I'm posting in this forum because the OPA657 problem affects anyone who attempts to simulate a high speed amplifier with an OPA657. The problem is with the macromodel but if this is posted there, it will not be noticed by anyone who is doing circuit design.
The OPA657 macromodel is missing its input capacitance! This gives very misleading results when simulating a high speed transimpedance amplifier, for example. The differential + common-mode capacitance is specified in the OPA657 data sheet as 0.7pF + 4.5pF but measuring it in a TINA simulation results in about 0.24pF.
I haven't checked but the OPA656 may have the same problems. Perhaps others...
There are consistent problems with most of TI's macromodels-- ditto for other vendors, too. Many macromodels omit important parameters such as voltage & current noise (and even if they do, the 1/f spectrum is missing), input capacitance, settling time, supply current, and on and on.
The only TI macromodels that I have confidence in were written by Bill (W.K.) Sands; these models are very complete and are as accurate as practical. Most models were written by a variety of engineers that were not modelling experts so the performance of simulations with these models vary widely. Without consistently well-written macromodels, a designer cannot have confidence that his circuit simulation is truly representative of the device's actual performance. Macromodels hurriedly written by amateurs to satisfy managers who pressure the engineer to "finish the project so that it can be introduced" is not the way to approach such an important part of a new device's introduction.
I contend that having a comprehensive, accurate macromodel for a new device is as important as having a clear, accurate data sheet.