Introducing 20 NM Technology in Microwind: 1. The Merge of Giants

MICROWIND APPLICATION NOTE
20 nm technology
Introducing 20 nm technology in Microwind

Etienne SICARD
Professor
INSA-Dgei, 135 Av de Rangueil
31077 Toulouse France
www.microwind.org
email: [email protected]
This paper describes the implementation of the CMOS 20 nm technology of a High Performance Bulk
Planar 20nm CMOS Technology proposed by the Joint Development Alliance (JDA) in Microwind38.
Power, performance and area (PPA) gains related to the 20 nm technology are illustrated, and new
concepts such as design for manufacturing, double-patterning, replacement metal gate process are
described. The performances of a ring oscillator layout and a 6-transistor RAM memory layout are
also analyzed.
1. The merge of giants

The Joint Development Alliance (JDA) (http://www.commonplatform.com ) gathers several IC
companies including IBM, Renesas, STMicroelectronics, Samsung Electronics, Toshiba and
GLOBALFOUNDRIES for combining the expertise and research resources of all of its partners, share
the enormous fab development costs (estimated to more than 5 B$), and propose joint nano-CMOS
technologies faster than by operating as individual companies. The JDA has recently released a High
Performance Bulk Planar 20nm CMOS technology [Huiling2012], as a result of a joint R&D
collaboration between partners. We recall in table 1 the main innovations over the past 15 years,
with an illustration of some key novelties in Fig. 1.
Technology node
180nm
Year of
introduction
2000
130nm
90nm
65nm
45nm
32/28nm
20nm
2002
2003
2004
2008
2010
2013
14nm
2015
Key Innovations
Cu interconnect, MOS options, 6 metal
layers
Low-k dielectric, 8 metal layers
SOI substrate
Strain silicon
2nd generation strain, 10 metal layers
High-K metal gate
Replacement metal gate, Double
patterning, 12 metal layers
FinFET
Application note
[Sicard2005]
[Sicard2006]
[Sicard2008]
[Sicard2010]
This application note
To appear
Table 1: most significant technology improvements over the past 15 years
Page 1/23
[email protected]
11/02/15
20 nm technology
Figure 1: Increased switch performances thanks to continuous innovations

The naming of the technology nodes (130, 90.. 14nm) comes from the International Technology
Roadmap for Semiconductors (http://www.itrs.net). The trend of CMOS technology improvement
continues to be driven by the need to
-
Integrate more functions within a given silicon area
Reduce the size of a given function
Dissipate less power
Reduce the IC fabrication cost
Increase operating speed
At each technology node corresponds a wide variety of performances, depending whether the
foundry is targeted to high performance devices (speed whatever the power consumption),
general purpose or Low Power (lower speed but power-efficient). As we may see, the 20-nm
technology proposed in Microwind is close to General Purpose characteristics.
2. Towards 100 Billion devices on a chip: what for?

The compilation of more than 100 device characteristics (Intel, AMD, ARM..) with associated
technology, year of introduction and number of transistors (expressed in million) is reported in Fig. 1.
It can be seen that the 20-nm technology node enables chip designs with around 10 Giga-transistors
on a die.
Page 2/23
[email protected]
11/02/15
20 nm technology
Figure 2: The 20-nm technology node offers design complexities approaching 10 billion devices
on a same silicon die
Figure 3: The 20-nm technology is targeted to mobile applications
Page 3/23
[email protected]
11/02/15
20 nm technology
Samsung and Apple have been the first major players in the mobile phone business to supply 20nmtechnology based platforms: Exynos 5430 for Samsung, and A8 processor from Apple. The 20nm
shrink reduces power consumption and extend battery life, in part to compensate for the additional
power required to support the 4G+ novel standard. The Apple A8 contains 2 billion transistors, that is
double the number in the A7 chip, with a die area less than 100 mm2, to improving CPU performance
(basic computing) and speeding up GPU performance (enhanced graphics, higher resolution). In
2020, the 14nm/7nm generation with extensive use of FinFET technology should be used to develop
mobile applications based on the 5G standard.
3. Power, performance and area gains

Expected benefits
The Power, performance and area (PPA) gains are an important metric for justifying a shift from older
technology nodes to new ones. Some of the key features of the 20-nm technology from the JDA
[Huiling2012][Goldberg2013] are summarized in Table 2. Compared to 32/28-nm technology node
[Sicard2010], the 20-nm technology offers:
30 % less power consumption
30 % increase in switching performance
2 times higher density
A comparison between 90nm, 45nm and 20nm technologies in terms of density and power savings is
proposed in Fig. 4. The IC surface is shrinked by a factor of 20 between 90nm and 20nm nodes, while
the power consumption is reduced by a factor of 5.
Figure 4: The 20-nm technology node as compared to 90-nm and 45-nm

Page 4/23
[email protected]
11/02/15
20 nm technology
Put more cores in a same area

Knowing that the silicon die size is limited by yield to approximatively 600 mm2 (3 x 2 cm) multi-core
processors benefit from the downsizing of cores, as illustrated in Fig. 5. While a 2-core die
approaches the maximum die size in 65nm, an octa-core processor fits in the same space with some
margin in 22nm.
Figure 5: Multi-core architectures benefit from lithography improvements
Pass the 1.0V barrier

The supply voltage, both internal to the cores and the external I/O supply have been continuously
decreased due to the thinning of the gate oxide and faster switching rates thanks to reduced voltage
swings. The 20-nm technology passes the 1.0V barrier and is fixed to 0.9V, while I/Os are supplied at
1.5V (Fig. 6).
Page 5/23
[email protected]
11/02/15
20 nm technology
Figure 6: Passing the mythic barrier of 1V as supply voltage: 0.9V in 20-nm
4. Key features of the 20-nm technology

Lambda
In Microwind, we use an integer unit for drawing, which is fixed to 11nm. The drawn gate length is
22nm. The lower metal pitch is 64nm in [Goldberg2013], very close to 66 nm in Microwind using
=11nm and historical 6 metal pitch. Fig. 7 presents the N-Channel MOS device using conservative
design rules. Note that nearly 5,000 design rules exist for the 20-nm technology, of which only 100
basic rules are checked by Microwind.
Parameter
20-nm
technology
Lambda
Minimum gate length
Minimum gate width
Metal pitch
20 nm
60 nm
64
In Microwind
11 nm
2 (22 nm)
6 (66 nm)
6 (66 nm)
Table 2: Using lambda-based design in Microwind
Page 6/23
[email protected]
11/02/15
20 nm technology
Figure 7: Example of basic n-channel MOS device in 20-nm technology. Microwind do not use
minimum dimensions to ease design and routing
Core MOS devices

Table 2 gives an overview of the key parameters for the 20-nm technological node according to
[Huiling2012], concerning the internal MOS devices and layers. In Microwind (Fig. 8), we select
the RVT (reduced Vt) as Low leakage MOS
the SLVT (super low Vt) devices as High Speed MOS.
Other MOS options are not supported in Microwind for shake of simplicity.
Parameter
VDD core (V)
Effective gate length (nm)
MOS variants
Ion N (mA/m) at VDD
Ion P (mA/m) at VDD
Ioff N (nA/m)
Ioff P (nA/m)
Gate dielectric
Gate stack
Equivalent oxide thickness (nm)
Value
0.9
20
5
0.7-1.2
0.7-1.4
0.06-200
0.06-200
HfO2
Al/TiN
1
In Microwind
0.9
20
2
0.9 (LL) 1.1 (HS)
0.8 (LL) 1.0 (HS)
1 (LL) 10 (HS)
1 (LL) 10 (HS)
HfO2
Al/TiN
1
Table 2: Key features of the core devices proposed in the 20 nm technology
Page 7/23
[email protected]
11/02/15
20 nm technology
Figure 8 MOS options of the Joint Development Alliance 20-nm and Microwind targets
IO MOS devices
Table 3 gives an overview of the key parameters for the 20-nm technological node according to
[Huiling2012], concerning the Input/output MOS devices and associated supply voltage. In
Microwind, we only consider 1.5V I/O supply and tune the High-Voltage (HV) MOS device on the
median performances.
Parameter
Value
VDD IOs (V)

Effective gate length (nm)
Ion N (mA/m)
IonP (mA/m)
Ioff N (nA/m)
Ioff P (nA/m)
1.2, 1.5 or 1.8

70-150
0.013-6
0.003-2
High Voltage (HV) MOS in

Microwind
1.5
100
0.3
0.22
0.1
0.1
Table 3: Key features of the I/O devices proposed in the 20 nm technology and corresponding
values in Microwind
Page 8/23
[email protected]
11/02/15
20 nm technology
Figure 9: Vertical cross-section of N-P MOS devices with associated metal layers
5. Transistor performances in 20-nm technology

Key features
In Microwind, only three types of MOS devices (3 nMOS, 3 pMOS, 6 MOS devices in total) exist in the
20-nm technology :
the low-leakage - Reduced Vt (RVT) MOS is the default MOS device, with reasonable
leakage (6nA/m). The main objective of this MOS device is to reduce the Ioff current,
that is the parasitic current that flows between drain and source with a zero gate
voltage.
the high-speed - Super Low Vt (SLVT) MOS has higher switching performance, thanks
to a shorter effective channel length, at the price of a leakage multiplied by 10 (60nA).
The designer may choose this high-speed MOS device for cells for which speed is the
critical point, at the price of an important leakage current.
the high voltage MOS used for input/output interfacing. In Microwinds cmos20nm rule
file, the I/O supply is 1.5 V.
N-channel MOS device characteristics

The 20-nm technology uses a stack of high-k dielectric and metal gate (HKMG). As mentioned in
[Huiling2012], it consists of a HfO2 oxide combined with TiN-based WorkFunction Metal, with a
central part filled with Aluminum (Al). The process is called Replacement Metal Gate (RMG), as a first
gate is constructed at early stages of MOS lithography, and later removed and replaced by the final
gate (Figs. 10 & 11).
Page 9/23
[email protected]
11/02/15
20 nm technology
Figure 10: Representative cross-sectional of the n-channel MOS (left) and p-channel MOS
(right) [Huiling2012, Fig. 3]
Figure 10: 2D cross-section of the N-channel MOS and P-channel devices [from Sato, 2013]
Page 10/23
[email protected]
11/02/15
20 nm technology
(a) Low leakage W=1m, L= 20nm
(b) High speed W=1m, L=20nm
Figure 11: Id/Vd characteristics of the low leakage and high speed nMOS devices.
(a) low leakage W=1m, L= 20nm
(b) high speed MOS W=1m, L= 20nm
Figure 12: Id/Vg characteristics (log scale) of the low leakage and high-speed nMOS devices
Page 11/23
[email protected]
11/02/15
20 nm technology
The I/V characteristics of the low-leakage and high-speed MOS devices (Figs. 11 and 12) are obtained
using the MOS model BSIM4 (See [Sicard2007] for more information about this model). The I/V
characteristics reported in Fig. 10 demonstrate that the low-leakage NMOS has a drive current
capability of around 0.9 mA for W=1.0 m at a voltage supply of 0.9V. For the high speed NMOS, the
drive current rises to 1.1 mA/m.
The drawback associated with this high current drive is the leakage current which rises from 1 nA/m
(low leakage NMOS) to 10 nA/m (high speed NMOS), as seen in the Id/Vg curve at the X axis
location corresponding to Vg= 0 V (Fig. 11).
From a design view-point, the option menu in the MOS generator enables to switch from low
leakage to high-speed. In terms of layout, the only difference is the option layer that contains the
MOS option information (Fig. 13).
Figure 13: Changing the option of the MOS device from low-leakage to high-speed using the
option layer. Double-click in one corner of the option layer to change its properties
P-channel MOS device characteristics

The PMOS drive current in CMOS 20-nm technology is around 0.8mA/m for the low-leakage MOS
and up to 1.0 mA/m for the high-speed MOS. Note that NMOS and PMOS performances are quite
comparable (Fig. 14), thanks to the eSiGe strain engineering for PMOS that nearly compensates the
intrinsic mobility degradation of holes (P-channel) vs. electrons (N-channel). The leakage current is
around 1 nA/m for the low-leakage MOS and nearly 10 nA/m for the high-speed device (Fig. 15).
Page 12/23
[email protected]
11/02/15
20 nm technology
Figure 14: Id/Vg characteristics (log scale) of the low leakage and high-speed pMOS devices
Figure 15: Id/Vg characteristics (log scale) of the low leakage and high-speed pMOS devices
Page 13/23
[email protected]
11/02/15
20 nm technology
6. MOS Design for Manufacturing

Process Variability
One important challenge in nano-CMOS technology is process variability. The fabrication of millions
of MOS devices at nano-scale induces a spreading in switching performances in the same IC.
The effect of process variability on the MOS Ioff/Ion characteristics is plotted using the menu Ioff vs.
Ion under the MOS I/V curve menu (Fig. 16). It can be seen that the MOS devices have a wide
variability in performances. The 3 MOS types (low leakage, high speed, high voltage) are situated in
well defined space in the Ioff/Ion domain. The low leakage is in the middle (medium Ion, low Ioff),
the high speed on the upper right corner (high Ion, high Ioff), and the high voltage is at the lower left
side of the graphics (low Ion, very low Ioff). Note that the exact locations of the dots will change for
each MOS characteristics plotted because it is a random process.
Figure 16: Ioff/Ion calculated by Microwind on 100 samples of n-channel MOS with random
distribution of VT, U0, and LINT with a Gaussian distribution around the nominal value
Dummy gates
One solution to reduce MOS performance variability is to design in a regular way the MOS gates. A
new option has been inserted in the MOS generator to add dummy gates around the active device
(Fig. 17). The variability of lithography depending on the environment will significantly affect the
dummy gate, but only little the central active gate, as illustrated in Fig. 18
Page 14/23
[email protected]
11/02/15
20 nm technology
Figure 17: The layout generation includes the dummy gate option at any side (or all sides) of
the active MOS device to reduce variability and increase manufacturability
Figure 18 : The sub 40-nm lithography induces severe pattern distortions that may be reduced
thanks to dummy components all around the sensitive area (20nm\nmos-dummy-gates.msk)
Page 15/23
[email protected]
11/02/15
20 nm technology
7. Interconnects
Metal Layers
As seen in table 4, the original 20-nm technology comprises 11 layers, 3 lower ones for short routing,
4 for long routing, 2 for local supply and 2 very thick layers for system supply. In Microwind, we
reassigned the metal1 to metal8 according to the table 4:
M1 & M2 are at 6 pitch for local routing
M3 & M4 are at 8 pitch for medium routing
M5 & M6 are at 32 pitch for long routing and local supply
M7 & M8 are dedicated to general supply
Parameter
Middle-of-the-Line
(MOL)
M1-M3
M4-M7
M8-M9
Pitch (nm)
64
Thickness (nm)
50
Pitch in Microwind
Not supported
Purpose
Intra-cell routing
64
80
358
68
80
150
M1-M2: 6 (66 nm)

M3-M4: 8 (88 nm)
M5-M6: 32
M10-M11
1000
200
M7-M8: 92
Short routing
Medium routing
Block supply and long
routing
System supply and IO
routing
Interconnect layer
permittivity K
2.5-2.7
Table 4: Key features of interconnects available in the 20 nm technology and corresponding

values in Microwind
Figure 19: The 8 metal layers of the Microwind implementation of the 20-nm, originally with 11
layers
Page 16/23
[email protected]
11/02/15
20 nm technology
Layers metal5 and metal6 are a little thicker and wider, while layers metal7 and metal8 are
significantly thicker and wider, to drive high currents for power supplies (Fig. 19).
Interconnect Resistance
At minimum width, the interconnect resistance of the M1-M2 lower metal layers is around 12 /m
(Fig. 20). Metal layers 2 to 4 have relaxed design rules, meaning lower resistance (7 /m).
Figure 20: Computing the wire resistance using 1m length metal tracks at minimum width and
spacing (20nm\metalM1-M4.MSK)
Double Patterning for M1-M2 Interconnects

The double patterning is requires for M1 and M2 as the pitch between tracks is smaller than 80nm.
Half the patterns go on the first patterning and half go on the second patterning, as illustrated in Fig.
21. In order to ensure an easy selection of metal tracks for the first and second patterning, regular
structures with straightforward orientation such as M1 east-west, and M2 south-north are requested
(Fig. 22). The other solution is to relax the pitch constraints for an improved manufacturability, at the
cost of an extended silicon area. M3-M8 with pitch of 80-nm and higher are still fabricated using
simple patterning.
Page 17/23
[email protected]
11/02/15
20 nm technology
Figure 21: Simple patterning of M1 and M2 may lead to bridges and shorts
Figure 22: Principles of double patterning (20nm\double-patterning-m1.MSK)
8. Ring Inverter Simulation

The ring oscillator made from 5,7,9,etc.. inverters, has the property of oscillating naturally. The
layout of one stage is reported in Fig. 23. It corresponds to the design proposed by [Scholze2011]
based on a fanout of 3, meaning that the equivalent capacitance of 3 gates is connected to each
node. Two inverters are in parallel for an improved current drive.
Page 18/23
[email protected]
11/02/15
20 nm technology
Figure 23: Redesigning a 20-nm inverter based on the layout proposed by [Scholze2011]
(20nm\invStage_FO3_Scholze.MSK)
Figure 24: Principles of the Inverter with Fanout 3 using capa cells as loads (2 on the input, one
on the output). Two inverters in parallel correspond to one single inverter with double current
drive
Page 19/23
[email protected]
11/02/15
20 nm technology
Figure 25: The simulation shows a 26 ps cycle period, that is approx.. 5ps/stage with Fanout 3,
which fits with [Scholze2011] (20nm\RingOsc_5Stage_FO3_Scholze.MSK)
Five oscillators are chained (Fig. 25) to create a free oscillation at a rate approaching 40 GHz (aound 5
ps/stage) for a High-speed option, very close to the results published in [Scholze2011]. The FO3
(Fanout3) oscillating frequency is lower than for FO1 (Fanout 1, no extra load capacitance), but closer
to real case situation where gates are usually connected to more than one single gate, with
interconnects of significant length.
Simulation of Process Variations (PVT)

Microwind gives access to Process-Votage-Temperature (PVT) simulation through the command
Simulate Simulation Parameters Process Variations. Direct access from the simulation
waveform window is also possible using the button Process Var.. The most usual simulation
consists of simulating extreme situations (Min and Max), as compared to typical conditions, as shown
in Fig. 26. More than 300% variation is observed between the oscillating frequency under Min and
Max situations (20 to 60 GHz).
Page 20/23
[email protected]
11/02/15
20 nm technology
Figure 26: PVT Min-Max simulation showing 300 % variation in oscillating frequency of the
ring-oscillator (20nm\RingOsc_5Stage_FO3_Scholze.MSK)
Page 21/23
[email protected]
11/02/15
20 nm technology
Notice that
In Min situation, VT is high, mobility U0 is low and the channel is long (LINT>0). The
supply is minimum and the temperature is maximum.
In Max situation, VT is low, mobility U0 is high and the channel is short (LINT<0). The
supply is maximum and the temperature is minimum.
Parameter
class
Process
Voltage
Temperature
Parameter
Threshold
Voltage
Mobility
Channel
length
reduction
Supply
Temperature
Symbol
(BSIM4)
VT
Unit
V
U0
LINT
cm.V
m
VDD
TEMP
V
C
-2
Min
(20-nm)
0.30
Typ
(32-nm)
0.25
Max
(20-nm)
0.20
500
-9
1e
600
0
700
-9
-1e
0.77
125
0.9
27
1.03
-50
Table 5: Variation of process parameters
9. 6-transistor static RAM

One of the most representative designs for comparing technology nodes is the static RAM cell
designed using 6 transistors (6T-SRAM). In our implementation in Microwind (see Fig. 27), the layout
size is 407 x 176 nm, with a surface area of 0.072 m2, which is close to [Huiling2012] 0.08 m2. Note
that the layout strictly obeys the basic design rules. Most contacts are shared with neighboring cells:
the VSS, VDD contacts, the Select and Data lines. It is usual to find more aggressive layout design
rules in RAM cell designs, in order to further decrease the cell area.
Figure 27: The 6-transistor RAM layout using 20-nm design rules (20nm\SRam_6T_20nm.MSK)
Page 22/23
[email protected]
11/02/15
20 nm technology
10.Conclusions
This application note has illustrated the trends in CMOS technology and introduced the 20-nm
technology generation, based on technology information available from the Joint Development
Alliance (JDA) regrouping major integrated circuit manufacturers. The key features of the 20-nm
CMOS technology have been illustrated, including the N-channel and P-channel MOS device
characteristics, design for manufacturing and double patterning. A 5-stage ring has been used for
calibration purpose. Future work will concern the 14-nm technology node which introduces the
FinFET device and extend massively the concept of design for manufacturability.
References
[Sicard2005] E. Sicard, Introducing 90-nm technology in Microwind, software application note,
August 2005, www.microwind.org
[Sicard2006] E. Sicard and S. M. Aziz, Introducing 65-nm technology in Microwind3, software
application note, July 2006, www.microwind.org
[Sicard2007] E. Sicard, S. Ben Dhia Basic CMOS cell design, McGraw Hill, 450 pages, international
edition 2007 ISBN 9780071488396
[Sicard2009] E. Sicard Microwind Users Manual, lite version 3.5, INSA editor, 2009,
www.microwind.org.
[Goldberg2013] Goldberg, 64nm Pitch Interconnects: Optimized for Designability, Manufacturability
and Extendibility, 2013 Symposium on VLSI Technology Digest of Technical Papers
[Huiling2012] Huiling Shang, , High Performance Bulk Planar 20nm CMOS Technology for Low Power
Mobile Applications, 2012 Symposium on VLSI Technology Digest of Technical Papers
[Sato2013] F. Sato, Process and Local Layout Effect interaction on a high performance planar 20nm
CMOS, 2013 Symposium on VLSI Technology Digest of Technical Papers
[Scholze2011] A. Scholze, Exploring MOL Design Options for a 20nm CMOS Technology using TCAD,
SISPAD 2011
Page 23/23
[email protected]
11/02/15

Introducing 20 NM Technology in Microwind: 1. The Merge of Giants

Uploaded by

Copyright:

Available Formats

Introducing 20 NM Technology in Microwind: 1. The Merge of Giants

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Introducing 20 NM Technology in Microwind: 1. The Merge of Giants

Uploaded by

Copyright:

Available Formats

MICROWIND APPLICATION NOTE