Adopting the Right Solution for Embedded Memory Test

By Steve Pateras

Logic BIST Scheme for Intra-/Inter-clock-domain At-Speed Testing

By Ravi Apte, Ph.D.

The Role of Memory Built-in Self Test and Built-in Self Diagnostics in Today's Overall Test Strategy

By Luigi Ternullo

This issue's focus is Built-In Self Test. 
You can view and add to our existing list of Test Products/Services, Test Literature, Test Definitions, Test Vendors, containing "Built-In Self Test," or "BIST."

You will be linked directly to the listings you select below. After checking two references you will be asked to log in before you can continue.  If you forgot your password, it will be emailed to you.  If you have not yet registered, you will be able to at NO COST.  You can forward this Newsletter to a Friend using the button on the top right margin.  We do not disclose your data to anyone!

Come to a Three-Day Course

Random Vibration and Shock Testing, HALT, ESS, HASS

in Tinton Falls, NJ on June 12-14, 2007 

Come to a Three-Day Course

Design for Testability and for Built-In Self Test

by Louis Y. Ungar in Houston, TX on June 25-27, 2007 

 By Steve Pateras

Embedded memories represent a significant and growing percentage of today’s systems-on-a chip (SoCs) and as a result, memory built-in-self-test (BIST) and yield optimization solutions have gained significant importance.

In its simplest form, memory BIST consists of an on-chip engine placed next to each embedded memory that writes algorithmically generated patterns to the memory and then reads these patterns back to discover and log any defects. Over the years, memory BIST solutions have evolved to very sophisticated automation and intellectual property (IP) solutions to deal with today’s complex design and test challenges. Several important capabilities should be evaluated when choosing a memory BIST solution to ensure a cost-effective memory test strategy.

Design Automation

Many of today’s designs have literally hundreds of memories of different types and sizes spread throughout the chip. Adding test and repair capabilities to these memories can be a daunting task, requiring a comprehensive automation flow. With such a large number of memories on a chip, it becomes necessary to share BIST engines across multiple memories. The optimum allocation of BIST engines to memories depends on several factors, including the physical location of the memories, the clock domains within which the memories reside, the power ratings of the memories, the size and type of each memory, and the maximum desired test time. A key automation capability is the ability to take these factors into consideration and configure each BIST engine to support the memories allocated to it and create an optimal allocation of BIST engines to memories.  Having this kind of automation can reduce memory BIST integration time on large designs from days to hours.

Test Algorithm Programmability

Continued decreases in process feature sizes and associated increases in memory densities are resulting in a growing number of memory defect types. Many of these new defect mechanisms are difficult to predict and are therefore difficult to test. These defects are increasingly being discovered during the production testing of a device or, worse, during the analysis of field returns. This can result in significant quality and cost issues if the predetermined test algorithm used by the BIST engine does not detect a newly discovered defect type. In many cases, functional tests have to be added to the manufacturing test flow at significant cost. Because of this growing problem, some commercial memory BIST solutions now provide programmable BIST engines. With these engines it is possible to download (on the tester or in system) program code that implements an arbitrary memory test algorithm that can be applied to specific memories to test for new defect mechanisms.

Built-in Self Repair

To improve yields, embedded memories are being fitted with spare elements – spare rows, columns or both. These spare elements can be used to replace corresponding faulty elements and salvage the memory. The manufacturing flow for supporting a memory repair strategy can be expensive. During wafer sort, memories not only have to be tested, but all failure information must be extracted and analyzed to determine if the memory is reparable and how to repair it. Once the necessary repair information is calculated, the wafer containing the bad die must be moved to a laser repair station to blow the necessary fuses for each bad die. The wafer must then be retested to ensure all the repairs were successful. Alternatively, the good and repaired die can be sliced and packaged and sent to final test for verification. In the latter case, there will be additional (and expensive) final test fallout, as some unsuccessfully repaired die will have been packaged. Advanced commercial memory BIST solutions such as the ETMemory solution from LogicVision now support a fully embedded memory repair flow. This approach, referred to as built-in self-repair (BISR), can significantly reduce the complexities and costs described above. The most advanced BISR solutions test and permanently repair all defective memories in a chip using no external resources. Central to ETMemory is the concept of a programmable fuse pool. Electrical or programmable fuses are smaller than laser-based fuses and can be programmed without the need of any external equipment. Pooling of fuses is also becoming popular to reduce overhead. Because most memories will typically need little to no repair on any given die, sharing a pool of fuses for all memories allows for much better fuse utilization. Memories needing little to no repair require little to no fuse information to be stored.

To simplify the fuse data allotment, standard data compression techniques are used to implicitly allocate the necessary amount of fuse storage per memory. With ETMemory a fuse box controller performs on-chip management of a centralized programmable fuse pool. This controller, along with one or more BIST engines, performs all necessary activities for testing and repairing memories. In this solution the BIST engines are enhanced not only to test the memories, but also to analyze how to repair faulty ones. This capability is typically referred to as built-in repair analysis (BIRA). The solution also supports incremental repair. That is, the fuse data stored in the fuse pool can be used as a baseline on every power-on-reset. The BIST engines are then used to detect any new failures, and the combined fuse data is stored in a local BISR register to repair the memories. This approach allows for improving long term device and end-system reliability.  

Diagnostics

The ability to precisely and efficiently diagnose memory failures is a key component of understanding and correcting yield issues. A highly automated memory diagnostic approach is a key ingredient of an embedded memory strategy. The most advanced diagnostic solutions consist of both an IP and software component. The IP component refers to additional capabilities within the BIST engine for the logging of failure data. The most common of these is a “stop-on-error” capability. This enables the BIST engine to stop on the detection of each error so all pertinent failure information, such as the failing row and column addresses, can be scanned out of the engine and logged on the tester. Making use of this embedded diagnostic capability can be a significant challenge without appropriate automation. Correctly generating the necessary tester patterns to control all of the stop-on error activity within a specific BIST engine connected to a specific memory can be both time consuming and error prone. For this reason, a diagnostic solution that also incorporates control software running directly on the tester is highly beneficial. For example, the ETDiagnostic solution from LogicVision automatically handles all interactions with the BIST engines and processing of failure data. No test patterns need to be generated and no tester fail data output has to be processed or analyzed.

Conclusion

A number of aspects must be considered when choosing and implementing an embedded memory test and yield management solution. Adopting the right solution can result in significant cost and time-to-market savings.

More information on test and diagnosis of embedded memories can be found at www.memorybist.com and www.logicvision.com

By Ravi Apte, Ph.D.

1. Introduction

Logic Built-In Self-Test (BIST) schemes based on STUMPS structure use on-chip circuitry to generate test stimuli and analyze test responses, with little or no help from an ATE. The STUMPS (Self-Test Using a MISR and Parallel Shift register sequence generator) structure applies pseudo-random patterns generated by a PRPG (Pseudo-Random Pattern Generator) to a full-scan circuit in parallel and compacts the test responses into a signature with a MISR (Multiple-Input Signature Register). This approach has such advantages as simple test interface, better test quality, lower test cost, and higher reliability.

Due to its conceptual simplicity, logic BIST has been used successfully for many years for designs with a single clock or with multiple synchronous clock domains. This usage has been primarily based on the true value of logic BIST that is in realizing at-speed testing of high-speed and high-performance circuits. When designs contain multiple, high-speed, asynchronous clocks, most logic BIST schemes face many practical hurdles, i.e. test frequency manipulation, control complexity, implementation difficulty in minimizing clock skew etc. SynTest patented proprietary technology resolved these practical hurdles and TurboBIST-Logic product was introduced in 2001.

With usage of submicron technologies growing rapidly, need for at-speed testing is also becoming more acute. The most critical yet difficult part of logic BIST is how to detect intra-clock-domain faults and inter-clock-domain faults thoroughly and efficiently with a clocking scheme for proper capture of results. There have been three major at-speed timing control methods in existence. The launch-from-shift with capture alignment method aligns the rising edges for all capture pulses. The launch-from-shift with last-shift alignment method aligns the rising edges for all last-shift pulses. And the one-hot method conducts capture for one clock domain at a time. These methods, however, suffer from at least one the following issues: (1) the frequencies for all clock domains are required to satisfy certain relations, (2) some scan enable signals must be designed as a high-speed signal, (3) capture-disabling circuitry has to be added to prevent crossing clock-domain logic from affecting each other in capture mode, (4) control circuitry is complex, (5) only a synchronous circuit is targeted, and (6) test time is long.

This article introduces a new, powerful method using SynTest patented fundamental technology, with flexibility of the logic BIST scheme and easy of physical implementation it provides, to achieves true at-speed test quality for intra-clock-domain faults and inter-clock-domain faults in any multi-clock circuit.

SynTest Logic BIST Architecture

The BIST architecture, illustrated in Fig. 1, (shown for a 2 clock domain design) for testing the BIST-ready core consists of a TPG (Test Pattern Generator) for generating test stimuli, an input selector for providing pseudo-random or top-up ATPG patterns for the core-under-test, a TRA (Test Response Analyzer) for compacting test responses, a clock gating block for generating test clocks from original or functional clocks, and a BIST controller for coordinating the whole BIST operation. The self-test operation is initiated by asserting the Start signal, its end is indicated by the Finish signal, and its result is shown by the Result signal. When required, a standard IEEE 1149.1 Boundary-Scan interface is used for loading initialization and configuration data or for downloading internal states for fault diagnosis.

Fig. 1  Logic BIST Architecture. 

The BIST-ready core is “full-scan” circuit that must meet all scan design rules with additional circuitry for preventing bus conflicts at tri-state buses and for disabling asynchronous set/reset signals as well as false paths. It must also meet all BIST-specific design rules, such as for X-blocking and for test point insertion (TPI).

The TPG and TRA circuitry consists of PRPG-MISR pairs for each clock domain and avoids additional design efforts for clock skew management. Linear phase shifters, PS1 and PS2, (or space expanders, SpE1 and SpE2) are used to reduce the length of PRPGs, whereas space compactors, SpC1 and SpC2, are used to reduce the length of MISRs.

The test timing control circuitry consists of a BIST controller and a clock-gating block. The inputs to the clock-gating block are system clocks CK1 and CK2, which become CCK1 and CCK2 after going through some buffers. In addition, the clock-gating block is controlled by signals from the BIST controller to generate test clocks TCK1 and TCK2. The waveforms of TCK1 and TCK2, especially in capture mode, play a critical role in determining the test capability and physical implementation easiness of the logic BIST scheme.

1.1     Intra-Clock-Domain Fault Detection

Intra-clock-domain fault detection is relatively easy by using an ordered sequence of capture clocks for all clock domains in each capture window. For each clock domain, a single clock pulse is used to detect structural faults in low-speed testing, while two at-speed clock pulses are used to detect timing-related faults in at-speed testing. TurboBIST-Logic uses the double-capture scheme as it detects not only timing-related faults but also structural faults.

An example of at-speed test timing control is shown in Fig. 2, where test clocks TCK1 and TCK2 are staggered and generated by the clock-gating block shown in Fig. 1. In the capture window shown in Fig. 2, two capture pulses are generated for each clock domain. The last shift pulse and the first capture pulse (C1 or C3) are used to create transitions at the outputs of some scan cells; responses to the transitions are then caught by the second capture pulse (C2 or C4), where d2 and d4 are set based on functional clock frequencies. Thus, true at-speed testing is guaranteed to detect timing-related delay faults since no test clock frequency manipulation is conducted. 

Fig. 2  Timing Control Using Staggered Double-Capture.

Note that delays d1 and d5 in Fig. 2 can be adjusted as long as needed, so that one can use a global, slow scan enable (SE) signal to drive all clock domains. This significantly eases the physical implementation of the logic BIST scheme.

1.2     Inter-Clock-Domain Fault Detection

Inter-clock-domain fault detection is more complex, especially for timing-related delay faults. Fig. 3 shows four timing waveforms for detecting inter-clock-domain faults from the clock domain driven by TCK1 to the clock domain driven by TCK2.

Fig. 3  Inter-Clock-Domain Fault Test Timing

For testing structural faults, delay d is adjusted to be larger than the clock-skew between the two clock domains. This adjustment is easy. For detecting timing-related delay faults, delay d is further adjusted to satisfy the specified timing relation between the two clock domains. The waveform of Fig. 3 can achieve higher inter-clock-domain fault coverage since a pattern of higher randomness is applied and no fault effect caught in the immediate test response is masked out.

1.3     Capture Clock Generation

In order to generate an ordered sequence of double-capture clocks, the daisy-chain clock-triggering technique is used since it is more suitable for testing asynchronous designs. The daisy-chain clock-triggering technique means that the completion of the shift-in operation triggers the SE signal to become 0, switching operation mode from shift to capture. This in turn triggers the generation of two at-speed clock pulses for the first clock domain, the rising edge of the second capture clock pulse triggers the generation of two at-speed clock pulses for the second clock domain, and so on. Finally, the rising edge of second capture clock pulse for the last clock domain triggers the SE signal to become 1, switching operation mode from capture to shift. An example timing waveform is shown in Fig. 4.

Fig. 4  Daisy-Chain Clock-Triggering.

Conclusions

An at-speed logic BIST scheme based on SynTest patented technology is presented using an ordered sequence of capture clocks for testing designs containing multiple clock domains. The scheme employed is most suitable for testing of intra- as well as inter-clock-domain faults in asynchronous designs to achieve true at-speed test quality without any clock frequency manipulation. Physical implementation becomes easier due to the use of a low-speed scan enable (SE) signal and reduced timing-critical design requirements.

References

L.-T. Wang, X. Wen, P.-C. Hsu, S. Wu, and J. Guo, “At-Speed Logic BIST Architecture for Multi-Clock Designs,” Proc. ICCD-2005, pp. 475-478, October 2005.

B. Cheon, E. Lee, L.-T. Wang, X. Wen, P. Hsu, J. Cho, J. Park, H. Chao, and S. Wu, “At-Speed Logic BIST for IP Cores,” Proc. IEEE/ACM Design Automation, and Test in Europe, pp. 860-861, Munich, Germany, March 2005.

 By Luigi Ternullo

Built-In Self-Test (BIST) has taken a larger role in today’s overall test strategy over time and has come to satisfy a significant need by enabling the testability of embedded components in System-on- Chip (SoC) designs in a very efficient manner.  Without the capabilities enabled by BIST, test coverage and/or test time of several embedded components would be severely impacted.  The need to facilitate testability for embedded components has lead to the development of interface standards and vector formats such as 1149.1 and WGL respectively, which in turn has helped to foster the continued adoption of BIST in the over all test strategy of SoCs. 

Leveraging BIST as part of a test strategy can have many advantages. In addition to improved coverage of embedded components and reduced test time due to efficient patterns, some of the advantages of BIST also include reduced pin count interface, reduced vector size, quick pattern bring up at the ATE, and system speed testing when used in conjunction with one chip clock multipliers such as a Phase Lock Loop. BIST satisfies several needs required for manufacturing test of embedded components, but manufacturing test is not the only component in the overall test strategy of a product.  In an ideal world, a design using a BIST strategy will enter manufacturing, work the first time, and never have any yield or testability issues.  Realistically this is not the case. Silicon issues always seem to occur and the majority of these silicon issues seem to manifest themselves in embedded memories. Embedded memories could be referred to as the canary in the coal mine, or the weakest link in SoC designs, because memories are typically developed with very aggressive design rules.  If there is a problem with the air in a coal mine, the canary will be the first to be affected.  Likewise, when there is an issue in silicon that affects yield, it will more than likely manifest itself in one or more embedded memories. The fact that memories are designed with aggressive design rules make them the weakest link, but the very nature of a memory is to have a regular structure, which also makes it the easiest to isolate silicon issues. 

Most silicon related issues a