1. iTop Cooling Analysis by Chris Ketter

2. eLog Analysis 18.6mA each amp 80.9mA each ASIC Average boardstack temp

3. 106mA per ASIC 18.9mA per amp

4. With Sanding and Indium Average board stack temp of all for stacks: 52.7°C Average power consumption of each board: 7.075 W

5. Checking HV Dependence

6. Thermal Simulation Calculated power consumption of major components –ASIC:.25A@2.5V =.625 W.25A@2.5V –Amps: 15.5mA@5V =.0775 W each15.5mA@5V –LT1529: 32*15.5mA@0.6V_drop =.2976 W32*15.5mA@0.6V_drop –LT1963:.25A@0.8V_drop = 0.2 W.25A@0.8V_drop Input power dissipations into HyperLynx ThermalSim. Left all other components at default value.

7. HyperLynx Thermal Environmental Conditions: –Ambient temp: 25°C –Air velocity: 0 –Sink: left and right edge –Sink temp: 31°C ( eLog729 says “entire stack about 6C warmer than cold plate.”) –Therm. Res. of sink = 6.02 (°C-in/W)---calculated from eLog729 –Location: in rack, w/adjacent boards of emissivity of 0.65(default),.356” apart, all dissipating 3.51W. 47.9 53.3 48.8 46.5

8. Further Considerations Noticed carrier02 RevD has several dark-looking vias on board edges. Inspected under microscope and found that many (30-40%) of the cooling vias were not properly filled with copper. –Checked carrier13 RevD, and 3 “fake” RevE carriers and did not find same issue. I’m curious as to how many other carrier boards have this same issue.

9. Pre-Conclusions It appears that the one-time high temperature of 71°C in eLog707 had poor thermal contact between the carrier boards and aluminum walls, temperatures came down significantly after the HV boards were sanded (eLog728). Even so, boards were still running too hot. Need further thermal analyses prior to production of RevE carrier board.

10. Carrier RevE Thermal Model Board changes: –High power density parts have been moved to board edge –power/GND planes increased to 1.5oz Cu –blind vias utilized to prevent “Swiss cheesing” of inner layers. Thermal model improvements: –Package sizes refined to include nominal air gaps, thermal pads, and pin dimensions. –Thermal conductivity of aluminum sidewalls changed from calculated value to that of aluminum. –Thermal screw placement evaluated

11. As Is (Using 4 Different FPGA Powers) MAX 50% 75% 25%

12. With Thermal Screws 24 6 8

13. Summary Recall that this model was based on 31 degree aluminum sidewalls, and thermal screws were defined at that same temperature. Using FPGA at 50% of max power consumption, T max = 41.1°C ; T – T sink = 10.1°C –2 screws: T max = 38.9°C ; T – T sink = 7.9°C –4 screws: T max = 37.1°C ; T – T sink = 6.1°C –6 screws: T max = 35.3°C ; T – T sink = 4.3°C –8 screws: T max = 34.3°C ; T – T sink = 3.3°C