on-chip esd protection with improved high holding current scr (hhiscr) achieving iec 8kv contact...
TRANSCRIPT
Conference paper On-Chip ESD Protection with Improved High
Holding Current SCR (HHISCR) Achieving IEC 8kV Contact System Level EOS/ESD symposium 2010
For the design of on‐chip ESD clamps against system level ESD stress three main challenges exist: reach a high failure current, ensure latch‐up immunity and limit transient overshoots. Bearing these in mind, high system level ESD requirements should be within reach. A novel improved high holding current SCR is introduced fulfilling all three requirements within drastically reduced silicon area.
Scan the code to get additional information.
On-Chip ESD ProtCurrent SCR (HH
Sorgeloos Bart, Sofics (formerly known a
phone: ++32-9-21-68This
Abstract – For the design of on-chip Ereach a high failure current, ensure lahigh system level ESD requirements introduced fulfilling all three requireme
I. IntroductionSystem level ESD robustness is impothe reliability for any product. ExteESD protection devices are widelyindustry, but for some innovatapplications there is a trend inducedfactor and operational performance toboard level devices and replace theprotection and thus connect I/O pinsharsh outside world. The on-chip ESD up to increase its failure current whichissues with traditional ESD solutionsleakage, increased load capacitance larger die area. Therefore, IC descustom ESD protection structures wiperformance but within a small silimaintaining the desired IC specificatio
Figure 1: Schematic representation of the cowith local protection between the Rx-input and
tection with Improved HighHISCR) Achieving IEC 8kV
System Level
Backers Ilse, Marichal Olivier, Keppens Bas Sarnoff Europe), Brugse Baan 188A, B-8470 Gistel, Be8-333; fax: +32-9-3-746-846; e-mail: [email protected] is co-copyrighted by Sofics and the ESD Association
Sofics BVBA BTW BE 0472.687.037 RPR Oostende
ESD clamps against system level ESD stress three match-up immunity and limit transient overshoots. Bshould be within reach. A novel improved high hoents within drastically reduced silicon area.
n ortant to ensure ernal (off-chip) y used in the tive consumer d by cost, form o remove these em by on-chip directly to the clamp is scaled
h leads to many s: increased pad
and especially signers request ith a high ESD con area while
ons.
ommunication chip VSS
This paper presents a case studyproduct where dedicated on-chiprotection is placed at the Rxtransmitter operates in an open dr
II. System Level ESTo fulfil system level ESD requESD pulse shape and test setup IEC 61000-4-2 standard [12], mar
1. High ESD current levels30 ns (According to leve
2. A fast ESD current peakrise time of 0.7ns-1ns.
3. The test must be doperation under powerfunctional component in
Figure 2: The current waveform as defin
h Holding V Contact
art elgium, om
main challenges exist: earing these in mind,
olding current SCR is
y of a communication ip system level ESD x and Tx pins. The rain configuration.
SD approach uirements, the applied
must conform to the rked by: s: 16A current peak at el 4 protection).
k value: 30A within a
done during normal red condition of the nside the system.
ned in the IEC standard
3A.1-1 EOS/ESD SYMPOSIUM 10-167
These standard system level test translate into 3 basic rules that an alevel ESD clamp must comply to:
1. Achieve high ESD current leve
2. Limit voltage overshoot
3. No latched state after the ESD
A misconception is that all the ESD cudirectly through the IO. The system ledefined to perform the test on a syston the application this means that a pcurrent will flow through the shieldipart will flow through the IO. Howepart can still be a lot larger than the recomponent ESD levels (HBM/MM). One prevailing approach in testing forapplying the system level ESD pulseIC, but this is not what the IEC 6100[12] describes. For example, connectorsshell will only be stressed with contathe metallic shell, while those with a ponly be stressed with air discharge. Designing on-chip ESD devices witcapability (16A) does not guarantee level specification will be met. Theduring system level ESD is to avoid asoft failures in the system. These failuby either EMI/EMC or due to IR dropswhich could cause triggering of the PReset) circuit, change state of some regIt is possible to conduct all the systethrough the on-chip ESD device, but a is to provide the customer with an protection featuring acceptable ESD vwith the necessary guidelines timplement the chip on the board in thguidelines ...). The standard approach to protect thesebased on the HHISCR (High holdingThe next 3 sections provide the reasdevice is an adequate protection for sys
1. SCR as a high current capable
Silicon Controlled Rectifiers (SCRs) many ESD protection devices used iThey combine a lot of advantagesleakage/capacitance, low clamping especially high current capability per the SCR into an excellent device for hilevels such as during system level ESD
specifications adequate system
els
pulse
urrent will flow evel standard is tem. Depending part of the ESD ing and another ever this second equired standard
r IEC on chip is e directly to the 00-4-2 standard s with a metallic act discharge on plastic shell will
th high current that the system real challenge
any latch-up and ures are induced s inside the chip
POR (Power On gisters ... em level current
better approach adequate ESD
values, and then o successfully he system (PCB
e applications is g current SCR). soning why this stem level
le ESD device
are one of the in the industry. s such as low voltage and area. This turns igh ESD current
D.
2. Reduction of overshoot circuit
The overshoot of an SCR is maintrigger speed of the clamp (ecircuit)[13]. As will be shown furproperty can be kept below the sensitive node. Especially for overshoot should be limited for to avoid any unrecoverable damprevent a disruption to the statedata).
3. Latch-up safe SCR soluCurrent SCR
Aside from a high current carrylimited overshoot voltage, latchthird requirement for a suiprotection. To improve the latchseveral techniques exist. One apvoltage engineering technique (fdiodes) to attain a holding volsupply voltage [1],[4],[5]. While thup safe, the clamping voltage abe higher, limiting its applicabrequirements. Another approachhigh clamping voltage, is the SCR (HHISCR) approach, pre[1],[2],[9],[10] . The basic idea trigger/holding current above tspec (e.g. 200mA) / normal spethrough external shunt resistorslayout.
Figure 3: HHISCR as presented in [2]
The JEDEC latch-up test is onlatch-up story: the ESD protectitrigger during this test. Howevoff-chip ESD protection, the onturn on during the system levefunctional circuits. To prevent condition changes so the device system level pulse has passe
with a proper trigger
nly determined by the especially the trigger rther in the paper, this failure voltage of the system level ESD,
2 reasons. First of all mage and secondly, to e of the chip (loss of
ution: High Holding
ying capability and a h-up immunity is the itable system level h-up safety of SCRs,
pproach is the holding f.i. by adding holding ltage larger than the his approach is latch-and on-resistance will ility for high current h, which solves this high holding current esented by Mergens
is to increase the the JEDEC Latch-up ecification (Figure 3) s and a special SCR
nly one aspect of the ion device should not er, in the absence of -chip protection must el test to protect the LU in that case, the must turn off after the ed. While the first
3A.1-2EOS/ESD SYMPOSIUM 10-168
condition generally relates to the trigger current of the HHISCR, the second is mainly governed by its holding current. To determine the relevant holding current and voltage of a stand-alone ESD clamp, an ESD stress pulse is superposed on a DC bias level. When the DC bias level is then stepwise decreased in subsequent measurements, the exact holding voltage and current can easily be established in the relevant time domain (Figure 4).
Figure 4: Test set up to determine real holding voltage/current
For the system level ESD protection under normal operation, the SCR must turn off after the ESD pulse has disappeared. In this case the holding current of the SCR must be above the maximum drive current of the PMOS (15mA). Applying the above measurement technique to the ESD clamp shows an effective holding current above 22mA at a holding voltage of 2.45V. This proves that the SCR will turn off after the system level ESD pulse is disappeared. The holding current can be tuned further by the resistors RSN and RSP and by the layout of the SCR body for other applications. Combining the high failure current and the demonstrated latch-up immunity, the HHISCR is a viable option for IO protection in this application. The two remaining constraints are related to area consumption and transient overshoot. The available area was not sufficient to implement the standard HHISCR, an improvement was needed to comply with all specifications. The next section presents an Improved HHISCR device with an area reduction of 22% and a faster SCR turn-on speed.
III. Integrated Trigger Approach The standard HHISCR uses fully silicided poly resistors or metal resistors (1-10 Ohm) and an NMOS
based trigger device. Instead of FS poly resistors, active resistors (N+/PWell) can be used. Figure 5 shows the top view of the implemented diffusion resistors. By applying a proper layout implementation, a parasitic NPN bipolar transistor is formed between the RSN (resistor between anode and G2) and RSP (resistor between G1 and cathode). This NPN bipolar acts as a trigger element replacing the NPN bipolar of the NMOS trigger device. The bipolar is evenly distributed across the resistors as shown in Figure 6 (b,c) for simplicity drawn as 3 separate devices. The upper NPN bipolar is formed between anode and G1, the lower NPN bipolar between G2 and GND and another NPN bipolar is formed somewhere in between. Note that each part of the NPN always sees the same resistance (RSN, RSP or a combination of these) given both resistors are of the same value.
Figure 5: Top view of the resistor implementation
The advantage of this distributed technique is that the SCR is not triggered in only one of the two trigger taps (G1, G2), but simultaneously in both trigger taps improving the turn-on speed and reducing the overshoot of the HHISCR. The basic device operation is as follows (Figure 6/Figure 7):
1. Avalanching starts uniformly over the total junction formed by the resistor RSN, but the current at the anode will increase more compared to the current at the G2 side of the junction: the avalanche current flows through the RSN resistor towards G2 which will induce a voltage drop. The avalanche current will turn on the NPNs between the 2 resistors but the bipolar at the G2/cathode side will conduct proportionally more current than the other bipolars (difference in base-emitter voltage due to resistive drops in RSP for each bipolar).
P+ diffusion
N+ diffusion
Contact
Coupled to anodeCoupled to G2
Coupled to cathodeCoupled to G1
0 1 x-axis
RS
NR
SP
3A.1-3 EOS/ESD SYMPOSIUM 10-169
Figure 6: a) the standard HHISCR with only bipolar) b) the NPN is formed distributed ovetrigger tap
2a. Conduction then spreads upmiddle bipolar
2b. Finally, the current will be dequally over the entire NPN.
3. When enough current flowresistor, to build up a voltage it, the base emitter of the NPNforward biased, inducing feedback and finally turning on
Figure 7: General IV curve of the Improved HHthe different regions 1, 2a, 2b and 3 related to different parasitic elements inside the resistors
It is important to note that elements/resistors must be separateguard rings from the SCR and other cithe formation of unwanted parasitic de
injection in G1 trigger tap (NMOS trigger device – also depr the 2 resistors (RSN, RSP) c) the NPN shown in b injects trig
pwards to the
distributed more
ws through the of 0.7 V across
N or PNP will be the positive
n the SCR.
HISCR highlighting current flow in the
the trigger ed with proper ircuit to prevent vices.
Simulations were performed to ibehaviour of the trigger ddistribution was simulated alindicated in Figure 5). The RSN adivided in different segments. FNPN model was added and a mavalanche current. The substratresistive network. (Figure 8)
DistributedSusbtrate Mo
Cathode
RSNG2
RSP
Iav IQ,c
Figure 8: Simulation model
In Figure 9 the simulated avalalong the x-axis are plotted foravalanche current is the higanode/substrate junction. All theRSN resistor will contribute to thbut the anode/substrate region
icted by its intrinsic NPN gger current in G1 and G2
investigate the current device. The current long the x-axis (as and RSP resistors were For each segment an model to simulate the e was modelled as a
d odel
G1
Anode
Iav IQ,c
lanche current levels r different times. The ghest at position 1: e junction area of the he avalanche current, will always see the
3A.1-4EOS/ESD SYMPOSIUM 10-170
highest voltage during the triggering and therefore conduct the most current. The avalanche current will pull the substrate voltage above ground and the distributed bipolar will start to conduct current.
Figure 9: Simulation of the current behaviour inside the trigger device: avalanche current (Iav) at the RSN resistor
Figure 10 shows the bipolar current distribution along the x-axis. First observation: the bipolar closest to the G2-cathode side of the junction will conduct initially the most current. This is because its base-emitter sees the largest voltage drop. Second observation: the bipolar closest to the anode junction will conduct proportionally more current.
Figure 10: Simulation of the current behaviour inside the trigger device: bipolar current (IQ,c) between RSN and RSP
IV. Analysis
1) Measurement results
Figure 11 compares TLP measurements of the standard and the improved HHISCR devices (100ns pulse width, 10ns rise time) showing a similar clamping behaviour with a small difference in the trigger voltage (Vt1). This is related to the different type of trigger bipolar that is formed. The increase from 8.5V to 8.8V stays within the design window for this design. The most sensitive element is a NMOS Gate oxide (gate source stress in inversion), with a failure level of 13.5V (including safety margins) well above the trigger voltage. Both devices remain within the design window of the I/O pins. A slightly higher failure current for the proposed device is observed due to an increased failure voltage of the trigger bipolar compared to the NMOS based trigger device.
Figure 11: TLP extracted IV curves
-04
-04
-03
-02
-02
-01
0 0.2 0.4 0.6 0.8 1
time 1
time 2
time 3
Ava
lanc
he C
urre
nt d
itst
ribu
tion
111110090908070706050504030202010000
0 0.2 0.4 0.6 0.8 1
time 1
time 2
time 3
time 4
time 5
time 6
time 7
time 8
Bipo
lar C
urre
nt d
itst
ribu
tion
0
1
2
3
4
5
6
7
0 2 4 6 8 10
10-11 10-10 10-9 10-8 10-7 10-6 10-5 10-4 10-3
HHISCRImproved HHISCR
I D
UT
(A)
V DUT(V)
ILEAKAGE
(A)
0
0.1
0.2
0.3
0.4
0.5
0.6
1 2 3 4 5 6 7 8 9
1 0-8
10-7
10-6
10-5
10- 4
10- 3
I D
UT
(A)
V DUT(V)
3A.1-5 EOS/ESD SYMPOSIUM 10-171
Figure 12 left: parallel implemented resistors; right: anti-parallel implemented resistors
2) Influence of the orientation of the resistors
This section investigates different resistor configurations. In the implementation of figure 5, the resistors were in a parallel orientation (Figure 12 left): the anode connection of RSN adjacent to the G1 connection of RSP and the G2 connection close to the cathode connection. A second possible implementation is to place the resistors anti-parallel (see Figure 12 right).
The difference between the two implementations is that in the parallel implementation (Figure 12a) each bipolar has RSN or RSP in series that will contribute to build up 0.7V over the resistor and over the base emitter junction of the bipolar transistors. In the anti-parallel implementation (Figure 12b), the cathode is adjacent to the anode instead of the G1 connection. This has some implications: not all of the 3 indicated bipolars will contribute equally to create the required voltage build up to turn on the SCR. As seen in Figure 12b, the current flow through bipolar 1 will not contribute to turn on the SCR (this current is flowing
direct from anode to cathode of the ESD clamp, so it will not contribute to forward the base-emitter junctions of the SCR). The TLP results of the two implementations are shown in Figure 13. The trigger current of the anti-parallel implementation has increased from 273 mA to 469 mA without any constraint on the area or value of the resistors.
Figure 13: Influence of the resistor orientation
Q3
12a2bcathode
G1
G2
RSN
RSP
anode
Current flow
Current flowRSN
RSP
Q1 Q2
anode
cathode G1
G2
Current flow
Current flow
RSN
RSP
anode
cathode G1
G2
2a2b 1
Parallel Implemented Resistors Anti-Parallel Implemented Resistors
cathode
G1
G2
RSN
RSP
anode
Q3Q1 Q2
top
vie
wC
urr
en
t flo
wQ1 Q2 Q3
Q1
Q2Q3
0
0.2
0.4
0.6
0.8
1
0 2 4 6 8 10
parallel implemented resistorsanti-parallel implemented resistors
I [A]
V [V]
Increase
of It1
3A.1-6EOS/ESD SYMPOSIUM 10-172
3) Trigger voltage engineering
The trigger voltage of the clamp is than the reference HHISCR (Figure 11the fact that the trigger element is no lobut an NPN formed between two N+ juPwell. One technique to lower this trito add a gate between the resistors Rprevent the formation of oxide in results in an improved and even fasreduction of trigger voltage).
Figure 14: Improving the NPN by adding aresistors
Figure 15:Trigger voltage engineering – TLP IV
0
0.05
0.1
0.15
0.2
4 5 6 7 8
Improved HHISCRImproved MOS-HHISCRHHISCR
I [A]
slightly higher ). This is due to
onger an NMOS unctions and the igger voltage is Rsp and Rsn to
between. This ster NPN (0.4V
a gate between the
V curves
4) Layout optimization The previous implementations adiscrete resistors. The device coufurther by splitting the resistor(Figure 16). The total width oflarge enough to handle the trimproved HHISCR to prevent dcircuit. . . ,With 1 , , The total length of trigger circuitto handle the trigger current, redesign constraint. . . ,With 1 , The third design parameter is thewhich determines the trigger curr
. .With these 3 formulas the tridesigned.
5) Overshoot investigation
The improved HHISCR as proposimilar behaviour as the standapulses. By adding a gate betwetrigger bipolar is improved (smalon-resistance). This is characteranalysis (Figure 17) by:
• An overshoot reduction
• A trigger speed enhancem
9 10V [V]
always consisted of 2 uld be optimized even s in different fingers f the device must be rigger current of the damage of the trigger
, (1) t must be long enough esulting in the second
, (2) e value of the resistor rent
(3)
igger circuit can be
osed in Figure 6 has a ard HHISCR for fast een the resistors, the ller base length, lower rized in the transient
ment
3A.1-7 EOS/ESD SYMPOSIUM 10-173
Figure 16: Optimized layout of the trigger devic
Figure 17: Transient behaviour extracted with system (2ns pulse width – 200 ps risetime)
V. Implementation
The optimal solution is the MOS-HHIparallel resistors drawn according thdescribed in previous section. Thsuccessfully implemented both itechnology (3.3 V domain) and in a 0technology (3.3 V domain). The improved HHISCR has a similcompared to the standard version but thwas decreased by 22%. This savingintegration of the trigger device in layout of the resistors. (Figure 18) The HBM level of the clamp exceedof the tester in line with the expecta
Wb
Wa
RS
N
Wa
RS
N
RS
P
An
od
e
An
od
e
G1
G2
G2
Cat
ho
de
ce
a standard vf-TLP
n
ISCR with anti-he design rules his clamp was in a 0.13um 0.35 um CMOS
lar performance he required area
g is due to the the optimized
ds the 8kV limit ations based on
SCR size. System level tests product level, under powered coany additional protection componpin passed 8kV contact and 15 kchip protection against 8kV HBMapplication since not all the currIO: a proper shield/ground imdischarges a part of the ESD curthe configuration of the boardcustomer requirements will havneeded on-chip ESD protection tsystem-level ESD protection anreal world ESD protection.
Figure 18: Top view of the ESD layout
Wa Wa Wa Wa Wa
RS
N
RS
N
RS
N
RS
N
RS
N
RS
P
RS
P
RS
P
RS
P
RS
P
An
od
e
An
od
e
An
od
e
An
od
e
An
od
e
G1
G1
G1
G1
G1
G2
G2
G2
G2
G2
Cat
ho
de
Cat
ho
de
Cat
ho
de
Cat
ho
de
Cat
ho
de
Sta
ndar
d H
HIS
CR
(IEC 61000-4-2) on onditions and without nents at the connector kV air discharge. On-M was enough for this rent flows through the
mplementation already rrent. The application, d/connectors and the ve an impact on the to achieve the desired d more important the
Impr
oved
HH
ISC
R
3A.1-8EOS/ESD SYMPOSIUM 10-174
The high current ESD level is achieved while the leakage is below 330 pA (VDD +10%) and 33 nA at elevated temperature (125°C). (Figure 19)
Figure 19: Low leakage at room and elevated temperatures
Conclusion High current levels, limited overshoot and latch-up safety are the key aspects for a successful system level implementation. This paper showed that optimized layout of the HHISCR [11] can reduce the needed area to reach system level ESD requirements by integrating the trigger device into the resistor layout.
Acknowledgements The authors would like to thank Natarajan M Iyer for his valuable suggestions to improve the structure of the paper.
References [1] M. Mergens, “Advanced SCR ESD Protection
Circuits for CMOS / SOI Nanotechnologies”, CICC 2005
[2] M. Mergens, “High Holding Current SCR (HHI-SCR) for ESD Protection and Latch-up Immune IC Operation.”, EOS/ESD 2002
[3] T. Smedes, “Relations between system level ESD and (vf-)TLP”, EOS/ESD 2006
[4] M. Ker, “Overview of On-Chip Electrostatic Discharge Protection Design With SCR-based Devices in CMOS Integrated Circuits”, IEEE Trans. Device and Materials Reliability, June 2005
[5] C. Russ , “GGSCRs: GGNMOS Triggered Silicon Controlled Rectifiers for ESD Protection in Deep Sub-Micron CMOS Processes.”, EOS/ESD 2001
[6] M. Roseeuw, “On-Chip ESD Protection Achieving 8kV HBM Without Compromising the 3.4Gbps HDMI Interface”, RCJ 2009
[7] V.A. Vashchenko, ”System Level and Hot Plug-in Protection of High Voltage Transient Pins“, EOS/ESD 2009
[8] D. Robinson-Hahn, “Is System-Level ESD Testing Valid for ICs?”, Power Electronics Rechnoloby, september 2009
[9] US 6,803,633, “Electrostatic Discharge Protection Structures Having High Holding Current For Latch-Up Immunity”
[10] US 7,064,393, “Electrostatic Discharge Protection Structures Having High Holding Current For Latch-Up Immunity”
[11] US 2010/0008002, “High Trigger Current Silicon Controlled Rectifier”
[12] IEC 61000-4-2, Edition 2.0
[13] G. Wybo. “Characterizing the Transient Device Behavior of SCRs by means of VFTLP Waveform Analysis”, EOS/ESD 2007
1.E-11
1.E-10
1.E-09
1.E-08
1.E-07
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
0 1 2 3 4 5 6 7 8 9 10
curr
ent (
A)
Voltage (V)
25C
85C
125C
3A.1-9 EOS/ESD SYMPOSIUM 10-175
THIS PAGE WAS INTENTIONALLY LEFT BLANK
3A.1-10EOS/ESD SYMPOSIUM 10-176
Sofics Proprietary – ©2011
About Sofics
Sofics (www.sofics.com) is the world leader in on‐chip ESD protection. Its patented technology is proven in more than a thousand IC designs across all major foundries and process nodes. IC companies of all sizes rely on Sofics for off‐the‐shelf or custom‐crafted solutions to protect overvoltage I/Os, other non‐standard I/Os, and high‐voltage ICs, including those that require system‐level protection on the chip. Sofics technology produces smaller I/Os than any generic ESD configuration. It also permits twice the IC performance in high‐frequency and high‐speed applications. Sofics ESD solutions and service begin where the foundry design manual ends.
Our service and support
Our business models include • Single‐use, multi‐use or royalty bearing license for ESD clamps • Services to customize ESD protection
o Enable unique requirements for Latch‐up, ESD, EOS o Layout, metallization and aspect ratio customization o Area, capacitance, leakage optimization
• Transfer of individual clamps to another target technology • Develop custom ESD clamps for foundry or proprietary process • Debugging and correcting an existing IC or IO • ESD testing and analysis
Notes
As is the case with many published ESD design solutions, the techniques and protection solutions described in this data sheet are protected by patents and patents pending and cannot be copied freely. PowerQubic, TakeCharge, and Sofics are trademarks of Sofics BVBA.
Version
May 2011
ESD SOLUTIONS AT YOUR FINGERTIPS
Sofics BVBA Groendreef 31
B‐9880 Aalter, Belgium (tel) +32‐9‐21‐68‐333 (fax) +32‐9‐37‐46‐846
[email protected] RPR 0472.687.037