Thin-film SOI CMOS for heterogeneous microsystems

1
RF Extraction Techniques for Series Resistances
of MOSFETs
J. C. Tinoco and J.-P. Raskin Université
catholique de Louvain Microwave Laboratory B-1348
Louvain-la-Neuve, Belgium
2
OUTLINE

• Introduction
• Bracales Method
• Bracales Modified Method
• Results
• Conclusions

3
INTRODUCTION
Different methods have been developed to
determine the extrinsic series resistances. They
can be divided in two groups DC and RF methods.
DC Methods RF Methods

• It is possible to extract independently the
  drain, source and gate resistances.
• Device biased under different conditions.
• Requires the equivalent circuit analysis.

• It is not possible to extract independently the
  drain and source resistances
• RT Rd Rs
• It is not possible to determine the gate
  resistance.

4
INTRODUCTION
The main RF methods are Lovelace, Torres-Torres,
Raskin and Bracale.

• Lovelace and Torres-Torres methods are quite
  sensitive to noise.
• Signal pre-treatments do not improve the
  extraction.

Torres-Torres
Lovelace
5
INTRODUCTION

• Raskins method also is quite sensitive to
  noise.
• Signal pre-treatments seem to improve the
  extraction.
• For deep-submicron devices its application seems
  limited.
• Bracales method is less sensitive to noise.
• Fails to determine the correct resistance values.

Raskin
Bracale

• Deep analysis is necessary for the Bracales
  method

6
Bracales Method
Bias VDS 0 V VGSgtVT Gmi ? 0

• Assumptions
• Perfectly symmetric Device Cgsi Cgdi C
• Constant mobility

7


INTRODUCTION
Bracales Method
Impedance Relationships
?
8
Bracales Method
Linear regression of the impedance relationship
respect to the inverse of the gate overdrive.
The intercept gives the corresponding series
resistance.
9
Bracales Method
Extraction Methods Rse Rde Rge
Classical Bracale 7.5 7.3 2.8
Used in Simulations 3 3 5

• The extracted values differ from the values used
  in the simulations.
• It is necessary to review the assumptions made
• Perfectly symmetric Device Cgsi Cgdi C
• Constant mobility

10
Bracales Modified Method
Mobility degradation coefficient
The inverse of the output conductance is a linear
function of the inverse of the gate overdrive
The slope s and the intercept b are
And thus
11
Bracales Modified Method
The impedance relationships will be expressed as
?
They follow linear function respect to the
inverse of the gate overdrive, the slope x will
be
Extraction Methods Rse Rde Rge
Classical Bracale 7.5 7.3 2.8
Mobility degradation included 3.4 2.77 4.84
Used in Simulations 3 3 5
12


Bracales Modified Method

• The mobility degradation strongly affects the
  extraction accuracy.
• Considering non-perfectly symmetry, the impedance
  relationships will be expressed as

Where k Cgs/Cgd is called the asymmetry
coefficient.
13

Bracales Modified Method
Thus, the extracted series resistances will be
obtained as

• Overcome the limitations of the classical method.
• Non-perfectly symmetry is considered.
• Mobility degradation coefficient is included (?).

Extraction Methods Rse Rde Rge
Classical Bracale 7.5 7.3 2.8
Mobility and asymmetry included 3 3.2 4.85
Used in Simulations 3 3 5
14
Bracales Modified Method
Asymmetry coefficient
The imaginary part of the impedance parameters
follow the next relationships
Thus, we can obtain k as
15
Results
ELDO software was used to simulate the
S-Parameters of Partially-Depleted 0.13 µm SOI
n-MOSFETs. The BSIM3SOI model from
ST-Microelectronics was used.
Rse Rde 3 W Rge 5 W
16
Results
q 0.6
Rse 3 W
17
Results
Rge 4.85 W
Rde 3.2 W
Extraction Methods Rse Rde Rge
Classical Bracale 7.5 7.3 2.8
Mobility degradation included 3.4 2.77 4.84
Mobility and asymmetry included 3 3.2 4.85
Used in Simulations 3 3 5
18
Conclusions

• Original Bracales method does not allow
  accurate extraction of the series resistances.
• The main limitations of this method are the
  carrier mobility degradation and transistor
  asymmetry.
• A new procedure was established, where the both
  effects are included.
• q is obtained from DC output conductance
  measurements.
• k is obtained as the ratio of the imaginary part
  of Z-parameters.
• The new procedure allows to determine the
  correct resistance values.