burn-in & test socket workshop · plating thickness has the biggest effect on load board life....

March 3-6 , 2002Hilton Phoenix East/Mesa Hotel

Mesa, Arizona

Sponsored By The IEEE Computer SocietyTest Technology Technical Council

Burn-in & TestSocket Workshop

IEEE IEEE

COMPUTERSOCIETY

COPYRIGHT NOTICE• The papers in this publication comprise the proceedings of the 2002BiTS Workshop. They reflect the authors’ opinions and are reproduced aspresented , without change. Their inclusion in this publication does notconstitute an endorsement by the BiTS Workshop, the sponsors, or theInstitute of Electrical and Electronic Engineers, Inc.· There is NO copyright protection claimed by this publication. However,each presentation is the work of the authors and their respectivecompanies: as such, proper acknowledgement should be made to theappropriate source. Any questions regarding the use of any materialspresented should be directed to the author/s or their companies.

Burn-in & Test SocketWorkshop

Session 2Monday 3/04/02 10:30AM

Managing High Frequency Requirements

“Optimizing Load Board Design And Modeling For High FrequencyContactors”

Jeff Sherry - Johnstech International Corporation

“The New YieldPro Array Series Contactor”Julius Botka - Agilent Technologies

“Electrical And Mechanical Performance Characterization Of High FrequencyTest Sockets”

Lisa Steckley - IBM MicroelectronicsDr. Hanyi Ding - IBM Microelectronics

Technical Program

1

Optimizing Load BoardDesign and Modeling

forHigh Frequency

Contactors

Jeff Sherry, RF EngineerJohnstech International

2

Discussion Topics Modeling & its importance

Designing load boards for optimalperformance

Load board effects

Grounding schemes

Crosstalk improvements

Load board pad size and placement

Model validation & performance examples

3

Modeling & Its Importance

Determine potential problems beforebuilding hardware

Determine expected performance

Determine trends and sensitivity of circuits

Determine effects of tolerances

Determine interaction between componentsin system (device, contactor, handler, etc.)

Design for the device, form, fit, andfunction of the end application

4

Designing Load Boards forOptimal Performance

Microstrip and Coplanar Effects Substrate thickness ↓↓↓↓ Impedance ↓↓↓↓

Trace width ↓↓↓↓ Impedance ↑↑↑↑

Permittivity (ε(ε(ε(εr) ↓↓↓↓ Impedance ↑↑↑↑

Trace thickness ↓↓↓↓ Impedance ↑↑↑↑

Coplanar Waveguide Effects Spacing (pitch) ↓↓↓↓ Impedance ↓↓↓↓

Adding ground plane Impedance ↓↓↓↓

5

Load Board Effects Loss tangent affects insertion loss more at higher

frequencies

Uniformity of εεεεr and loss tangent vs. frequency

Substrate thickness affects inductance to ground

Increased frequency means thinner substrates

Substrate thickness and dielectric constant (εεεεr),along with line width, are major parameters incontrolling impedance

Plating thickness has the biggest effect on loadboard life

6

Load Board Effects Things to improve load board design

Eliminate right angles

Eliminate changes in line width

Separate high frequency traces with ground

Move clock traces away from other signallines

Place decoupling components close to thedevice

Use matched impedance traces up to thedevice or test contactor

7

Grounding Schemes Length and area of the ground path is very

important

Ground inductance should be minimized

Resistance of ground should be low (tominimize power supply voltage drop)

Grounding controls crosstalk between signals

8

Crosstalk Improvements

Increase the spacing (S) between traces

Minimize substrate height (H) while achievingmatched impedance

Route signals orthogonally between layers

Minimize parallel runlengths between signals

Use differential routing for clock or critical nets

9

Load Board Pad Size andPlacement

10

Load Board Pad Size andPlacement

11

Model Validation &Performance Examples

Validating the Model

Explain test results

Confirm if contactor will meet customer needs

Identify improvements to contactor

Identify improvements to load board (i.e. padsizes)

Determine equivalent circuits

Investigate changes to contactors to meetcustomer specific needs

12

Model Validation &Performance ExamplesLeaded Series Return Loss (S11) Data

-50

-45

-40

-35

-30

-25

-20

-15

-10

-5

0

0 1 2 3 4 5 6 7 8 9 10Frequency (GHz)

Ret

urn

Loss

(dB)

Measured Test Data Taken by GigaTest Labs for.010" Thick ContactEquivalent Circuit Modeled Data for .010" ThickContactOptimum Load Board Layout Match Using HFSS

-20 dB Point 10.6 GHz from Measured Data

13

Model Validation &Performance ExamplesLeaded Series Insertion Loss (S21) Data

-2

-1.75

-1.5

-1.25

-1

-0.75

-0.5

-0.25

0

0 1 2 3 4 5 6 7 8 9 10Frequency (GHz)

Inse

rtion

Los

s (d

B) Measured Test Data Taken by GigaTest Labs for.010" Thick ContactEquivalent Circuit Modeled Data for .010" ThickContactOptimum Load Board Layout Match Using HFSS

14


10GBit/s BGA Non-Optimized Load Board Layout

15


10GBit/s BGA Optimized Load Board Layout

16


10GBit/s BGA With Non-Optimized Load BoardLayout

17


10GBit/s BGA With Optimized Load Board Layout

18


-40

-35

-30

-25

-20

-15

-10

-5

0

1 2 3 4 5 6 7 8 9 10Frequency (GHz)

Ret

urn

Loss

(dB)

Previously Standard Pad with Optimum Contact

Optimum Load Board Pad and Contact

Return Loss (S11) of Enhanced Pad Series Contact withPrevious Standard Pad and Optimal Load Board Pad

19


-1

-0.9

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0

1 2 3 4 5 6 7 8 9 10Frequency (GHz)

Inse

rtion

Los

s (d

B)

Previously Standard Pad with Optimum ContactOptimum Load Board Pad and Contact

Insertion Loss (S21) of Enhanced Pad Series Contact WithPrevious Standard Pad and Optimum Load Board Pad

-1 dB point @ 20+ GHzAll data modeled in HFSS

20

Conclusion Modeling trends can determine how to

improve load board layout

Load board layout can greatly affect test data

Good microwave design principles apply toload board design

Modeling and test data have shown thatsignificant improvement can be attained bybetter matching load board and contactorimpedance to the Device Under Test (DUT)

Ball Series contributions from QuakeTechnologies indicated with the Quake logo

The New YieldPro Array Series Contactor

Patent USP# 6,299,459

Julius Botka, Master [email protected]

Julius BotkaBITS.ppt 10/31/01 Page 2

Brief History• Semiconductor manufacturers and

test houses are required to testdevices with multiple functions

• Packages vary in type, size and pitch• Contactors provide the final crucial

link to testers• First YieldPro contactor patented in

1997• Lowest parasitics with

independently compliant wiping contacts

• Path length is 0.037”,performance up to 18GHz Simple

SOIC-8Housing

Frame andSlider

mounted onelastomer

Frame withslider

inserted


Original YieldPro Contactor for LeadedComponents



Today’s evolution of thebusiness:• Size and pitch keep getting smaller, number of

contacts increasing• Move from leaded to leadless BGA/LCC packages (Ball

Grid Array and Leadless Chip Carrier)

Leaded YieldPro Contactor YieldPro Array ContactorUSP#6,299,459


The New YieldPro Array Contactor• A new solution is needed for

contacting new chip scalepackages, such as BGA andLCC

• Agilent’s new YieldPro Arraydesign addressestechnologies, Bluetooth®,Wireless LAN and highspeed digital components

• All are heading towardhigher frequencies andrequire good performance inhigh speed digital, and at the3rd harmonic of the RFfundamental



Problem:Lack of reliable/repeatable contact between tester andcomponent to be tested

• Two parallel contactpaths from solderball/contact pad to DUTboard

• Contacts do not wear theDUT board

Solution:Slider

Design Considerations


Problem:

Unwanted and unmanageable inductive or capacitanceparasiticsSolution:

CSPSLIDER

BGASLIDER

• Parasitic inductanceand capacitance arereduced bycontrollingimpedance to becloser to 50ΩΩΩΩ

• Parasitics are furtherreduced byminimizing height ofthe contactor

• If increasedimpedance isrequired, adjacentground contacts canbe removed


Problem:Excessive andundefined contactresistance sensitive tocontaminants

• Resistance is lowered byhaving two parallelcontact paths

• The slider can follow theball, always maintainingtwo contacts to pad orthe sides of the ball

Solution:

Contact at onepoint/line only or

through the spring

Top View

BreakawaySide View

*Initial Agilent Design

Consideration


Problem:No wiping contact to ball/pad

• Slider wipes the sides ofthe ball, while barbspierce through the oxidelayer without gougingeither the ball or the pad

• CSP Head contacts flexand wipe contact padtoward each other

Solution:

Flexibleheads

providewipingaction

Slider wipes thesides of the ball,barbs pierce the

oxide layer


Problem:Damage to the bottom of the solder ball/pad due todirect hit and excessive pressure from contact

• No sharp pointsassociated with the topof the contact.Therefore, no gougingor re-shaping canoccur. The ball/pad isnot damaged by theYieldPro Arraycontactor.

Solution:


Large gapsin betweencontacts

Problem:Given there is always contamination in a test operation,it becomes very difficult to clean the contactor due tonarrow open spaces on the top surface wherecontamination can lodge

Solution:• Housing designed for easy

removal of contaminatesusing low pressurecompressed air

• Complete contactor can bethoroughly cleanedultrasonically


Problem:After 100,000 plunges, excessive wear requires costly andfrustrating contactor component maintenance or replacement

Solution:

• New designsupportsoperation withoutreplacement ofparts throughhundreds ofthousands ofcycles

Slider TorlonHousing

Solder ball on bottom ofDUT (Device under Test)

One of two wipe andpierce contacts to ball

One of twowipingcontactinteractionsof frame andslider

Frame One of two contactlines to PC Boardpad

Elastomer LCC Contact


Problem:Life of some contactors is limited to a few tens ofthousands plunges, offsetting the initial lower price

• Typical life of a newYieldPro Arraycontactor is well overa million cycles

• Contacts can beeasily replaced onsite

Solution:


Problem:Plating will wear away, oxidation may occur

• Metal contactcomponents of theYieldPro Arraycontactors are madeof solid precious andsemi-precious metals

• No increase in contactresistance with wear

Solution:


Problem:Part handler alignment is difficult.Solution:

• Sliders shown in fully compressedposition

BGA SLIDERCSPSLIDER

• Generous individualindependent contactcompliance, up to0.011”

• With the contactorattached to the DUTboard, there is noforce between thebottom of thehousing and theDUT board. Thiseliminates housingdistortion over time


Problem:Usability issues revolvingaround scaled down versions oflarge designs, first developedfor lower frequency applications

Solution:

• The YieldPro Array designconcept is not subject toproblems associated withscaled down designs

• Performs at high RF andmicrowave frequencies indemanding applications

Scaled down springuncoils with use, and cansplit the housing sleeveretaining contaminantsand losing compliance

Initial Agilent designconsideration


Fixture for Contactor EvaluationThe two opposingcoplanar waveguide linesoverlap for the length ofthe contacts

0.275” wide and 0.010”thick alumina coplanarwaveguide line with0.015” air gap under themid 0.230” of width

FR4 supportboard


Coplanar waveguide substrate with gated short at contactor plane.Measurement yields 2 x loss and dispersion. (Shift phase 180° before saving)

FrequencyDomaindata

TimeDomaindata Gate

Markers


Fixture S-parameters measured

MARKER 13 GHz

MARKER 26 GHz

MARKER 312 GHz


Fixture time domain is computed from S-ParametersHousing and contacts gated

GateMarkers


S-Parameter of fixture with gates “on”

MARKER 13 GHz

MARKER 26 GHz

MARKER 312 GHz

MARKER 3(Return Loss)


S11 Gated, coplanar substrate electrical length isremoved with port extension, no loss or dispersioncorrection

m1freq=1.000GHzS(1,1)=0.041 / -102.526impedance = Z0 * (0.979 - j0.079)



freq (1.000GHz to 12.00GHz)

S(1,

1)

m1

m2

m3

Scale = 0.25

Port Extension = 348 ps, 104.33 mm, 125.196 degrees


m1freq=1.000GHzData_minus_loss=0.044 / -98.862impedance = Z0 * (0.983 - j0.085)m2freq=3.090GHzData_minus_loss=0.121 / -104.742impedance = Z0 * (0.916 - j0.217)m3freq=11.78GHzData_minus_loss=0.087 / -124.213impedance = Z0 * (0.898 - j0.130)


Dat

a_m

inus

_los

s

m1

m2

m3

Scale = 0.25

S11 gated of contacts and housing with 2x loss anddispersion removed by vectorially dividing measuredand gated S11 bystored data from Page 18, thereby correcting for lossand dispersion

Use this data togenerate model


CSP 0.5 mm

TLINTL2

F=1 GHzE=2.2 noopt 1.4 to 2 Z=34.5 Ohm opt 32 Ohm to 36 Ohm

OptimOptim1

Use AllGoa ls=ye sUse AllOptVa rs=ye sSa ve AllIte ra tions=noUpda teDa ta se t=ye sSa ve OptimVa rs=noSa ve Goa ls=ye sSa ve Solns=ye sSe e d= Se tBe stVa lue s=ye sSta tusLe ve l=4De s iredError=0.0P=2Ma xIte rs=25ErrorForm=L2OptimType =Gra die nt

OP TIM

TLINTL3

F=1 GHzE=-9.18889 opt -13 to -7.5 Z=50.0 Ohm

TLINTL5

F=1 GHzE=9.38041 opt 8 to 11 Z=43.0871 Ohm opt 40 Ohm to 50 Ohm

TLINTL4

F=1 GHzE=7.11405 opt 6 to 13 Z=48.5497 Ohm opt 40 Ohm to 50 Ohm

Goa lOptimGoa l1

Ra nge Max[1]=Ra nge Min[1]=Ra nge Var[1]=We ight=Ma x=.000001Min=SimIns ta nce Na me ="SP1"Expr="mag(Da ta _minus_loss -S55)"

GOAL

Te rmTe rm6

Z=50 OhmNum=6

Te rmTe rm5

Z=50 OhmNum=5

Model of Contacts with Effects of Housing Overlay

TL3: Its negative electrical lengthresets the phase to zero fromthe contacts to the edge ofhousing.

TL4 and TL5: Represent lowerimpedance of coplanarwaveguide lines underhousing to contacts.

TL2: Is the transmission linerepresenting the contacts,one signal and two groundson either side


Data to 1 GHz and above 12 GHz is discarded due to expected inaccuracies of time domain conversion.Good agreement is shown over 3.5 octaves. Evidence of a “good” model!

Measured S11 of gated housing and contactsand computed S11 (S55) from model is shown.

Scale = 0.2


S(5,

5)D

ata_

min

us_l

oss

2 4 6 8 100 12

-26

-24

-22

-20

-18

-16

-28

-14

freq, GHzdB

(S(5

,5))

dB(D

ata_

min

us_l

oss)

CSP 0.5mm

Measured Data Model


S11 of contact region from model

2 4 6 8 10 12 14 160 18

-35

-30

-25

-20

-15

-40

-10

freq, GHz

dB S

11

Scale = 0.25

Z1 = 34.5 OhmsElec trical Length @ 10 GHz = 22 degreesC (distributed over the contact length) = 0.177 pFL (distributed over the contact length) = 0.211 nHMax Current = 1A

CSP 0.5 mm Contactor


S11

Use this data toset specs with

guardband


Best DUT Board Practices at High Frequencies

• Coplanar waveguide to contactor is best transmissiontype, use on top surface of the board above 6 GHz

• In microstrip, minimize lengths of vias to pads undercontactor

• Use blind vias, not to have the signal via extend downto or through the ground plane

• Control impedance between pads under contactor• Compensate signal path extending below contactor

housing, to achieve desired impedance/compensationfor best contactor match


*

Accounting for housing effects

*

• This transmission line represents a reduced impedance segment of theZo transmission line due to the overlay of the housing from the edge tothe contacts

• A reduction from 50 Ohms to ~ 40 Ohms was seen due to the overlay ofthe housing on a coplanar waveguide transmission line on the top layerof the DUT board; Less reduction is expected for microstrip lines

• The effect of the housing overlay can be mitigated by accounting for thehigher effective overall permitivity under the housing material

• Adjusting the dimensions of the transmission line accordingly canmaintain Zo to the contacts

• This same region’s impedance Z2 can be adjusted to values other thanZo to transform/improve the contact’s specified impedance over thedesired frequency range

Con

tact

Pr

essu

re

Con

tact

Wid

th

Rec

omm

ende

d B

all S

izes

Rec

omm

ende

d Si

ze P

ad M

in.

Con

tact

C

ompl

ianc

e

DC

Res

ista

nce

Leak

age

Impe

danc

e

Zo

Elec

tric

al

Leng

thM

ax. C

urre

nt

Car

ryin

g C

apac

ity

Cap

acita

nce

Indu

ctan

ce

Phys

ical

Le

ngth

Effe

ctiv

e D

iele

ctric

C

onst

ant,

Contactor Type & Pitch

3 GHz 10 GHz 18GHz

Alo

ng th

e Le

ngth

of

Con

tact

s

@ 1

0 G

Hz

in D

egre

es

Cen

ter

Con

tact

to

Gro

und

Note 1 Note 2 Note 4 Note 6

YieldPro 1.25 mm 40 gr 0.020" 0.007" <40m Ohm 21 dB >12 dB >8 dB 26.3 Ohm 18 3 A 0.190 pF 0.131 nH 0.927 mm / 0.0365" 2.65Note 7 70 gr max

YieldPro 0.8 mm 25 gr, 0.015" 0.007" <40m Ohm 18.5 dB >9.5 dB >5.9 dB 22 Ohm 18 2 A 0.227 pF 0.11 nH 0.927 mm / 0.0365" 2.65Note 7 40 gr max

YieldPro 0.5mm 20 gr, 0.010" 0.007" <40m Ohm 16.25 dB >7.25 dB >4.5 dB 17.6 Ohm 18 1 A 0.284 pF 0.088 nH 0.927 mm / 0.0365" 2.65Note7 28 gr max

YieldPro 0.4mm 15 gr, 0.007" 0.007" <40m Ohm 18.45 dB >9.45 dB >6.15 dB 21 Ohm 18 0.75 A 0.238 pF 0.105 nH 0.927 mm / 0.0365" 2.6520 gr max

For CSP: BGA/LGA PackagesYieldPro Ultra 30 gr, 0.6-0.762 0.5 x 0.5 0.011" <40m Ohm 27.5 dB >17.8 dB >14 dB 38.5 Ohm 22 2 A 0.159 pF 0.235 nH 1.438 mm / 0.0566" 1.65BGA/LGA 1.0mm 50 gr max mm mm

YieldPro Ultra 25 gr, 0.48-0.54 0.35x0.35 0.011" <40m Ohm 20.75 dB >11.5 dB >8 dB 28.6 Ohm 22 2 A 0.213 pF 0.175 nH 1.438 mm / 0.0566" 1.65BGA/LGA 0.8mm 40 gr max mm mm

YieldPro Ultra 20 gr, 0.3-0.42 0.2 x 0.2 0.007" <40m Ohm 24.5 dB >15 dB >11.2 dB 34.5 Ohm 22 1 A 0.177 pF 0.211 nH 1.438 mm / 0.0566" 1.65BGA/LGA 0.5mm 28 gr max mm mm

Note 1: Resistance measured on a clean contactor. Note 5: Capacitance and Inductance are distributed over the contact length.Note 2: Leakage current measured w ith 10V applied to signal contact. Note 6: Physical length is show n fully compressed.Note 3: Assumes 50 Ohm environment. Note 7: Agilent YieldPro contactors for leaded/LCC packages support: SOT, SOIC, SSOP, Note 4: Apply current only after making contact. TSSOP, MSOP, QFN, TQFP, MLF, and MLP device package requirements.

For leaded packages/LCC

Values in this table pertain to center contact with ground contacts on two sides

Return Loss

Note 3 Note 5

εε εε eef


Economic Benefits:• Increase the yield by not

rejecting good parts. Smallerguard bands can be usedbecause of less uncertainty inthe measurement

• Contacts are made of preciousmetal and will not oxidize.Good performance ismaintained due to lack ofcorrosion

• The longer life of the contactorreduces cost per parts tested

Electrical and MechanicalPerformance

Characterization of HighFrequency Test SocketsAuthors: Dr. Hanyi Ding

and Lisa Steckley,IBM Microelectronics

Presented by: Lisa Steckley

March 3-6, 2002 BiTS Workshop 2

Introduction

This presentation summarizes the ongoingwork at IBM to understand the factorsaffecting both the electrical performance andmechanical durability of several commercialRF test sockets.


Objective

Purchase from a single source to reduceCOST

Choose best socket for specific application;PERFORMANCE

Deliver ROBUST manufacturing solution


Socket Requirements for Testing RFModules

High frequency – to 12 GHz Low inductance Repeatability High manufacturing precision – fine pitch High volume test – ease of maintenance Heat dissipation


Socket Qualification Process

Determinepackage

Collect multiple socketcontactor technologies

Test socketsmechanically

Test electricalcontinuity

Perform RFanalyses

Compare withelectrical

simulations


Qualification Process:Determine package

Leadless Plastic Chip Carrier (LPCC) 20lead-tin leads, 4x4mm, 0.5mm pitch, exposedpaddle ground


Qualification Process:Contactor technologies studied

Pogo-style Spring loaded, gold plated, BeCu pogo, torlon housing

S-geometry Gold plated, BeCu contact, torlon and elastomer housing

Low-profile interconnect Conductive interconnect simulating plunge-to-board, torlon

housing Conductive elastomer

Gold plated contact set plus conductive elastomer, torlonhousing


Qualification Process:Test Sockets Mechanically

Build packages with shorted die Build board for continuity test Inspect packages, contacts, and board Cycle packages through handler, testing for

DC contact Re-inspect for contact and board wear


Mechanical Analysis:Pogo-style socket before test


Mechanical Analysis:Pogo-style socket after test

(picture after 5000, 10000 parts cycled)

Lead-tin from package collects on gold pogo

Inexpensive maintenance

Further testing required to determinecleaning frequency


Mechanical Analysis:S-contact socket before test


Mechanical Analysis:S-contact socket after test


Lead-tin debris on elastomer; potential forshorting

Little to no damage on contacts

Inexpensive maintenance



Mechanical Analysis:Low-profile interconnect socket before test


Mechanical Analysis:Low-profile interconnect socket after test


Lead-tin debris visible

Contact scrubbing points show little wear

Expensive maintenance; further testing todetermine mechanical life



Mechanical Analysis:Conductive elastomer socket before test


Mechanical Analysis:Conductive elastomer socket after test


Lead-tin collects on gold contact set

Expensive maintenance; further testing todetermine mechanical life



Mechanical Analysis:Board wear after test Pogo-style socket

Little wear on board; consistent witness marks

S-contact socket Evident wear from wiping action

Low-profile interconnect socket Visible wear on board; can be improved by usingelastomer added for compliance

Conductive elastomer socket Little wear on board


Qualification Process:Test socket continuity

Yield using pogo socket:



Yield using s-contact socket:



Yield using low-profile interconnect socket:



Yield using conductive elastomer socket:


Electrical Performance Simulations

Socket companies typically providecharacterization reports based on a verysimple equivalent circuit.

Our goal is to verify the performanceparameters they claim using:Momentum, a 2.5D electromagnetic simulatorAnsoft HFSS, a 3D modeler and simulator



Equivalent circuit for socket pins



Structure setup for pogo pin type socket modeling using Momentum



Structure setup for pogo pin type socket modeling using HFSS


Recognition of Support

John Blondin – IBM, Front-End-Hardware Ronald Clarke – IBM, Test Applications Eng. Dr. Hanyi Ding – IBM, RF Board Design John Ferrario – IBM, Manager RF Test

Development Gene Patrick – IBM, Front-End Hardware Kevin Potasiewicz – IBM, RF Test Engineer Randy Wolf – IBM, RF Board Design

burn-in & test socket workshop · plating thickness has the biggest effect on load board life....

Documents