426
Compact Models for Integrated Circuit Design
of the shortest device. Note that in most cases, L min represents the nominal
device of the target technology node.
12.2.1.2 Selection of Device Characteristics
As discussed in Chapter 4 and 5 for MOSFETs and Chapter 11 for bipolar junctions transistors, the device operation is typically characterized by distinct
regions, and compact models are developed to mathematically describe each
region by separate equations or a single model. In most compact models, the
device performance in each mode of device operation is described by a set
of model parameters. Thus, a set of device characteristics such as currentvoltage (I–V) and capacitance-voltage (C–V) is required to fit the model
equations to each operating region of device characteristics and extract the
corresponding model parameters of the selected set of devices.
Again, to illustrate the selection of device characteristics for compact modeling, let us use BSIM4 regional model. The MOSFET device is primarily characterized by subthreshold, linear, and saturation regions. Thus, to model the
entire device characteristics for the selected set of devices of a target technology, the device data are obtained under the appropriate biasing conditions
determined by the target supply voltage, V dd , at the target operating range of
the ambient temperature (T) of the devices in the IC chips. Table 12.1 shows
a typical set of device characteristics required for compact modeling of the
selected MOSFET devices of a specific technology under different gate-source,
drain-source, and body-source voltages V gs , V ds , and V bs , respectively.
A similar set of device data are obtained for PMOS (n-type body with p+
source-drain) devices by changing the sign of the operating applied voltages.
In addition, the source-drain pn-junction I–V and C–V characteristics are
obtained for both NMOS (p-type body with n+ source-drain) and PMOS
(n-type body with p+ source-drain) devices to extract the source-drain diode
model parameters. The diode model is an integral part of MOSFET compact
model, and therefore, diode characteristics are part of data collection for
compact MOSFET modeling. For analog/RF modeling, the additional set of
required device characteristics is obtained.
The rev0 compact model of a target technology can be generated from the
numerical device simulation data for the feasibility study and early IC design
evaluation [5–7]. However, in order to generate the final compact device
model of a technology for product design in CAD environment and release
to process design kit (PDK), the measurement data are collected from a single
die, referred to as the golden die of a specific silicon wafer, referred to as the
golden wafer. The golden die of the golden wafer provides the target device
performance of the target technology node [6].
In order to develop a statistical model library, the required set of data shown
in Table 12.1 is collected from different silicon die, wafers, and different
wafer lots over a period of time. Then from the statistical distribution of data
Compact Models for Integrated Circuit Design
of the shortest device. Note that in most cases, L min represents the nominal
device of the target technology node.
12.2.1.2 Selection of Device Characteristics
As discussed in Chapter 4 and 5 for MOSFETs and Chapter 11 for bipolar junctions transistors, the device operation is typically characterized by distinct
regions, and compact models are developed to mathematically describe each
region by separate equations or a single model. In most compact models, the
device performance in each mode of device operation is described by a set
of model parameters. Thus, a set of device characteristics such as currentvoltage (I–V) and capacitance-voltage (C–V) is required to fit the model
equations to each operating region of device characteristics and extract the
corresponding model parameters of the selected set of devices.
Again, to illustrate the selection of device characteristics for compact modeling, let us use BSIM4 regional model. The MOSFET device is primarily characterized by subthreshold, linear, and saturation regions. Thus, to model the
entire device characteristics for the selected set of devices of a target technology, the device data are obtained under the appropriate biasing conditions
determined by the target supply voltage, V dd , at the target operating range of
the ambient temperature (T) of the devices in the IC chips. Table 12.1 shows
a typical set of device characteristics required for compact modeling of the
selected MOSFET devices of a specific technology under different gate-source,
drain-source, and body-source voltages V gs , V ds , and V bs , respectively.
A similar set of device data are obtained for PMOS (n-type body with p+
source-drain) devices by changing the sign of the operating applied voltages.
In addition, the source-drain pn-junction I–V and C–V characteristics are
obtained for both NMOS (p-type body with n+ source-drain) and PMOS
(n-type body with p+ source-drain) devices to extract the source-drain diode
model parameters. The diode model is an integral part of MOSFET compact
model, and therefore, diode characteristics are part of data collection for
compact MOSFET modeling. For analog/RF modeling, the additional set of
required device characteristics is obtained.
The rev0 compact model of a target technology can be generated from the
numerical device simulation data for the feasibility study and early IC design
evaluation [5–7]. However, in order to generate the final compact device
model of a technology for product design in CAD environment and release
to process design kit (PDK), the measurement data are collected from a single
die, referred to as the golden die of a specific silicon wafer, referred to as the
golden wafer. The golden die of the golden wafer provides the target device
performance of the target technology node [6].
In order to develop a statistical model library, the required set of data shown
in Table 12.1 is collected from different silicon die, wafers, and different
wafer lots over a period of time. Then from the statistical distribution of data