Re: TC1025 comparator spice model

From: Mike Engelhardt (nospam_at_spam.org)
Date: 12/31/04 

Date: Fri, 31 Dec 2004 20:06:58 GMT

Jim,

>>> Hi Great One Jim,
>>>
>>> Looks fine. Now add power supply CURRENTS ;-)
>>>
>>> Thank you.
>>> Could you show me how to do that please?
>>
>> You might just use the LTspice device "1pole".
>> It's a behavioral opamp with a tanh() xfer
>> function. Using the rail-to-rail opamp without
>> feedback approximates a comparator. The 1pole
>> does steer current from the supplies to the
>> output. Also, the 1pole uses the tanh() xfer
>> function that's part of an intrinsic SPICE
>> device that knows how to help the solver find
>> the solution with higher than just the 1st
>> order information that's usually available to
>> the solver's Newton iteration. The behavioral
>> modeling like Jim's is using can't do that.
>>
> Tanh works just ducky in PSpice, or anywhere
> else for that matter...it allows bounding
> without discontinuities.

Yes, it can run "just ducky" in simple
simulations. But as the complexity increases,
especially if the behavioral devices are
found in feedback loops, you've have a
progressively more difficult time with
convergence problems when using a function
like atanh() as a behavioral expression
as contrasted with using it as part of a
properly written intrinsic SPICE device.

There's a difference between using a
transcendental like tanh() in an intrinsic
device and using it in a behavioral expression.
When implemented in an intrinsic device; almost
any SPICE, at least both PSpice and LTspice; is
aware of how to fix problems that arise in the
Newton iteration method. Newton-Raphson is a
1st order method that can run into trouble
finding the solution of multi-dimensional non-
inear systems like circuits. The types of
techniques used in intrinsic SPICE devices like
semiconductors can't be applied to general
behavioral expressions using atanh().

--Mike