State-of-the-art equipment does not meet the combined resolution and
measurement speed requirements of NBTI assessment. Instruments either meet
(and exceed) the required accuracy, but are too slow to capture the fast
NBTI degradation transients (e.g. parameter analyzers), or deliver the
necessary time resolution, but are limited by their inherent coarse amplitude
resolution (e.g. digital storage oscilloscopes, DSO). Since in the latter case the
amplitude resolution can be enhanced by averaging, while in the former
there is no remedy for a too slow measurement, a DSO is used to record
multiple stress/relaxation-cycles and take the average of these. Care
has to be taken to conform to the preconditions of proper averaging,
namely to record the same process many times. Only in this way, the
measurement noise is reduced, while the ‘hidden’ deterministic process is
reproduced without introducing systematic errors. In the measurements
this is provided by very short stress times, and a very low duty cycle
in order to achieve nearly relaxation in-between stresses; the
characterization due to such a measurement yields a
-shift of less than
.
The method developed in [12] and shown in Fig. 2.6 is related to the previous
work of Kerber et al. and Shen et al. [19, 27] and also used a pulse generator
and a digital storage oscilloscope but is able to perform even shorter
stress measurements than the previously mentioned methods. Reisinger et
al. conceived a bridge circuit containing two differential amplifiers. To
suppress the noise the of the device under test (DUT) is compared to a
reference current, giving only differences, which can then be captured
with higher resolution. To furthermore obtain the required resolution of
better than
in
, the equipment was designed to deliver a settled
gate stress voltage
within
in
. For this reason, a
battery using a passive voltage divider and a fast electronic switch are
used.