In addition to combinations of temperature range and power supply voltages (usually more than one), the process design kit which landed on your desk yesterday presented a bewildering alphabet soup of device corner combinations which you need to consider when verifying your circuit design.
Fast/slow, high/low, pre/post. If I have to spend the time to run all the possible combinations, I won't have much time left to tune my design. But how do I really know which combinations are going to be the worst case for each of my specifications?
This is exactly the problem addressed with the Worst-Case Corners (WCC) run mode in the Virtuoso Analog Design Environment GXL (ADE GXL).
The approach is straightforward, the setup is simple, and the results are powerful.
The Approach-Framing the Problem
Worst-Case Corners follows a basic Design of Experiments (DoE) methodology. That is, given a set of input variables (your corner parameters and their values), and a set of measured output responses (your testbench measurements), a number of trial sets of input values are intelligently chosen and simulated. The results are observed and analyzed and then one corner is created for each specification containing the combination of variable values which has been determined to give the "worst" result. (Two worst corners are created for range and tolerance type specs.)
In general, process corners are given special treatment during trial creation, since there is inherently no natural sequence to the list of model sections. Thus, all process corners are considered in simulation.
The Setup-What Goes In?
WCC works with your existing ADE XL testbenches, measurements and specifications. You can use the corners you already have defined in ADE XL as input for the WCC setup, or you can specify any combination of variable values to be considered.
Several methods are provided to trial experiment selection and corner creation. You choose the method based on the linearity of your circuit response and inter-relationships of your input variables. Standard DoE and response surface methods (RSM), such as one-factor-at-a-time (OFAT), Central Composite and Factorial, are supported to give a full spectrum of analysis capabilities and performance.
The number of simulations required to determine the worst-case corners depends on the number of variables and the number of values for each variable. Typically, on the order of 30-50 simulations are run for most common corner setups.
The Results-What Comes Out?
At the conclusion of a run, worst-case corners are created for each specification in your ADE XL setup. Depending on the method used, the created corners will be simulated and validated against the models used to identify the worst-case variable values.
In addition, the results of the simulations are analyzed for sensitivity and displayed in an interactive table, allowing you to further examine the characteristics of your circuit's responses to each variable as well as how each factor contributes to the overall variation in circuit performance.
Now that you have created the worst-case corners for your design, you are well-positioned to efficiently optimize your design, either manually using ADE XL, or by taking advantage of advanced ADE GXL analyses such as Sensitivity Analysis and Circuit Optimization.