12.12.2012 Views

characterization, modeling, and design of esd protection circuits

characterization, modeling, and design of esd protection circuits

characterization, modeling, and design of esd protection circuits

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

110 Chapter 4. Simulation: Calibration <strong>and</strong> Results<br />

<strong>and</strong> does not diminish the value <strong>of</strong> the simulation because the results <strong>of</strong> interest all occur<br />

at current levels above 100µA. The thermal boundary conditions consisted <strong>of</strong> overlapping<br />

the electrical contacts with constant-temperature (297K) thermal contacts with no thermal<br />

resistance. Although the simulations examined here are referred to as calibration<br />

simulations, if the mobility <strong>and</strong> impact-ionization models have already been fixed by the<br />

MOSFET-characteristic calibration, then comparing the measured <strong>and</strong> simulated Vt1 , Vsb ,<br />

<strong>and</strong> Rsb is really a verification procedure rather than a calibration procedure.<br />

An example <strong>of</strong> the I-V curve <strong>of</strong> a dc snapback simulation is shown in Fig. 4.42. The<br />

horizontal line in the log curve shows where the solutions began incorporating the<br />

thermal-diffusion equation. Note that although the lattice temperature does not<br />

significantly increase above 300K until after snapback, the breakdown voltage is<br />

substantially lower without including the thermal diffusion equation because the<br />

temperature-dependent impact-ionization model cannot be used. Two things were<br />

immediately noticeable from the initial snapback simulations. First, the snapback<br />

resistance appeared to be a reasonable value (compared to experiment) immediately after<br />

snapback, but the curve quickly rolled over at higher currents, indicating a much higher<br />

resistance than in the experimental structures. Second, even when the snapback voltage<br />

was extrapolated from the initial, steep part <strong>of</strong> the snapback portion <strong>of</strong> the curve, i.e.,<br />

using a value <strong>of</strong> Rsb equal to the measured value, the snapback voltage was about 1.8V too<br />

high. It was apparent from these simulations that calibration <strong>of</strong> the mobility <strong>and</strong> impactionization<br />

models using the st<strong>and</strong>ard MOSFET curves was inadequate for snapback<br />

simulations <strong>and</strong> thus that further manipulation <strong>of</strong> the model coefficients was needed.<br />

Since the problem regarding the high snapback voltage was the simplest to underst<strong>and</strong>, it<br />

was dealt with first. The high Vsb value indicates that the impact-ionization generation rate<br />

is too low for a given electric field in the snapback region <strong>of</strong> the I-V curve because the<br />

simulated voltage (<strong>and</strong> electric field) needed to sustain a given current level is too high. As<br />

shown by Eq. (3.27) <strong>and</strong> Fig. 3.22, the impact-ionization rate for electrons is determined<br />

∞ crit<br />

by two model coefficients, αn <strong>and</strong> En (or λn , which by Eq. (3.28) is inversely<br />

crit<br />

proportional to En ), assuming βn is constant. In the calibration <strong>of</strong> the MOSFET<br />

∞<br />

substrate characteristic, αn was held constant <strong>and</strong> λn was varied until the effective II rate<br />

resulted in the proper amount <strong>of</strong> substrate current. A good fit <strong>of</strong> the substrate characteristic<br />

was attained because, as Fig. 3.22 shows, if the spread in peak electric field values<br />

throughout the stress conditions <strong>of</strong> the substrate-current test is relatively narrow, the

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!