Fabless designers rely on the accuracy of the Process Design Kit (PDK) supplied by the foundry in accounting for the statistical nature of the offset voltages. However a foundry PDK does not exactly model all choices of devices. At process geometries of 130nm and below, the designer requires a tool, such as PDKChek to accurately determine the relevant parameters for the specific devices used in the design (width and length choices).
Matching performance is a key parameter for CMOS processing. Mismatch in transistor threshold voltage, resistance, and capacitance are dominant factors in IC performance. Mismatch in VT can increase offset voltage, CMRR and poor performance. Differences in turn on current in digital circuits produce timing errors, reduce design margins, and can impact yield. Process-related parameter variance is inversely proportional to transistor area and therefore becomes increasingly important as the dimensions of the transistors are reduced.
The influence of local process-related variations in device characteristics on electrical parameters of an IC is becoming a critical issue as device geometries and power supplies continue to decrease.
Ridgetop’s PDKChek measures die-level process-induced variations, both random and systematic, in MOS transistor threshold voltage (VT), resistance, capacitance, and turn on/off current. PDKChek is an unobtrusive, stand-alone, IP Block designed to accurately and precisely measure the variation in parameters resulting from the randomness inherent to processing.
Ridgetop’s PDKChek IP block provides circuit designers with independent verification data to improve the accuracy of process design margins, increased process yield, expedited problem resolution (design- or process-related), reduce design iterations (duration and frequency), and shorter time to market.
PDKChek allows for faster testing than traditional scribeline transistors. Testing can take place before or after packaging the die. Die level testing takes advantage of test structures and bonding pads that are already present on the die, so there is no additional error introduced by the contact resistance of a probe station.