Transcript propdefense3

Three-Dimensional
Microelectronics Integration:
Design, Analysis and
Characterization
Zeynep Dilli
Ph.D. Program Dissertation Proposal
Introduction & Motivation: 3-D Integration

Current trend in electronics: Tighter integration at every integration
level:




Device  Gate  Chip  …  Board  Main Board  System
Still Planar!
Limitations: Speed, compactness, signal clarity, robustness…
…with sidewall
“Smart Dust” systems



Ideally self-contained, self-powered and small
May require mixed-signal integration
3-D stacking might be the ideal answer
Stacks with cap chips…
…with intertier vias
connections
3-D vs. 2-D Integration:
Advantages & Disadvantages
1.
2.
Net system size reduction
Increased active Si area/chip
footprint
Delay reduction: faster clocks
or higher bandwidth
3.
•
•
4.
5.
6.
7.
8.
1.
2.
Increased heat-dissipation
problems
Increased design complexity.
Shorter interconnects
Lower parasitic & load
impedance
Potential intra-system noise
reduction
Potential substrate noise
reduction
Heterogeneous integration
More freedom in geometric
design
Lower power consumption.
Challenge to the Designer
3-D integration is a subject that
ties together chip design, chip
physics, device design, circuit
design, electromagnetics, and
geometrical layout problems.
Outline
Proof of concept: A 3-D integrated selfpowering system
 Circuit performance in 3-D integration
 Passive devices for self-contained 3-D
systems

Self-Powered Electronics by
3-D Integration: Proof of Concept
3D system concept: Three tiers



Sensor (Energy harvesting: Photosensor)
Storage (Energy: Capacitor)
Electronics (Local Oscillator and Output Driver)
Process & Circuit

0.18 μm fully-depleted SOI process



ip
3 metals
p-type substrate, ~1014
Implants: Threshold adjustment, CBN and CBP (5x1017),
source-drain, PSD and NSD (0.5x1019, 1x1019)
Process information





Silicon islands
50 nm thick
Three-metal
process
Three tiers
stacked
Through-vias
Top two tiers
turned upsidedown
Figure adapted from MIT_LL
3D01 Run Application Notes
Photodiodes: Design Issues
Major


Photocurrent=Responsivity [A/W] x Incident Power
Responsivity= Quantum efficiency x λ [μm] /1.24


problem: The material depth is very small
For red light, λ [μm] /1.24 = 0.51
Incident Power = Intensity [W/μm2] x Area
[μm2]
Sunlight intensity ≈ 1x10-9 W/μm2
 Quantum Efficiency: η = [# electron-hole pairs]/ [# incident photons]


Depends on reflectance, how many carrier pairs make it to the outer circuit,
and absorption
 At 633 nm (red light), absorption coef. ≈3.5e-4 1/nm
 amount of photons absorbed in 50 nm depth is (1-exp(-αd)) ≈ 0.017


η = 0.017 x reflectance x ratio of non-recombined pairs
≈ 0.017 x 0.75=0.013
x 0.51 x 1x10-9 x Area[μm2]
= 6.63 pA/μm2
 Photocurrent=0.013
Photodiodes: Design Issues
x 0.51 x 1x10-9 x Area[μm2] = 6.63 pA/μm2
Photosensitive area: pn-junction depletion region width (Wd) times length
Available implants: Body threshold adjustment implants (p-type CBN and
n-type CBP, both 5x1017 cm-3); higher-doped source-drain implants and
capacitor implants; undoped material is p-type, ~1014 cm-3.
Two diode designs: CBN/CBP diode and pin diode (CBP/intrinsic junction)
 Photocurrent=0.013






CBN/CBP diode Wd=0.0684 μm; A=0.5472 μm2
Pin-diode Wd ≈ 1.5 μm; A=15 μm2; possibly problematic
To increase: Higher-intensity light; optimal wavelength (higher wavelength
increases λ/1.24, but decreases absorption)
Layout Constraints: As many diodes as possible;
diodes in regular arrays; need three bonding pads of a
certain size; assigned area 250 μm by 250 μm only
Layout: 2062 CBN/CBP diodes: 7.48 nA; 52 pin diodes: 5.17 nA
 Expect about 12 nA
Photodiodes: CBN/CBP Diode Layout
Photodiodes: pin diode layout
Layout: Tier 3, Diodes and Pads
“GND”
Oscillator
output
“VDD”
Layout: Tier 2, Capacitor
Top plate: Poly
Bottom plate: N-type
capacitor implant, CAPN
Extracted value: 29 pF
Expected value: 30 pF
Layout: Tier 1, Local Oscillator
Circuit Operation

Cstorage is charged up to a stable level depending on iph.
Circuit Operation

Vrail=270 mV, fosc=1.39 MHz for iph=15 nA.
3-D System, Further Research, 1

Test the device once fabrication is completed:
 Self-contained

system, 250 μm x 250 μm x 700 μm
Design a version to be fabricated at LPS
 Greater
active photosensor depth
 Voltage regulator to prevent diode forward bias
current overtaking the photocurrent

Investigate rectifying antennas as alternate
power source
 Preliminary
investigation done on circuit board level
 See further work on passive structures
3-D System, Further Research, 2


Compare yield of 3-D system with planar system of the
same footprint area
Codify self-powering system design methodology

Photodiode-based:




Rectifying-antenna based:




Photodiode power generation ability: Diode design and chip optical
design (microlenses/AR coatings…)
Charge storage system: Capacitor, high-k dielectric use
Power regulation circuit requirement
Antenna properties: Possible low-k dielectric use
Need for a transformer
Rectifier diode design
Tied to load circuit characteristics
Outline
Proof of concept: A 3-D integrated selfpowering system
 Circuit performance in 3-D integration
 Passive devices for self-contained 3-D
systems

3-D Integration: Performance Study

Speed
 Intra-chip

communication
Heat
 Generation

and dissipation
Noise
 Substrate
noise
 External Interference


Signal Integrity
Compare performance with planar integrated
circuits and connections
Performance: Speed

Bonding pads, wires: Extra load + parasitics:
Slow things down
Left: “External”
ring oscillator, 11
stages (two
stages are shown)
Right: Internal ring
oscillator, 31 stages,
output to divide-by64 counter
Both are comprised of minimumsize transistors, simulated speed
for 31 stages: 132 MHz.
Internal Osc.
External Osc.
One-stage delay
112
MHz
(31stage)
(equivalent to 1.16
GHz for 3 stages)
398
KHz
(11stage)
(equivalent to 1.46
MHz for 3 stages)
~330
ps
for
internal, ~330 ns
for external dev.
Speed ratio: 794.5
Load ratio: ~1000
3-D Connections vs. Planar Off-chip Connections
Chip-to-chip communication between different chips with vertical vias that
require 12m x 12m metal pads
Cadence-extracted capacitance for a pad 9.23 fF: Same order of magnitude
as inverter load cap
in2
out1
out2
in1
3-D Connections: “Symmetric” Chip
Structures that can
be connected in
3D and planar
counterparts for
comparison
3-D Connections: “Symmetric” Chip
•A 31-stage planar ring oscillator and
•A 31-stage 3-D ring oscillator (In the figure, groups of 5-5-5-5-5-6).
The proper pairs of pads have
to be connected to each other
through vertical through-chip
vias post-fabrication for the
circle to close.
Simulation results:
Planar: 142 MHz
3-D, six “layer”s: 122 MHz (six
vertical pads as extra load)
To counter input
“symmetry” axis
3-D Integration: Performance Study

Speed
 Intra-chip

communication
Heat
 Generation

and dissipation
Noise
 Substrate
noise
 External Interference


Signal Integrity
Compare performance with planar integrated
circuits and connections
Performance: Heat

Coupled simulation at device and chip level to characterize chip
heating: Generation, distribution and dissipation
Performance: Heat
T
C
  (To ) 2 T  H
t
Modified heat flow equation:
T 6  o T f S f
CV

 H
t f 1 l f
V
Integrated over a volume to
obtain a “KCL” eqn.:
“C(dV/dt)
Discretized:
th
i , j ,k
C
(T
l
i , j ,k
T
t
l 1
i , j ,k
) (T

l
i , j ,k
T
l
i 1, j ,k
Rith 1 , j ,k
2
) (T

l
i , j ,k
T
l
i , j 1,k
Rith, j  1 ,k
2
+
(V2-V1)/R
) (T

l
i , j ,k
T
l
i , j ,k 1
Rith, j ,k  1
=
)
I”
 Iil, j ,k (Ti ,l j ,1k )
2
General Algorithm: Solve device equations for a range of temperatures; set up the
chip thermal network including heat generation; assume initial temperatures and
solve the thermal network; obtain heat generated by each transistor at that
temperature; reevaluate heat generated by each transistor, repeat until convergence
Also Possible: Use solver to suggest layout solutions for heat dissipation
Performance: Heat


Device operation characteristics depend
on temperature (e.g. I-V characteristic of
a diode)
Affects circuit operation (e.g. foscof a ring
oscillator)
3-D Integration: Performance Study

Speed
 Intra-chip

communication
Heat
 Generation

and dissipation
Noise
 Substrate
noise
 External Interference


Signal Integrity
Compare performance with planar integrated
circuits and connections
Performance: Noise

Substrate Noise
 Modeled
as substrate currents [1] or lumpedelement networks [2]
 Characterized with test circuits [3] or Sparameter measurements [4]
[1] Samavedam et al., 2000
[2] Badaroglu et al., 2004
[3] Nagata et al., 2001, Xu et al., 2001
[4] Bai, 2001
Performance: Signal Integrity



On-chip interconnects on lossy substrate: capacitively
and inductively coupled [1]
Characterized with S-parameter measurements
Equivalent circuit models found by parameter-fitting
[1] Zheng et al, 2000, 2001; Tripathi et al, 1985, 1988…
Performance: Signal Integrity

Substrate properties and return current paths
affect interconnect characteristics
 Three
modes of operation, affecting loss and
dispersion [1]
 Mutual inductance from a return current with a
complex depth to calculate interconnect p.u.l
inductance [2]


Effect of a second substrate stacked in proximity
not investigated
Effect of vertical interconnects not investigated
[1] Hasegawa et al, 1971
[2] Weisshaar et al, 2002
Performance: Further Research, 1
Speed: 3-D integrated chip in design
revision
 Heating: Planar heating characterization
chip in fabrication; 3-D heating
characterization chip being designed
 Noise: Design a planar chip to model and
characterize the substrate noise coupling
within; design a 3-D integrated chip to
compare noise performances

Performance: Further Research, 2

Signal Integrity: Adapting the heat-elements
network to coupled interconnect models
 Evaluate
the sensitivity of different interconnect layouts
to external pulses



Set up a coupled network
Use random current sources as external pulses to generate a
coupling map
Design interconnect chips for experimental verification
 Interconnect
equivalent circuit model/parameter
alterations for 3-D integration



Investigate capacitive and inductive coupling to a nearby
conductive, electrically disconnected substrate in addition to the
associated substrate
Derive transmission line parameters
Design chips for experimental verification
Outline
Proof of concept: A 3-D integrated selfpowering system
 Circuit performance in 3-D integration
 Passive devices for self-contained 3-D
systems

On-Chip Inductors and Transformers

Self-contained integrated
systems may require passive
structures:



Communication with other
units in the network: on-chip
antennas
Power harvesting: Rectifying
antennas and transformers
Various analog circuits--mixers, tuned amplifiers,
VCOs, impedance matching
networks: Inductors,
transformers, self-resonant LC
structures
On-Chip Inductors and Transformers

Planar inductor design





Number of turns
Total length
First/last segment length
Trace width
Planar inductor
modeling

Define an equivalent
circuit, parametermatch to
measurements
 Separate physicsbased approaches for
serial or shunt
parameters





Trace separation
Metal layer
Substrate doping
Substrate shields
Stacked or coiled
structure
De-embedding and Extraction
L( ) 
m(1/ Y11 )

m(1/ Y11 )
Q( ) 
e(1/ Y11 )
An on-chip inductor has frequency
ranges where it behaves
inductively or capacitively:
----DUT----
-----Ref. frame after Open is taken out-------
--------Measured reference frame for DUT_full------------
A Qualitative Look at the
Inductance Curve
On-Chip Inductors: Measurements
Different Metal Layers
M3: Lowest capacitance, highest Q factor
Planar vs. Stacked Inductors
58072 micron2 vs. 22500 micron2
Transformers
Tuning an Inductor with Light
Bottom right: Shunt capacitance
increases; fsr drops ~150-200 MHz
Top right: Substrate resistance
decreases: peak impedance increases
Exploiting Self-Resonance

Certain circuits need LC tanks
 Mixers, tuned
 LC filters


LNAs (as a tuned load)
There is some interest on intentionally
integrating inductors with capacitors to obtain an
LC tank
Use the equivalent circuit model in a circuit
design to exploit this effect and investigate
optimization; verify experimentally
Passives: Further Research

3-D inductors in chip stacks
 Investigate
multiple-substrate inductor
geometries
 Electromagnetic modeling, experimental
verification
 Low-k dielectrics

Tunable Self-Resonant Structures
 Controlled
tuning methodology
Photoelectrical
 Electrical
 Electromagnetic

 Circuit
applications
Summary
3-D Self-contained System Design
 3-D System Performance Analysis

 Speed
 Heat
 Noise
 Signal

integrity
Passive Structures for Self-Contained
Systems
 3-D
passives
 Tunable passives
…And on another track…

Engineering education research

Past work:

Helped direct ECE program for Maryland Governor’s Institute of Technology
program, Summer 2001



Contributed to the design of ENEE498D, Advanced Capstone Design


Presented paper in ITHET 2004
Participated in departmental outreach programs


Co-authored textbook
Benjamin Dasher Best Paper Award in FIE 2002
GE program for high school teachers, MERIT students, WIE summer programs
Attending PHYS 708: Physics Education Research Seminar


Engineering education is an open research field as well
Considering several questions adapted from PER: Student knowledge body
characterization; concept lists for EE…