Re: Diode Modeling in SPICE

Silvia,

It's been a while since I did this work, so my memory might be a little rusty...

I remember extracting Is and n parameters from lab data for bipolar transistors
a few years ago; this is essentially the same measurement. We put the data on a
log-log plot. From the slope you can get n. The y-axis intercept gives you Is.

Start with the diode equation:

I = Is * {exp [(Vq)/(nkT)] - 1}

If you assume the -1 is negligible and take the natural log of both sides,

ln(I) = ln(Is) + (qV/nkT)

You'll also need some way to measure T. For a nice graph, see "The PN Junction
Diode," by Gerold Neudeck, Addison-Wesley, 1989, figure 4.7, section 4.2, page
83 in my copy.

I applaud you for your efforts. I wish all my model providers did this kind of
work!

(Shameless plug for a committee I've been working on: you can find some useful
information on making accurate IBIS datasheets and documenting your results.
Look under "accuracy" on the IBIS web page. It's called the IBIS Accuracy Spec,
although nobody is interested in using it as a spec just yet. It has some great
information on what kind of measurements you can make to validate the accuracy
of the models you provide.)

Greg Edlund
Advisory Engineer, Critical Net Analysis
IBM
3650 Hwy. 52 N, Dept. HDC
Rochester, MN 55901
gedlund@us.ibm.com

"Montoya, Silvia" <Silvia.Montoya@actel.com> on 05/13/99 06:50:24 PM

To: "'ibis-users@eda.org'" <ibis-users@eda.org>
cc: "Montoya, Silvia" <Silvia.Montoya@actel.com> (bcc: Gregory R
      Edlund/Rochester/IBM)
Subject: Diode Modeling in SPICE

Hello IBIS model makers !

   I have a 5V CMOS I/O tristate-able buffer I'm trying to model. The IBIS
curves generated by s2ibis2 are a very good match
   to measured data between 0.7v up to about 5.9v. But between -5V to 0.7v
and 5.9v to 10v the curves deviate quite a bit.
   I've disabled the parasitic diodes in the nmos and pmos device models (
IS=0 ) and added n+ and p+ parasitic diodes to the
   netlist. I've been playing with the n+ and p+ diode model (level 3)
parameters ( n, js, ivb, area, pj ) trying to match SPICE
   results to measured data with limited success.

   Are there any suggestions for other model parameters to tweek, or maybe
there is something going on with the s2ibis2 SPICE
   sim limits/parameters I'm not aware of ?

  Thanks,
  The time you spend helping me out is greatly appreciated !

   Silvia Montoya
  email : silvia.montoya@actel.com
Received on Fri May 14 06:41:59 1999