Modification improves VCA's Spice simulation

Alex Rysin, RSVI Acuity CiMatrix, Canton, MA -- EDN, 9/1/2000

The Burr-Brown VCA610 is a good wideband voltage-controlled amplifier (VCA) that you can use successfully for your application. The Spice model from Burr-Brown simulates normally when the VCA uses dual power supplies (Figure 1). However, when you attempt to use the Spice model in a single-supply application, which is equivalent to moving the GND symbol to the negative V_CC pin of IC₁ (Figure 2), you get an error message indicating that the circuit fails to converge. To find the source of the problem, restore the original schematic used for the Spice-model creation of the VCA610 and carefully analyze it. The model has several problems that prevent you from using it in single-supply simulations. The main problem arises from the faulty usage of "0" nodes in the model description. Spice considers subcircuit nodes as "local" and assigns them new names for simulations. This trait of Spice holds true for all subcircuit nodes except for ground nodes named "0," which are global for the entire simulation file. You can use the ground node "0" in subcircuit descriptions, but you must be careful that the node does not create unwanted dc references in a circuit that is intended to be floating. To avoid problems, do not to use "0" nodes in the model description but rather use different names, such as GND or AGND. Otherwise, you must carefully separate them the "0" nodes from dc connections that may affect the internal voltages and currents of the subcircuit. You can use control sources to separate the nodes. Listing 1 is Burr-Brown's model of the VCA610.

In Listing 1, all entries with problematical usage of the global node "0" appear in boldface. The line "E1 11 7 POLY(1)(3,0)0.45 -0.11911" is a polynomial voltage-controlled voltage source controlled by the voltage applied between nodes 3 (the gain-control input Pin 3) and the global "0." If you use the subcircuit in the dual-supply circuit of Figure 1, and Pin 2 (GND) connects to the common-point ground, then the voltage at the node (Pin 3) creates the correct output signal in the controlled source, E1. If you use the model for single-supply applications (Figure 2), then Pin 3 receives a significant dc-bias component, and E1 produces erroneous results. A similar problem occurs with other components and their nodes, designated by boldface type in Listing 1. To resolve the problem, give the "0" nodes different names. In this case, you can replace them with the other ground node of the model—the node "2." Note that not every "0" node creates a problem. You need take into account only those "0" nodes that have a dc reference to external connections. Other "0" replacements are unnecessary. After the cited corrections, the VCA610 model in Listing 2 yields acceptable results in simulation. Unfortunately, it still has a convergence problem when you include it in a large schematic. You can significantly improve convergence by adding a 0.1 to 0.5W series resistor in the DX and DZ diode models. This addition makes the model more realistic and helps the Newton-Raphson algorithm that Spice uses to find an initial solution for the simulation. (DI#2572)