AT-speed computer model testing methods

Details AT-speed computer model testing methods AT-speed computer model testing methods

1999-01-27

2001-12-04

Teska, Kevin J. (Department: 2123)

Data processing: structural design, modeling, simulation, and em

Simulating electronic device or electrical system

C703S014000, C716S030000, C702S117000, C714S738000

Reexamination Certificate

active

06327556

ABSTRACT:

BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates generally to integrated circuits, and more particularly to methods for performing computer model at-speed testing of integrated circuit designs.
2. Description of the Related Art
Testing integrated circuits that are ultimately fabricated onto silicon chips has over the years increased in complexity as the demand has grown, and continues to grow for faster and more densely integrated silicon chips. In an effort to automate the design and fabrication of circuit designs, designers commonly implement hardware descriptive languages (HDL), such as Verilog, to functionally define the characteristics of the design. The Verilog code is then capable of being synthesized in order to generate what is known as a “netlist.” A netlist is essentially a list of “nets,” which specify components (know as “cells”) and their interconnections which are designed to meet a circuit design's performance constraints. The “netlist” therefore defines the connectivity between pins of the various cells of an integrated circuit design. To fabricate the silicon version of the design, well known “place and route” software tools that make use of the netlist data to design the physical layout, including transistor locations and interconnect wiring.
When testing of the digital model, various test vectors are designed in order to test the integrated circuit's response under custom stimulation. For example, if the integrated circuit is a SCSI host adapter chip, the test vectors will simulate the response of the SCSI host adapter chip as if it were actually connected to a host computer and some kind of peripheral device were connected to the chip. In a typical test environment, a test bench that includes a multitude of different tests are used to complete a thorough testing of the chip. However, running the test vectors of the test bench will only ensure that the computer simulated model of the chip design will work, and not the actual physical chip in its silicon form.
To test a silicon chip
12
after it has been packaged, it is inserted into a loadboard
14
that is part of a test station
10
, which is shown in FIG.
1
A. Although the model of the chip design was already tested using the test vectors of the test bench, these test vectors are not capable of being implemented in the test station
10
without substantial modifications, to take into account the differences between a “model” and a “physical” design. In the prior art, the conversion of a test model test vector into test vectors that can actually be run on the test station
10
required a very laborious process that was unfortunately prone to computer computational errors as well as human errors. Of course, if any type of error is introduced during the generation of the test vectors that will ultimately be run on the silicon chip
12
, the testing results generated by the test station
10
would indicate that errors exist with the part, when in fact, the part functions properly. This predicament is of course quite costly, because fabrication plants would necessarily have to postpone release of a chip until the test station indicated that the part worked as intended.
As mentioned above, the prior art test vector generation methodology was quite laborious, which in many circumstances was exacerbated by the complexity of the tests and size of the chip being tested. The methodology required having a test engineer manually type up the commands necessary to subsequently generate a “print-on-change” file once executed using Verilog. Defining the commands for generating the print-on-change file includes, for example, typing in the output enable information for each pin, defining pin wires, setting up special over-rides for power-on reset pins, etc. At this point, the print-on-change file would then be generated using a Verilog program, which in turn uses the commands generated by the test engineer.
In addition to manually producing these commands, a separate parameter file having timing information is separately produced in a manual typing-in fashion by the engineer. The generated print-on-change file and the parameter file are then processed by a program that is configured to produce a test file, which is commonly referred to as an AVF file. However, the production of the AVF is very computationally intensive because the generated print-on-change file can be quite large. The size of the print-on-change file grows to very large sizes because every time a pin in the design changes states, a line of the print-on-change file is dumped. Thus, the more pins in the design, more CPU time is required to convert the print-on-change file into a usable AVF file. In some cases where the test is very large or complex, the host computer processing the print-on-change file is known to crash or in some cases lock-up due to the shear voluminous amount of data.
Unfortunately, as mentioned above, the generated AVF file may have defects, such as timing errors, which may translate into errors being reported by the test station
10
. The problem here is that the test station
10
will stimulate the part differently than the stimulation designed for the digital version. This problem therefore presents a very time consuming test and re-test of the part by the test station
10
. When re-testing is performed, many modifications to the parameter file, containing timing information, are performed in an effort to debug errors with the AVF file. Although some parts are in fact defective in some way, the test engineer is still commonly required to re-run the tests to determine whether the errors are due to a defective AVF file or the physical device.
Furthermore, most conventional testing techniques require that the physical part actually be placed into the physical test station. At that time, the physical test station only allow test engineers to test parts at speeds that are a fraction of the true operating speeds of the integrated circuit design part. As a result, many times integrated circuit designs that are believed to work properly under test conditions will fail once they are exposed to their true functional operating speeds. This of course increases the cost of developing, testing, re-testing, and redesign of integrated circuit devices.
In view of the foregoing, there is a need for methods and computer readable media for testing integrated circuit designs via a computer model that enables testing at speeds that resemble actual functional operating speeds in order to reduce testing uncertainties, customer returns, and thereby increase customer satisfaction.
SUMMARY OF THE INVENTION
Broadly speaking, the present invention fills these needs by providing techniques for testing integrated circuit design computer models at speeds that resemble those of normal operating physical integrated circuit designs. The present invention further discloses test vector conversion techniques that take into account temperature and silicon variations which lead to variations in operating speeds. The converted test vectors are thus capable of being run on both integrated circuit design models having slow speed characteristics and those having fast speed characteristics without producing erroneous test data. It should be appreciated that the present invention can be implemented in numerous ways, including as a process, an apparatus, a system, a device, a method, or a computer readable medium. Several inventive embodiments of the present invention are described below.
In one embodiment, a computer implemented method for performing testing of a computer model of an integrated circuit design is disclosed. The method includes generating a first AVF test file for a first integrated circuit design having slow characteristics and generating a second AVF test file for a second integrated circuit design having fast characteristics. The method then proceeds to comparing test file parameters from the first AVF test file and the second AVF test file. After the comparing, the method moves to generating a modified AVF test file that

Affiliated with

Also associated with

Rating

AT-speed computer model testing methods does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with AT-speed computer model testing methods, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and AT-speed computer model testing methods will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2598553