Hi,
I have for many years been very happy to see TI offer so many none encrypted and extensive models for various products. I also noted that some time ago a clear change too place that involved the introduction of the, "Green-Williams-Lis Op Amp Macro-model Architecture" within many macro models used to model TI's opmaps. I have seen older models become obseleted and models improved with this new architecture. There have even been some basic descriptions of the methodology used by this model in a TI article. More would always be welcome.
Having said this I have become somewhat disappointed. I do not use Pspice or TINA but like undoubtedly many many other engineers I make use of LTspice. We can argue the market realities of this, but it is a defacto truth that I am far from alone in using LTspice which is a well respected and established product of considerable worth which is both mature and well supported by other vendors in terms of available LTspice specific models. I make this statement not to in any way promote the software, but rather to avoid the 'easy' response which might be, TI does not support LTspice.
To the issue. It seems that all the models that make use of the "Green-Williams-Lis Op Amp Macro-model Architecture" have an unfortunate habit is very strange behavior during transient simulation startup. They will often sink/source totally unrealistic currents while the IC power supply rails have not yet become active. Often times these currents may be in the order of mA sometimes many 10s of amps, which can drive nodes in the simulation to hundreds of KV. I am very careful in my simulations to generate most of my voltage sources using current models. They are soft sources with realistic start-up behaviors, no badly behaved discontinuities and realistic internal resistances. I ramp everything from time = 0 to clearly define circuit start-up behavior.
I am writing to you because I wish to report my day-to-day truths. I have started avoiding the use of any TI op-amp models, instead being forced to used alternative vendors, or in some cases using simplified op-amp macro models to estimate the performance of TI parts. If I download a op-amp model from TI, I will open its subckt definition and look to see its it is using the "Green-Williams-Lis" model and then simply stop using it. I have wasted days of effort trying to get them to play well that I have essentially abandoned all hope. I am frustrated by this because it seems clear from reading the work done by the engineers in developing this model, that the approach has a great deal of merit. It is consistent and accurate and I would very much like to take full advantage of the capabilities.
Has anyone in the modeling team at TI taken some time to take a careful look at each of the blocks used inside the "Green-Williams-Lis" model to try and avoid these corner cases and start-up behaviors. I have made some basic investigations myself, but the models are based on a large number of predefined subckt blocks, they are quite generic and interconnected complexity is almost impossible to follow as no consistent and clear node naming methodology is used, rather everything is simply numbered, which makes reverse reading the spice code almost impossible. I am convinced these models must be graphically generated and then the spice output has no need to concern itself with numbered node names. Of course this is fine if you have access to the original schematic layouts used to generate the code, but impossible as the end-user when working backwards.
I don't expect much about my comments will make much difference, but I am throwing this over the fence anyway. Guys please fix these issues, for me at least they are crippling the good efforts of your previous team and I am sure that having one of your customers abandon your op-amp models was never really your intention.
I am very much open to any well considered suggestions and I would be happy to try any modification that you might offer. I would add finally that I mean this with the very best intentions. I do not try to suggest what the issues might be exactly and I am sure that my modeling is not perfect, but what I can say that these problems I have, keep coming back and pointing the finger at this TI op-amp macro model. When I replace the parts in my simulations with older TI models or other vendor model, often the problems are effortlessly resolved.
All the best.
Aidan