UCSB Encapsulation Studies UC Santa Barbara Based upon a 6 week study by F. Garberson in collaboration with A. Affolder, J. Incandela, S. Kyre and many others

Slide 2UCSB Encapsulation Study, CERN - March 15, incandela Encapsulation Studies 96 modules encapsulated with Sylgard 186 Modules tested before/after encapsulation à No channel failures were found For thermal cycling, an environmental chamber was leased. Cycles were run in 3 modes: à SEVERE: -30 C to 50 C at 45 min/cycle and then tested à VERY SEVERE -40 C to + 60 C or -20 to +80 C at 65 min/cycle and re-test. à EXTREME -40 C to +80 C at 95 min/cycle and re-test. à Between 6 and 50 thermal cycles in all cases

Slide 3UCSB Encapsulation Study, CERN - March 15, incandela Humidity Humidity not well controlled à Environmental chamber does not have a dedicated dry air supply à The heater allows outside air to enter –Humidity spikes upwards and condensation appears at beginning of every heating cycle à At the end of every set of cycles, modules are held at high temperatures until humidity drops. –Modules come out dry Overall a more severe test. 1 hour 2 hours

Slide 4UCSB Encapsulation Study, CERN - March 15, incandela Severe Thermal Cycles 96 modules tested à At least 7 cycles each between -30 C and +50 C at 45 min per cycle 5 had new failed channels after cycles à One had 9 new 2 sensor opens, three others had 1 new 2 sensor open, and one had 1 new 1 sensor open. –2 bond lift-offs visible In total, 0.03% channels were affected à Note that a few opens were found in the PA itself –Not underneath encapsulant –It appears that micro-cracks or scratches in the PA opened up with multiple thermal cycles (possibly due to residual humidity freeze cycles - expand the cracks)

Slide 5UCSB Encapsulation Study, CERN - March 15, incandela Very Severe Thermal Cycles An intermediate range à Thermal cycles intended to be extreme, were not quite as extreme as expected because the temperature ramp intervals were too short. 9 modules cycled -40 C to +60 C at 65 minutes per cycle à 3 modules had total of 5 new 2 sensor opens (2SO): 0.11% of channels. 46 modules cycled -20 C to +80 C at 65 minutes per cycle à 15 modules had 60 new opens: 0.24% of channels à 3 APVS on 2 modules fail: 1.5% of channels (discussed later)

Slide 6UCSB Encapsulation Study, CERN - March 15, incandela Extreme Thermal Cycle 34 modules cycled -40 C to +80 C at 95 minutes per cycle 16 modules had total of 58 opens: 0.29% of channels Notes of possible concern: à 1 APV with dead FE

Slide 7UCSB Encapsulation Study, CERN - March 15, incandela 4 Modules Irradiated Report from T. Affolder. à On the 4 SS4 modules (16 chips), 3 dead chips and 1 broken wire bond developed during the irradiation or afterward. –One of the dead chips was NOT seen at Karlsruhe. Details: à Module 5049: –Chips 3 & 4 dead. –No charge inject response, noise consistent with a dead or saturated chip. Fluence=3.7E14 à Module 5102: –Chip 2 dead. Same symptoms as above. Fluence=3.9E14 à Module 5071: –One new 1 sensor-to-sensor open. Fluence=4.8E14 à Module 5050: –No new problems. Fluence=5.3E14

Slide 8UCSB Encapsulation Study, CERN - March 15, incandela APV Failures Did the encapsulant cause the APV failures? à In chip failures of thermal-cycled and irradiated modules, there was less current drawn from the preamp than normal. à No indication that back-end bonds were affected. –Currents moved as expected when changing initialization of the readout circuitry. Encapsulant over FE bonds removed, bonds remade for 4 APV à All 4 chips had the expected currents when varying the chip initialization. à This points to the encapsulation shearing bonds to the FE somewhere. warm

Slide 9UCSB Encapsulation Study, CERN - March 15, incandela APV Failures cold Did the encapsulant cause the APV failures? à In chip failures of thermal-cycled and irradiated modules, there was less current drawn from the preamp than normal. à No indication that back-end bonds were affected. –Currents moved as expected when changing initialization of the readout circuitry. Encapsulant over FE bonds removed, bonds remade for 4 APV à All 4 chips had the expected currents when varying the chip initialization. à This points to the encapsulation shearing bonds to the FE somewhere.

Slide 10UCSB Encapsulation Study, CERN - March 15, incandela Opens Overall results à Total of 106 two sensor opens and 30 one sensor opens –90% from Extreme cycles à Many caused by bond liftoffs –Notably on module edges where bonds are difficult à Most opens were not visible Hypothesis à One issue may be that these were old modules that had a number of wirebonding issues. à Decided to try a control sample of recent production modules

Slide 11UCSB Encapsulation Study, CERN - March 15, incandela Control Sample 9 Modules without encapsulant put through both Very Severe and Extreme cycles, with electronic tests done in between. à 6 with ST silicon and type 20 hybrids, 3 with HPK silicon 23 cycles from -30 to +50 C, 15 cycles from -20 to +80 C à No modules damaged Modules were then encapsulated à Repeat same set of cycles à Again, no modules damaged Lastly they were extreme cycled: -40 C to +80 C in 95 minute intervals, and made it through a full 19 cycles (far more than any other set of modules). à 8 of 9 were perfect à 1 module had three new two-sensor opens.

Slide 12UCSB Encapsulation Study, CERN - March 15, incandela All Results Rates of opens à Severe cycles: 0.03% (old modules) 0.00% (new) à Very Severe: 0.15% (old modules) 0.00% (new) à Extreme: 0.30% (old modules) 0.02% (new) Locations of opens: (old modules) à 80% pitch adaptor to sensor and 20% sensor-to-sensor à No known Backend Hybrid wirebond breaks à Several power bonds (see below) Chip Failures à Irradiated OB2 modules (4 old ones) –3 dead chips due to broken power bonds à Thermal cycles –4 dead chips due to broken power bonds

Slide 13UCSB Encapsulation Study, CERN - March 15, incandela Conclusions In older modules - Encapsulation causes 0.03% to 0.3% of wirebonds to fail in SEVERE and EXTREME thermal cycles, respectively à Most failures PA to SENSOR Wirebond failure drops to completely negligible level in a sample of newer modules Most serious concern is that power bonds are pulled due to location between APV and PA Notable non-problems à No back end bonds failed in thermal cycles or irradiation à Almost no SENSOR-SENSRO bonds failed either

Slide 14UCSB Encapsulation Study, CERN - March 15, incandela Follow up studies What happens to noise performance of modules as a function of bias voltage with encapsulation? à We see no change in noise due to presence of encapsulant Can encapsulation be done in a dry environment? à To some extent, yes, but not so easy to do in an super low humidity environment à Should be considered in comparison with possible humidity/corrosion effects over time for non- encapsulated bonds.

Slide 15UCSB Encapsulation Study, CERN - March 15, incandela Recommendations Either: à Wirebond back end bonds of hybrids If these bonds have something fall on them or short them, we lose whole chips in many cases if they are not encapsulated. There appears to be no risk with encapsulation. Do not encapsulate any other bonds Could conceivably encapsulate sensor-sensor without much risk, but the gain is marginal and it would mean that all bonding centers would need encapsulation equipment. If only BE bonds encapsulated, then we would need encapsulation capability only at a few locations (possibly only CERN) OR Drop encapsulation altogether May mean less work but maybe not, and Would leave most control bonds prone to shorting or damage