Warning! This is the archived 1999 Fabweb site! Here is the latest site
The Integrated Circuit Characterization and Analysis Program - ICCAP
Manufacturer: Hewlett-Packard
CONTENTS:
Related files:
The Integrated Circuit Characterization and Analysis Program from Hewlett-
Packard will help you collect performance data and find optimized
models for the devices you make "from scratch". These models will be
useable by most implementations of SPICE, thus connecting design
methods utilizing SPICE taught in several other courses with
physical device fabrication.
You will also use ICCAP to
gather data without becoming bogged down in the details of operating the instruments.
This should allow you to focus more on understanding the fabrication process and the
devices themselves.
Since the overwhelming majority of ECE 344 students are ECE majors, testing
will not be as "cook-booked" as the chemistry type procedures which were used to create
the devices. We have prepared some basic, "safe" models and setups for you to use, but in
almost every case, there are modifications which can, and therefore should, be made to the
setups in order to exhibit as much useful information about your particular device as
possible. The object will be to apply your knowledge of the devices to make these
adjustments. For example, don't settle for plots in which all the "action" happens
along one edge.
THE STRUCTURE OF ICCAP
ICCAP was written in C++ and takes advantage of the power of object oriented
programming. At the top of the hierarchy is the MODEL which is intended to be a
collection of measurement setups, processing variables and equivalent circuit parameters.
In short, everything necessary to determine a SPICE model for an integrated circuit.
Several Devices Under Test (DUTs) with their own set of variables often make up a
circuit. We will utilize the DUT level to separate fundamental device concepts or
components. For example, a BJT has two distinct pn diodes within it which will be
individually characterized as separate DUTs.
Each DUT level can have any number of measurement setups (for all practical
purposes) since many instrument and probe configurations may be needed to determine
each DUT's parameters (equivalent circuit). Each setup can also have its own variables
and parameters. Equations can be used to EXTRACT equivalent circuit parameter values
from measurements. Such calculations can come close to creating a good SPICE model,
but they are inevitably based on approximations. SIMULATIONS (by SPICE) of the
SPICE model based on the extraction results or default values can be compared to the real
data for some or all of the measured data sets and OPTIMIZED for best "fit" by iterative
techniques (Levenberg-Marquardt algorithm). SPICE will be used by ICCAP several
times during an optimization to generate the predicted results from the
measurement setup and current SPICE parameters. Iterative adjustment of the
parameters between each SPICE call improves the "fit" with the actual measured
data set. The user determines when the aggreement between actual and simulated
data (from the model) is good enough. The resulting model is called "optimized",
as opposed to "optimal".
This program was designed to help the circuit designer take full advantage of ever
improving IC processing capabilities. Interfaces to TCAD's implementations of Supreme
and Pices, processing and physical device simulation programs, can similarly provide the
process engineer with feedback between simulation and real measurement results.
Someday you may be working with versions of these programs and so will ECE 344
students. Perhaps on future PCs!
A vauge new type of specialization has been forming within the broad scope of
engineering. The kind of person likely to be in Electrical Engineering is often well suited
to excel in this unnamed disipline. Skill at utilizing new computers, software and instruments
in a minimal learning time, together with a solid understanding of the chemistry
and physics involved in a marketable
product can help one make significant contributions to the large team it takes to
manufacture competitive products.
Modern software packages such as ICCAP trade off program size, complication,
and speed for intuitive ease of use. Consequently, there are many ways you can
accomplish a given task. Since navigating within menu driven software is quickly
becoming an essential skill in and of itself for engineers, you will basically be turned loose
on the software after a demonstration by your instructor. Below are some of the basics
you will see in the demo. Excerpts from the ICCAP manual
and training course viewgraphs are available to *.uiuc.edu machines.
Complete documentation is available in the lab.
X-Windows must be running before ICCAP can be loaded. Type x or
x11start to start X-Windows if necessary. Like most tasks pertaining to ICCAP,
there are several methods to start it. The recommended method in the ECE 344 lab is to use the motif
window manager root menu. ICCAP will appear as an option when you "click left" and
hold anywhere in the X-server background (outside windows) area. Releasing on the
ICCAP sub menu item "load my_models" will load ICCAP and the basic ECE 344 model
set. Your my_models.mdl file will be a copy of the basic_344.mdl file in the
linked_models directory below your iccap directory. The iccap directory will
be set up the first time you run iccap.
If you are using a
remote X-Window session, you will have
to type iccap in the window to the remote machine.
The Engineering WorkStations (EWS) made by Hewlett-Packard can now
run IC-CAP locally, but you must use ftp to move the model files between
your accounts. Be careful not to overwrite your work.
This is the basic procedure, but you should always refer to the
World Wide Web help file for the specific model for extra details. There will normally
be a macro called HELP which will print the URL when executed.
Therefore, you will probably want to be running Mosaic or Netscape too.
It will help if you study the model by running IC-CAP before you come to
lab to make measurements. There are almost always a few things
you should be able find which should be changed to better suit your devices.
- Start IC-CAP and load the model file (usually this will be my_models.mdl).
- Probe
a device. The probe assignments will be in the help file, but you can
also determine them from the setups. Note the
Device Cell number.
- Acquire measured data for each setup or DUT:
- Check the cable connections and the coax/triax switch positions. Note
that the CM(L) and possibly the CM(H) tees may have to be moved for
some measurements.
- Select Dispaly Plots for each setup or DUT.
- Select Measure for each setup or DUT. Note: the first capacitance
measurement will require you to raise the probes for a calibration.
- Decide if the plots look reasonable. For the SPICE models there should
be examples that you can load for comparison, but be sure to delete
them before submitting your my_models.mdl file.
- If you have trouble getting reasonable characteristics, try the following:
- Adjust the probes. Often just raising each probe tip and re-contacting
the aluminum contact pad can help.
- Check the switches on the prober.
- Check the xterm that started IC-CAP to see if there were any errors listed.
- scrutinize the setup to see if changes are needed.
- Try a device from a cell in a much different location. There may have
been a localized fabrication problem.
- Verify continuity to the probe tips. You may get your instructor to help
if this is necessary.
- Final data should normally be acquired with the microscope light off
and the prober cover down. Don't waste time lowering the lid
if a plot doesn't look promising with the lid up first.
- SAVE YOUR MODEL FILE OFTEN!!!!
- Save it to the normal file name (usually my_models.mdl) with
Model List | Write All Models to File every time you get
additional good data.
- Occasionally save it to a backup filename (such as my_models.bu.mdl).
- Occasionally
submit it to your instructor (who can then restore if you need it back).
- Perform a write all models to a file every time you get
additional good data.
- For most efficient use of lab time, complete all the measurements
before continuing.
- Completely fill in or verify the entries under:
- Parameter Set | Edit.
- Model | Edit Variables.
- Edit and Edit Variables for each DUT name.
- The lab report may require new calculations to be made and plotted.
- SAVE YOUR MODEL FILE OFTEN!!!!
Not all models are set up to do extractions, simulations, and optimization.
Those that are set up for full modeling will be used in the following general
manner. Check with the help file for the specific model for further detail.
- Verify that the following are filled in properly:
- all measured data sets.
- Entries under Parameter Set | Edit.
- Entries under Model | Edit Variables.
- Entries under Edit and Edit Variables for each DUT name.
- Bring the xterm that launched IC-CAP to the top.
- Begin with the top-most setup and proceed in order down to the bottom setup
doing the following for each setup:
- Open the plots.
- Perform an Extract. Extractions specified in the "Transform" section
of the setup will be run. These will calculate some of the Parameters for the model.
- Check the xterm for messages.
- Perform a Simulate. This will call SPICE with the current parameters and
measurment stimuli. The results will be plotted using dashed lines.
- Check the xterm for messages.
- Perform an Optimize. This will methodically, adjust the parameter set
to seek better agreement between the actual measured data and the simulated data.
- Check the xterm for messages.
- Repeat the optimizations if you feel it's necessary.
- If you have problems, you may want to select Open Simulation Debugger
on the Utilities menu and watch for messages as you retry.
- Close the plots.
- SAVE YOUR MODEL FILE OFTEN!!!!
It will, of course, help a great deal if you understand each device's operation well
enough to know what to expect from each measurement and how to make the most of a
given setup by adjusting START and STEP values, for instance. Reviewing basic device
operation in your ECE340 text may help you make these adjustments intelligently and
without destroying the device. We expect you to have retained a fair amount of
knowledge from that course.
The details of operating ICCAP are too complicated and yet intuitive to make it
worthwhile explicitly describing each step. Besides, it would be incredibly boring and
after all, almost all of you are ECE students. The demo will be worth at least a thousand
words. If you happen to miss the demo for your section, try to attend the one for another
section (check the calendar in the gown room for dates).
You are to measure 3 each of 3 devices. Depending on how the overall class
progresses, another round of more advanced testing will be made later. Since your
devices are likely to be similar to each other, "copy" the Models after you are satisfied with
the setups and measurements. Name the copies with numbers in place of the digits "344". For
example, copy the first completed PMOSFET to pmos3_1 and the second one to pmos3_2 and so forth.
Be sure to perform Model <modelname>" "edit variables" to enter the location, etc. of the device just
measured. Use the *_344.mdl models to gather new measurements on the next device type.
It is suggested that you use the probe assignments listed in the Macro window of
each device model. This will minimize the amount of table entry modifications that will
need to be made.
I like to edit this file with the following command:
vuepad /ece344/Software/ICCAP/Instructions.html &
This screen was created by Mike Fitzsimmons - U of Illinois ECE Dept. -
mikef@uiuc.edu
E-mail comments and suggestions to ece344@uiuc.edu or use the
FEEDBACK FORM.
Warning! This is the archived 1999 Fabweb site! Here is the latest site
