Re: Problems with SPICE models from vendors

From: Robert Baer (robertbaer_at_earthlink.net)
Date: 03/25/05 

Date: Fri, 25 Mar 2005 22:38:18 GMT

Helmut Sennewald wrote:
> "Robert Baer" <robertbaer@earthlink.net> schrieb im Newsbeitrag
> news:1PQ0e.3481$gI5.1145@newsread1.news.pas.earthlink.net...
>
>> The LM324 model from TI works fine,but the one from National
>>Semiconductor is junk.
>> I tried numerous Analog Devices models for various rail-to-rail opamps,
>>and found that almost all i tried gave me the same kind of cryptic square
>>root error.
>> Those tried: AD8605, AD531, AD541, AD8552, and the AD8571; the only one
>>tried that did work was for the AD8614.
>> Now i would dearly like to have a set of models that were known to work
>>(and more or less correctly), but i need to get a working AD8605 model.
>> The sets i have came from the manufacturer created in the latter of
>>1992, and thus are not quite up-to-date.
>> However, the model for the AD8605 was downloaded via the web just
>>yesterday - implying the problem is not fixed.
>>
>> Can anyone help?
>
>
> Hello Robert,
> I don't believe that you really can judge the quality of these
> models as a beginner with SPICE simulations.
> I agree with you that most models have difficulties with convergence.
> Many of them are really over complicated and sometimes generated
> by stupid programs or "roboters" and not by engineers.
>
> I assume that the listed models will work with some tweaking
> of the convergence parameters.
>
> What simulator do you use?
> If it's LTspice then send me your files and I will make
> you a working example with your AD8605.
> I always want to see the schematic, because I know that
> people sometimes have errors in their circuit.
> One important thing is to have a DC path to ground(0).
>
>
> Best Regards,
> Helmut
> Moderator of the LTspice user group
>
>
>
>
   Well, in a sense you are correct in labellling be as a beginner; i
rarely use SPICE, but that useage has covered over 30 years.
   When one models a simple voltage follower, with the NI input half way
between the poser supplies for the op-amp, one expects it to work, and
not give a cryptic square root error.
   Furthermore, replacing the model used to a different one (eg replace
the call from the AD8605 to the AD8614 (and changing *nothing* else) and
have it work begs the question: what is wrong with the AD8605 model?
   The same can be said about the models for the LM324; the TI model
works and the NatSemi does not.

   And speaking of bad models that DO "work", the AD8614 is rather poor
(from the .OUT file):

  .OPTIONS ACCT LIST NODE OPTS NUMDGT=6 RELTOL=0.00001 NOPAGE
  .TEMP 27
  .LIB ANLG_DEV.LIB ; most rail-to-rail opamps die with error
  .DC VBAT 4.499 4.501 0.001
   VBAT 01 00 DC 4.5
   VSET 10 00 0.209171
   VIN 05 00 0.018051
   R2 05 07 18.4K
   R3 10 08 18.4K
   R4 09 07 100K
  * NI I OUT
   XAMP2 08 07 01 00 09 AD8614/AD
  .PRINT DC V(05) V(07) V(08) V(09)
  .PLOT DC V(05) V(07) V(08) V(09)
  .SAVE

        V(5) V(7) V(8) V(9)

      1.80510E-02 2.22746E-01 2.21746E-01 1.26487E+00

   Look at the poor results: large input currents, large Vos. Almost
useless; certainly not representative of the part.