International Business Machines Corporation

Globalfoundries Inc.

Vikram Iyengar

Gary Douglas Grise

Peter Anton Habitz

David E Lackey

Jeanne P Bickford

Jinjun Xiong

Mark R Taylor

Chandramouli Visweswariah

Vladimir P Zolotov

Steven Frederick Oakland

Brian A Worth

Pamela Sue Gillis

David James Hathaway

Eric A Foreman

David W Milton

Pascal A Nsame

Monty Montague Denneau

Phillip J Nigh

Rahul K Nadkarni

Douglas E Sprague

Robert W Bassett

Kenneth Pichamuthu

Malede Wondmagegne Berhanu

Andrew Ferko

Animesh Khare

Douglas C Pricer

Konda R Baalaji

Operating speeds of integrated circuit devices are tested to establish maximum and minimum frequency at maximum and minimum voltage. The devices are sorted into relatively-slow and relatively-fast devices to classify the devices into different voltage bins. A bin-specific voltage limit is established for each of the voltage bins needed for core performance at system use conditions. The bin-specific voltage limit is compared to core minimum chip-level functionality voltage at system maximum and minimum frequency specifications. The method correlates system design evaluation of design maximum and minimum frequency at design maximum and minimum voltage conditions with evaluation of tested maximum and minimum frequency at tested maximum and minimum voltage conditions. A chip-specific functionality voltage limit is established for the device. Initial system voltage for all devices from a voltage bin is set at a greater of the bin-specific voltage limit and the chip-specific functionality voltage limit consistent with the evaluation conditions.

Method and structure for multi-core chip product test and selective voltage binning disposition

Methods and systems for qualifying a single cell with product path delay analysis are provided. A method includes designing a product using a model from an initial test site. The method also includes creating performance path tests for one or more paths on the product. The method further includes measuring performance path parameters of the product. The method includes determining that the measured performance path parameters match predicted performance path parameters.

Systems and methods for single cell product path delay analysis

Disclosed are embodiments of a method, system and computer program product for optimizing integrated circuit product yield by re-centering the manufacturing line and, optionally, adjusting wafer-level chip dispositioning rules based on the results of post-manufacture (e.g., wafer-level or module-level) performance path testing. In the embodiments, a correlation is made between in-line parameter measurements and performance measurements acquired during the post-manufacture performance path testing. Then, based on this correlation, the manufacturing line can be re-centered. Optionally, an additional correlation is made between performance measurements acquired during wafer-level performance testing and performance measurements acquired particularly during module-level performance path testing and, based on this additional correlation, adjustments can be made to the wafer-level chip dispositioning rules to further minimize yield loss.

Integrated circuit product yield optimization using the results of performance path testing

Aspects of the invention provide for reducing power consumption during manufacturing testing of an IC. In one embodiment, aspects of the invention include a method for reducing power consumption during a manufacturing test of an integrated circuit (IC), the method including: providing a plurality of domains, each domain associated with a clock phase; grouping, based on each domain, a first plurality of scan chains into a first test group; grouping, based on each domain, a second plurality of scan chains into a second test group, wherein the grouping of the first test group and of the second test group includes determining which domains can be tested simultaneously; and performing the manufacturing test of the IC.

Reducing power consumption during manufacturing test of an integrated circuit

A method and system is provided for automatically generating valid at speed structural test (ASST) test groups. The method includes loading a netlist for an integrated circuit into a processor. The method further includes determining a plurality of clock domain crossings between a plurality of clock domains within the integrated circuit. The method further includes generating a first test group. The method further includes adding a first clock domain of the plurality of clock domains to the first test group. The method further includes adding a second clock domain of the plurality of clock domains to the first test group when the second clock domain does not have a clock domain crossing into the first clock domain.

Automatic generation of valid at-speed structural test (ASST) test groups

A method of test path selection and test program generation for performance testing integrated circuits. The method includes identifying clock domains having multiple data paths of an integrated circuit design having multiple clock domains; selecting, from the data paths, critical paths for each clock domain of the multiple clock domains; using a computer, for each clock domain of the multiple clock domain, selecting the sensitizable paths of the critical paths; for each clock domain of the multiple clock domain, selecting test paths from the sensitizable critical paths; and using a computer, creating a test program to performance test the test paths.

Test path selection and test program generation for performance testing integrated circuit chips

A method of grouping clock domains includes: separating a plurality of test clocks into a plurality of domain groups by adding to each respective one of the plurality of domain groups those test clocks that originate from a same clock source and have a unique clock divider ratio; sorting the domain groups in decreasing order of size; and creating a plurality of parts by adding the respective one of the plurality of domain groups to a first one of the plurality of parts in which already present test clocks have a different clock source, and creating a new part and adding the respective one of the plurality of domain groups to the new part when test clocks present in the respective one of the plurality of domain groups originate from a respective same clock source and have a different clock divider ratio as test clocks present in all previously-created parts.

Clock edge grouping for at-speed test

Disclosed are embodiments of a method, system and computer program for optimizing system yield based on the results of post-manufacture integrated circuit (IC) chip performance path testing. In these embodiments, a correlation is made between IC chip performance measurements, which were acquired from IC chips specifically during post-manufacture (i.e., wafer-level or module-level) performance path testing, and system performance measurements, which were acquired from systems that incorporate those IC chips previously subjected to performance path testing. Based on this correlation and a target system performance value, a post-manufacture (i.e., wafer-level or module-level) chip dispositioning rule can be adjusted to optimize system yield (i.e., to ensure that subsequently manufactured systems which incorporate the IC chip meet the target system performance value). Optionally, simulation of such processing can be performed during design of the IC chip for incorporation into the system in order establish the initial chip dispositioning rule in the first place.

System yield optimization using the results of integrated circuit chip performance path testing

A method and system for dispositioning integrated circuit chips. The method includes performing a performance path test on an integrated circuit chip having one or more clock domains, the performance path test based on applying test patterns to selected sensitizable data paths of the integrated circuit chip at different clock frequencies; and dispositioning the integrated circuit chip based on results of the performance path test.

Disposition of integrated circuits using performance sort ring oscillator and performance path testing

A system and method for detecting hold path faults in an integrated circuit is provided in exemplary embodiments. These exemplary embodiments introduce a method of identifying data paths within an integrated circuit with statistically the highest timing slack among the data paths within the integrated circuit that cover the entire process space of the circuit. By identifying these paths (i.e., shortest data paths), a robust test pattern can be generated that directly tests for hold path faults on short data paths within the integrated circuit using one functional clock pulse.

Method to test hold path faults using functional clocking

Disclosed are embodiments of a system and method for automatically selecting and generating test patterns for an at-speed structural test of an integrated circuit device. Specifically, a test pattern generation pass is started and proceeds until the 'knee' of the simulated test coverage curve is observed. Next, the test patterns are optionally reordered and some are removed. Then, another test pattern generation pass is started. The process is repeated iteratively until some predetermined final stopping criterion is met. By performing multiple test pattern generation passes and reducing the number of available test patterns that can be generated with each pass, the method exploits the initial increase in the test coverage curve inherent in each pass and limits the overall test pattern count.

System and method for automatically generating test patterns for at-speed structural test of an integrated circuit device using an incremental approach to reduce test pattern count

A method of hold fault modeling and test generation. The method includes first modeling a fast-to-rise and a fast-to-fall hold fault for a plurality of circuit nets. Testing a fast-to-rise hold fault is accomplished by: setting up a logic value on each of the plurality of circuit nodes to 0; transitioning each of the plurality of circuit nodes from 0 to 1 with a single clock pulse; and determining if at least one downstream node was inadvertently impacted by the transitioning from 0 to 1. Testing a fast-to-fall hold is accomplished by: setting up a logic value on each of the plurality circuit nodes to 1; transitioning each of the plurality of circuit nodes from 1 to 0 with a single clock pulse; and determining if at least one downstream node was inadvertently impacted by the transitioning from 1 to 0.

Hold transition fault model and test generation method

A multiprocessor system comprising a plurality of processors is disclosed. The plurality of processors includes a first processor including first monitor on-chip and a second processor including a including a second monitor on-chip. The first monitor on-chip is configured to measure load on the second processor and the second monitor on-chip is configured to measure load on the first processor. The first monitor on-chip is configured to cause the second monitor on-chip to perform a self-test on the second processor if the load on the second processor is below a second processor load threshold value and the second monitor on-chip is configured to cause the first monitor on-chip to perform a self-test on the first processor if the load on the first processor is below first processor load threshold value.

Online multiprocessor system reliability defect testing

Disclosed are embodiments of a clock generation circuit, a design structure for the circuit and an associated method that provide deskewing functions and that further provide precise timing for both testing and functional operations. Specifically, the embodiments incorporate a deskewer circuit that is capable of receiving waveform signals from both an external waveform generator and an internal waveform generator. The external waveform generator can generate and supply to the deskewer circuit a pair of waveform signals for functional operations. The internal waveform generator can be uniquely configured with control logic and counter logic for generating and supplying a pair of waveform signals to the deskewer circuit for any one of built-in self-test (BIST) operations, macro-test operations, other test operations or functional operations. The deskewer circuit can selectively gate an input clock signal with the waveform signals from either the external or internal waveform generator in order to generate the required output clock signal.

Growing community of inventors

Pittsburgh, PA, United States of America

Vikram Iyengar