Subject: RE: Number of Significant Digits
From: Lynne Green (lgreen@cadence.com)
Date: Wed Feb 27 2002 - 11:27:35 PST
If the digits are random noise (whether measurement or simulation),
then they should be dropped. Random fluctuations in insignificant
digits cause slope changes in the tables, which can degrade simulation
run times. For example, in SPICE, you can print out 8 digits, but
probably only 3 or 4 are meaningful if you use the default options for
tolerances, so everything past the 4th digit should be dropped.
Exponents should be kept reasonable. For example, including the
diode's series resistance in the diode model in SPICE will limit most
clamp currents to less than 10A. In measurements, the clamp diode
will physically break down (open or short, depending on the failure
mechanism). If neither of these limits the current in your model, then
as a last resort you can simply replace all currents above 100A with
the value 100A. SiQual has reported on the reflector that this should
not cause any problems in simulation, as long as the model is accurate
in the physically-real operating voltage range.
How you combine digits and exponents is up to you, since the tools
convert table entries to floating point numbers (float/double/...).
- Lynne
-----Original Message-----
From: Lewis, Tony L [mailto:tony.l.lewis@intel.com]
Sent: Wednesday, February 27, 2002 8:57 AM
To: 'ibis@eda.org'
Subject: Number of Significant Digits
Hello Everyone,
I am wondering if there are any rules or common methods for deciding
how many significant digits should be used when creating an IBIS model?
Thank you,
Tony Lewis
Desktop Platform Group
Email: <mailto:tony.l.lewis@intel.com>
This archive was generated by hypermail 2b28 : Wed Feb 27 2002 - 11:45:02 PST