Piped ADS Simulations

Specifying CAN_PIPE (the default) in your usersimulators file for the ADS simulator enables IC-CAP to take advantage of the tune mode built into the ADS simulator. This mode permits changing parameters of a simulation without requiring the simulator to be relaunched. This greatly reduces the time required for optimizations to run. However, each setup requires a new simulator to be launched. By default, IC-CAP permits up to 3 ADS simulators to be running at once so that an optimization across as many as 3 setups can be completed in the fastest time possible. Certain large simulations may require a great deal of system resources and having 3 simulations currently active can degrade system performance. If you encounter this problem, you can set the MAX_PARALLEL_SIMULATORS system variable to 1 or 2. If your system can handle more than 3 simulators in parallel and you need to optimize across more than 3 setups at a time, the value of MAX_PARALLEL_SIMULATORS can be increased.

When CAN_PIPE mode is used, the ADS simulator will bring up a status window during simulation. The first time the simulator is launched it can take several seconds for this window to appear. Once it is open, successive simulations will attach to the same status window. Each time a new setup is simulated, a new simulator must be started. There is a certain start-up delay associated with each invocation. This will be much shorter than the very first invocation which needed to launch the status window. Successive simulations of a setup which has been previously simulated will return in the shortest time as the simulator does not need to be reinvoked.

Opening the Simulation Debugger will terminate all running simulators, and close the ADS status window. Simulations done with the Simulation Debugger window open are performed in non-piped mode and thus the ADS status window is not opened.

In situations when you want to use the $mpar or $dpar feature in #echo lines for MNS and ADS netlists, you must enter names properly. The proper ADS name syntax is a dot-separated name, such as NPN.Bf. If you fail to use a proper name, simulations will yield incorrect results when you try to use the simulator in CAN_PIPE mode. If names cannot be revised, use CANNOT_PIPE.

This was especially problematic for userdefined models requiring many #echo lines using the $mpar feature in order for IC-CAP to parse it properly. This problem occurs when the technique used to implement userdefined models in ADS is declaring 2 new components, 1 a modelform and another an instance. This implementation of user-defined models led to the requirement for #echo lines. The modelform component looked like any other ADS netlist component, but it had no nodes. The parser is modified for IC-CAP 2001 to recognize a nodeless component as a userdefined model; however, only in the context of a subcircuit. If you want to create this type of userdefined model in ADS, then you must use a subcircuit. Doing so eliminates the need for #echo lines in the netlist and the subcircuit will parse and simulate properly.