Transcript EM Effects on Nanoscale Circuits and Devices

Microwave Effects and Chaos in
21st Century Analog & Digital
Electronics
V. Granatstein, S. Anlage, T. Antonsen, Y. Carmel,
N. Goldsman, A. Iliadis, J. Melngalis, P. O’Shea,
E. Ott, O. Ramahi & J. Rogers, University of
Maryland
and R.J. Baker,
Boise State University
PART A
• Study Vulnerability of Low-Voltage,
Submicron-Scale Semiconductor Devices
Circuits and Systems
• Diagnostics of Failure Modality Using
Focused Ion Beams
MW Effects on Nanoscale Circuits and Devices:
Background
• Modern IC’s can contain 10s of millions of deep submicron
transistors.
• Transistors are connected by complex network of metal
interconnects containing millions of nodes.
• Network forms extremely complex active RLC,
transmission line circuit structure.
• Future IC’s are likely to contain billions of very fragile
nanoscale transistors and billions of metal interconnects in
a complex 3-D network.
• Future IC supply voltages are likely to be as low as 25mV.
MW Effects on Nanoscale Circuits and Devices:
Problems
• Such complex circuits and small devices are likely to be
extremely vulnerable to temporary upsets and permanent
damage resulting from MW radiation.
• Unwanted MW radiation may enter circuits, especially at
I/O ports.
• Radiation can then induce high voltage and current levels
on interconnect network.
• High voltage levels can cause bit errors in digital circuits
leading to circuit failure and undetected errors.
• Very high levels can cause permanent damage to nanoscale
devices resulting from oxide breakdown and electrostatic
discharge (ESD).
MW Effects on Nanoscale Circuits and Devices:
Goals
• Determine IC subcircuits and physical layouts most
vulnerable to MW coupling.
• Understand and predict characteristics of transmission line
and RLC network formed by interconnect structure.
• Understand and predict behavior of nanoscale devices
(CMOS, BJT, HBT) with terminal voltages rapidly varying
due to MW coupling.
• Understand and predict interaction of device and
interconnect network for typical circuit topologies. Find
most vulnerable locations for soft and hard errors.
• Offer improved designs and models.
MW Effects on Nanoscale Circuits and Devices:
Experimental Approach
• Design and fabricate IC’s with test structures containing
typical circuit blocks (LNA, PLL, DAC, ADC, Clock,
Processors, ALU, Register, Control Logic). The block
diagram CPU shown represents one of our test IC’s.
• Design and incorporate test equipment directly into IC’s
being investigated.
• Expose circuits to MW radiation. Have internal IC test
equipment measure levels of induced voltages and currents,
and determine locations of failure.
MW Effects on Nanoscale Circuits and Devices:
Theoretical and Modeling Approach
• Extract RLC network resulting from interconnects using
both new and existing methods.
• Calculate MW induced currents and voltages in RLC
interconnect network.
• Use SPICE to model complex circuit. Incorporate MW
induced voltages as sources to determine effects of MW on
entire circuit.
• Use existing and develop new methods to model nanoscale
devices based on the Boltzmann and Schrodinger
equations.
• Use device simulation methods to predict internal behavior
of transistors coupled to rapid MW transitions.
Example of Detailed Internal Analysis of Submicron
MOSFET Using Boltzmann and Schrodinger Equations
Electron Concentration
MOS Cross Section
Distribution Function
Y=0.0001mm
Y=0.4mm
Use of focused ion beams to diagnose effects
of microwaves on integrated circuits
• determine how the threshold for soft failure
depends on neighboring circuits or structures by
cutting (or reconnecting) conductors
• analyze the nature of hard failures by cross
sectioning the burned out component (e.g.
transistor)
• isolate a damaged transistor and connect probe
pads to measure its characteristics
Cross section of two conductors in an IC
showing FIB steps to make a connection
Focused ion beam cut of conductor and
deposited connection
Focused ion beam milled cross section of a
part of an IC illustrating defect analysis
PART B, CHAOS STUDIES
• Chaos offers at least 2 uniquely useful perspectives for
study of the influence of microwaves on electronic
circuits.
• Wave Chaos uses new ideas on the properties of solutions
of wave equations in complex topologies to develop
statistical descriptions of fields and resonances in
enclosures containing circuits.
• Chaos in Circuits considers the possibility that RF
radiation might cause upset or damage by inducing chaotic
behavior in circuits. This is expanded upon in the
following slides.
The Resistor-Inductor-Diode Circuit
Varactor diode
(capacitance
depends on voltage)
Approximate equation for charge flowing in the circuit:
d 2Q
dQ
Q
L 2 R

 V0 sin(t )
dt
dt C (V )
As the amplitude of the drive (V0) increases, the system will
show period doubling and chaos
Setup for the Experiment
Examples of Period 1,2,3 and Chaotic Voltage Waveforms
Period-2 with 50 MHz and 11 dBm input
15
17
Q
O
0.03
16
p
q
o
P
p
17
Q
O
16
15
o
P
18
Rr
S
18
19
s
q
19
R
13
S
N
Tt
14
N
n
20
t
14
n
13
m
M
U
12
K
m
M
21
u
l
T
21
U
12
Ll
k
11
0.02
K
V
v
k
22
v
10
J
22
V
10
Jj
I
9
23
i
j
W
w
iI
W
9
23
w
h
8
h
H
0.01
G
x
7
g
24
X
x
24
G
g
X
7
F
6
F
f
y
f
6
Y
25
Y
25
z
e
E
e
5
0.00
z
26
Z
E
D
d
26
C
d
4
D
4
AA
27
aa
c
af
AA
27
AH
34
ah
c
3
28
ae
AB
C
A
30
ad
28
ab
ac
B
2
b
AB
AC
29
AD
30
31
B
b
ab
29
ac
AD
ad
31
ae
ai
AF
AG
ag
33
AH
35
AI
ai
aj
32
AG
ag
34
35
36
AJ
aj
AK
AI
36
AJ
AK
AF
32
AE
af
37
38
37
ak
38
ak
1
a
-0.01
AL
al
AL
al
A
1
a
2
AM
am
39
an
39
AN
40
AM
40
an
AN
-0.02
0
10
20
30
40
Output across the diode in RLD circuit in voltage
Output across the diode in RLD circuit in voltage.
Perion-1 with 50 MHz and 4 dBm input
0.04
0.4
0.2
o
9Ii
H
8
v
6
B
0.2
e
l
7
a
18
e
p
17
P
Q
14
AI
19
T
24
AD
7
K
20
D
W
AK
13
1
10
J
2
o
j
40
r
23
Y
16
AI
O
3
19
39
14
21
32
G
Z
Q
U
u
af
Z
AB
F
AE
M
0.0
an
S
15
5
E
38
AG
8
b
X
R
11
H
D
20
15
N
n
13
G
l
t
10
ae
S
AF
F
aa
9
q
25
Y
16
v
U
u
d
C
4
T
E
ab
w
AB
23
28
n
AD
AF
ac
22
-0.2
B
r
A
21
V
W
AC
s
17
z
R
L
w
d
o
x
y
26
24
y
z
35
AA
ad
31
AM
32
AN
af
ak
V
37
an
al
am
8
b
P
27
28
i
x
26
27
ab
29
12
-0.4
h
K
aa
29
22
30
I
25
ag
ai
s
X
AC
i
k
m
L
q
31
AA
-0.6
ac
h
3
2
C
ad
f
38
-0.8
am
30
AL
AE
5
11
34
ag
35
4
ah
AJ
33
AJ
34
A
12
AH
36
ai
40
ak
33
AG
-1.0
37
ae
I
1
a
AH
H
ah
6
36
aj
39
AK
AM
AN
c
9
aj
c
al
AL
-1.2
-1.4
0
10
20
30
Time (40 is equal to 20 nano-seconds.)
40
Output across the diode in RLD circuit in voltage
Output across the diode in RLD circuit in voltage
18
p
O
N
M
O
0.0
P
17
Q
q
p
q
18
r
s
19
p
T
14
t
n
21
u
N
U
16
22
V
16
P
m
5
23
n
7
E
17
Rr
Q
R
12
18
e
S
A
C
c
4
3
D
d
W
v
S
s
19
d
22
V
15
O
o
15
13
m
M
W
x
w
24
X
20
T
12
L
24
x
20
t
25
Y
X
y
U
u
3
Y
z
21
1
a
AJ
l
23
34
A
1
a
-0.2
ag
k
v
26
AF
af
AB
25
35
AK
ak
37
36
AI
AH
39
AN
40
an
ai
34
ah
AG
AE
ae
28
H
AG
32
28
w
33
ag
ac
AC
AD
30
ad
ae
31
8
h
G
g
Zz
h
y
6
-0.4
Ff
G
7
g
27
5
E
AA
aa
e
26
Z
f
AB
ab
aa
AA
27
-0.6
C
D
c
4
29
ab
B
2
b
ac
AC
29
B
AD
30
2
ad
b
31
AE
AN
AF
32
-0.8
33
40
an
af
am
AH
ah
AM
39
38
AI
ai
al
AL
AJ
36
aj
AK
38
aj
ak
AL
al
AM
am
-1.0
0
10
20
30
40
Period-3 with 50 MHz and 22 dBm input
f
m
l
Time (40 is equal to 20 nano-seconds.)
0.4
k
L
11
K
14
N
Chaos with 50 MHz and 13 dBm input
Jj
11
K
k
j
13
M
Time (40 is equal to 20 nano-seconds.)
g
10
Jj
9iI
F6
3.5
n
15
o
16
p
14
3.0
q
17
13
m
r
18
19
s
12
l
20
t
k
u
11
21
22
v
10
2.5
j
23
w
9
24
x
i
8
2.0
O
o
h
p
q
r
Q
R
P
25
S
s
y
n
T
t
N
26
m
M
7
g
U
u
z
l
L
v
V
aa
f
1.5
6
K
k
27
W
w
Jj
e
I
x
X
28
ab
5
i
4
d
1.0
18
R
H
h
Q
c
16
3
15
G
g
17
y
Y
19
S
21
U
P
29
V
O
22
14
N
Zz
23
W
2
b
30
X
13
a
0.5
ac
T
20
Ff
24
M
1
E
e
D
11
d
K
C
c
25
ad
Y
AA
aa
12
Z
L
26
AA
27
31
ae
an
AB
ab
AB
28
10
J
AN
ac
32
AC
29
af
am
AM
AL
al
B
b
0.0
9
AD
ad
30
I
AE
A
a
AD
H
8
33
31
AE
ae
7
F
AK
ak
ag
32
AF
af
G
6
AJ
aj
AG
34
ah
AG
ag
33
AH
ah
AH
34
40
AI
ai
AM
39
AN
am
an
39
40
AL
38
35
AI
ai
35
AK
37
AJ
36
E
5
-0.5
D
4
A
36
aj
C
37
3
ak
B
2
al
38
1
-1.0
0
10
20
30
Time (40 is equal to 20 nano-seconds.)
40
Results of the Test, so far
L = 39 mH, C = 510 pF, Narda GC3207 varactor diode,
National LM741CN op-amp
Period doubling seen, e.g., at 50 MHz, +16 dBm in RLD
No failure of op-amp observed (10->60 MHz, 0->20 dBm)
L = 390 nH, C = 510 pF, NTE 610 varactor diode,
Motorola MC1741SCP1 and MC1741CU op-amps
Period doubling seen, e.g., at 20 MHz, +20 dBm in RLD
No failure of op-amps observed (10->60 MHz, -20->20 dBm)
Conclusions
We have just begun work, so our results are preliminary…
Period doubling and chaos are common in RLD and
RLD/Op-Amp circuits driven with rf signals
No evidence of op-amp failure under circumstances similar
to those employed by Wallace
PART C
MICROWAVE TESTING
•
•
•
•
Frequency Range 100 MHz to 100 GHz
Single and Repetitive Pulses
Clock Frequency Effects
Coupling Through Apertures
Direct Injection
Purpose: to evaluate and understand the test circuit response under well-controlled
conditions; to identify components that are particularly susceptible to RF upset.
RF Frequency and power level
Microprocessors and associated circuitry may be sensitive to upset from relatively low
power RF at the clock frequency or harmonics thereof.
Initial Experiment
To test the susceptibility of generic dynamic random access memory (DRAM) RF was
injected directly into its 33MHz clock input. The DRAM module was modified to include a
capacitance coupler and was installed in a personal computer running a memory-checking
program.
CW and pulsed signals from an RF synthesizer were applied to the coupler and the power
was increased until the program detected data faults. The procedure was repeated for
frequencies of 10-400 MHz in 5 MHz steps.
Experimental Measurement of Data Corruption
in Computer Memory by RF Injection
1 Mb DRAM
Memory Module
in Computer Running
Memory Checking
Program
Synthesized
RF Generator
Pulse
Modulaton
0
0
0
0
&
0
0
0
0
0
RF
Coupling
Capacitors
Threshold Power to Corrupt RAM Data
10
9
Injected Power [dBm]
8
7
6
5
4
3
2
1
0
0
50
100
150
200
250
Frequency [MHz]
300
350
400
450
Results
Integrated circuits may be an order of magnitude more susceptible to upset at specific
frequencies. In the above case the frequencies correspond loosely to harmonics of the
DRAM’s 33 MHz clock frequency. Device sensitivity may be highly frequency-dependent
and suggests that it is important to know the upset characteristics of constituent electronic
devices before complete systems can be analyzed for upset thresholds.
For pulsed RF waveforms upset occurred at about the same power as in the CW
case, however, pulsed RF caused the computer system to halt (operating system failure)
while CW injection corrupted only the digital data, and the error detecting program
continued to function.
Understanding Energy Penetration Mechanism
Externally
coupled devices
“imperfect” source (antenna)
diffusion through shield
source
Electronic Circuitry
Apertures
(screens)
Conducted and radiated
coupling through attached power cables
Numerical Simulation of EM
Coupling through Slots
Understanding Behavior of Currents
Near Slots