IP Library Granted Patent US 12,658,280
Granted Patent B2
US 12,658,280 · App. 18/648,655 · Granted Jun 16, 2026

Method for performing aging test on semiconductor used for neural network

Inventor: Lok Won Kim (Seongnam-si, KR)
Assignee: DEEPX CO., LTD.
G11C29/78G05B19/41875G06F11/26G11C29/10G05B2219/32368G05B2219/45031
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,658,280
App. No.
18/648,655
Granted
Jun 16, 2026
Kind
B2
Abstract

Provided is a method for performing an aging test on a neural processing unit (NPU) with a capability of a runtime test. The method may comprise: performing an aging test on the NPU which comprises a plurality of functional components. The plurality of functional components may comprise at least one memory and plural processing elements. The performing of the aging test may include: performing a scan test on the NPU to verify whether at least one functional component in the NPU is defective or not; and performing a memory test on the at least one memory. At least one of the scan test and the memory test may be repeatedly performed to put a stress on the NPU for the aging test. The aging test may be repeated by a predetermined number.

Claims (64)

1 . A method for performing an aging test on a neural processing unit (NPU) with a capability of a runtime test, comprising:

performing an aging test on the NPU having a plurality of functional components,

wherein the plurality of functional components comprise at least one memory and a plurality of processing elements,

wherein the performing of the aging test includes:

performing a scan test on the NPU to verify whether at least one functional component of the plurality of functional components in the NPU is defective or not; and

performing a memory test on the at least one memory,

wherein at least one of the scan test and the memory test is repeatedly performed to put a stress on the NPU for the aging test, and

wherein the aging test is repeated by a predetermined number.

2 . The method of claim 1 ,

wherein the memory test includes a Memory Built-In Self-Test (M-BIST).

3 . The method of claim 1 ,

wherein the aging test is performed when the NPU is in a test mode.

4 . The method of claim 3 ,

wherein if the aging test is finished, the NPU is switched from the test mode to a runtime mode.

5 . The method of claim 1 ,

wherein the scan test includes:

forming one or more scan chains by connecting a plurality of flip-flops to each other in each functional component,

injecting a test input into at least one of the plurality of flip-flops, and

acquiring a test result from operations of combinational logics of the at least one of the plurality of flip-flops to analyze whether each functional component is defective or normal.

6 . The method of claim 1 ,

wherein the performing of the aging test further comprises: performing a function test on the NPU.

7 . The method of claim 6 ,

wherein the performing of the function test comprises:

inputting a test input data into the NPU which operates a first artificial neural network (ANN) model pre-set for the function test; and

checking whether the NPU operates normally or erroneously by verifying whether a test result acquired from the NPU into which the test input data is inputted matches a preset test result.

8 . The method of claim 1 ,

wherein the performing of the scan test on the NPU comprises:

selecting, as a component under test (CUT), the at least one functional component of the plurality of the functional components, by an in-system component tester (ICT); and

performing the scan test on the at least one functional component selected as the CUT, by the ICT.

9 . The method of claim 1 ,

wherein the NPU further comprises: one or more wrappers, each being connected to the at least one functional component of the plurality of functional components.

10 . The method of claim 9 ,

wherein the one or more wrappers allow the at least one functional component to be isolated during the scan test.

11 . The method of claim 9 ,

wherein the one or more wrappers prevent the at least one functional component from being disrupted by any access during the scan test.

12 . The method of claim 11 ,

wherein the any access is requested by another functional component of the plurality of functional components in the NPU.

13 . A method for performing an aging test on a system on chip (SoC) with a first capability for an artificial neural network and a second capability of a runtime test, the method comprising:

performing an aging test on the SoC having a plurality of electronic circuitry having a plurality of functional components,

wherein the plurality of functional components comprise at least one neural processing unit (NPU), at least one memory and a system bus,

wherein the performing of the aging test includes:

performing a scan test on the SoC to verify whether at least one functional component of the plurality of functional components in the SoC is defective or not; and

performing a memory test on SoC,

wherein at least one of the scan test and the memory test is repeatedly performed to put a stress on the SoC for the aging test, and

wherein the aging test is repeated by a predetermined number.

14 . The method of claim 13 ,

wherein the memory test includes a Memory Built-In Self-Test (M-BIST).

15 . The method of claim 13 ,

wherein the aging test is performed when the SoC is in a test mode.

16 . The method of claim 15 ,

wherein if the aging test is finished, the SoC is switched from the test mode to a runtime mode.

17 . The method of claim 13 ,

wherein the scan test includes:

forming one or more scan chains by connecting a plurality of flip-flops to each other in each functional component,

injecting a test input into at least one of the plurality of flip-flops, and

acquiring a test result from operations of combinational logics of the at least one of the plurality of flip-flops to analyze whether each functional component is defective or normal.

18 . The method of claim 15 ,

wherein the performing of the aging test further comprises: performing a function test on the SoC.

19 . The method of claim 18 ,

wherein the performing of the function test comprises:

inputting a test input data into the SoC which operates a first artificial neural network (ANN) model pre-set for the function test; and

checking whether the SoC operates normally or erroneously by verifying whether a test result acquired from the SoC into which the test input data is inputted matches a preset test result.

20 . The method of claim 13 ,

wherein the SoC further comprises: one or more wrappers, each being connected to the at least one functional component of the plurality of functional components.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Apr 29, 2024
From: KIM, LOK WON
To: DEEPX CO., LTD.
Reel/Frame 067250/0185 →
Priority Claims (2)
KR 10-2024-0031476 · Mar 5, 2022 · national
KR 10-2022-0054878 · May 3, 2022 · national
Continuity (3)
Continuation In Part 18193313 · Mar 30, 2023
Continuation 17886463 · Aug 12, 2022
Related Publication 20240274225A1 · Aug 15, 2024
References Cited (38)
US 6357020B1 · Bohizic et al. · 2002 [cited by applicant]
US 6571359B1 · Padwekar et al. · 2003 [cited by applicant]
US 11237894B1 · Baum et al. · 2022 [cited by applicant]
US 11263077B1 · Seznayov et al. · 2022 [cited by applicant]
US 11609269B2 · Whetsel · 2023 [cited by applicant]
US 20030125908A1 · Wynn et al. · 2003 [cited by applicant]
US 20030196146A1 · Padwekar et al. · 2003 [cited by applicant]
US 20120226942A1 · Gangasani et al. · 2012 [cited by applicant]
US 20120233514A1 · Patil et al. · 2012 [cited by applicant]
US 20140096097A1 · Prasanna et al. · 2014 [cited by applicant]
US 20150039950A1 · Monroe et al. · 2015 [cited by applicant]
US 20150115994A1 · Graas et al. · 2015 [cited by applicant]
US 20160349320A1 · Laisne et al. · 2016 [cited by applicant]
US 20170193154A1 · Sul · 2017 [cited by applicant]
US 20180357143A1 · Panambur et al. · 2018 [cited by applicant]
US 20190088515A1 · Jeong et al. · 2019 [cited by applicant]
US 20190154757A1 · Lujan et al. · 2019 [cited by applicant]
US 20210018590A1 · Grau · 2021 [cited by examiner]
US 20210286693A1 · Alben et al. · 2021 [cited by applicant]
US 20220012164A1 · Grymel et al. · 2022 [cited by applicant]
US 20220100601A1 · Baum et al. · 2022 [cited by applicant]
US 20220101042A1 · Kaminitz et al. · 2022 [cited by applicant]
US 20220101043A1 · Katz et al. · 2022 [cited by applicant]
US 20220101887A1 · Han et al. · 2022 [cited by applicant]
US 20220103186A1 · Kaminitz et al. · 2022 [cited by applicant]
US 20220113889A1 · Park et al. · 2022 [cited by applicant]
US 20220171668A1 · Quigley et al. · 2022 [cited by applicant]
US 20220206068A1 · Kim · 2022 [cited by applicant]
US 20220206978A1 · Kim · 2022 [cited by applicant]
US 20240211740A1 · Jeong · 2024 [cited by examiner]
CN 112506690A · 2021 [cited by applicant]
CN 114689961A · 2022 [cited by applicant]
EP 1923788A2 · 2008 [cited by applicant]
WO 2005029329A2 · 2005 [cited by applicant]
WO 2016191013A1 · 2016 [cited by applicant]
WO 2016200718A1 · 2016 [cited by applicant]
WO 2017017691A1 · 2017 [cited by applicant]
Lok-Won Kim et al., “Runtime Testability on Autonomous System”, IEEE Transaction on Reliability, Apr. 29, 2022. [cited by applicant]