In addition to lowering the required capacitor value, Miller
compensation entails a very important property: it moves the output pole
away from the origin. This effect is called pole
splitting
image-20230103223438823
The 1st stage is replaced with Thevenin equivalent circuit , Vi ≅ Vi ⋅ gm1Ro1
$$\begin{align}
\frac{V_i-V_{o1}}{R_{o1}} &= V_{o1}\cdot sC_{o1}+(V_{o1}-V_o)\cdot
sC_c \\
V_{o1} &= \frac{V_i+sR_{o1}C_cV_o}{1+sR_{o1}(C_{o1}+C_c)}
\end{align}$$$$
(V_{o1}-V_o)sC_c=g_{m2}V_{o1}+V_o(\frac{1}{R_{o2}+sC_L})
$$ substitute Vo1, we get
s3 terms in
denominator H3 = s3 ⋅ (Ro1Ro2Rc+Ro1Ro2Rsw+Ro1RcRsw) ⋅ Co1CcCLs2 terms in
denominator $$\begin{align}
H_2
&=s^2\cdot(R_{o1}R_{o2}C_{o1}C_c+R_{o1}R_{o2}C_{o1}C_L+R_{o2}R_cC_cC_L+R_{o1}R_{o2}C_cC_L+R_{o1}R_cC_{o1}C_c\\
&+R_{o2}R_{sw}C_cC_L+R_{o1}R_{sw}C_cC_L\cdot
g_{m2}R_{o2}+R_{o1}R_{sw}C_{o1}C_L+R_{sw}R_cC_cC_L+R_{o1}R_{sw}C_cC_L)
\end{align}$$
s1 term in
denominator H1 = s(Ro1⋅gm2Ro2Cc+Ro1Co1+RcCc+Ro1Cc+Ro2Cc+Ro2CL+RswCL)s0 term in
denominator H0 = 1
set Rc = 0
and Rsw = 0,
the H* reduced to
$$\begin{align}
H_3 &= 0 \\
H_2 &=s^2R_{o1}R_{o2}(C_{o1}C_c+C_{o1}C_L+C_cC_L) \\
H_1&=s(R_{o1}\cdot
g_{m2}R_{o2}C_c+R_{o1}C_{o1}+R_{o1}C_c+R_{o2}C_c+R_{o2}C_L) \\
H_0&=1
\end{align}$$ That is H = s2Ro1Ro2(Co1Cc+Co1CL+CcCL) + s(Ro1⋅gm2Ro2Cc+Ro1Co1+Ro1Cc+Ro2Cc+Ro2CL) + 1
which is same with our previous analysis of Generic circuit in
textbook
And we know $$
\frac{V_o}{V_{o2}}=\frac{1}{1+sR_{sw}C_L}
$$ Finally, we get $\frac{V_o}{V_i}$$$\begin{align}
\frac{V_o}{V_i} &= \frac{V_{o2}}{V_i} \cdot \frac{V_o}{V_{o2}} \\
&= -g_{m2}R_{o2}\frac{\left[ sC_c(R_c-1/g_{m2})+1
\right](sR_{sw}C_L+1)}{H_3+H_2+H_1+1} \cdot \frac{1}{1+sR_{sw}C_L} \\
&= -g_{m2}R_{o2}\frac{ sC_c(R_c-1/g_{m2})+1}{H_3+H_2+H_1+1}
\end{align}$$
The loop transfer function is $$
\frac{V_o}{V_i} =-g_{m1}R_{o1}g_{m2}R_{o2}\frac{
sC_c(R_c-1/g_{m2})+1}{H_3+H_2+H_1+1}
$$
The poles can be deduced $$\begin{align}
\omega_1 &= \frac{1}{R_{o1}\cdot g_{m2}R_{o2}C_c} \\
\omega_2 &= \frac{1}{1+g_{m2}R_{sw}}\cdot \frac{g_{m2}}{C_L} \\
&= \frac{1}{(gm_2^{-1}+R_{sw})C_L}
\end{align}$$
The pole $\omega_2=\frac{1}{gm_2^{-1}C_L}$ is changed
to $\omega_2=\frac{1}{(gm_2^{-1}+R_{sw})C_L}$
In order to cancell ω2 with ωz, Rc shall be
increased
Following demonstrate how derive fnd
from Razavi’s equation. We copy ω2 here $$
\omega_2 = \frac{R_{o1}C_c\cdot
g_{m2}R_{o2}+R_{o2}(C_c+C_L)+R_{o1}(C_{o1}+C_c)}{R_{o1}R_{o2}(C_cC_{o1}+C_LC_{o1}+C_LC_c)}
$$ which can be reduced as below
Any dc current flowing through a diode generates the
so-called “shot noise” due to the random nature of the hole and
electron transitions across the pn junction
Shot noise is not relevant in CMOS devices since it
is only present in bipolar transistors and junction diodes
TIA stage allows for improved gain with better
linearity, as mostly signal current passes through RFTODO
📅 ??? Quantitative analysis
Switched-Capacitor Resistor
$$
R_{eq} = \frac{1}{f_sC}
$$
image-20240905202145206
Channel-Length Modulation
& Pinched off
$\lambda \propto
\frac{1}{L_g}$
$\lambda \propto
\frac{1}{V_{DS}}$
image-20241116080122184
If VDS is
slightly greater than VGS − VTH,
then the inversion layer stops at x ≤ L, and we say the
channel is “pinched off”
Upon passing the pinchoff point, the electrons simply shoot through
the depletion region near the drain junction and arrive at the drain
terminal
L′ is the
function of VDS
with $\frac{1}{L^{'}} = \frac{1}{L-\Delta
L}=\frac{L+\Delta L}{L^2-\Delta L^2}\approx
\frac{1}{L}\left(1+\frac{\Delta L}{L}\right)$, we have $$
I_D \approx \frac{1}{2}\mu_n C_{ox}\frac{W}{L}\left(1+\frac{\Delta
L}{L}\right)(V_{GS}-V_{TH})^2 = \frac{1}{2}\mu_n
C_{ox}\frac{W}{L}(V_{GS}-V_{TH})^2 (1+\lambda V_{DS})
$$ assuming $\frac{\Delta L}{L} =
\lambda V_{DS}$
λ represents the
relative variation in length for a given increment in VDS.
Thus, for longer channels, λ
is smaller
In reality, however, rO varies with
VDS. As
VDSincreases and the pinch-off point moves toward the source, the
rate at which the depletion region around the source becomes
wider decreases, resulting in a higher incremental
output impedance.
The dependence of offset voltage and current mismatches upon the
overdrive voltage is similar to our observations for corresponding
noise quantities
differential pair
image-20240624222306837
In reality, since mismatches are independent statistical
variables
image-20240624222417564
Above shows that the input transistors must be designed for high
gain ($g_mr_o =
\frac{2}{V_{OV}\lambda}$), which means they must be designed for
smallVGS − VTH.
It is desirable to minimize VGS − VTH
by lowering the tail current or increasing the transistor
widths
Ceq and Req are obtained $$\begin{align}
C_\text{eq} &= \frac{1+|A|^2-2A_r}{1-A_r}\cdot C_f \\
R_\text{eq} &= \frac{A_i}{1+|A|^2-2A_r}\cdot \frac{1}{\omega C_f}
\end{align}$$
D/S small signal model
image-20240106161059584
The Drain and Source of MOS are determined
in DC operating point, i.e. large signal.
That is, top of M2 is drain
and bottom is source, $$\begin{align}
R_\text{eq2} &= \frac{r_\text{o2}+R_L}{1+g_\text{m2}r_\text{o2}} \\
& \simeq \frac{1}{g_\text{m2}}
\end{align}$$
PMOS small signal model
polarity
The small-signal models of NMOS and PMOS transistors are
identical
A negative ΔVGS leads to a
negative ΔID.
Recall that ID, in the
direction shown here, is negative because the actual current of holes
flows from the source to the drain.
image-20240106170315177
Conversely, a positive ΔVGS produces a
positive ΔID, as
is the case for an NMOS device.
image-20240106164923917
Leakage in MOS
image-20241109195527005
Subthreshold leakage
Drain-Induced Barrier Lowering (DIBL)
Reverse-bias Source/Drain junction leakages
Gate leakage
two other leakage mechanisms
Gate Induced Drain Leakage (GIDL)
Punchthrough
image-20241110001311117
W. M. Elgharbawy and M. A. Bayoumi, “Leakage sources and possible
solutions in nanometer CMOS technologies,” in IEEE Circuits and Systems
Magazine, vol. 5, no. 4, pp. 6-17, Fourth Quarter 2005, doi:
10.1109/MCAS.2005.1550165.
X. Qi et al., “Efficient subthreshold leakage current optimization -
Leakage current optimization and layout migration for 90- and 65- nm
ASIC libraries,” in IEEE Circuits and Devices Magazine, vol. 22, no. 5,
pp. 39-47, Sept.-Oct. 2006, doi: 10.1109/MCD.2006.272999.
P. Monsurró, S. Pennisi, G. Scotti and A. Trifiletti, “Exploiting the
Body of MOS Devices for High Performance Analog Design,” in IEEE
Circuits and Systems Magazine, vol. 11, no. 4, pp. 8-23, Fourthquarter
2011, doi: 10.1109/MCAS.2011.942751.
Andrea Baschirotto, ISSCC2015 “ADC Design in Scaled Technologies”
As a result of DIBL, threshold voltage is reduced
with shorter channel lengths and, consequently, the subthreshold leakage
current is increased
image-20240901231532412
impact on output impedance
The principal impact of DIBL on circuit design is the degraded output
impedance.
In short-channel devices, as VDS
increases further, drain-induced barrier lowering becomes significant,
reducing the threshold voltage and increasing the drain
current
image-20240901232709711
Impact Ionization and GIDL are different, however both
increase drain current, which flowing from the drain into the
substrate
image-20241120210915254
Gate induced drain leakage
(GIDL)
image-20241110001118250Figure 4.3
The large current flows from the drain to bulk and this
drain leakage current is named gate-induced drain leakage
(GIDL) since it is due to a gate-induced high electric
field present in the gate-to-drain overlap region
gate-induced drain leakage (GIDL) increases exponentially due to the
reduced gate oxide thickness
image-20240902000820459
Chauhan, Yogesh Singh, et al. FinFET modeling for IC simulation
and design: using the BSIM-CMG standard. Academic Press, 2015.
image-20240901225754731
$$
\frac{g_m}{I_D} = \frac{2}{V_{GS}-V_{TH}}
$$ Decrease of gm/Id results from decrease in VT.
GIDL (Gate induced drain leakage) as at weak
inversion may results in a weak lateral electric field causing leakage
current between drain and bulk, which degrade the efficiency of the
transistor (gm/ID).
REF: D. A. Yokoyama-Martin et al., “A Multi-Standard Low Power
1.5-3.125 Gb/s Serial Transceiver in 90nm CMOS,” IEEE Custom Integrated
Circuits Conference 2006, 2006, pp. 401-404, doi:
10.1109/CICC.2006.320970.
Power/Ground and I/O Pins
Power / Ground Pin
Information
In both digital and analog I/O, power and ground pins appear at the
sub-circuit definiton, allowing user to use the I/O in voltage islands.
They follow certain naming conventions.
digital I/O sub-circuit
VDD: pre-driver core voltage (supplied by PVDD1CDGM)
VSS: pre-driver ground and also global ground (supplied by
PVDD1CDGM)
VDDPST: I/O post-driver voltage, i.e. 1.8V (supplied by PVDD2CDGM or
PVDD2POCM)
VSSPOST: I/O post-driver ground (supplied by PVDD2CDGM or
PVDD2POCM)
POCCTRL: POCCTRL signal (supplied by PVDD2POCM)
analog I/O placed in a core voltage domain, the convention is
TACVDD: analog core voltage (supplied by PVDD3ACM)
TACVSS: analog core ground (supplied by PVDD3ACM)
VSS: global core ground
analog I/O placed in an I/O voltage domain, the convention is:
TAVDD: analog I/O voltage, i.e. 1.8V (supplied by PVDD3AM)
TAVSS: analog I/O ground (supplied by PVDD3AM)
VSS: global core ground
Power/Ground Combo Cells
power/ground combo pad cell
pins to be connected to bump
to core side pin name
PVDD1CDGM
VDD VSS
VDD VSS
PVDD2CDGM PVDD2POCM
VDDPST VSSPST
N/A
PVDD3AM
TAVDD TAVSS
AVDD AVSS
PVDD3ACM
TACVDD TACVSS
AVDD AVSS
Note for the retention mode
At initial state, IRTE must be 0 when VDD is
off.
IRTE must be kept >= 10us after VDD turns on again (from the
retention mode to the normal operation mode).
IRTE can be switched only when both VDD and VDDPST are on.
rention_seq.drawio
When the rention function is needed, IRTE signal must come from an
“always-on” core power domain. If you don’t need the rention function,
it is required to tie IRTE to ground. In other words, no matter
the rention feature is needed or not, it is required to have PCBRTE in
each domain.
PCBRTE_in_digital_domain.drawio
Note: PCBRTE does not need PAD
connection.
Internal Pins
There are 3 internal global pins, i.e. ESD,
POCCTRL, RTE, in all digital domain
cells.
In real application,
ESD pin is an internal signal and
active in ESD event happening
POCCTRL is an internal signal and active in
Power-on-control event.
However, these special events (i.e. ESD event and Power-on-control
event) are not modeled in NLDM kit (.lib), only normal function is
covered, so ESD and POCCTRL pins are
simply defined as ground in NLDM kit (.lib).
These 3 global pins will be connected automatically after
cell-to-cell abutting in physical layout.
Power-Up sequence in
Digital Domain
Power up the I/O power (VDDPST) first, then the core power (VDD)
pocctrl_seq.drawio
PVDDD2POCM cell would generate Power-On-Control signal (POCCTRL) to
have the post-driver NMOS and PMOS off, so that the crowbar current
would not occur in the post-driver fingers when the I/O voltage is on
while the core voltage remains off. As such, I/O cell would be in the
Hi-Z state. when POCCTRL is on, the pll-up/down resistor is disabled and
C is 0.
The POCCTRL signal is transmitted to I/O cells through cell
abutment. There is no need to have routing for POCCTTRL nor
give a control signal to the POCCTRL pin any of I/O cells. Note that the
POCCTRL signal would be cut if inserting a power-cut (PRCUT) cell.
power-on-control-ciruit.drawio
Power-Down sequence in
Digital Domain
It’s the reverse of power-up sequence.
Use model in Innovus
1 2 3 4 5 6 7 8 9 10 11 12
set init_gnd_net "vss_core vss DUMMY_ESD DUMMY_POCCTRL"
When a flip-flop samples an input that is changing during its
aperture, the output Q may momentarily take on a voltage between 0 and
VDD that is in the forbidden zone. This is called a metastable state.
Eventually, the flip-flop will resolve the output to a stable state of
either 0 or 1. However, the resolution time required to reach the stable
state is unbounded
M. Tian, V. Visvanathan, J. Hantgan and K. Kundert, “Striving for
small-signal stability,” in IEEE Circuits and Devices Magazine, vol. 17,
no. 1, pp. 31-41, Jan. 2001, doi: 10.1109/101.900125.
Ali Sheikholeslami, Circuit Intuitions: Thevenin and Norton
Equivalent Circuits, Part 3 IEEE Solid-State Circuits Magazine, Vol. 10,
Issue 4, pp. 7-8, Fall 2018.
—, Circuit Intuitions: Thevenin and Norton Equivalent Circuits, Part
2 IEEE Solid-State Circuits Magazine, Vol. 10, Issue 3, pp. 7-8, Summer
2018.
—, Circuit Intuitions: Thevenin and Norton Equivalent Circuits, Part
1 IEEE Solid-State Circuits Magazine, Vol. 10, Issue 2, pp. 7-8, Spring
2018.
—, Circuit Intuitions: Miller’s Approximation IEEE Solid-State
Circuits Magazine, Vol. 7, Issue 4, pp. 7-8, Fall 2015.
Mismatch between the pole and zero frequencies leads to the
“doublet problem”. If the pole and the zero do not
exactly coincide, we say that they constitute a
doublet
Problem 10.19 in Razavi 2nd book
Suppose the open-loop transfer function of a two-stage op amp is
expressed as $$
H_{open}(s)=\frac{A_0(1+\frac{s}{\omega_z})}{\left( 1+
\frac{s}{\omega_{p1}}\right)\left( 1+ \frac{s}{\omega_{p2}}\right)}
$$ Ideally, ωz = ω2
and the feedback circuit exhibits a first-order behavior, i.e., its step
response contains a single time constant and no overshoot.
Then the transfer function of the amplifier in a unity-gain
feedback loop is given by $$\begin{align}
H_{closed}(s)
&=\frac{A_0\left(1+\frac{s}{\omega_z}\right)}{\frac{s^2}{\omega_{p1}\omega_{p2}}+\left(
\frac{1}{\omega_{p1}} +
\frac{1}{\omega_{p2}}+\frac{A_0}{\omega_{z}}\right)s+A_0+1} \\
&=\frac{\frac{A_0}{A_0+1}(1+\frac{s}{\omega_z})}{\frac{s^2}{\omega_{p1}\omega_{p2}(A_0+1)}+\left(
\frac{1}{\omega_{p1}} +
\frac{1}{\omega_{p2}}+\frac{A_0}{\omega_{z}}\right)\frac{s}{A_0+1}+1}
\end{align}$$
The denominator part of Hclosed(s)
is $$
D(s) = \frac{s^2}{\omega_{p1}\omega_{p2}}+\left( \frac{1}{\omega_{p1}} +
\frac{1}{\omega_{p2}}+\frac{A_0}{\omega_{z}}\right)s+A_0+1
$$
Assuming two poles (ωpA ≪ ωpB)
of Hclosed(s)
are widely spaced, $$\begin{align}
D(s) &= \left( 1+ \frac{s}{\omega_{pA}}\right)\left( 1+
\frac{s}{\omega_{pB}}\right)\\
&\cong \frac{s^2}{\omega_{pA}\omega_{pB}}+\frac{s}{\omega_{pA}} + 1
\end{align}$$
Thus, the two poles of the closed-loop transfer function of system
are $$\begin{align}
\omega_{pA} &= \frac{A_0+1}{\frac{1}{\omega_{p1}} +
\frac{1}{\omega_{p2}}+\frac{A_0}{\omega_{z}}} \\
&= \frac{(A_0+1)\omega_{p1} \omega_{p2}}{\omega_{p1} + \omega_{p2}
+ \frac{A_0}{\omega_z}\omega_{p1} \omega_{p2}} \\
\omega_{pB} &= \omega_{p1} + \omega_{p2} +
\frac{A_0}{\omega_z}\omega_{p1} \omega_{p2}
\end{align}$$
Assumingωz ≃ ωp2
and ωp2 ≪ (1+A0)ωp1ωpA = ωp2
and ωpB = (1+A0)ωp1
The closed-loop transfer function is $$
H_{closed}(s) =
\frac{\frac{A_0}{A_0+1}\left(1+\frac{s}{\omega_z}\right)}{\left(1+\frac{s}{(1+A_0)\omega_{p1}}\right)\left(
1+\frac{s}{\omega_{p2}} \right)}
$$
The step response of the closed-loop amplifier
Consider the Laplace transform function of step response, $X(s)=\frac{1}{s}$$$
Y(s)=\frac{1}{s}\times H_{closed}(s)
$$ Thus, the small-signal step response of the
closed-loop amplifier is $$
y(t)=\frac{A_0}{A_0+1}\left[1-e^{-(A_0+1)\omega_{p1}t}-\left(1-\frac{\omega_{p2}}{\omega_z}\right)e^{-\omega_{p2}t}
\right]u(t)
$$ Since, ωp2 ≪ (1+A0)ωp1.
Therefore, rewrite the y(t)$$
y(t)\cong
\frac{A_0}{A_0+1}\left[1-\left(1-\frac{\omega_{p2}}{\omega_z}\right)e^{-\omega_{p2}t}
\right]u(t)
$$ The step response contains an exponential term of the form
$\left(1-\frac{\omega_{p2}}{\omega_z}\right)e^{-\omega_{p2}t}$.
This is an important result, indicating that if the zero does not
exactly cancel the pole, the step response exhibits an exponential with
an amplitude proportional to $\left(1-\frac{\omega_{p2}}{\omega_z}\right)$,
which depends on the mismatch between ωz and ωp2 and a time
constant τ of $\frac{1}{\omega_{p2}}$ or $\frac{1}{\omega_{z}}$
B. Y. T. Kamath, R. G. Meyer and P. R. Gray, “Relationship between
frequency response and settling time of operational amplifiers,” in IEEE
Journal of Solid-State Circuits, vol. 9, no. 6, pp. 347-352, Dec. 1974,
[https://sci-hub.se/10.1109/JSSC.1974.1050527]
P. R. Gray and R. G. Meyer, “MOS operational amplifier design-a
tutorial overview,” in IEEE Journal of Solid-State Circuits, vol. 17,
no. 6, pp. 969-982, Dec. 1982, [https://sci-hub.se/10.1109/JSSC.1982.1051851]
—. 2024. Analysis and Design of Analog Integrated Circuits, 6th
Edition. Wiley Publishing
The most accurate method to calculate the degradation of transistors
is the SPICE-level simulation of the whole netlist with application
programming interface (API) and industry-standard stress process
models
MOSRA: MOSFET reliability analysis Synopsys
RelXpert: Cadence
TMI: TSMC Model Interface, TSMC
OMI: Open Model Interface, Si2 standard,
The Silicon Integration Initiative (Si2) Compact Model Coalition has
released the Open Model Interface, an Si2 standard, C-language
application programming interface that supports SPICE compact model
extensions.OMI allows circuit designers to simulate and analyze such
important physical effects as self-heating and aging,
and perform extended design optimizations. It is based on TMI2, the TSMC
Model Interface, which was donated to Si2 by TSMC in 2014.
BTI occurs predominantly in PMOS (or p-type or p channel)
transistors and causes an increase in the transistor’s absolute
threshold voltage.
Stress in the case of NBTI means that the PMOS transistor is
in inversion; that means that its gate to
body potential is substantially below 0 V for analogue circuits
or at VGB = −VDD for digital circuits
Higher voltages and higher temperatures both have
an exponential impact onto the degradation, induced by NBTI.
NBTI will be accelaerated with thinner gate oxide, at a high
temperature and at a high electric field across the oxide region.
During recovery phase where the gate voltage of pMOS is high and
stress is removed, the H atoms in the gate oxiede diffuse back to
Si-SiO2 interface and the recombination of Si-H bonds reduces the
threshold voltage of pMOS.
image-20230513111525657image-20230513111657285
The net result is an increase in the magnitude of the device
threshold voltage |Vt|, and a degradation of the
channel carrier mobility.
Caution: The aging model provided by fab may
NOT contain recovry effect
Short-channel MOSFETs may exprience high lateral electric
fields if the drain-source voltage is large. while the average
velocity of carriers saturate at high fields, the instantaneous velocity
and hence the kinetic energy of the carriers continue to increase,
especially as they accelerate toward the drain. These are called
hot carriers.
In nanometer technologies, hot carrier effects have
subsided. This is because the energy required to create
an electron-hole pair, Eg ≃ 1.12eV,
is simply not available if the supply voltage is around 1V.
FE = E ⋅ q
$$\begin{align}
E_k &= F_E \cdot s \\
&= E \cdot q \cdot s
\end{align}$$
Electrons and holes gaining high kinetic energies in
the electric field (hot carriers) may be injected into
the gate oxide and cause permanent changes in the
oxide-interface charge distribution, degrading the current-voltage
characteristics of the MOSFET.
The channel hot-electron (CHE) effect is caused by electons flowing
in the channel region, from the source to the drain. This effect is more
pronounced at large drain-to-source voltage, at which the lateral
electric field in the drain end of the channel accelerates the
electrons.
Four different hot carrier injectoin mechanisms can be distinguished:
- channel hot electron (CHE) injection - drain avalanche hot carrier
(DAHC) injection - secondary generated hot electron (SGHE) injection -
substrate hot electron (SHE) injection
HCI is more of a drain-localized mechanism, and is
primarily a carrier mobility degradation (and a Vt
degradation if the device is operated bi-directionally).
image-20230512213236023
For smaller transistor dimensions, CHE dominates the hot
carrier degradation effect
The hot-carrier induced damage in nMOS transistors has been found to
result in either trapping of carriers on defect sites in the oxide or
the creation of interface states at the silicon-oxide interface, or
both.
The damage caused by hot-carrier injection affects the transistor
characteristics by causing a degradation in transconductance, a shift in
the threshold voltage, and a general decrease in the drain current
capability.
HCI seems to have just a weak temperature
dependency. Unlike BTI, it seems to be no or just little recovery. As
holes are much “cooler” (i.e. heavier) than electrons, the channel hot
carrier effect in nMOS devices is shown to be more significant than in
pMOS devices.
image-20231106224938502
Degradation saturation
effect
HCI model can reproduce the saturation effect if stress time is long
enough
image-20230513112108262
TDDB
TDDB effect is also related to oxide traps. In general, TDDB refers
to the loss of isolating properties of a dielectric layer. If this
dielectric layer is the gate oxide, TDDB will initially lead to an
increase in the gate tunnelling current.
This soft breakdown can already lead to a parametric degradation.
After a long accumulation period, TDDB leads to a catastrophic reduction
of the channel to gate insulation and thus a functional failure of the
transistor.
image-20230513105908505
Scaling drive more concerns in TDDB
imgimg
waveform-dependent nature
The figure below illustrates the waveform-dependent nature of these
mechanisms – as described earlier, BTI and HCI depend upon the region of
active device operation. The slew rate of the circuit inputs and output
will have a significant impact upon these mechanisms, especially
HCI.
Negative bias temperature instability (NBTI). This
is caused by constant electric fields degrading the dielectric,
which in turn causes the threshold voltage of the transistor to degrade.
That leads to lower switching speeds. This effect depends on the
activity level of the circuits, with heavier impact on parts of the
design that don’t switch as often, such as gated clocks,
control logic, and reset, programming and test circuitry.
Hot carrier injection (HCI). This is caused by
fast-moving electrons inserting themselves into the gate and
degrading performance. It primarily occurs on higher-voltage modes and
fast switching signals.
image-20230513110202915
longer channel length help both BTI and HCI
largerVds
help BTI, but hurt HCI
lower temperature help BTI of core device, but hurt that of
IO device for 7nm FinFET
MOSRA
MOSRA is a 2-step simulation: 1) Age computation, 2) Post-age
analysis
A. Zhang et al., “Reliability variability simulation methodology for
IC design: An EDA perspective,” 2015 IEEE International Electron Devices
Meeting (IEDM), Washington, DC, USA, 2015, pp. 11.5.1-11.5.4, doi:
10.1109/IEDM.2015.7409677.
W. -K. Lee et al., “Unifying self-heating and aging simulations with
TMI2,” 2014 International Conference on Simulation of Semiconductor
Processes and Devices (SISPAD), Yokohama, Japan, 2014, pp. 333-336, doi:
10.1109/SISPAD.2014.6931631.
Article (20482350) Title: Measure the Impact of Aging in Spectre
Technology URL:
https://support.cadence.com/apex/ArticleAttachmentPortal?id=a1O0V000009ESBFUA4
Karimi, Naghmeh, Thorben Moos and Amir Moradi. “Exploring the Effect
of Device Aging on Static Power Analysis Attacks.” IACR Trans. Cryptogr.
Hardw. Embed. Syst. 2019 (2019): 233-256.[link]
Y. Zhao and Y. Qu, “Impact of Self-Heating Effect on Transistor
Characterization and Reliability Issues in Sub-10 nm Technology Nodes,”
in IEEE Journal of the Electron Devices Society, vol. 7, pp. 829-836,
2019, doi: 10.1109/JEDS.2019.2911085.
When set to aocv_multiplicative, the derating factor
will be calculated as AOCV derating * OCV derating, which is set using
the set_timing_derate command.
When set to aocv_additive, the derating factor will be
calculated as AOCV derating + OCV derating values.
When you use this global variable, the report_timing
command shows the total_derate column in the timing report
output, which allows you to view and cross-check the calculated total
derate factor.
To set this global variable, use the set_global
command.
image-20221210143256639
reference
Genus Attribute Reference 22.1
Innovus Text Command Reference 22.10
Article (20416394) Title: Analysis with Advanced On-chip Variation
(AOCV) derating in EDI system and ETS URL:
https://support.cadence.com/apex/ArticleAttachmentPortal?id=a1Od000000050NxEAI
The root cause of the delay mismatch is related to how parasitic
extraction tools distribute coupling capacitances over the nodes of the
resistive networks
The most likely reason for such asymmetry is the anisotropy of
computational geometry algorithms used by extraction tools.
figure-4
PERC
CD: current density checks
P2P: point to point resistance checks
LDL: logic driven layout checks, latch up related
TOPO: topology, circuit connection and device size
checks
database
CD, P2P, LDL : dfmdb
TOPO: svdb
Frank Feng. New Approach For Full Chip Electrical Reliability
Verification [pdf]
The antenna effect is a common name for the effects
of charge accumulation in isolated nodes of an
integrated circuit during its processing.
This effect is also sometimes called “Plasma Induced Damage”,
“Process Induced Damage” (PID) or “charging effect”.
antenna ratio
The antenna rule specifies the maximum tolerance for the ratio of a
metal line area to the area of connected gates.
metal jumping
Long metal can be taken to higher metal routing layer. This
is known as metal jumping.
This metal jumping will break the long interconnect and hence the
charge collected on the long interconnect will not discharge through
gate oxide because the higher metal layer is not yet fabricated.
so, if the gate immediately connects to the highest level by jump-up
metals, large amount of charges can not be collected, while the poly
finally connected to the diffusion part by highest level, thus no
antenna violation will normally occure.
Diode Insertion
Diode helps dissipate charges accumulated on metal. Diode should be
placed as near as possible to the gate of device on low level of
metal.
Diode should always be connected in reverse bias, with
cathode connected to gate electrode and anode connected to ground
potential.
During processing, even if the diodes are reversely biased, because
of the elevated wafer temperature (200oC plus) it
will provide a much conductive path
In the reverse bias region, the reverse saturation current of Si and
Ge diodes doubles for every 10oC rise in
temperature
main-qimg-c3fe57dfac5fd5e5b5616ddf4f89f08a-pjlq
Tuvia Liran, Antenna effect (PID): Do the design rules really protect
us? [link]
Place well tie and substrate tie where they are needed. Redundant
guard ring consume area and increase the routing of critical signal
net.
guardring_stypes.drawio
Continuous OD
Performance & Matching
image-20220219223723289
current mirror
split diffusion with dummy transistors
mirror_continuous_OD_split_with_dummy.drawio
cascode structure
off transistor split diffusion
cascode_continuous_OD_split_with_dummy.drawio
sharing source & drain
sharing_SD.drawio
Stacked MOSFETs
Matching
Common Centroid
The common centroid technique describes that if there are n blocks
which are to be matched then the blocks are arranged symmetrically
around the common centre at equal distances from the centre. This
technique offers best matching for devices as it helps in avoiding
cross-chip gradients
HTML5 Icon
Inter-digitation
Interdigitation reduces the device mismatch as it suffers equally
from process variations in X dimension. This technique was used to
layout current mirrors and resistors in PTAT and BGR circuits. In the
Figure-15 below each brown stick represents a PFET of uniform length.
This representation is termed as an inter-digitated layout.
As shown in Fig. 35 in older planar technology nodes, gate pitch is
so relaxed such that S/D contacts and gate contacts can easily be placed
next to each other without causing any shorting risk (see Fig.
35(a)).
As the gate pitch scales, there’s no room to put gate
contacts next to S/D contacts, and gatecontacts have been pushed away
from the active region and are only placed on the STI
region.
image-20230708221916716
In addition, at tight gate pitch, even forming S/D contact
without shorting to gate metal becomes very challenging.
The idea of self-aligned contacts (SAC) has been
introduced to mitigate the issue of S/D contact to gate shorts.
As shown in Fig. 35(b), the gate metal is fully encapsulated by a
dielectric spacer and gate cap, which protects the gate from
shorting to the S/D contact.
image-20230708230238362
A dielectric cap is added on top of the gate so that if the contact
overlaps the gate, no short occurs.
MD layer represent SACs in PDK
image-20230709005334372
self-aligned gate contacts
(SAGCs)
Self-aligned gate contacts (SAGCs) have also been
implemented and Denser standard cells can be achieved by eliminating the
need to land contacts on the gate outside the active area.
SAGCs require the source/drain contacts to be capped with an
insulator that is different from both contact and gate cap dielectrics
to protect the source/drain contacts against a misaligned gate contact
etch.
image-20230708233009568image-20230708232429240
According to the DRC of T foundary, poly extension > 0 um and
space between MP and OD > 0 um., which demonstrate self-aligned gate
contact is not introduced.
Contacted-Poly-Pitch (CPP)
Wider Contacted-Poly-Pitch allows wider MD and VD size, which help
reduce MEOL IRdrop
Schematic representation of a logic
standard cell layout (CPP = contacted poly pitch, FP = fin pitch, MP =
metal pitch; cell height = number of metal lines per cell x
MP).
Naoto Horiguchi. Entering the Nanosheet Transistor Era [link]
A native layer (NT_N) is usually added under
inductors or transformers in the nanoscale CMOS to define the non-doped
high-resistance region of substrate, which decreases eddy currents in
the substrate thus maintaining high Q of the coils.
For T* PDK offered inductor, a native substrate region is created
under the inductor coil to minimize eddy currents
image-20230810000702597
OD inside NT_N only can be used for NT_N potential pickup purpose,
such as the guarding-ring of MOM and inductor
Derived Geometries
Term
Definition
PW
{NOT NW}
N+OD
{NP AND OD}
P+OD
{PP AND OD}
GATE
{PO AND OD}
TrGATE
{GATE NOT PODE_GATE}
NP: N+ Source/Drain Ion Implantation
PP: P+ Source/Drain Ion Implantation
OD: Gate Oxide and Diffustion
NW: N-WELL
PW: P-WELL
CMOS Processing Technology
Four main CMOS technologies:
n-well process
p-well process
twin-tub process
silicon on insulator
Triple well, Deep N-Well (optional):
NWell: NMOS svt, lvt, ulvt …
PWell: PMOS svt, lvt, ulvt …
DNW: For isolating P-Well from the substrate
The NT_N drawn layer adds no process cost and
no extra mask
The N-well / P-well technology, where n-type diffusion is done over a
p-type substrate or p-type diffusion is done over n-type substrate
respectively.
The Twin well technology, where NMOS and
PMOS transistor are developed over the wafer by simultaneous
diffusion over an epitaxial growth base, rather than a substrate.
Deep N-well
Chew, K.W., Zhang, J., Shao, K., Loh, W., & Chu, S.F. (2002).
Impact of Deep N-well Implantation on Substrate Noise Coupling and RF
Transistor Performance for Systems-on-a-Chip Integration. 32nd European
Solid-State Device Research Conference, 251-254. URL:[slides,
paper]
Kuo-Tsai LiPaul ChangAndy Chang, TSMC, US20120053923A1, “Methods of
designing integrated circuits and systems thereof”
Substrate noise
A variety of techniques can be used to minimize this noise, for
example by keeping analog devices surrounded by guard rings, or using a
separate supply for the substrate/well taps.
However guard rings alone cannot prevent noise coupling deep in
the substrate, only surface currents.
PMOS are less noisy than NMOS since PMOS has its nwell which isolates
the substrate noise, but such is not valid for NMOS .
DNW
The N-channel devices built directly into the P-type substrate are
not as effectively isolated as P-channel devices in their N-wells. This
is because despite creating a P+ guard ring around the devices, there
remains an electrical path below the guard ring for charge to flow.
To overcome this issue, a deep N-well can be used to more
effectively isolate these N-channel devices.
image-20230529001556060image-20230529010836003BM_SS_Together at Last_Fig1
the P-well is separated, allowing the voltage to be controlled
because the circuit within the deep N-well is separated from the
p-substrate in this structure, there is the benefit that this circuitry
is less susceptible to noise that propagates through the
p-substrate.
reference
Mikael Sahrling, Layout Techniques for Integrated Circuit Designers
1st Edition , Artech House 2022
JED Hurwitz, ISSCC2011 “T4: Layout: The other half of Nanometer CMOS
Analog Design” [slides,
transcript]
Tom Quan, TSMC, Bob Lefferts, Fred Sendig, Synopsys, Custom Design
with FinFETs - Best practices designing mixed-signal IP
Jacob, Ajey & Xie, Ruilong & Sung, Min & Liebmann, Lars
& Lee, Rinus & Taylor, Bill. (2017). Scaling Challenges for
Advanced CMOS Devices. International Journal of High Speed Electronics
and Systems. 26. 1740001. 10.1142/S0129156417400018.
Joddy Wang, Synopsys “FinFET
SPICE Modeling” Modeling of Systems and Parameter Extraction Working
Group 8th International MOS-AK Workshop (co-located with the IEDM
Conference and CMC Meeting) Washington DC, December 9 2015
A. L. S. Loke et al., “Analog/mixed-signal design challenges in 7-nm
CMOS and beyond,” 2018 IEEE Custom Integrated Circuits Conference
(CICC), San Diego, CA, USA, 2018, pp. 1-8, doi:
10.1109/CICC.2018.8357060.[slides]
Prof. Adam Teman, Advanced Process Technologies, [pdf]
Luke Collins. FinFET variability issues challenge advantages of new
process [link]
Loke, Alvin. (2020). FinFET technology considerations for circuit
design (invited short course). BCICTS 2020 Monterey, CA
Alvin Leng Sun Loke, TSMC. Device and Physical Design Considerations
for Circuits in FinFET Technology”, ISSCC 2020
A. L. S. Loke, C. K. Lee and B. M. Leary, “Nanoscale CMOS
Implications on Analog/Mixed-Signal Design,” 2019 IEEE Custom Integrated
Circuits Conference (CICC), Austin, TX, USA, 2019, pp. 1-57, doi:
10.1109/CICC.2019.8780267.
A. L. S. Loke, Migrating Analog/Mixed-Signal Designs to FinFET Alvin
Loke / Qualcomm. 2016 Symposia on VLSI Technology and Circuits
Lattice Semiconductor, 16FFC Process Technology Introduction December
9th, 2021[pdf]
The parameter that shows the dependence of the reference voltage on
temperature variation is called the temperature coefficient and is
defined as: $$
TC_F=\frac{1}{V_{\text{REF}}}\left[
\frac{V_{\text{max}}-V_{\text{min}}}{T_{\text{max}}-T_{\text{min}}}
\right]\times10^6\;ppm/^oC
$$
$$\begin{align}
V_{os}[n] &= V_{os}[n-1] - \frac{\Delta I_1}{g_m} \\
V_{os}[n] &= I_\Delta[n] R_E \\
\beta I_\Delta &= I_1[n] + I_2[n-1]
\end{align}$$ where IΔ is the
variation of Ie1 + Ie2
due to Vos and
$R_E = \frac{R_1R_2}{R_1+2R_2}$
Using Red Hat Enterprise Linux 8, Rocky
Linux 8 and the GNOME 3 window manager, the
new Virtuoso Schematic/Layout/ADE windows and forms sometimes pop up
under or below the Library Manager or on the desktop in
the background instead of the foreground and cannot be seen. Sometimes,
they are iconized; they do not come on the top in front, though it is
the most recent window opened.
oxide capacitance (aka gate-channel
capacitance) between the gate and the channelC1 = WLCox
divided between CGS and
CGD
depletion capacitance between the channel
and the substrateC2
overlap capacitance: direct overlap and fringing
field
junction capacitance between the
source/drain areas and the substrate
The value of CSB and
CDB is
a function of the source and drain voltages with respect to the
substrate
image-20240727134110758image-20240727134150216
The gate-bulk capacitance is usually neglected in
the triode and saturation regions because the inversion layer acts as a
“shield” between the gate and the bulk.
classification with Intrinsic and
Extrinsic MOS capacitor
Base Band MOSCAP model (nmoscap) is built without effective series
resistance (ESR) and effective series inductance (ESL) calibrations,
which is for capacitance simulation only
LC-Tank MOSCAP model (moscap_rf) is for frequency-dependent Q factor
and capacitance simulations
capacitance of MOS gate varies nonmonotonically
with VGS
“accumulation-mode” varactor varies
monotonically with VGS
Inverter capacitance
invCap
reference
R. L. Bunch and S. Raman, “Large-signal analysis of MOS varactors in
CMOS -G/sub m/ LC VCOs,” in IEEE Journal of Solid-State Circuits,
vol. 38, no. 8, pp. 1325-1332, Aug. 2003, doi:
10.1109/JSSC.2003.814416.
T. Soorapanth, C. P. Yue, D. K. Shaeffer, T. I. Lee and S. S. Wong,
“Analysis and optimization of accumulation-mode varactor for RF ICs,”
1998 Symposium on VLSI Circuits. Digest of Technical Papers (Cat.
No.98CH36215), 1998, pp. 32-33, doi: 10.1109/VLSIC.1998.687993. URL: http://www-smirc.stanford.edu/papers/VLSI98s-chet.pdf
R. Jacob Baker, 6.1 MOSFET Capacitance Overview/Review, CMOS Circuit
Design, Layout, and Simulation, Fourth Edition
B. Razavi, Design of Analog CMOS Integrated Circuits 2nd