p3 Correlation Verification between Transistor Variability Model with Body Biasing and Ring Oscillation Frequency in 90nm Subthreshold Circuits

ABSTRACT

This paper presents modeling of manufacturing variability and body bias effect for subthreshold circuits based on measurement of a device array circuit in a 90nm technology.

1. INTRODUCTION

Subthreshold circuits operate at lower supply voltage than threshold voltage (Vth) of MOSFETs

(p1, col 2)
Subthreshold circuits are sensitive to manufacturing variability,
as mentioned above.

This work is the first work that explicitly verifies the correlation
between device variability modeling and performance variation of
subthreshold circuit considering body biasing.

2. DEVICE ARRAY CIRCUIT
2.1 Circuit Structure Overview
Figure 1 shows the device array circuit designed to measure variations
of MOSFET I-V characteristics and RO frequencies in subthreshold
region.

2.2 Measurement Circuit for MOSFET characteristics
Figure 3 shows the measurement circuit for MOSFET I-V characteristics.

The transistor for measuring its I-V characteristics is selected
by a column selection signal and a row selection signal

2.3 Measurement Circuit for Ring Oscillator
Frequencies

3. MEASUREMENT RESULTS AND CHARACTERIZING
VARIATIONS
3.1 Variation of RO Frequencies

3.2 Characterizing Variations
We here discuss how to model MOSFET variations from measured
I-V characteristics. The

(page 3)
A parameter of DELVTO is used to change Vth.

Figure 6 shows an example of measured and simulated Ids-Vgs
characteristics at Vds=0.3V with the extracted DELVTOand NFACTOR
parameters.

Performing the extraction for all transistors, we can obtain the
distributions of DELVTO and NFACTOR variations that express
Vth and n variations.

(p3 col 2)
In handling measurement data, exclusion of outliers is critically
important to make a reasonable statistical model, since some outliers
lead the average and standard deviation to totally inappropriate
values.