WP1 - Pixel detectors for HEP tracking and Imaging Richard Plackett (SR) and Timo Tick (ESR) ACEOLE 12 Month Meeting, 1st October 09

Richard Plackett – 1st 6 months 6 week secondment to PANalytical Formal internal training on XRD/XRF On the job training with Pan experts implementing new techniques based on Medipix2 chip Automatic Known Good Die (KGD) testing and classification – MATLAB based wafer probing system Prerequisite to large area tiling Formal training course on MATLAB On the job training on chip behaviour and classification by local experts A significant number of wafers sucessfully tested Reports in Collaboration meetings

Automatic Wafer Characterisation A KS PA200 probestation was automated to allow the automatic testing of readout chips at the wafer level A custom optical feedback system allows automatic alignment and chip finding to test wafers and individual assemblies Karl Suss PA200 Probestation Microscope video feed Target image Correlation Reconstructed chip positions on a Medipix1 Peak finding Position reconstruction

Richard Plackett – LHCb activities Timepix / Pixel VELO Testbeam Activity Testbeam Results Plans for Future Timepix Telescope RICH Photon Detector Upgrade Proposal Medipix3 radiation sensitivity measurements Slides from presentations to VERTEX09 and LHCb RICH upgrade meetings.

Recent Testbeam Activity A Timepix like chip is a candidate for a Pixel VELO upgrade Required to demonstrate suitability for tracking Measure efficiency and resolution Provide information for VELOPIX design 3 testbeams at CERN SPS with 120GeV Pions June: as Medipix Group to test Telescope concept July: Running parasitically from CMS SiBit telescope August: Running parasitically from EUDET/LCFI testbeam Significant improvements to telescope design at each testbeam

Testbeam Telescope V1 and V2 The first testbeam in June proved we could do tracking with Medipix2 chips and the synchronisation scheme worked The July testbeam took two weeks of data across all areas of interest. This experience allows us to significantly improve the design of the Timepix Telescope

Timepix Telescope 4 Timepix, 2 Medipix planes in telescope Symmetric positioning of planes around Timepix DUT Telescope planes mounted at nine degrees in x and y DUT position and angle controlled remotely by stepper motors Measurements of resolution with angle, threshold, sensor bias, 3D sensor and timewalk

Angled Planes to Boost Resolution Hits that only affect one pixel have limited resolution (30um pixel region) Angling the sensor means all tracks charge share and use the ToT information 55um 300um 10o Timepix (ToT) tracked position vs cluster reconstructed position

Results – Resolution Vs Track Angle Rotation about Y axis 2.5um Estimated track contribution to residual PRELIMINARY: UNCALIBRATED DATA!! The tracking uncertainty of the telescope is very competitive, with uncalibrated, uncut data, 2.5um is comparable to EUDET running with 30um pixels (Timepix is 55um)

Results – 3D Sensors Efficiency Glasgow double sided CNM sensor Perpendicular particles passing through a doped hole will deposit less charge in the silicon The sub-pixel resolution of the telescope allows us to see the efficiency losses due to the anode and cathode holes in the silicon.

Design Lessons for VELOPIX Analogue IS better than Binary, even with 55um pixels, even with clusters and charge sharing Clustering WILL significantly increase hit multiplicity Interactions within the sensor will occur often and need to be handled to avoid overflowing the chips for that event From this data we will determine how many bits ToT we need and optimise on chip the data transport Timepix (ToT) 5um Medipix (binary) 11um

Plans for a future Timepix Telescope We plan to add fast readout systems and software synchronisation High resolution (<3um) as current telescope Track timing down to 10ns Link to PMT/external device for trigger and integration of DUTs Maximum track rate increased up to ~200kHz Convenient, flexible device as current telescope Should become baseline LHCb telescope for upgrade work DUT New Readout New Readout Timing unit

LHCb RICH System LHCb RICH must also upgrade to 40Hz electronics Requires at least HPD photon detector planes to be replaced Possible to use VELOPIX as base for new MCP photon detector The Medipix group already has collaborators with experience in this area Requires high density through Silicon Via technology Photon Quartz window and photocathode as HPD and MAPMT Photoelectron 200V/mm drift field MCP cascade amplification (~1kV) Electron shower Bare readout chip array (no bump bonds) A Proximity focused MCP tube could perform as well as an HPD with many additional advantages… 40MHz digital readout Magnetic field tolerance Low vacuum volume Higher resolution Avoids 20kV supply simplifying services

400Mrad Irradiation 60 MRad 460 MRad 400 MRad Used a calibrated X-ray machine (Seifert RP149) Beam profile is smaller than the Medipix3 → Two runs: On the Pixel Matrix 60Mrad Threshold Variation Gain Variation Noise Increase 60 MRad On the Periphery 400Mrad Check DACs E-fuses Logic functionality 460 MRad 400 MRad

Performance after 460MRad Threshold Noise Yield artifact σ=1.72 ke- µ=9.3 ke- σ=12.9 e- µ=71.6 e- Threshold can be re-tuned using 5 bit equalisation

Performance after 460MRad Row[0:255] Yield artifact Qin=2ke- After 460MRad there is essentially no gain variation observed

Richard Plackett - Conclusions First months devoted to learning materials analysis techniques, MATLAB and Medipix2/Timepix chip operation Automatic wafer probe test system developed and successfully operated A series of test beams were carried out using a large number of Timepix assemblies Excellent test beam results led LHCb VELO to adopt that approach for the upgrade Along the way we developed a very accurate and convenient particle telescope we would like to take further, currently being used in SPS crystal collimator studied New ideas for a large area tiled photon detector were developed There is the possibility of the VELO chip being used in MCPs or HPDs in the RICH upgrade. Measurements on the behaviour of the Medipix3 chip 0.13um indicate its suitability for sLHC and other high dose applications (such as x-ray materials analysis and CT)

Timo Tick - Outline Low-cost bumping and flip chip solutions Through Silicon Via (TSV) process development Large area tiling GOAL: TSVs Read out chip Work carried out together with Sami Vaehaenen in the framework of WP6 of PH detector R and D

Low-cost bump bonding - motivation Bump bonding (BB) costs for a single detector unit have been over € 200 Readout chip (ROC) : sensor chip (SC) : bump bonding (cost ratio) = 1:2:7! Imagine building a square meter sized area from detectors with unit area of 1.5 cm2 with 200 € BB unit costs  667 detectors required for and the total BB cost is 1.3 M€! More pixel detectors are desired to be used in SLCH and the total area will cover tens of square meters - the cost issues is evident Reliable and cheap bumping technologies do not exist yet for ultra-fine-pitch (UFP) applications, such as Alice/Atlas/CMS/Medipix chips Here’s a cost estimation how the costs should be shared between readout chip (ROC) bumping, sensor chip (SC) bumping and FC bonding In the calculated graph, sensor bumping costs are dominating, but in reality flip chip bonding services have become more expensive

Low cost bump deposition solutions Way to realize the goal is to benchmark the manufacturing technologies used in commercial electronics and adapt them for pixel wafers. Study the technologies and find the essential limits (minimum bump size and pitch) for a low-cost technology to be realized. To provide cost savings processes have to become faster and the use of expensive and non-flexible process equipment should be avoided or minimized. Reliable and cheap bumping technologies do not exist yet for ultra-fine-pitch (UFP) applications, such as Alice/Atlas/CMS/Medipix chips. Focus is directed at applying electroless deposition process (Ni/Pd/Au) under bump metallization (UBM), which is a low-cost solution and enables various solder transfer and flip chip bonding technologies. Solder transfer tools combined with electroless UBM’s has a very high potential to be the ultimate low-cost bumping solution. Electroless Ni/Au UBM deposition was demonstrated with 55 µm pitch.

Flip chip development work Flip chip bonding can be done using with or without solder bumps. Anisotropically conductive adhesives (ACF) enable assembly without solder. The flip chip work can be divided to two main categories: Optimization of tooling and tuning of software parameters for existing FC bonders (FC150’s). Search for new bonding equipment which could boost up the throughput, but could still fulfil the resolution criteria. ± 5 µm accuracy should be good enough for pixel detector assembly. Nickel and carbon nanofibers in polymer matrix will be tested. The films with aligned fibres are the best candidates for achieving ultra-fine pitch with area array pixels. Chip-to-wafer bonding process is desired to be developed, because it increases the cost efficiency of flip chip assembly. A report on low cost bumping solutions (D11) is almost ready for publication (due month 10) ACF’s are interesting because no solder is needed.

TSV development Joint development project started with VTT 1.7.2009, duration 12 months Step 1: (3 months) Study and evaluation of the available techniques at VTT for the TSV process Selection of candidate processes Definition process parameters for individual process steps/equipment Step 2: (6 months) Optimize process and manufacture prototype vias on dummy wafers Preparation the 2 medipix2/3 wafers for processing Step 3: (3 months) Manufacture TSV’s on real medipix2/3 wafers Time (month) Deliverable Partner Layout of test vehicle CERN 1 Masks for test vehicle VTT 3 Detailed TSV process description 8 Supply of 2 Medipix2/3 wafers 9 TSV Assemblies based on test vehicle 10 Test report on test vehicle assemblies 11 Delivery of Medipix2/3 assemblies 12 Test report on Medipix2/3 assemblies Done

Test vehicle Test vehicle was designed to serve the flip chip and the through silicon via (TSV) development work. Test vehicle has daisy chain and Kelvin test structures to characterize the interconnection yields and resistances at three interfaces: Flip chip bumps (20852/chip) TSV’s (196/chip) BGA joints (100/chip) Test chips can be used for gathering reliability data during accelerated thermal stress testing. Layout has been designed to be used also in wafer-level assembly tests Chip-to-wafer-bonding Wafer-to-wafer bonding 24 silicon wafers were processed at VTT to evaluate electroless UBM deposition, low-cost flip chip techniques and BGA connections Probe card for test vehicle has been designed and acquired - ready for measurements

TSV development Detailed process description has been agreed with VTT and CERN Development of process is started with 6” dummy wafers although CERN wafer are 8” Test vehicle layout is used to process test chips for yield and quality evaluation Via etching test are currently being made at VTT Plan is to move to 8” processing as soon as possible – process description has been done by Sami and Timo Formal training on TSV at ECTC, san Diego Informal on the job training at CERN with S. Vaehaenen and during 2 visits (3 weeks in total) with experts at VTT

Large area tiling Study of large area tiling possibilities of detectors (with TSV’s) has been made: Report by the end of Nov. (D13 - 12 month deliverable) Two concepts exist: Research focused on BGA mounted detector modules Work is done in parallel with TSV development develop solutions to BGA mount detectors with TSV’s as soon as they arrive Carrier board selection and evaluation BGA bump type selection and evaluation Set up BGA bumping and assembly procedures GOAL: to be ready to design the backside layout of the detector chip and to BGA mount the detectors as soon as chips with TSVs come out

Carrier board and BGA bump selection Carrier board requirements for area array solder mounting large silicon chips Low CTE, high density signal wiring (multilayer wiring), high current wiring for powering and good heat conductivity Selected candidates: Multilayer ceramics (LTCC), Carbon composite laminates (CCL) and silicon 3 different BGA bump types were chosen for evaluation: Non collapsible plastic core BGA Light weight solution, radiation hardness to be proved Collapsible BGA Standard technique, reference Land Grid Array (no bump) Simplest and cheapest solution, reliability needs to be proved Silicon LTCC CCL FR-4 CTE [ppm/K] 2.6 6 2-10 16 Thermal Conductivity [W/mK] 13 2-3 ~100 (bulk 600) 0.25 Density [g/cm2] 2.3 3 1.55-1.9 1.9 Young’s modulus [GPa] 130 350 > 100 15 PCBGA BGA LGA

Carrier board and bump type evaluation LTCC – Test board has been designed To evaluate the reliability of the three chosen bump types with temperature cycling tests Test the radiation hardness of the plastic core balls Set up a prototype bumping and assembly procedure in CERN SMD lab. Status: waiting for UBM deposition on dummy chips CCL – Suppliers have been contacted to acquire reliability data Mechanical properties received from the supplier need to be verified Board reliability and radiation tests need to be done If these preliminary tests are passed, BGA joint evaluation can be done Status: Negotiating with material supplier Silicon – Collaboration with Gigatracker project BGA reliability is not the issue with silicon carriers Cooling and high current wiring is and solutions to these issues are being investigated in gigatracker project Gigatracker is not using TSVs but same solutions apply Timo investigated the processing possibilities at VTT and presented them to Gigatracker cooling group Status: Agreement to collaborate, Gigatracker group contacting VTT

Conclusions - Timo Tick First months devoted to understanding needs in HEP and other applications Low cost bumping approaches studied together with S.Vaehaenen TSV project established with VTT Large area tiling based on BGA – candidate solutions identified Test vehicle designed with S. Vaehaenen for low cost bumping, TSV and BGA evaluation Test vehicle wafers processed at VTT by Timo and Sami Formal training undertaken at ECTC conference PhD thesis in print – defence 27th November