b117_wilson_s
Download
Report
Transcript b117_wilson_s
Continuous-Time Laser Programmable
Analog Array for Radiation Environments
MAPLD
September 8 – 10 2004
Anthony L. Wilson
Ji Luo, Joseph B. Bernstein, J. Ari
Tuchman, Hu Huang, Kuan-Jung
Chung
ATK Mission Research
5001 Indian School Road NE
Albuquerque, NM
[email protected]
University of Maryland
2100 Marie Mount Hall
College Park, MD
Outline
• Introduction – Goals and Objectives
• Overview of Technical Approach
• Process Description
• “Switch” Description
• Amplifier Design
• Configurable Analog Block (CAB) Architecture
• Laser Programmable Analog Array (LPAA) Architecture
Description
• How to get from Point A to Point B - Router Development
• User Interface
• Lessons Learned
Wilson
2
B117/MAPLD 2004
Introduction
“… a monolithic collection of analog building blocks, a usercontrollable routing network used for passing signals between
building blocks…”
E. Pierzchala, et al., “Field-Programmable Analog Arrays”, Kluwer Academic Pub., 1998
• Definitions
•
•
•
•
•
•
Wilson
LPAA: Laser-Programmable Analog Array
CAB: Configurable Analog Array
FBFTFN: Fully Balanced Four-Terminal Floating Nullor
PRA: Programmable Resistor Array
PCA: Programmable Capacitor Array
SOS: Silicon-on-Sapphire
3
B117/MAPLD 2004
Goals/Objectives
• Design of a radiation tolerant programmable
analog array
• Design of an high performance analog array
• Demonstration of the laser via technology
• Demonstration of Peregrine UTSi™ CMOS SOS
technology for analog applications
• Development of an analog router for analog arrays
Wilson
4
B117/MAPLD 2004
Overview of LPAA Design
• Array style architecture
• Fully Differential Circuit Topology
• Voltage and Current Mode Capability
• CMOS Silicon-on-Sapphire Technology
• Continuous-Time Operation
• Parasitic Aware Analog Router
• Low Resistant Laser Via Technology
Wilson
5
B117/MAPLD 2004
Process Description
•
Fully depleted CMOS/SOS
(silicon on sapphire) processing
technology
•
Has only a small source/drain to
body. Reduces parasitic
capacitance and produces a
higher speed technology
•
The threshold voltage can be
easily tailored and values of 0.0,
0.3, and 0.8 volts are available in
the technology
•
The thin gate oxide has high
intrinsic radiation hardness
•
Fully depleted body exhibits no
back channel effects
•
There is no possibility of
radiation induced leakage
current between adjacent
devices
Wilson
• Linear MIMCAP available
• High Value resistor available
6
B117/MAPLD 2004
Radiation Performance of the UTSi Transistors
Peregrine SOS 20 x 20 NChannel #5
10
10
10
10
10
10
-5
-6
10
'10krad'
'160krad'
'640krad'
'1040krad'
'1540krad'
-7
10
-8
10
-9
ID
10
Peregrine SOS 20 x 20 PChannel #5
10
10
-7
-8
-9
-10
-11
-11
-12
10
10
'10krad'
'160krad'
'640krad'
'1040krad'
'1540krad'
-10
10
10
-6
-12
-13
-1
0
1
2
3
-3
Pre/Post Radiation Performance UTSi NMOS transistor. (Data
courtesy of Peregrine Semiconductor)
Wilson
-2
-1
VG
0
1
Pre/post Radiation Performance of UTSi PMOS transistor. (Data
courtesy of Peregrine Semiconductor)
7
B117/MAPLD 2004
“Switch” Description - Laser Via Technology
•
The most reliable type of link that can be
formed is based on the principle that two
adjacent (horizontally or vertically) metal lines
can be exposed to an infrared (IR) laser pulse
and thermal expansion of the metal fractures
the intervening dielectric. Rapidly expanding
molten metal quickly flows through the
resultant crack and electrically connects the
lines
•
The structure is designed using the top two
layers of metal in the target process
•
Laser via structures can be located directly
over transistors since the active area is
shadowed by the upper metal annulus and, at
the laser energy levels needed for link
formation
•
•
Cross-section B
Laser beam
SiO2/Si3N4
Upper
metal
(M2)
Link sheet
CrossSection A
Lower metal (M1)
(a)
Link sheet
Cross-section A
(b)
W. Zhang, et al., “Reliability of Laser-Induced Metallic Vertical Links”, IEEE
Trans. on Adv. Pack., vol. 22, no. 4, pp. 614-619, 1999.
Hermeticity, radiation hardness, and
compatibility with advanced CMOS fabrication
processes
M2
Laser vias exhibit the same hardness to total
ionizing dose and to energetic heavy ions as
normal vias and normal interlevel dielectric
M1
link
FIB Cross-section View
Wilson
8
B117/MAPLD 2004
Laser Via Energy Profile
•
This energy window corresponds to the
amount of laser energy deposition is
needed to form the via
•
Formed via has less than 0.8 ohms of
resistance
Wilson
•
•
8 different test patterns
•
The preliminary evaluation achieved
fewer than one failed link per 20,000
links across the entire laser energy
window from 0.30 to 0.70 mJoules
9
Each block was a continuous chain of
links that would be connected using the
laser formed vias
B117/MAPLD 2004
Amplifier Description
•
Fully balanced fully differential
architecture
•
Fully Balanced Four-Terminal Floating
Nullor (FBFTFN) Amplifier Structure
•
Class AB output stage
Wilson
•Differential difference input stage
•Inner and outer CMFB circuits (not shown)
•Bias circuit included (not shown)
•Provides voltage and current output nodes
simultaneously
10
B117/MAPLD 2004
FBFTFN Flexibility
• Support configurations of a “universal active element”
H. Schmid, “Approximating the Universal Active Element”, IEEE TCAS-II, vol. 47, no. 11, pp. 1160-1169, 2000
Wilson
11
B117/MAPLD 2004
FBFTFN AC Performance
• Simulation plot of the
FBFTFN circuit
• Shows gain-bandwidth
= 15 [MHz]
• Shows phase margin =
67 [deg]
Wilson
12
B117/MAPLD 2004
Voltage Amplifier Simulation
•
FBFTFN configured as a
voltage amplifier
•
Feedback resistors are
varied for various gains
•
Shows reduced bandwidth
for increased gain (as
expected)
Wilson
13
B117/MAPLD 2004
Transconductor Amplifier Simulation
•
FBFTFN configured as a
transconductor amplifier
•
•
Variable gains
•
Bandwidth smaller than
voltage amplifier
Show constant bandwidth
for various gains
Wilson
14
B117/MAPLD 2004
Sallen-Key Filter Simulation
• Bandpass Filter simulation
• FBFTFN configured as a fully differential 2nd generation
current conveyor
Center Frequency = 837 kHz
Q = 7.07
Wilson
15
B117/MAPLD 2004
PRA and PCA Descriptions
•
Programmable Resistor
Array (PRA)
• Allows of single, parallel
and series connnection
• Can select multiple
individual resistors if
needed
• Can short paths for direct
connections
•
Programmable Capacitor
Array (PCA)
• Similar to PRA
• No short connectors
Wilson
16
B117/MAPLD 2004
CAB Design
• Each CAB uses 1
FBFTFN
• 4 inputs/4
outputs
• 4 PRAs
• 4 PCAs
• Lines in read = 8
track buses
• Boxes represent
laser formed via
arrays used for
routing
Wilson
17
B117/MAPLD 2004
Architecture Description
• 4x2 array of CABs
• 8 tracks per channel
• Switch Box
• Connection Box
• Dedicated routing for power supplies
• Dedicated routing for bias and common mode voltages
• Input pads on left and bottom
• Output pads on right and top
Wilson
18
B117/MAPLD 2004
Analog Array Architecture Description
Assumptions:
•
The Analog Array has the coordinate system defined from (0, 0) to (3, 5).
The four corner positions, (0, 0), (0, 5), (3, 0), (3, 5), are blank areas.
•
Each X or Y directed channel belongs to the pad or CAB right below it, or
on the left to it, having the same coordinates.
•
The legal routing connections are:
• LHS pads can connect to all the tracks in channel y (0, 1); RHS pads can connect to
all the tracks in channel y (2, 1).
• Bottom row pads can connect to all the tracks in channel x (1, 0); top row pads can
connect to all the tracks in channel x (1, 4).
• Input CAB pins can connect to tracks in the channels immediately on the left, top
and bottom of the CAB; output CAB pins can connect to tracks in the channels
immediately on the right, top and bottom of the CAB.
• Tracks in the horizontal channel can connect to tracks in vertical channel if a
switch is available at the intersection.
• Direct connections between CAB pins are not allowed; direct connections between
PADs and CAB pins are not allowed.
• Dogleg is not allowed, i.e., CAB pin cannot be acted as intermediate vertex to route
a net.
Wilson
19
B117/MAPLD 2004
Analog Array Architecture Diagram
A
uniform architecture
Channel Width: 8
Connection Block Flexibility:8
Switch Block Flexibility: 4
Segment Length: 4
Wilson
20
B117/MAPLD 2004
Connection and Switch Boxes
•
Connection box used to
achieve a connection flexibility
of 8
•
Switch box chosen to achieve a
connection flexibility of 4
•
Alternate parity for switch
boxes (Pattern 1 and Pattern 2)
used to increase routability
Wilson
21
B117/MAPLD 2004
Router Development
•
Definition of routing: Given a netlist along with a placement of the CABs and IO
cells, to route all the nets on the given FPAA architecture without exceeding the total
available routing resources and without overly degrading the performance of the
circuit (Goals: (1) Avoid congestion; (2) Satisfy a set of performance constraints)
•
Routing Problem - Graph Problem
Routing Resource Graph (RRG)
a b
1
3
4
12
CAB1 3
CAB3 11
9
c
d
a
2
10
2
9
b
c
10
5
CAB2
6
d
16
CAB415
7
13
g h
14
7
Er : feasible connections
Routing Resource type: pin,
pad and track (PPT number)
f
g
8
Vr : I/O pins, wire segments
e
e
f
8
h
16
13
find a set of disjoint trees T={T1,…Tn}. Each tree spans all the terminals of
the net with “minimum” cost:
Minimum Steiner Tree (MST) Problem: NP-Complete
Wilson
22
B117/MAPLD 2004
LPAA Router Description
•
Analog array router based on the
Pathfinder Negotiated Routing
Algorithm
•
•
Completed a routability-driven router
•
Therefore the accumulated parasitic
(especially the loading capacitance and
serial resistance) are automatically
kept to a near minimum value, along
with the wave expansion process
(breadth-first search).
•
The extremely small parasitic
associated with the laser via, this router
is sufficient for this relatively smallscaled analog array
Although called routability-driven, the
router resolves the congestion and
attempts to find the “shortest path”.
Wilson
LPAA Design Flow
23
B117/MAPLD 2004
Pathfinder Negotiated Routing Flow
Loop over all sinks
Init
Loop over all nets
PQ <= RT
Pathcost(n)= base(n)
Y
RT(i)=φ?
Sink in RT?
N
Y
N
Rip_up RT(i)
Update p(n)
RT(i)<=net(i) Source
Dequene (m) PQ
Enquene fanout (m)
Update pathcost=pathcost(m) + Base(n)
Route Net i
N
N
Sink found ?
All nets done?
Y
Backtracing
Update p(n)
Update h(n)
Y
Y
N
Overuse?
MAX_ITER?
Finish all sinks?
Y
N
END
Wilson
24
B117/MAPLD 2004
User Interface
• JavaTM chosen as language – cross platform
• Direct input from user
• Allows user to perform local connections within CAB
• Allows user to perform CAB-to-CAB connections
• Spice netlist creation
• Router compatible netlist file creation
Wilson
25
B117/MAPLD 2004
Screenshot of User Interface
• Early Development Phase
Wilson
26
B117/MAPLD 2004
Lessons Learned
• Reticle Size Problem
• Floor planning a must
•Optimizations:
• Add additional parameters to the router code cost
function
• Segmentation would need further addressing for larger
arrays
• Adding switched-capacitor implementation for increased
flexibility
• Programmable amplifier bias
• Inclusion of voltage reference
• Separation of amplifier and buffer stage
Wilson
27
B117/MAPLD 2004