This paper has been awarded "Best Tutorial Paper" at BiTS 2012 Workshop.

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