Evaluation and Optimization of the Thermal Performance of a Socketed Device for an HTOL Application
- 1. Evaluation and Optimization of the Thermal Performance of a Socketed Device for an HTOL Application Considerations in the selection of a socket for a plastic molded, thermal enhanced package Nathanaël Loiseau+, Marco Michi* and James Forster* + Presto Engineering, * WELLS-CTI 2012 BiTS Workshop March 4 - 7, 2012
- 2. Introduction Today’s device: • More functionality • Higher frequency • More power / thermal dissipation • Smaller size DESIGN OFFICE Today’s HTOL: H.T.O.L. ! No, I don’t see that • From static to true application in my design specification environment simulation HTOL OFFICE • Higher frequency A.P.P.L.I.C.A.T.I.O.N. ! • Higher current No, I don’t see that in my reference • More power / thermal dissipation norm Thermal considerations of socket and test board become more important 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 2
- 3. Content • Technology challenge • Environment challenge • Thermal resistance definition • DUT impact • PCB impact • Chamber impact • Self heating impact • Socket impact (DOE) • Conclusion Definition: HTOL High Temperature Operating Life Typically 1000 hours at Tambient 125 deg C or Tjunction 150°C. But other times and temps are performed 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 3
- 4. Technology challenge Package • Smaller, integration of several functions, dies • Power dissipation concentrated • Smaller surface for thermal exchange Advance process • Smaller size, higher frequency • IC Voltage ↘ IC current ↗ Current flow ↗ Socket & board self heating ↗ SOURCE: Semiconductor Industry Association. The International Technology Roadmap for Customer willingness Semiconductors, 2009 Edition. SEMATECH, Austin, TX, 2009. • Static HTOL does not always reveal relevant failure mechanism • Dynamic HTOL closer to real life conditions 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 4
- 5. Environment challenge Principle for HTOL Tjunction DUT = Ta + Rja x PDISS Chamber « Ta » • Ta is regulated locally or globally in order to Tjunction other Tjunction DUT have the correct Tjunction DUT component s Ta must be compatible with Tboard Tsocket • Other elements temperature (board, socket, components) • Chamber specification (ex: Tamin = Troom + 30°C if no cooling capability) 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 5
- 6. Thermal resistance definition Thermal resistance for heat transfer Electrical resistance for charge transfer Thermal Electrical Important Equation Important Equation R1-2 = T1 - T2 R1-2 = V1 - V2 Heat/t Charge/t Thermal Electrical resistance - resistance + HOT COLD LOW - + HIGH T1 T2 V1 - + V2 - + HEAT Charge(-) 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 6
- 7. Thermal resistance definition Thermal resistance is Thermal resistance is not • The characterization of heat • A universal standard transfer between 2 points – JESD 51 17 documents – MIL-STD-750E 13 methods • The key point to get thermal equilibrium and Tj under control • A standard value available in IC data functionality & reliability of the sheet or package specification IC Rjca 1/Rja = 1/Rjca + 1/Rjba Tj Ta Rjba 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 7
- 8. Thermal resistance definition Soldered device vs Socketed device • Thermal resistance is typically given in device datasheet for the device soldered to a board in a certain environment • Thermal resistance of a socketed device is expected to be higher than one soldered to a PCB in same environment. Mostly unknown. Heat dissipation path: • Die – Case – Ambient • Die – IC Pads – Board – Ambient Heat dissipation path: • Die – Case – Socket – Ambient • Die – IC Pads – Socket – Board – Ambient 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 8
- 9. Thermal resistance definition Thermal resistance circuitry for a device in a socket Tambient Rca Temp of ambient Ta DUT Tcase Rjc Temp of junction Tj Tjunction Temp of case Tc Qin Temp of pad Tp Tsignal pad Rjs T Rjp die pad SOCKET Rsb Rpb Signal pad Package pad Temp of contact Tct to contact to contact Temp of the interface interface board, PCB Tb Tboard′ Top of board to Temp of pad Tp BOARD Rbb contact / socket Tboard interface Temp of the board Rba Bottom of board to under the socket, Tb’ Tambient ambient interface 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 9
- 10. DUT impact Thermally enhanced package = package with a pad dedicated for • Thermal conduction to ground (always) • Electrical connection to ground (sometimes) Package configuration Leadframe based package Laminate based package = die attached on metal plate = die attached on a pcb 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 10
- 11. DUT impact Die assembly Stacked dies + Wire bonding Stacked dies + Flip-chip Die / lid ratio 7mm Measure done on 2 socketed device (same package, same HTOL board, difference on die size only) Device Package Die pad Die Rja A 2.7x2.5mm 31°C/W 7x7mm 5.1x5.1 mm 7mm B 3.3x2.5mm 22°C/W Chip designer choices Rjc, Rjs, Rjp Often unknown values when designing an HTOL setup 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 11
- 12. PCB impact Example of Rja HVQFN56 soldered on application boards in same environment Application board #1 Application board #2 PCB PCB Nbr of PCB size Nbr of PCB size thickness thickness layer (mm) layer (mm) (mm) (mm) 1.17 10 30x60 1.43 10 35x80 Rja = 36.4 °C/W Rja = 31.6 °C/W Material, size, thickness, stack up Rbb, Rba In HTOL, depend of chamber size and DUT requirements 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 12
- 13. Chamber impact From previous BiTS presentations, HTOL chamber often seen like: • A simple box • A “tunnel” between BIB’s • “Fresh” air at one side • Inconsistent air flow between boards • Extraction of heat capability For HTOL, JESD22-A108D: The environmental chamber shall be capable of maintaining the specified temperature within a tolerance of ± 5 °C throughout the chamber while parts are loaded and unpowered. For test cost reduction, most often: • Maximize the number of BIB’s in chamber • Maximize the number of site per BIB 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 13
- 14. Chamber impact Experimental results with a 1.3W DUT • Chamber and BIB populated in • Air flow modified on a BIB position order to keep maximum air flow by adding a « box » on the socket around sockets Ta = Ta = Ta = Ta = 85.9°C 86.0°C 86.0°C 85.6°C Tj = Tj = Tj = Tj = 146.9°C 146.9°C 149.3°C 147.6°C Chamber & Life-test Engineer choices Rba, Rca 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 14
- 15. Self heating impact With Joule effect, materials heat up according to I² x R For PCB Max current in internal PCB trace 4 30°C elevation - 35µm Copper 3,5 10°C elevation - 35µm Copper 3 30°C elevation - 17,5µm Copper 2,5 10°C elevation - 17,5µm Copper I (Amp) 2 1,5 For socket contacts Current Carrying Capacity at Ambient 1 150 140 0,5 130 Contactor Temperature, °C 120 Derated 25°C 0 110 0 500 1000 1500 2000 2500 100 Trace larger (µm) Derated 85°C 90 80 70 Derated 125°C 60 50 40 30 20 10 0 0 0,2 0,4 0,6 0,8 1 1,2 Individual Contactor Current, Amps 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 15
- 16. Self heating impact If the contacts or board traces Tambient generate heat, the heat transfer Rca from the DUT to ambient will Tcase change DUT Rjc Tjunction Qin Tsignal pad Rjs T Rjp die pad SOCKET Rsb Rpb Qpin Qpin Signal pad Package pad to contact to contact interface interface New contributors Tboard′ Top of board to BOARD Rbb contact / socket Qboard Tboard Qboard interface Rba Bottom of board to Tambient ambient interface 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 16
- 17. Self heating impact Measure on HTOL board also illustrates mutual heating between sites • with 1.3W DUT Junction self heating with 1.3W at Ta = 85°C • 780mA per DUT 64 49,5 • "1" used as Tj ref 49 63 Self heating 48,5 10 11 12 13 + Tj-Ta 48 62 Rth(j-a) mutual heating Rth(j-a) (°C/W) Tj-Ta (°C) 47,5 2 6 3 14 61 47 Self heating 46,5 9 1 7 15 60 only 46 45,5 59 5 8 4 16 45 58 44,5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Nb of site loaded Self heating, mutual heating Rsb, Rpb, Rbb, Rba 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 17
- 18. Socket impact DOE in order to evaluate impact of socket design on global Rja DUT Vehicle : Socket Vehicle : • Thermal enhanced HVQFN56 • Clamshell, 8x8mm • Surface mount • Laminate based • 10 different thermal designs • 1.3W dissipation • Always the same IC used Board Vehicle : HTOL chamber Vehicle : • HTOL board 540x245mm, • 16 BIB position available, • Full application mode simulated • Board vehicle always loaded in (800MHz signals in), same position • 16 sites available • Measure of Tj / Rja at minimum • 1 site loaded (always the same) 3 temperatures (only 1 reported in this document) 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 18
- 19. Socket impact Method used for simulation Tambient Rca • Based on simplified thermal Tcase circuitry DUT Rjc Tjunction Qin • No self heating effect considered Tsignal pad Rjs T Rjp • Package estimation done for Rjc, die pad Rjs, Rjp based on die/lid ratio = SOCKET 0.67 Rsb Rpb Signal pad Package pad • Chamber air flow unspecified but to contact to contact interface interface considered as high value, estimation done for Rca, Rba BOARD Tboard′ Top of board to contact / socket Tboard Rba interface • For socket: Bottom of board to – Rsb = Rth(pin) / nb of signal pin (56) Tambient ambient interface – Rpb = Rth(pin) / nb of epad pin 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 19
- 20. Socket impact Method used for Tj / Rja measurement on HTOL board Measure of junction temperature performed in true HTOL condition thanks to an embedded diode • 1st step: calibration of the diode vs temperature with unbiased IC Then for a given Ta • 2nd step: heat up IC (power supplies ON) during 1hr • 3rd step: measure diode and IC power dissipation. Calculate Tj and Rja 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 20
- 21. Socket impact Socket / pin variations (1) 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 21
- 22. Socket impact Socket / pin variations (2) P-Pin S-Pin Z-Pin Rth = 2,080°C/W Rth = 12,464 °C/W Rth = 1,100 °C/W pogos on Thermal button on die BART = heatsink on the die pad pad (Rth = 9 °C/W) cover (Rth = 25°C/W) 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 22
- 23. Socket impact Different signal pin type • Measurements confirm that thermal performance of B is worse than A • This is due to the difference in pin performance • For B, simulation is more conservative than actual 230 A simulation A measure 205 B simulation B measure Tjunction Temperature, oC 180 A B 155 130 105 80 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 Device Power, W P-Pin S-Pin 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 23
- 24. Socket impact Number of pins on the die pad • Increasing the number of pins on the die pad improves the thermal performance • For C, simulation close to actual. Signal path become negligible due to number of P-pin on die pad 230 B simulation B measure 205 C simulation C measure Tjunction Temperature, C o 180 155 B C B C 130 105 80 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 Device Power, W S-Pin P-Pin 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 24
- 25. Socket impact Different pin type for both signal & die pad • Global thermal performance is influenced by the thermal characteristics of the pin chosen • Variation between simulation and actual data but trend correct 230 A simulation A measure 205 D simulation D measure H simulation H measure Tjunction Temperature, C o 180 155 A D H 130 105 80 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 Device Power, W P-Pin S-Pin Z-Pin 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 25
- 26. Socket impact Different pin structure for the die pad • Thermal button improve global Rja • But was not able to get all the gain expected by thermal button (too few vias between die pad and ground plane) 230 A simulation A measure 205 E simulation E measure Tjunction Temperature, oC 180 155 A E 130 105 80 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 Device Power, W P-Pin Thermal Button 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 26
- 27. Socket impact Open-top vs Clamshell design • Simulation sees no difference between clamshell and open top socket • Actual data shows that open top is worse than clamshell 230 B simulation B measure 205 I simulation I measure Tjunction Temperature, oC 180 Rjca is through 155 • package surface only with open top 130 socket 105 • by whole socket 80 surface with 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 clamshell Device Power, W 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 27
- 28. Socket impact Integrated heatsink • Even though thermally enhanced package is designed to dissipate heat through the thermal pad on the bottom, adding a heatsink improve Rjca thus global Rja • This is confirmed by simulation and actual 150 140 A simulation A measure F simulation F measure 130 G simulation G measure Tjunction Temperature, C o 120 110 A F G 100 90 80 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 Device Power, W P-Pin Heatsink 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 28
- 29. Socket impact Cumulative impact of button & heatsink • Using best performance Rjba (thermal button) and Rjca (heatsink) improves global Rja 150 140 E simulation E measure F simulation F measure 130 J simulation J measure Tjunction Temperature, C o 120 E J F J 110 100 90 80 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 Device Power, W Thermal Button Heatsink 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 29
- 30. Socket impact Simulation ranking 220 I 200 B D C 180 A Tjunction Temperature, C H o G 160 F E 140 J 120 100 80 0,0 0,5 1,0 1,5 Related to Device Power, W pcb design Socket thermal design Rsb, Rpb, Rca 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 30
- 31. Socket impact And what about a “standard” socket based on package mechanical parameters only ? From D to “Standard” 220 B simulation B measure • B socket is the 200 C simulation C measure same as C with 180 D simulation D measure Tjunction Temperature, C o less P-pins on die 160 pad Tj = 150°C 140 From C to B • “standard” would 120 be the same as D 100 with less S-pins on die pad 80 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 Device Power, W “Package Outline Drawing” is not enough for socket selection 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 31
- 32. Conclusion • DOE shows that simulations and actual data follow the same tendency. But amplitudes appear different due to incomplete/unknown information such as accurate design/package information or considering self heating of the contacts. • In fact, the real simulation « fine-tuning » is all in the details ie the more detailed the information which is considered (such as precise package structure, die/lid ratio, lid material, precise socket/pcb setup…) always improves the accuracy of the simulation data • Since device and chamber contribution in thermal performance can not be modified, socket thermal design becomes an increasingly important factor to consider and optimize the overall thermal performance (Rja) of the device-socket-board structure during an HTOL test. • But … even with an « optimal socket design » the benefit of this upfront accurate design will be lost if other aspects such as board design and population, chamber population and ventilation…are not taken into account by the lifetest engineers. 03/2012 Evaluation and Optimization of the Thermal Performance of a Socketed Device for an H 32