Page 232 - A Practical Guide from Design Planning to Manufacturing

204 Chapter Seven

The length of the transistor is the spacing between the source and
drain. This can be specified as L gate , the actual width of the poly gate,
or as L eff , the effective separation between the source and drain dopants
under the gate. L gate is a physical parameter that can be measured with
an electron microscope, but it is L eff that will actually determine how
much current the transistor produces. Because the dopants of the source
and drain diffuse through the silicon underneath the gate, L is always
eff
smaller than L gate . When we speak of Moore’s law driving the semicon-
ductor industry forward by making transistors smaller, we are prima-
rily taking about reducing the minimum transistor length.
As the transistor length is reduced, the thickness of the gate oxide (T )
ox
must also be reduced for the gate to have good control of the channel.
If T is not scaled, there could be significant current even in the cutoff
ox
region. Moving the gate physically closer to the channel makes sure
the transistor can be effectively turned off. The width of the transistor
is determined by the size of the source and drain regions. A wider chan-
nel allows more current to flow in parallel between the source and drain.
The factor e ox measures the electric permittivity of the gate oxide
material. This will determine how strongly the electric field of the gate
penetrates the gate oxide and how strong it will be when it reaches
the silicon channel. Finally, the factor m is a measure of the mobility of
the charge carriers. The drift velocity of the carriers across the channel
is estimated as their mobility times the electric field between the source
and drain. Together these five factors, the transistor length, width,
oxide thickness, oxide permittivity, and carrier mobility, determine the
b value of a transistor.
Of these five, the circuit designer has only control of the transistor’s
length and width. Most transistors on a die will be drawn at the mini-
mum allowed channel length in order to produce the most current and
provide the fastest switching. Some sequential circuits will make use of
transistors as “keepers,” which only hold the voltage level of a node when
it is not being switched by other transistors. These keeper devices may
be intentionally drawn with nonminimum lengths to keep them from
interfering with the normal switching of the node.
The circuit designer increases a transistor’s switching current by increas-
ing the width, but this also increases the capacitive load of the gate.
Increasing the width will make the drain of this transistor switch faster,
but the transistor driving the gate will switch more slowly. Also, a wider
transistor requires more die area and more power. Much of a circuit
designer’s job is carefully choosing the widths of transistors to balance
speed, die area, and power.
The manufacturing process determines the oxide thickness, oxide per-
mittivity, and charge mobility. Modern processes use gate oxides less than
10 molecules thick, so further scaling of the oxide may be impractical.