Gate Pulse Settings

6.1 Gate Pulse Settings

In order to automatically perform the required averaging of the recorded $ID$ , rectangular gate pulses were used for short-term NBTI stresses in the range of $1μs$ to $1s$ , as illustrated in Fig. 6.3. Each gate pulse was followed by a 100 times longer recovery sequence which allowed for full recovery of the built up degradation [14].

Consequently, a pulse train with $tlead = ttrail = 5ns$ , a width $tW = tstr$ , and a period of $tP = 100tstr$ , consisting of $N$ pulses is used. The product $NtP$ is only limited by the overall measurement time $tM = N tP$ . A compromise between the recovery time in-between pulses ( $≈ tP$ ) to let the device fully recover and a reasonably high $N$ has to be found in order to gain sufficient measurement accuracy through averaging.

Figure 6.3: Short-term NBTI stress is performed using $1μs$ up to $1s$ long rectangular gate pulses. The total stress time is split into three sub-intervals including some overlap. To record the same process many times, the above mentioned long recovery time is required. Averaging enhances the amplitude resolution. The number of used pulses is shown in the legend.

Figure 6.4: The very low duty cycle is necessary to achieve full relaxation between stresses.

Since the oscilloscope uses a linear time scale, but NBTI stress must be assessed on a logarithmic scale spanning at least 3 to 4 decades, the stress time of $1s$ had to be split into three intervals, cf. Fig. 6.3. This allows higher time resolution at the beginning of the stress phase and lower resolution at its end. Since the measurement noise decays with the inverse of the time resolution, with the slower sequences a lower averaging number is necessary to achieve a given amplitude resolution. The according values of $tstr$ , $trel$ , $tP$ , and $N$ are shown in Tab. 6.1, as well as the resolution, which also equals the minimum stress time of the respective stress sequence.

Sequence	$tW = tstr$	$trel$	$tP$	$N$	Resolution

1	$1ms$	$99ms$	$0.1s$	1000	$0.16μs$
2	$100ms$	$9.9s$	$10s$	10	$16μs$
3	$1000ms$	$99s$	$100s$	5	$160μs$

Table 6.1: Details of the rectangular stress pulses used to maximize the amount of recorded information together with the resolution.

In order to combine the three sequences into a single degradation curve with a maximum effective resolution from $1μs$ to $1s$ , the three stress sequences are chosen to overlap for at least one decade of time. Since only differences of currents ( $I D$ ) are recorded, the overlap regions provide information to align the sequences to a single stress characteristic. An example is displayed in Fig. 6.5. The offset in $ID ∕ID,0$ depends on the different amplification factor in each measurement sequence.

Figure 6.5: Different amplification factors in the DSO settings are responsible for the vertical offset (top). This has to be corrected to make the stress sequences coincide. Merged stress sample (bottom) using a log-fit and shifted to the reference time $t0,ref = 2μs$ .

[next] [front] [up]