000-3000 euros for Mask NRE (x2 for Sensor+Timepix3)• 2500-4000 euros per wafer for processing• 1500 euros per sensor wafer for processing• 100-300 euros for each flip-chip