Upload
others
View
7
Download
0
Embed Size (px)
Citation preview
A Test Circuit Based on a Ring Oscillator
Array for Statistical Characterization of
Plasma-Induced Damage
Won Ho Choi1, Saroj Satapathy1, John Keane2,
and Chris H. Kim1
1 University of Minnesota, Minneapolis, MN
2 Intel Corporation, Hillsboro, OR
2/21
Agenda
• Plasma-Induced Damage (PID)
• Proposed PID characterization circuit
Based on a Ring Oscillator Array
• Antenna Design
• Statistical Experiment Results
• Conclusions
3/21
Plasma-Induced Damage (PID)
• Plasma charge generated during the fabrication process
leads to damage in the gate dielectric manifesting as
latent device lifetime issue.
• The contiguous metal structure referred to as “antenna”
Receiver Driver
Gate dielectric Junction
M1
M2
M3
M4
Plasma charge
Charge build up during back-end process
Long metal interconnect (Antenna)
Z. Wang, et al., ICICDT 2005
4/21
(a) PID in Inverter Chain
PID
Driver Receiver
M1
M2
M3
Long metal wire
Receiver
Receiver
Reverse
biased
P-diode
Reverse
biased
N-diode
(b) PID Solution:
Jumper Insertion
(c) PID Solution:
Protection Diode
(1) Increased Vth shift
>> Delay↑
(2) Aggravated TDDB
>> Lifetime↓
Inserting vias
>> R↑, Delay↑
>> EDA tool support
>> Time to market ↑
Inserting diodes
>> C↑, Delay↑
>> Leakage current↑
>> EDA tool support
>> Time to market ↑
Long
metal wireLong
metal wire
M4
Jumper
Circuit Impact and Mitigation Techniques
P. H. Chen, IEEE Circuits & Devices Magazine 2004
• Mitigation techniques incur speed, power, cost, and
time-to-market overhead
• PID impact on circuits need to be accurately assessed
5/21
Pros
Cons
Impact of
latent PID
BTI Test
High sensitivity
Short test time
Difficult to collect high quality data
(fast BTI, unwanted recovery)
Increased ∆Vth
Mechanism
n+ p+ n+ p+Si
H
+Si
H
Si
H
VDD
NBTI Stress
+
n+ p+ n+ p+Si
H
Si
H
Si
H
VDD
NBTI Recovery
+
VDD
VDD
Characterizing Latent PID:
BTI (Bias Temperature Instability) Test
6/21
Increased Vth Shift by “Latent” PID
0
1
2
3
4
5
6
7
0 10E+3 15E+3
No
rma
lize
d Δ
Vth
(a
.u.)
Stress Time (a.u.)5E+3
PID
-In
du
ce
d Δ
Vth
Increased
ΔVth
w/ PIDw/o PID
• Initial Vth shift occurs on the device by “latent” PID
• After the device is being used, the Vth shift increases
7/21
Comparison with Prior Art
2D Array 2D ArraySingle device
Parameter of
interestTDDB
Increased ΔVth
TDDBIncreased Δf
Schematic
Measurement
resolutionLimitedHigh High
Sampling
Time
Short
(Parallel stress)
Long
(Serial stress)
Short
(Parallel stress)
Circuit based systemDevice probing
Measurement
time
Depends on tester speed
(several milliseconds)N/A
Short time interruption
(>1μs*)
An
ten
na
BLIGCurrent-to-digital converter
Counting number
* For a 0.01% frequency shift Resolution and a ROSC period of 10ns, enable to minimize the unwanted BTI recovery
Beat frequency system
Counting number
SmallLarge SmallSilicon area
1D Array
Initial ΔVth
High
-
(No stress purpose)
Depends on I/O BW
(several milliseconds)
Small
An
ten
na
S
G
D
An
ten
na
ID
Note: one stress cell is shown
D
G
S
An
ten
na
G
An
ten
na
`
An
ten
na
An
ten
na
An
ten
na
f BL
[2] [3][1] This work
Cell feature
[1] T.B. Hook, et al., IRPS 2000 [2] P. Simon, et al., TSM 2000 [3] W.H. Choi, et al., IRPS 2013
8/21
Proposed PID Characterization Circuit
• 10x4 stress ROSC cells (two types: PID protected,
damaged) array allows parallel stress/serial
measurement capability
• 3 reference ROSCs trimmed to various points in
stressed ROSC frequency distribution
Ro
w P
eri
ph
.
Column Periph.
Ref. ROSC 1
Three
Beat Frequency (BF)
Detection Systems
SCANOUT
RESULTS
FSM
+
Scan
Chain
Ref. ROSC 2
Ref. ROSC 3
BITLINE
NOTE: Ref. ROSC identical to PID Protected ROSC
NI DAQ Board, LabVIEWTM
interface
4
10
Protected
ROSC
Damaged
ROSC
Protected
ROSC
Damaged
ROSC
Protected
ROSC
Damaged
ROSC
Protected
ROSC
Damaged
ROSC
9/21
PID Protected and Damaged ROSCs
with Antenna Structures
6 bit trimming caps
Antenna Antenna Antenna Antenna Antenna AntennaAntenna
Output Stages
SUPPLY SUPPLY SUPPLY SUPPLY SUPPLY SUPPLY SUPPLY
SH
AR
ED
BIT
LIN
E
I/O device (tox: 5nm)
1 (measurement)
0 (stress)
SUPPLY
VDDVSTRESS
SUPPLY Switch Block
PBTI stress on NMOS
NBTI stress on PMOS
{M/S
M/S
M/SM/S M/S
RECOVER
M/S
P: 800/280nm
N: 400/280nmROSC DUT Dimensions {
• 5V DC stress, nominal 2.5V for the measurement
• The primary degradation mechanism is NBTI
• PID for 7 DUTs, NBTI degradation on PMOSs of 3 DUTs
10/21
Cross-sectional View of a Single Stage
• AR values of 4.4k (Metal) and 0.7k (VIA) were
implemented in a single inverter stage
M1
M2
M3
M4
M5
M6
M7
Plate antennas
PID Protected ROSC
VIASmall Jumper
Plate antennas
PID Damaged ROSC
total surface area of antenna structure
gate areaAR =
11/21
Simulated Frequency Difference
Between Two ROSC Types
• Simulations showing negligible freq. difference between
two ROSC types
292
294
296
298
300
302
0 0.05 0.10 0.15 0.20 0.25 0.30
Fre
qu
en
cy
(M
Hz)
Time (μs)
PID Protected ROSCs (Avg.: 297.58MHz)
PID Damaged ROSCs (Avg.: 297.60MHz)
Post Layout, VDD=2.5V, TT, 26ºC
Avg. Frequency Difference w/ No PID:
0.02MHz (0.0067%)
12/21
Beat Frequency Detection System
(PID Protected or Damaged ROSC)
Phase
Comp.
Beat
Frequency
CounterB
A
A
B
fstr : 0.99GHz
fref : 1.00GHz
Stressed ROSC
Reference ROSC(PID Protected ROSC)
• Phase comparator is used to generate the beat frequency
• 3 benefits: high resolution (>0.01%), insensitive to
common mode noise, fully-digital scan-based interface
T.H. Kim, et al., VLSI Circuits 2007
13/21
Beat Frequency for a High Resolution
(PID Protected or Damaged ROSC)
Phase
Comp.
Beat
Frequency
CounterB
A
A
B
PC_OUT
(fbeat = fref - fstress)
C
C
Count:
N=100
Stressed ROSC
Reference ROSC(PID Protected ROSC)
• Phase comparator output: fbeat=fref-fstress
• Counter counts the number of reference cycles in one
period of the beat signal
14/21
Beat Frequency for a High Resolution
(PID Protected or Damaged ROSC)Stressed ROSC
Reference ROSC(PID Protected ROSC)
Phase
Comp.
Beat
Frequency
CounterB
A
PC_OUT
(fbeat = fref - fstress)
C
Count:
0.99
0.98
fStress
(GHz): fref : 1.00GHz
N=100
N=50
• 1% frequency difference before stress N=100
• 2% frequency difference after stress N=50
• Δf or ΔT sensing resolution is 0.01%
15/21
Measured Fresh Frequency Distributions
• 1.15% degradation in average frequency as a result of
PID
0
5
10
15
20
25
30
35
40
236 238 240 242 244 246 248 250
Oc
cu
rre
nc
es
(%
)
ROSC Frequency (MHz)
μ =241.6
μ=244.4
PID Protected ROSCsPID Damaged ROSCs
Fresh, 26 C,O
1.15% Avg.
frequency (μ)
degradation
Chip #1 ~ #4
16/21
Chip-to-Chip Variation
• Degradation trend in the average fresh frequency is
consistent across different chips
0
0.4
0.8
1.2
1.6
2
2.4
#1 #2 #3 #4 #5 #6 #7 #8 #9
Chip #
PID
-In
du
ce
d Δ
f (%
)
Max: 1.39%
Min: 0.88%
Fresh, 26 CO
17/21
Measured Freq. Degradation
after 15000s, 3.6V DC Stress
Oc
cu
rre
nc
es
(%
)FreshAfter 15000s Stress
3.6V DC Stress, 26 C
ROSC DUT Frequency (MHz)
μ =227.8
FreshAfter 15000s Stress
μ=241.9
PID Protected ROSCs
PID Damaged ROSCs
O
3.2% freq. shift
0
5
10
15
20
25
30
35
40
220 225 230 235 240 245 250
μ =236.9
μ=244.8
0
5
10
15
20
25
30
35
40
220 225 230 235 240 245 250
5.8% freq. shift
18/21
Measured BTI-Induced Frequency Shift
0
1
2
3
4
5
6
7
0 10E+3 15E+3
No
rma
lize
d Δ
f, M
ea
n (
%)
Stress Time (s)
3.6V DC Stress, 26 C, 80 ROSCsO
5E+3
1.1
5%
2.6
%
PID Damaged ROSCsPID Protected ROSCs
• The initial frequency shift of 1.15% between both ROSC
types increases to 2.6% after 15000s, 3.6V DC stress
19/21
PID-Induced BTI Lifetime Projection
from Measured Statistics
101
102
103
104
105
106
107
108
109
1010
3 4 5
PID
-In
du
ce
d B
TI L
ife
tim
e (
s)
Voltage (V)
PID Damaged ROSCsPID Protected ROSCs
Operating Lifetime: 10yrs
VMAX, DAMAGED
DC Stress, 26 C, measured at 2.5VO
VMAX, PROTECTED
• The proposed test circuit would be equally useful in the
prediction of PID-induced BTI lifetimes with different
ARs in any type of device with any topology of antenna
structure under any fabrication process
20/21
65nm Die Photo and Chip Features
Measure
Interrupt
Process65nm LP
CMOS, 7M
Core / IO
Supplies1.2V / 2.5V
2.5V
Δf
Resolution> 0.01%
> 1μs
Stress
Voltage
Measurement
Voltage
P: 800/280nm
N: 400/280nm
ROSC DUT
dimensions
3.6V, 3.8V, 4.0V
496x767 μm2Total
Area
IO devicesDUT Type
VSTRESS DECAP
Ro
w p
eri
ph
.+ V
DD
H D
EC
AP
10x4 ROSC Array
Column periph.
VDDL
DECAPFSM +3 ref. ROSCs
+ 3 BF Systems
PID Protected ROSC
PID Damaged ROSC
• Measured with LabVIEW & NI DAQ board
21/21
Conclusions
• ROSC array-based PID characterization circuit with
antenna structures fabricated in a 65nm process
– To collect high quality massive PID-induced BTI statistics
– It provides a high measurement precision (>0.01%) with a short
measurement interrupt (>1μs)
• Measurement results confirm that circuits undergo
PID-induced frequency degradation even before they
are being used and therefore, the BTI lifetimes are
reduced down the road.