Introduction and Comparison of an Alternate...

1

Kevin Hinckley, Sun Microsystems, Burlington, MADoug Winterberg, Sun Microsystems, Burlington, MAMike Ballou, Agilent TechnologiesGustavo Blando, Sun Microsystems, Burlington, MAJason R. Miller, Sun Microsystems, Burlington, MARoger Dame, Sun Microsystems, Burlington, MAAlexander Nosovitski, Sun Microsystems, Burlington, MAGregory Truhlar, Sun Microsystems, Burlington, MAShelley Begley, Agilent TechnologiesIstvan Novak, Sun Microsystems, Burlington, MA

Introduction and Comparison of an Alternate Methodology for Measuring Loss Tangent of PCB Laminates

1

DesignCon 2010, Santa Clara, CA. February 2, 2010 2

AGENDA• Introduction• The Capacitance Gradient Method (CGM)• Laminate Study

• Subject, purpose, scope• Copper-clad laminates: Low Frequency• Bare laminates: High Frequency• Unreinforced laminates• Composite test results

• Potential sources of errors• Conclusions, acknowledgement


INTRODUCTION

• Laminate loss is becoming more important• Df measurement options

• direct impedance measurements• resonance-based methods • wide-band model-based signature tests

• Multitude of IPC standards for Df testing• No agreed-upon method followed by the majority

• Data (if available) may be conflicting and confusing


THE CAPACITANCE GRADIENT METHOD (1)What is it and what are the steps• Df derived from change in Capacitance with

frequency• Measure the impedance of a CCL (Copper

Clad Laminate) DUT sample• Extract capacitance vs frequency• Establish the trendline of C(f)• Calculate Df(f) from Wideband Debye model

Impedance magnitude and phase [ohm, deg]

1.E-02

1.E-01

1.E+00

1.E+01

1.E+02

1.E+03

1.E+06 1.E+07 1.E+08 1.E+09 1.E+10Frequency [Hz]

-150

-100

-50

0

50

100

Magnitude

Phase

Measured and estimated capacitance and Df [F, -]

1.00E-10

1.25E-10

1.50E-10

1.75E-10

2.00E-10

6 7 8 9Log frequency [Hz]

0

0.01

0.02

0.03

0.04Estimated Df

Measured capacitance

Estimated capacitance


THE CAPACITANCE GRADIENT METHOD (2)The theory behind it• Capacitance can be

extracted from Im{Z}• The real and imaginary

parts of impedance are linked through causality constraints

• Integral wide-band Debye model needs only one Df(fo) point to define the entire curve

• Df is proportional to the slope of Dk

)10ln(

1ln)(1

2

12

''

ωωωωεεωε

jj

mmr ++

−Δ

+= ∞

Integral Wide-band Debye model:

Dk(f), Df(f) [-]

0.0E+0

1.0E+0

2.0E+0

3.0E+0

4.0E+0

5.0E+0


0.0E+0

4.0E-3

8.0E-3

1.2E-2

1.6E-2

2.0E-2

DkDf

ωC

Re{Z}

Im{Z}

Impedance Vector


THE CAPACITANCE GRADIENT METHOD (3)Assumptions, benefits• Relies on Wideband Debye

model: >> input data is OK from any limited frequency band

• C(f) is closer related to the magnitude of Y, relative measurement error is usually lower

• C(f) is measured in a convenient low frequency band >> there are fewer error factors

Dk(f), Df(f) [-]

0.0E+0

1.0E+0

2.0E+0

3.0E+0

4.0E+0

5.0E+0


0.0E+0

4.0E-3

8.0E-3

1.2E-2

1.6E-2

2.0E-2

DkDf

Error vector

Admittance Vector ωC

G

Df = G / ωC


THE CAPACITANCE GRADIENT METHOD (4)Limitations• Does not work if/where

Wideband Debye model is not valid

• Lowest frequency is set by impedance limit

• High-frequency limit is set by lowest resonance frequency

-150

-100

-50

0

50

100

1.0E-02

1.0E-01

1.0E+00

1.0E+01

1.0E+02

1.0E+03


Impedance magnitude and phase [ohm, deg]

Magnitude

Phase

1.0E+011.0E+021.0E+031.0E+041.0E+051.0E+061.0E+071.0E+081.0E+091.0E+101.0E+11

1.E+01 1.E+02 1.E+03 1.E+04 1.E+05Size of square [mil]

Frequency limits of measurements [Hz]

Impedance Limit

Resonance Limit

• DUT size limitation• Effective working area


LAMINATE STUDY: Subject, purpose, scope (1)

• Subject• Df (no Dk)

• Purpose• Correlate CGM against other methods• Find a Df test method that fits our needs

• Scope• Glass-reinforced laminates• 45% - 55% glass-resin ratio• 4-5 mil thickness• One glass style


LAMINATE STUDY: Subject, purpose, scope (2)

Material Thickness (mil)

Glass Style

Resin Content

(%) Vendor Df Freq (GHz) Method (IPC)

Laminate A FR408HR 5.0 #2116 55

0.0072 0.0086 0.0093

0.1 1.0

10.0

2.5.5.3 2.5.5.9 2.5.5.5

Laminate B R1566V 5.0

#2116 55

0.012 0.012 0.018

0.001 1.0

10.0

2.5.5.9 2.5.5.9 2.5.5.5

Laminate C LGC-451HR 4.0

#2116 44

0.0118 0.0124 0.159

0.001 1.0

10.0

2.5.5.9 2.5.5.9

2.5.5.13

Laminate D 370HR 4.0

#2116 46

0.0150 0.0161 0.0250

0.1 1.0

10.0

2.5.5.3 2.5.5.9 2.5.5.5


Loss tangent at mid-side points [-]

0.00E+00

2.50E-03

5.00E-03

7.50E-03

1.00E-02


COPPER CLAD LAMINATES: Low Frequency (1)

Capacitance at mid-side points [F]

1.60E-09

1.62E-09

1.64E-09

1.66E-09

1.68E-09


• Laminate A measured with E4294A Impedance Analyzer and 16192A SMD (Surface Mount Device) fixture

• Measurements taken on unconditioned samples


COPPER CLAD LAMINATES: Low Frequency (2)Comparing the four laminates

0.00E+00

5.00E-03

1.00E-02

1.50E-02

2.00E-02

2.50E-02

1.00E+02 1.00E+03 1.00E+04 1.00E+05 1.00E+06 1.00E+07Frequency [Hz]

Df(f) of Copper-Clad Laminates

Laminate A

Laminate B

Laminate C

Laminate D


COPPER CLAD LAMINATES: Low Frequency (3)

Laminate A, 3”x3” size

Correlation to Integral Wideband Debye model is poor

)10ln(2;

ln)(

)()()(π

=

⎟⎟⎠

⎞⎜⎜⎝

⎛ Δ+Δ+−

= a

fff

afC

ffCfCfDf

Differential form of Wideband Debye model correlates well

0.00E+00

2.50E-03

5.00E-03

7.50E-03

1.00E-02



Df(f) from Differential Debye Model

0.00E+00

2.50E-03

5.00E-03

7.50E-03

1.00E-02



Df(f) from Debye @ 10kHz

Df(f) from Debye @ 100kHz

Df(f) from Debye @ 1MHz


BARE LAMINATES: High FrequencyInstrumentation for 1MHz – 1GHz:

E4991A and 16453A fixture (Parallel Plate)Instrumentation for 10GHz:

E8363A and 85072A Split Cylinder Resonator (SCR)

Sample

• Bare laminate sample• Fixture creates plate capacitor• C(f) and D(f) directly measured

• Cylinder Q is measured with/without DUT• Dk and Df are calculated from shift of

resonance frequency and Q


B-STAGE BARE LAMINATES: High Frequency (1)

0.00E+00

5.00E-03

1.00E-02

1.50E-02

2.00E-02


Df of B-Stage Laminates

Laminate A

Laminate B

Laminate D

Laminate C

• All 4 laminates measured with PP and SCR• Measurements taken on unconditioned samples

• B-stage: not fully cured resin

• Df < 1% under 1MHz

• All laminates show similar trends

• Significant changes in slope

• Dashed line indicates frequency range with no data


B-STAGE BARE LAMINATES: High Frequency (2)

0.00E+00

4.00E-03

8.00E-03

1.20E-02

1.60E-02

2.00E-02


Df B-Stage Laminate B [-]

Df(f) from differential Debye model

• Integral Wideband Debye model does not match • Differential Debye model correlates well

Debye @ 10GHz

Debye @ 1GHz

Debye @ 100MHz

Debye @ 10MHz


0.00E+00

5.00E-03

1.00E-02

1.50E-02

2.00E-02


Df of C-Stage Laminates

Laminate A

Laminate B

Laminate D Laminate C

C-STAGE BARE LAMINATES: High Frequency

• C-stage: fully cured resin• Trend significantly different• Multiple inflection points• Neither integral Wideband

Debye nor Differential Debye model correlates well

Df(f) from differential Debye model

• All 4 laminates measured with PP and SCR• Measurements taken on unconditioned samples


C-STAGE BARE LAMINATES: Etching

8.0E-03

1.2E-02

1.6E-02

2.0E-02

1.E+06 1.E+07 1.E+08 1.E+09 1.E+10

Frequency [Hz]

Df, Laminate D, inner and outer layer etching [-]

Outer Process Parallel PlateInner Process Parallel PlateOuter Process SCRInner Process SCR

Etching Experiment• Why did trend change from

B-stage to C-stage?• C-stage were etched cores• Impact of inner vs. outer

layer etching process is minimal


C-STAGE BARE LAMINATES: Soaking

0.0E+00

2.0E-03

4.0E-03

6.0E-03

8.0E-03

1.0E-02

1.2E-02

1.4E-02

1.E+07 1.E+08 1.E+09 1.E+10Frequency [Hz]

Loss tangent of Laminate A [-]

Parallel Plate Fixture “as is”

Split Cylinder “as is”

Parallel Plate Fixture soaked

Split Cylinder soaked

• 2 Hour Soak in pressure cooker

• Moisture absorption increases Df but doesn’t change trend

• Is moisture responsible for Df(f) signature?

• Qualitative measurements by Parallel Plate and SCR methods


C-STAGE BARE LAMINATES: Baking

Df of Laminate B [-]

0.0E+00

4.0E-03

8.0E-03

1.2E-02

1.6E-02

2.0E-02


Step 1

Step 2

Step 3

Step 4

Step 130min @ 110C

Step 22 hours @ 110C



• Baking DOES change Df signature

• Step 4 shows previous C-stage trend

• Additional baking had no effect


UNREINFORCED LAMINATES (1)

Low and high-frequency Df signature of DuPont FR0121A acrylic laminate

0.00E+00

4.00E-03

8.00E-03

1.20E-02

1.60E-02

2.00E-02


Df High Frequency [-]

0.00E+00

1.00E-02

2.00E-02

3.00E-02

4.00E-02

5.00E-02

6.00E-02


Df Low Frequency [-]

• Negative Df slope • Inflection point in Df(f) is still present

Purpose: checking to see if difference in field orientation shows up in isotropic laminates as well


UNREINFORCED LAMINATES (3)

Df of Oak Mitsui Technologies BC24M 1/1 (left graph) and experimental laminate (right graph). Blue data points: SUN Microsystems; red data

points: courtesy of Oak Mitsui Technologies.

0.0E+00

5.0E-03

1.0E-02

1.5E-02

2.0E-02

2.5E-02

1.E+06 1.E+07 1.E+08 1.E+09 1.E+10

Frequency [Hz]

Df BC24M Laminate

0.0E+00

2.0E-03

4.0E-03

6.0E-03

8.0E-03

1.0E-02

1.E+06 1.E+07 1.E+08 1.E+09 1.E+10

Frequency [Hz]

Df Experimental Laminate

• Maybe no inflection point

• Maybe no inflection point


COMPOSITE TEST RESULTS (1)Df of Laminate B for two different resin contents, measured with Short

Pulse Propagation (SPP) method. Data courtesy of Compeq.

0.00E+00

5.00E-03

1.00E-02

1.50E-02

2.00E-02

2.50E-02

1.00E-04 1.00E-03 1.00E-02 1.00E-01 1.00E+00 1.00E+01

Frequency [GHz]

Df variation based on resin content [-]

Fitted 68.6% resin

Fitted 46.7% resin

SPP:• Laminated interconnect is

measured with a narrow pulse

• Different length traces are measured

• Complex propagation constant is calculated from far-end received pulse

• From cross section data, DC resistance and field-solver data, a first order model is created

• R(f), L(f), C(f) and G(f) are fitted to match measured response


COMPOSITE TEST RESULTS (2)

Df of Laminate A (left graph) and Laminate B (right graph) with different measurement methods. SPP data courtesy of Compeq and GCE

1.00E-02

1.20E-02

1.40E-02

1.60E-02

1.80E-02

2.00E-02

1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 1.E+01Frequency [GHz]

Df of Laminate B [-]

GCE 50.5%Compeq 48.66%SUN PPSUN SCR

0.00E+00

5.00E-03

1.00E-02

1.50E-02

2.00E-02

1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 1.E+01Frequency [GHz]

Df of Laminate A [-]

GCE 50.5%SUN PPSUN SCR


POTENTIAL SOURCES OF ERRORS (1)

Impact of electrode pressure of 16453A fixture on the measured capacitance and Df

Capacitance and Df [F, -]

0.0E+00

2.0E-12

4.0E-12

6.0E-12

8.0E-12

1.0E-11

1.2E-11

1.E+07 1.E+08 1.E+09Frequency [Hz]

0.0E+00

5.0E-03

1.0E-02

1.5E-02

2.0E-02

2.5E-02Capacitance

Df

Low pressure

Medium pressure

High pressure



Impact of point averaging in E4991A Impedance Analyzer

1.0E+061.5E+06

2.2E+063.3E+06

5.0E+067.4E+06

'100 pts''50 pts'

'20 pts''10 pts'

'5 pts''2 pts'

'1 pt'0.E+00

1.E-02

2.E-02

3.E-02

4.E-02

Frequency [Hz]

Df(f)



Impact of stacking on Df. Laminate D samples were measured in 16453A Parallel-Plate fixture.

1.5E-02

1.7E-02

1.9E-02

2.1E-02

2.3E-02

2.5E-02


Df of Laminate D

Sample 1Sample 1&2

Sample 1&2&3Sample 1&2&3&4

Sample 1&2&3&4&5


SUMMARY AND CONCLUSIONS• Wideband Debye model does not match measured data

• Multiple inflection points on Df(f) curves

• CGM can not be used to extrapolate to higher frequencies with no data

• Differential Wideband Debye model matches measured Df data wherever capacitance can be extracted reliably

• CGM can be used within the measured frequency range to cross-correlate data

• Short Pulse Propagation, Parallel-plate and Split-cylinder methods provide different results


ACKNOWLEDGEMENT

The authors wish to express their thanks to the following companies and individuals for their valuable comments and suggestions, for providing samples and measurement results: LG Innotek (Jay Juon, Mickey An), GCE (Dan Slocum Jr, Dino Chen, Joe Beers), Compeq (Michael Spencer, Richard Tu, Jesse Tsai), Viasystems (Greg Lucas), DuPont (Gerry Sinks, Glenn E. Oliver, David McGregor), Oak Mitsui Technologies (John Andresakis, Jin-Hyun Hwang), Amphenol (Bob McGrath, Tony King), Panasonic (Antonio Senese), Northeastern University (Nian Sun, Xing Xing), Agilent Technologies (Yasuhiro Mori), IBM (Roger Krabbenhoft and Alina Deutsch), CCNi (Don DeGroot), Sun Microsystems (Karl Sauter, Mike Freda).

37

THANK YOU

37

Introduction and Comparison of an Alternate...

Documents

Transcript of Introduction and Comparison of an Alternate...