investigation of accelerated alpha testing with...
TRANSCRIPT
![Page 1: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/1.jpg)
Investigation of Accelerated Alpha Testing
with Vacuum
10-27-2011
![Page 2: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/2.jpg)
2
Agenda
� Motivation
� Accelerated Alpha SER (AASER) vs Source/DUT Spacing
� Vacuum System
� AASER vs. Source/DUT Spacing in Vacuum
� Geometric Effects
� Summary
![Page 3: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/3.jpg)
3
JESD 89A Section 5.4.5
� A spacing with less than 1mm is recommended
� How does spacing between source and device affect the FIT rate
measurements?
![Page 4: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/4.jpg)
JESD 89A 5.6.3 – Geometric Effect
�Simple analytical model
�Flux highest when source is
right on top of die
�30% drop in flux at 1 mm
spacing if source is same
size as DUT
�10% drop in flux at 1 mm
spacing if source >> DUT
![Page 5: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/5.jpg)
Am241 Alpha Source
� Sealed Am241 source with 0.1uCi activity from Eckert & Ziegler
� 5cm diameter with active area diameter of 4.7cm
![Page 6: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/6.jpg)
Difficulty in Controlling the Spacing
�Direct soldered on device vs. devices mounted in socket
�Various configurations could have 1 to 3 mm differences in the spacing
between die and source.
�The die is recessed from the top of the package or socket.
Soldered Device
< 1mm
Custom Designed
Socket < 2mmDebug Socket
< 3mm
![Page 7: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/7.jpg)
Alpha SER Rate vs Source and DUT Spacing Study
� 1 mm higher in spacing should have reduced the flux by ~10%
� The observed upset event rate jumped by more than 35%
200
250
300
350
400
450
0 2 4 6 8 10
Am241 Distance from Top of Test Socket in mm
Alp
ha
Pa
rtic
le E
ve
nt
Ra
te (
A.U
.)
![Page 8: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/8.jpg)
200
250
300
350
400
450
0 2 4 6 8 10
Am241 Distance from Top of Test Socket in mm
Alp
ha
Pa
rtic
le E
ve
nt
Ra
te (
A.U
.)
Alpha SER Rate vs Source and DUT Spacing Study Cont’d
� As the spacing is increased by another 1 mm, the observed rate is
still increasing
![Page 9: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/9.jpg)
200
250
300
350
400
450
0 2 4 6 8 10
Am241 Distance from Top of Test Socket in mm
Alp
ha
Pa
rtic
le E
ve
nt
Ra
te (
A.U
.)
Alpha SER Rate vs Source and DUT Spacing Study Cont’d
� The observed rate is almost 2X higher at 1cm above the socket than
when it is right on top of the socket
![Page 10: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/10.jpg)
Geometry and Absorption Effects
� The SER vs spacing trend is
opposite to published data
� Rectangular source
� Die area is 4% of source area
� Difficult to separate the
absorption and geometric
effects
� Absorption in air and Si top
layers can shift the Bragg
peak to affect error rate
� Vacuum environment could
enable the study of geometry
effect by eliminating
absorption by air
Baumann et. al., IEEE TRANSACTIONS ON NUCLEAR SCIENCE,
VOL. 54, NO. 6, DECEMBER 2007
![Page 11: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/11.jpg)
11
Vacuum Chamber
� Laco Technologies 18” diameter, 12” height cylindrical chamber with an
acrylic top
� Custom designed cable feed-through via an ISO 100 port with epoxy cement seal
� Vacuum level down to below 100 mtorr
Digital Pressure
Gauge Cable feed-thruAlpha
Source
![Page 12: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/12.jpg)
SRAM Test Chip
� 40nm SRAM test chip with a mixture of SRAM arrays
� Die size is approximately 6mm x 6mm
� Total memory bits are greater than 6 Mbits
� Data pattern is solid 1’s at nominal voltage
� Custom designed socket minimizing the spacing between DUT and source
� Two pressure conditions are used
� Atmospheric pressure (760 torr)
� Vacuum: < 200 mtorr
� Source position varied from directly on top of socket to 1cm above
socket via 1mm thick Aluminum ring shape spacers
� Test program does continuous read until 5000 errors are collected
� Multiple-bit upsets are identified via a post-processing script
� The upset event counts are generated to eliminate variations in error rate due to
change in multi-cell upset rate
![Page 13: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/13.jpg)
Aluminum Spacer for Separation Control
![Page 14: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/14.jpg)
Upset Event Rate vs. Spacing between Source and DUT (Socket)
� Event rates below 2mm spacing are quite similar
� There are significant differences after the spacing is increased beyond 2mm
� There is a peak event rate between 5 and 6 mm but the curve drops off after 6 mm
� At 10mm, the atmospheric event rate is higher than vacuum
� The vacuum data can be attributed to geometric effects only
200
250
300
350
400
450
500
0 2 4 6 8 10
Am241 Distance from Top of Test Socket in mm
Alp
ha
Pa
rtic
le E
ve
nt
Ra
te (
A.U
.)Vacuum
ATM
![Page 15: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/15.jpg)
Single and Double-bit Upset Events
� In vacuum, double-bit upset rate trends higher from 0 to 5 mm spacing
� More contribution from the outer radial area of the source
� As the source is pulled further away, only near normal incident angles alpha particles
would hit the die and so the vacuum DBU trends lower
� Multi-cell upsets from alpha are mostly caused by alpha particles with shallow angles of incident
� Under atmospheric pressure, the collision with air molecules scattered the alpha
particles and make the DBU more uniform
86%
87%
88%
89%
90%
91%
92%
93%
94%
95%
96%
97%
0 2 4 6 8 10
Am241 Distance from Top of Test Socket in mm
% o
f S
ing
le-b
it U
ps
et
Ev
en
t
Vacuum
ATM
0%
2%
4%
6%
8%
10%
12%
0 2 4 6 8 10
Am241 Distance from Top of Test Socket in mm%
of
Do
ub
le-b
it U
ps
et
Ev
en
t
Vacuum
ATM
![Page 16: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/16.jpg)
Geometric Effects with Recessed Die Configuration
� Increasing distance will reduce the flux due to a larger area of exposure
� As the source is raised from the surface of the socket, more area of the
source will come into the effect.
� In example below, area A2 is much larger than A1.
� The angle distribution will also change as more alphas are coming from the outer radial area
of the source.
� Effect will peak when the entire source is within the line of sight of the DUT.
h2
X1
A1
h1
h2
X1X2
X3
A2
![Page 17: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/17.jpg)
Simple Model for Geometric Effects with Recessed Die
� Flux divergence approximation
� Area of exposure increases as h is
increased
� Observed area of source is a
concentric square that increases in
size as the spacing is increased
� As square corners extend outside the
active area of the source, the exposed area
is near the peak
� When the square completely engulfs the
active area, the peak is reached
h
ro
r1=ro + tan(angle) * h
A0/A1 = ro^2 / (ro + tan(angle) * h ))^2
yx
rL/2
z
![Page 18: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/18.jpg)
Geometric Effect Modeling
� Combination of two effects using parameters from the 40nm SRAM test chip, the model
shows that if there is no absorption by air, there should be an increase in flux as the
source is raised above the socket.
� The geometric factor would peak at about 5 mm spacing and gradually decrease due the
divergence effect dominating at larger spacing.
� The model showed similar trend as the experimental data.
0.0000
0.1000
0.2000
0.3000
0.4000
0.5000
0.6000
0.7000
0.8000
0 1 2 3 4 5 6 7 8 9 10
Distance between Source and Surface of Die in mm
Ge
om
etr
y F
ac
tor
![Page 19: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/19.jpg)
Atmospheric Data Corrected for Geometric Effect
� Alpha particles lose more energies as the spacing is increased
� Higher event rate could be due to the Bragg peak shifted for higher energy
transfer to the active regions of the device
0
100
200
300
400
500
600
700
0 2 4 6 8 10
Am241 Distance from Top of Test Socket in mm
Alp
ha
Pa
rtic
le E
ve
nt
Ra
te (
A.U
.)
![Page 20: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/20.jpg)
Summary
� Accelerated alpha SER testing is easy to perform but difficult to get accurate
results due to limitations in real world set up
� Spacing control of < 1mm between source and DUT is prohibitively expensive
� Observed SER rate as a function of spacing depends on
� Source and DUT configuration
– Measured alpha SER rate of de-capsulated device with recessed die configuration has different spacing
dependence than standard geometric model.
� DUT Process Technology
� Vacuum chamber can be used to characterize the geometric effect
� At < 2mm spacing, vacuum and atmospheric error rates are found to be very similar.
� At larger spacing distances, the results could be different by up to 2X.
� Multiple-upset events from alpha particles are mostly due to particles from
shallow angles of incident passing through multiple-diffusion areas.
![Page 21: Investigation of Accelerated Alpha Testing with Vacuumewh.ieee.org/soc/cpmt/presentations/cpmt1110w-5.pdf · Accelerated alpha SER testing is easy to perform but difficult to get](https://reader034.vdocuments.site/reader034/viewer/2022050206/5f5955be5ce7594f213c4a10/html5/thumbnails/21.jpg)
Follow-up Questions
� Is JESD 89A the most appropriate number to be used for the
“measured” alpha SER rate?
� What does the alpha particle incident angle distribution look like inside the
package? Would it be different for wire-bonded package and flip-chip?
� Vacuum vs. atmospheric pressure measurement?
� Should it be the average over a range of spacing between source and DUT, or
the worse case?