Drain Engineering Optimization for Nanometer nMOSFETs on Mixed Signal Application
3.3 Experimental result and discussion
The threshold voltage independence on Lg is excellent Vth characteristics. Fig1
and Fig2 show Vth vs. drawn gate length (Lg) and Ion vs. Ioff respectively for diverse
pocket implant doping concentration. The heavy pocket doping could lead to high
31
could compress off current while keeping Ion current fixed
In analog circuit applications, the voltage gain is directly proportional to Rout .In [8], it’s shown existing analytical models for MOSFET Rout are not adequate, because only
channel length modulation effect is included. Here, we notice, The Rout curve can be
clearly divided into four regions with each region dominated by a mechanism in Fig3 be
shown .It is worth noting that the second region is the near-saturation region dominated by
CLM. The third region is dominated by DIBL and the fourth is the high field region in
which Rout is greatly reduced by the substrate current induced body effect (SCBE).[8] In
this case, Rout is continual raise so that SCBE could be neglected.
In view of foregoing concept, let us then consider that the variation of Rout with
different pocket and LDD process devices to Lg at the same Vgst (Vgs-Vt=0.1 V) and
Vd bias which Vd=0.7 V for IO and Vd=0.6v for Core. It is shown in Fig4 and Fig5. It
clearly shows that cross phenomenon. The different devices structure and process
could move the cross points to gate length. Hence, the Rout at least should be
differentiated into two physic effect. We believe that different channel length could
lead to change of physic mechanism boundary of Fig3. The fore and back cross due to
CLM and DIBL respectively. The concept of Rout to variable Lg should be changed as
well. The most important characteristics of Rout in analog circuit designs are the
maximum Rout which determines the maximum available gain from the device. Fig6
shows the gain of wide part dependence on the Rout. To obtain the high gain devices
have different notions in distinct Lg. The lighter pocket implant concentration could
induce high Rout and gain at after of cross. Adversely, that could induce poorer Rout
and gain at before of cross.
VA has three components, i.e., VACLM, VADIBL and VASCBE, corresponding to CLM, DIBL
and SCBE, respectively. Each component can be evaluated separately.[8]
]
The individual component of VA together with the resultant VA is shown in Fig7. The
dominant mechanism is the one with the smallest Early voltage in each region.
In this case of VA which obtaining by intersection of X axis with tangent of Vd
bias at 0.7V and 0.6V for IO and Core could exclude VASCBE because of this Vd bias is
not enough to induce substrate current. For experimental data in Fig8 and Fig9 these
are similar to the Rout that VA appears cross points around 3 time technology node to
33
The concept of VA of fore cross due to CLM dominated. The length ∆L of the pinch-off region increases by an expansion in the direction of the source end with
lighter pocket doping concentration and effective channel length has been reduced by
∆L at the same bias condition. In other word, the ∆L increase with decreasing pocket
dose concentration could reduce VA. CLM is not a special short-channel phenomenon,
but its relative importance becomes distinctly more pronounced at short gate lengths.
This is traditional notion for design VA , but the concept of back cross is in conflict with
pocket implant concentration. For long channel devices of back cross, despite CLM
and DIBL should slowly be negligible. However, large residual DIBL exists at back
cross for the pocket devices in Fig8, dominating the VA values which function (1). In
these dimension, higher pocket doping concentration lead to Vt,sat shift even more.
The conventional explanation of channel length modulation is not entirely valid in
pocket devices. Hence, these new observations have a significant impact on design of
VA by cross phenomenon of the pocket device. Anyway, the experimental results from
pocket device are shown in Fig8 and Fig9. It can be seen that these devices have lower
DIBL leading to higher VA at back cross. Adversely, leading to poorer VA at fore cross.
3.4 Summary
We have presented experimental data showing that certain pocket implant
concentration for digital CMOS technology and then found out the interesting cross
points, which imply pocket implants affect at least have two physic mechanisms CLM
and DIBL. The conventional explanation of channel length modulation is not entirely
valid in pocket devices. Hence, these new observations have a significant impact on
design of VA and Rout. The concept of back cross differs from CLM so that the
analog device design for pocket implant must be notice. The lighter pocket implant
concentration could induce higher Rout, gain and VA at back of cross. On the other
hand, the heavier pocket implant concentration could induce higher Rout, gain and
VA at fore of cross.
35
0.1 1 10
300 350 400 450 500
550 NMOS IO Tox=32A
for different pocket implant concentraion
Vt(mv)
Vd
heavy In and BF light In and BF light BF
Fig1. Vt vs. Lg for different pocket implant concentration
1E-6 1E-5 1E-4 1E-10
1E-9 1E-8
1E-7 NMOS IO Tox=32A
for different pocket implant concentration
Ioff
Ion heavy In and BF light In and BF light BF
Fig. 2 Ion vs. Ioff for different pocket implant concentration
37
Fig. 3 Typical drain current and output resistance.
1 0
5 10 15 20 25
30 NMOS IO Tox=32A Vd=0.7v Vgt=0.1v for diffent pocket devices
Rout (kohm)
Lg heavy In and BF light In and BF light BF
Fig4. Rout vs. Lg for NMOS IO
39 light In and medium As heavy In and light As
NMOS Core Tox=20A
for different PKT and LDD concentration
Rout (kohm)
Lg
light In and heavy As light In and medium As heavy In and light As
Fig5.Rout vs. Lg for NMOS Core
0.1 1
Fig6. Gain and Rout vs. Lg for different pocket implant devices.
41
Fig7 Early voltage and its components versus Vds
0
43
light In medium As heavy In light As