2.2.1 Fast Pulsed -characteristics

Kerber et al. [19] were the first to circumvent the problem of slow response times by developing the fast pulsed -method shown in Fig. 2.4. They adapted the MSM-technique and used a digital storage oscilloscope (DSO) to quickly measure the voltages and currents of the device under test (DUT) and a programmable pulse-pattern generator. The basic principle of the fast pulsed -method is depicted in Fig. 2.5 for NBTI (top) and PBTI (bottom) and works as follows: During initialization, stress or relaxation, is set to the corresponding constant values , or . The pulse generator triggers the fast -measurement by sending a gate-pulse reaching from accumulation to inversion when in relaxation-mode, respectively from inversion to accumulation when in stress-mode.

Figure 2.4: Kerber’s setup to simultaneously record and . Since the digital storage oscilloscope (DSO) can not measure directly, is calculated via the voltage drop across to finally obtain the -characteristic. The corresponding triangular -pulses shown in Fig. 2.5 are supplied by the pulse generator.

Figure 2.5: Top: The fast-pulsed--characteristics are performed via a superposition of a constant gate level (stress or relaxation) with triangular gate pulses. Switching from the requested NBTI stress of into the measurement mode ranging from to should be carried out as fast as possible in order to avoid undesired relaxation defects. Bottom: The same sequence for a PBTI stress of .

Since the DSO can only measure voltages, the actual drain current is calculated via the voltage drop across (Fig. 2.4), assuming small enough so that the transistor stays in the ohmic region.

With standard equipment, pulse times between [19], [20, 21, 22, 23, 24, 25], down to [26] can be achieved. The form of used pulses varies from trapezoidal [19, 21, 26], over rectangular with only very small rise and fall times compared to the pulse width itself [21], up to triangular [20, 21, 22, 24, 25]. By varying the rise and fall times of the pulses the trapping and detrapping kinetics can be analyzed [21]. To avoid spurious hystereses (parasitic capacitances) in the -characteristics between the rising and falling edges of the pulses, the cable length has to be adjusted in order to ensure the synchronized signal transmission to the DSO [25, 20].

The major issue with this method is that the gain in speed is partly consumed by the fact that the resolution of the DSO is too limited for real ‘single’-pulse-measurements [12]. After the necessary averaging of a few () pulses, the measurement time increases by the averaging factor. Furthermore, the synchronization between the pulse-pattern generator and the DSO turns out to be tricky.