characterization, modeling, and design of esd protection circuits

2.2. Transmission Line Pulsing 33
above a predefined level. Reasonable time-to-failure measurements can be made down to
about 50ns. For times of 1µs and greater, a pulse generator can be used in place of the
charged transmission line. After a curve is experimentally determined, the dimensions of
the theoretical box can be extracted by fitting the model to the experimental curve.
Since a Pf vs. tf curve reveals circuit failure thresholds over a wide spectrum of stress
times, it suggests how robust a device is throughout the ESD and EOS regimes. It has been
suggested that Pf vs. tf and If (failure current, or It2 ) vs. tf curves be used to qualify EOS/
ESD reliability in addition to or in place of standard tests such as the HBM because
reliability is then defined over a large range of stress events [24]. This attribute is
attractive because it may show that a protection-circuit design performs relatively well in
one domain of the EOS spectrum but performs poorly in another. Retesting after design
modification would reveal what portions of the spectrum are affected by a certain device
parameter. Some correlation has been drawn between TLP failure levels and HBM
robustness [23], but further qualification must be done before IC manufacturers accept the
Pf vs. tf method as a valid reliability measure. The value of the method ultimately depends
on how well the accepted classical models are represented by the constant-current stresses
of TLP.
2.2.3 Leakage Current Evolution
The previous section mentioned the measuring of device leakage current after a TLP stress
to verify that second breakdown has taken place. If a device exhibited a second snapback,
it would probably not create a large increase in leakage and thus could be distinguished
from the thermal second breakdown. It is in fact very useful to monitor the leakage
evolution after each stress step of a TLP experiment. This can be done by removing the
transmission-line connection from the input of the device under test, applying a voltage to
the input (typically the supply voltage, VCC ), measuring the current with a multimeter in
series with the VCC supply, then reconnecting the transmission line. The voltage should be
applied as briefly as possible to avoid corrupting the TLP experiment by further stressing
the device. In contrast to the single leakage measurement made after a HBM stress, this
technique reveals how the increased leakage evolves as a device is stressed through the
various levels of the snapback curve. Before snapback, the leakage current is typically in
the pA range. A jump in leakage above the 1µA level is usually observed after second
breakdown due to diffusion of dopants from source to drain, filament formation across the