Authored by Andrew Olney, Brad Gifford,
John Guravage, Alan Righter - Analog Devices
Presented by Jay Hamlin – Medtronic
(with Andrew’s authorization)
Real-World Charged Board
Model (CBM) ESD Failures
Purpose of this Presentation
Document damage on real-world ESD failures
Describe a CBM test method for replicating damage on real-world failures
Explore relationship between PCB design and IC CBM robustness
Discuss corrective actions for CBM failures
Provide guidelines for minimizing real-world CBM damage
Outline Causes of Real-World ESD Failures
Background on Charged Board Model
2 Case StudiesReal-world failure signaturesCBM test methodCBM simulation resultsEffects of varying PCB parametersFixes for case study failuresDiscussion of results
Conclusions
Prevalence of EOS / ESD FailuresReview of ADI’s ’99-’08 Failure Analysis Database:
Excluding No Trouble Found, #1 Cause of Customer Returns is EOS / ESD damage
<1% of ESD failures due to HBM!
>99% of ESD failures due to CDM & CBM
~50% of “EOS” failures may be CBM damage!
HBM Becoming Legacy ESD ModelEffective controls by IC manufacturers & users
Highly automated PCB mf’g lines
Latest packages have mm-range dimensions
- mBGAs, SOTs, SC70s, LFCSPs, Bumped Die, etc.
- Effectively too small to touch just one pin / solder ball
55-Ball3.6 x 2.6mm Bumped Die
Background Causes of most real-world IC ESD failures
Charged Device Model
Charged Strip Model [Olney 2002] – now rare
Charged Board Model [Olney 2003]
HBM & CDM robustness ≠ CBM robustness
CBM damage more severe than CDM damage
May be mistaken for EOS damage
Field-Induced Charged Device Model
Replace IC (Device Under Test) with Printed Wiring Assembly
Real-World CDM Damage Same subtle damage simulated by 700V FICDM testing
•Fix: Redesign IC so capacitors are not tied directly to
package corner pins; increase customer use of ionizers
Charged Board Model (CBM)Similar to the CDM, but discharge occurs on PCB
Assumes the PCB is charged either:
(1) Directly via the triboelectric effect (frictional charging between the PCB & another object) or
(2) Indirectly due to an external electric field
Metal on the PCB (pins, edge connectors, etc.) subsequently contacts a conductive surface at or near ground potential
Result: Charge stored by the relatively large PCB capacitance is dissipated in an ultra-fast, high-energy spark discharge
Case Study 1 – Dual Op Amp IC 6.7% customer PCB failure rate
Five Dual Op Amp ICs per boardSubmicron CMOS; 2-level metal 8-lead SOIC package
All ICs operated within Absolute Max. Ratings
Failures always occurred at IC Position #5
Failure Mode: <100 short Pin 2 (-IN A) to Pin 4 (V- / substrate)Could not simulate this via HBM or CDM
Case Study 1 – Dual Op Amp IC
Customer board failure
- Damage at anode of input protection diode
CBM Test Results – Case Study 1Evaluation Board Pad 2 CBM stressed at -500 V
- Damage at anode of input protection diode
CBM Test Results – Case Study 1
Evaluation Board Pad 2 CBM stressed at -625 V
- Catastrophic damage at same protection diode
Effect of Smaller PCB Power Planes Full-size Evaluation Board for Dual Op Amp
- 3” x 3” ground plane
- CBM ESD withstand voltage = ±250 V
Half-size Evaluation Board for Dual Op Amp
- 1.5” x 3” ground plane
- CBM ESD withstand voltage = ±375 V
IC CBM ESD withstand voltage
- Inversely related to power plane area
- Inversely related to charge on power plane
Fix for Case Study 1 – Dual Op Amp Plastic CPU socket next to IC Position #5
- Charged up to 1400 V
- Induced charge on adjacent Dual Op Amp
- Subsequent IR reflow caused CBM damage
Customer PCB production line fix
- Added ionizer just before IR reflow
Result
- No more Dual Op Amp IC failures
Results Discussion – Case Study 1
FICBM vs. FICDM discharge waveforms
375 V charge voltage
Case Study 2 – DSP IC
~200 PPM customer PCB failure rate
One DSP IC per board Deep-submicron CMOS; 4-level metal 208-lead Plastic Quad Flat Pack 35 GND pins & 33 Vdd pins
DSP operated within Absolute Max. Ratings
DSP located at corner of PCB
Failure Mode: typically functional failures
Could not simulate this via HBM or CDM
Case Study 2 – DSP IC
Customer board failure
- Cracked glassivation (optical view)
Case Study 2 – DSP IC
MET4 AlCu melt / reflow “fingers” (FIB view)
Similar damage to [Banerjee et al. 2000]
CBM Test Method - Case Study 2
Cut-down customer PCB with DSP IC
Test pad discharged in ±125 V increments
CBM Test Results – Case Study 2
Cut-down customer PCB CBM stressed at -250 V
- MET4 AlCu melt / reflow “finger” (FIB view)
CBM Test Results – Case Study 2Cut-down customer PCB CBM stressed at 1 kV
- Catastrophic damage at supply buses & pads
- Most FA Engr’s would incorrectly conclude EOS
Fix for Case Study 2 – DSP IC
Large plastic edge connector on PCB
- Some connectors charged to >1000 V
- Induced charge up to ±300 V on DSP IC
- Subsequent wave solder caused CBM damage
Customer PCB production line fix
- Added ionizer just before wave solder
Result
- No more DSP IC failures
Results Discussion – Case Study 2
FICBM vs. FICDM discharge waveforms
- 250 V charge voltage
g g
-2
0
2
4
6
8
10
0.00
0.25
0.50
0.75
1.00
1.25
1.50
Time (nanoseconds)
Pe
ak
Cu
rre
nt
(Am
ps
) GND test pad FICBM
GND pin FICDM
PCB-Level EOS/ESD Protection
CBM testing was repeated on a customer PCB with a TVS + Schottky Diode across the 3.3V supply planes
No CBM damage occurred to the DSP when zapped up to 2000VNo CBM damage occurred to the DSP when zapped up to 2000V
[Without[Without T VS + Schottky, damage occurred at 250V.]
Conclusions - 1
Charged Board Model testing conducted
Test method similar to FICDM
However, higher energy & faster discharges
Successfully replicates real-world failures
CBM ESD damage
More severe than CDM damage
Occurs at lower voltages than CDM damage
Can be easily mistaken for EOS damage
Conclusions - 2
CBM ESD withstand voltages Inversely proportional to PCB capacitance
Larger power planes higher capacitance Inversely proportional to # of IC supply pins
More connections lower R & L
Minimizing real-world CBM discharges Use ionizers at key PCB assembly steps
e.g., just prior to wave solder / reflow
Device-level robustness ≠ board-level robustness DSP: Pass 2000 V CDM; fail 250 V CBM
Conclusions - 3• ADI data on real-world EOS/ESD failures:
<1% of ESD failures due to HBM!
>99% of ESD failures due to CDM & CBM
~50% of “EOS” failures may be CBM damage!
Difficult to distinguish CBM damage from EOS
CBM testing required to help distinguish the two
CBM damage typically off supply pins
Requires understanding of customer application
Need thorough ESD audit of customer PCB lines
End of Presentation
Thank you